Six versus 12 months' adjuvant trastuzumab in patients with HER2-positive early breast cancer: the PERSEPHONE non-inferiority RCT

HERA

HERA was an international, multicentre, randomised trial that compared 1 year or 2 years of trastuzumab given every 3 weeks with observation in patients with HER2-positive and either node-negative or node-positive breast cancer who had completed locoregional therapy and at least four cycles of neoadjuvant or adjuvant chemotherapy. A total of 1694 patients were randomised to 2 years of trastuzumab, 1694 were randomised to 1 year of trastuzumab, and 1693 were in the observation arm with chemotherapy alone. Results were first reported in the New England Journal of Medicine10 for 1 year of trastuzumab compared with chemotherapy alone. With a median follow-up of 1 year, 347 disease-free survival (DFS) events (recurrence of breast cancer, contralateral breast cancer, second non-breast malignant disease, or death) were observed: 127 events in the trastuzumab group and 220 events in the observation group. The unadjusted hazard ratio (HR) for a DFS event in the trastuzumab group compared with the observation group was 0.54 [95% confidence interval (CI) 0.43 to 0.67; p < 0.0001], representing an absolute benefit in terms of DFS at 2 years of 8.4%. Symptomatic congestive heart failure (CHF), including severe CHF, developed in 1.7% of the women treated with trastuzumab.

Subsequent follow-up confirmed the beneficial effect. The 2-year analysis of the HERA trial demonstrated a benefit for overall survival (OS) as well as for DFS (HR for OS 0.66, 95% CI 0.47 to 0.91; p = 0.0115: updated HR for DFS 0.64, 95% CI 0.54 to 0.76; p < 0.0001). 11 These results confirmed that trastuzumab given sequentially after adjuvant chemotherapy has a significant benefit with a median follow-up of 2 years. The emergence of benefit after a median follow-up of only 1 year suggests that there is a significant effect from the first months of trastuzumab therapy. In addition, it is important to note that during the HERA trial trastuzumab was given sequentially after at least four cycles of standard chemotherapy, demonstrating the same degree of benefit as seen in the trials that used concurrent trastuzumab and chemotherapy.

Longer-term follow-up results of HERA have been published, confirming significant benefit for 12 months’ trastuzumab compared with observation at a median follow-up of 8 years. 12 This is despite nearly 50% of patients in the observation arm ‘crossing over’ to receive adjuvant trastuzumab after the early results were reported in 2005. There is no dispute about the significant benefit of adjuvant trastuzumab in HER2-positive breast cancer.

The HERA trial also tested 24 months’ trastuzumab compared with 12 months’ trastuzumab. The results for the 24-month arm showed no additional benefit in the 2013 publication after 8 years’ follow-up,12 or in the further publication in 2017 after 11 years’ follow-up. 13 Therefore, 12 months remained the standard of care throughout the world. In the accompanying editorial to the 2013 publication,14 Heikki Joensuu discussed the biology of HER2-positive breast cancer that leads to aggressive clinical behaviour and early cancer recurrence. He went on to say ‘The HERA results lend support to the hypothesis that patients with HER2 amplification do not benefit from long treatment durations with HER2-targeted therapy, but might be managed best with effective regimens of short duration’. 14 Interestingly, had the HERA trial been analysed and reported after only 3 years, a benefit for 2 years of trastuzumab would have been found. However, with longer follow-up,13 both 1- and 2-year DFS and OS are identical, and any earlier difference has disappeared.

NSABP B-31 and NCCTG N9831

The second paper published in the New England Journal of Medicine15 included the combined results of two US trials that compared adjuvant chemotherapy with and adjuvant chemotherapy without concurrent trastuzumab in women following surgery for HER2-positive breast cancer. The National Surgical Adjuvant Breast and Bowel Project trial (NSABP B-31) compared group 1 (doxorubicin and cyclophosphamide followed by paclitaxel either every 3 weeks or weekly) with group 2 (the same chemotherapy plus 52 weeks of trastuzumab beginning with the first dose of paclitaxel). The North Central Cancer Treatment Group (NCCTG) trial N9831 compared three regimens: group A (doxorubicin and cyclophosphamide followed by weekly paclitaxel), group B (the same chemotherapy followed by 52 weeks of trastuzumab after paclitaxel) and group C (the same chemotherapy plus 52 weeks of trastuzumab initiated concurrently with paclitaxel). The studies were amended to include a joint analysis comparing groups 1 and A (the control groups) with groups 2 and C (with trastuzumab given concurrently with taxanes). Group B was excluded from this first analysis because trastuzumab was not given concurrently with paclitaxel. The trial was reported when 394 DFS events had occurred, of which 133 were in the trastuzumab group and 261 were in the control group (HR 0.48; p < 0.0001). Three-year DFS was 87.1% in the trastuzumab group, compared with 75.4% in the control group, showing an absolute difference of nearly 12% in favour of trastuzumab. Trastuzumab therapy was associated with a 33% reduction in the risk of death (p = 0.015). The 3-year cumulative incidence of severe congestive heart failure or death from cardiac causes in the trastuzumab group was 4.1% in trial NSABP B-31 and 2.9% in NCCTG N9831.

Subsequently, longer-term follow-up16,17 confirmed the benefits of trastuzumab given for 12 months. At a median follow-up of 8.4 years, there was a 37% relative improvement in OS with the addition of trastuzumab (HR 0.63, 95% CI 0.54 to 0.73; p < 0.001) and an increase in the 10-year OS rate from 75.2% to 84%. DFS rates also improved by 40% (HR 0.60, 95% CI 0.53 to 0.68; p < 0.001), with an increase in the 10-year DFS rate from 62.2% to 73.7%. 17

Analysis of NCCTG N9831 on its own was able to address the comparison of the three arms. 16 The non-trastuzumab control arm was AC-paclitaxel (doxorubicin, cyclophosphamide and paclitaxel) chemotherapy, the second arm was AC-paclitaxel with sequential trastuzumab for 12 months, and the third arm was AC-paclitaxel with concurrent trastuzumab starting with paclitaxel and continuing afterwards to complete 12 months of weekly treatment. With a median follow-up of 6 years, the 5-year DFS rates were 80.1% and 84.4% for the second and third arms, respectively. There was an increase in DFS with concurrent trastuzumab and paclitaxel relative to sequential administration (HR 0.77, 99.9% CI 0.53 to 1.11), but the p-value of 0.02 did not cross the prespecified O’Brien–Fleming boundary (0.00116) required for this arm of the trial to be stopped early in the interim analysis. However, following the publication of these results in 2011, concurrent as well as sequential trastuzumab and chemotherapy were adopted as a ‘standard of care’ and, subsequently, concurrent treatment became the preferred option.

BCIRG-006

The Breast Cancer International Research Group (BCIRG) study of trastuzumab in women with HER2-expressing early breast cancer (BCIRG-006) examined concurrent treatment of trastuzumab with a non-anthracycline-containing chemotherapy regimen (docetaxel and either carboplatin or cisplatin). 18 In BCIRG-006, 3222 patients were recruited, and similar numbers of node-positive and high-risk node-negative women were randomised to each of three arms: adjuvant doxorubicin, cyclophosphamide and then docetaxel plus trastuzumab (AC-TH) or doxorubicin, cyclophosphamide and then docetaxel (AC-T), or docetaxel, carboplatin (or cisplatin) and trastuzumab (TCH). After a median follow-up of 65 months, there was significant benefit in DFS and OS in both groups treated with trastuzumab, compared with the group who received AC-T, who had a 5-year DFS of 75% and an OS of 87%. Among patients receiving AC-TH, the 5-year DFS was 84% (HR for the comparison with AC-T 0.64; p < 0.001) and OS was 92% (HR 0.63; p < 0.001). Among patients receiving TCH, the 5-year DFS was 81% (HR 0.75; p = 0.04) and OS was 91% (HR 0.77; p = 0.04). There was no difference between the two trastuzumab-containing arms, although the study was not sufficiently powered from a statistical point of view to confirm non-inferiority. There was a numerical trend for TCH showing a lower 5-year DFS (81% compared with 84% in the AC-TH arm), although there was significantly less cardiotoxicity among patients receiving TCH than among those receiving AC-TH.

FNCLCC-PACS-04

This was the only study that failed to demonstrate an advantage of adjuvant trastuzumab. 19 Five hundred and twenty-eight patients with operable node-positive breast cancer were randomised (1 : 1) to anthracycline-based chemotherapy with or without trastuzumab and, in a secondary randomisation, with or without docetaxel. The 3-year DFS rates were 78% (95% CI 72.3% to 82.5%) and 81% (95% CI 75.3% to 85.4%) in the observation and trastuzumab arms, respectively. There was a 14% reduction in the risk of relapse (HR 0.86, 95% CI 0.61 to 1.22; p = 0.41, log-rank stratified on pathologic node involvement), which was non-significant. All patients in this trial received sequential trastuzumab, and this result, together with the NCCTG N9831 analysis,16 which was the only trial directly comparing concurrent with sequential trastuzumab, resulted in a gradual shift in UK practice to increase the numbers of patients receiving trastuzumab concurrently with chemotherapy. This coincided with the increasing use of anthracycline-with-taxane chemotherapy combinations, which were approved by the National Institute for Health and Care Excellence (NICE) for the treatment of node-positive breast cancer during the course of the trial.

FinHer

Soon after the results of the registration trials, the FinHer trial was published. 20 This study compared docetaxel with vinorelbine for the adjuvant treatment of patients with early breast cancer. In this trial, women with HER2-positive tumours were also randomised to receive chemotherapy with or chemotherapy without concurrent trastuzumab. Two hundred and thirty-two women were recruited who had axillary-node-positive or high-risk node-negative HER2-positive cancer. They were randomised to receive three cycles of docetaxel or vinorelbine with or without concurrent weekly trastuzumab for 9 weeks, followed by three cycles of 5-fluorouracil, epirubicin and cyclophosphamide (FEC). In the subgroup of patients who had HER2-positive cancer, those who received trastuzumab had significantly better 3-year recurrence-free survival than those who did not (89% vs. 78%; HR for recurrence or death 0.42, 95% CI 0.21 to 0.83; p = 0.01). Docetaxel was associated with more adverse effects than was vinorelbine. Importantly, despite trastuzumab being given immediately before anthracyclines, this arm was not associated with reduced left ventricular ejection fraction (LVEF) or cardiac failure. The FinHer trial demonstrated that a short course of trastuzumab administered concurrently with docetaxel or vinorelbine was effective in women with HER2-positive breast cancer.

Subsequently, at a median follow-up of 62 months,21 the benefit of 9 weeks’ trastuzumab was maintained in the exploratory comparison of docetaxel and trastuzumab followed by FEC with the same treatment without trastuzumab (HR 0.32, 95% CI 0.12 to 0.89; p = 0.029). However, this was not the case in the comparison of vinorelbine and trastuzumab followed by FEC compared with the same treatment without trastuzumab (HR 0.92, 95% CI 0.47 to 1.83; p = 0.82).

Hence, when the PERSEPHONE trial started in 2007, whether or not similar outcomes could be achieved with durations of trastuzumab treatment of < 12 months was an important question worldwide.

E2198

This was a trial of the Eastern Cooperative Oncology Group (ECOG)22 that was set up to compare the cardiotoxicity of a short course of trastuzumab with that of the conventional 12 months of trastuzumab. This was not published until 2015 and therefore did not inform the design of the PERSEPHONE trial. In addition, it was not powered to demonstrate non-inferiority in the 12-week trastuzumab arm. The trial included 227 women with stage II or IIIa HER2-positive breast cancer. Patients were randomised to 12 weeks of paclitaxel and trastuzumab followed by four cycles of doxorubicin and cyclophosphamide or to the same treatment with a total of 1 year of trastuzumab (standard arm). The primary objective was to assess the safety of the different durations of trastuzumab therapy, in particular with regard to cardiac toxicity, which was defined as CHF or LVEF decrease of ≥ 10%. DFS and OS were secondary end points. The frequency or severity of cardiac toxicity did not increase; three patients in the experimental arm and four in the standard arm experienced CHF. There was no difference in 5-year DFS, which was 76% and 73% for the short and standard arms, respectively, with a HR of 1.3 (95% CI 0.8 to 2.1; p = 0.3). There was also no statistically significant difference in OS (HR 1.4; p = 0.3). The trial was not powered for efficacy; however, the longer duration of trastuzumab therapy did not demonstrate a signal for marked superiority. Retrospectively, this trial provided some additional evidence for the efficacy of shorter-duration trastuzumab.

PHARE

Based on the FinHer results, the French group Institut National du Cancer developed the PHARE clinical trial to compare 6 months’ trastuzumab with 12 months’ trastuzumab (Table 1). Randomisation into the PHARE trial occurred between 3 and 6 months after a patient had started trastuzumab, which was one of the differences between PHARE and PERSEPHONE. At the start of PERSEPHONE, all patients were randomised before starting trastuzumab, but subsequently randomisation was permitted at any time up to and including the ninth cycle of trastuzumab. In PHARE, patients were stratified by oestrogen receptor (ER) status (positive or negative) and concurrent or sequential trastuzumab and chemotherapy. It mapped on to standard practice in France and was carried out as a multicentre study recruiting 3384 patients: 1690 randomised to 6 months’ trastuzumab and 1690 randomised to 12 months’ trastuzumab. In the original statistical plan outlined in the first publication in 2013,23 a margin of non-inferiority of 2% was calculated on an assumed 2-year DFS of 85%. This was the 2-year DFS figure that had been predicted from the first results of the HERA trial. 10 The number of patients required was 7000 to have a 5% level of significance and 80% power to confirm non-inferiority of the experimental arm with an upper confidence limit below the HR, which was calculated at 1.15. In an international collaboration with the PHARE group, PERSEPHONE will carry out a pre-planned joint analysis of results, which will allow a non-inferiority comparison of 6 and 12 months’ trastuzumab with a margin of non-inferiority of 2%. This requires a total of at least 7000 patients to be entered into the two trials. The PHARE trial defined the HR of non-inferiority in the statistical analysis plan before the start of the trial, and then applied this when the analysis was carried out. This is an important difference between the statistical analysis of PHARE and that of PERSEPHONE. In the latter we defined non-inferiority as an absolute difference of no worse than 3% below the standard arm’s 4-year DFS rate, and the HR limit was to be calculated at the time of analysis using the real, observed DFS of the standard arm.

HORG

The Hellenic Oncology Research Group (HORG) study24 randomised HER2-positive patients between 6 and 12 months (see Table 1). Randomisation occurred prior to the commencement of chemotherapy, which was dose-dense FEC given every 2 weeks for four cycles with granulocyte-colony stimulating factor support, followed by docetaxel given every 2 weeks with granulocyte-colony stimulating factor. A total of 481 patients were randomised between trastuzumab for 6 or 12 months, and this was started concurrently with docetaxel chemotherapy. The non-inferiority margin was set with an absolute 8% boundary, and an estimated control arm 3-year DFS of 85%. The authors state that this produced a HR for the non-inferiority boundary of 1.53. With a type I error of 5% and 80% power, 239 patients were required to enrol in each arm within an accrual period of 3 years. The trial recruited 481 patients over nearly 8 years (between June 2004 and May 2012), which was longer than intended.

SOLD

The SOLD (Synergism or Long Duration) trial25 was co-ordinated from Finland and included international centres (see Table 1). This trial arose directly out of the FinHer trial and patients were randomised between standard 12 months’ trastuzumab and a shorter course of 9 weeks’ trastuzumab. All patients received single-agent docetaxel for three cycles concurrently with trastuzumab. This was followed by three cycles of FEC chemotherapy. Patients randomised to the standard arm completed 12 months’ trastuzumab following completion of chemotherapy, whereas patients randomised to 9 weeks’ trastuzumab did not receive more after chemotherapy was completed. Patients were randomised prior to any adjuvant treatment and all patients received the same chemotherapy; all therefore received trastuzumab concurrent with chemotherapy and upfront. Over nearly 7 years (January 2008 to December 2014), 2176 patients were recruited from seven countries. The SOLD trial was originally powered as a superiority trial, as it was anticipated that the standard 12 months’ treatment would lead to a significant increase in cardiac deaths, which was not the case. The original sample size calculation was for a trial of 3000 patients. The primary end point was changed to non-inferiority of DFS in the experimental arm, and a non-inferiority boundary of 4% on the true (observed) 5-year DFS rate of 88.7%. This resulted in a HR limit of 1.385, and the final sample size was 2168 patients (1084 patients in each group). The primary analysis was planned to be an event-driven analysis after 366 DFS events or when the last patient entered had been followed up for 2.0 years after randomisation, whichever occurred first.

Short-HER

Short-HER26 was an Italian government-sponsored trial with a similar design to the SOLD trial (see Table 1). The standard arm consisted of four cycles of anthracycline chemotherapy [doxorubicin and cyclophosphamide (AC) or epirubicin and cyclophosphamide (EC)] followed by four cycles of docetaxel with trastuzumab given 3-weekly, followed by completion of a full 12 months’ trastuzumab. The experimental arm was similar to the SOLD trial, with concurrent docetaxel and weekly trastuzumab for three cycles followed by anthracycline chemotherapy for three cycles (FEC) and then no further trastuzumab. This study was designed to assess if a shorter trastuzumab administration is non-inferior to the standard of 12 months with respect to DFS. Non-inferiority was defined before the trial started as an absolute difference of no worse than 3%, and the non-inferiority HR limit of < 1.29 was set. The sample size was estimated by setting an alpha of 5% and a power of 0.80, which resulted in a requirement for 372 events and 2332 patients. However, accrual of patients was slower than expected and, to comply with timelines set by the government funding body, the trial had to complete recruitment after 1256 patients had been enrolled. The data analysis was carried out after 198 events but, because of lower recruitment than planned, the study was underpowered for its DFS end point (statistical power 56%). As well as the standard frequentist approach, a Bayesian analysis was planned.

Primary end point

• To assess DFS non-inferiority of 6 months’ (nine cycles) compared with 12 months’ (18 cycles) trastuzumab in patients with HER2-positive early breast cancer.

Secondary end points

• To assess OS non-inferiority of 6 months’ (nine cycles) compared with 12 months’ (18 cycles) trastuzumab in patients with HER2-positive early breast cancer.

• To assess the expected incremental cost-effectiveness [cost per quality-adjusted life-year (QALY)] of 6 months (nine cycles) of trastuzumab compared with 12 months (18 cycles) of trastuzumab.

• Cardiac function as assessed by LVEF during trastuzumab therapy, and analysis of predictive factors of development of cardiac damage.

Secondary objectives: substudies

• Trans-PERSEPHONE: tumour blocks (paraffin-embedded) will be collected prospectively from patients in the study for molecular and candidate gene analysis as prognostic and predictive markers (separate protocol).

• Trans-PERSEPHONE-SNPs: blood samples will be collected prospectively from patients in the study for single nucleotide polymorphism analysis to research genetic/pharmacogenetic determinants of inherited susceptibility to HER2-positive breast cancer, prognosis and trastuzumab response and toxicity (separate protocol).

Sponsorship

Cambridge University Hospitals NHS Foundation Trust and the University of Cambridge are joint sponsors of PERSEPHONE.

Ethics, regulatory and research and development approvals

The trial was approved by the Medicines and Healthcare products Regulatory Agency (MHRA) on 10 May 2007 and received a favourable opinion from the North West – Haydock Research Ethics Committee (REC) (previously named North West REC and then North West 5 REC – Haydock Park) on 9 August 2007. Local research and development department approval was obtained at each participating NHS trust before patients were randomised. The trial was conducted in accordance with the principles and guidelines of the International Council for Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use,27 Good Clinical Practice (GCP),28 UK legislation,29,30 Cambridge and Warwick Clinical Trials Units Standard Operating Procedures and the REC and MHRA approved protocol. The current trial protocol is available online at the Warwick Clinical Trials Unit website. 31

Management of the trial

The Trial Management Group (TMG) was a multidisciplinary team of clinicians, statisticians, translational scientists and patient advocates who had considerable expertise in all aspects of trial design, conduct, safety, quality assurance and analysis. This group was in charge of running the trial. The TMG met regularly by teleconference to discuss site set-up, recruitment targets, safety and all matters pertaining to efficient conduct of the trial. After recruitment was completed, the focus of the TMG was data collection and case report form (CRF) completion.

The overall supervision of the trial was provided by an independent Trial Steering Committee (TSC) appointed by the National Institute for Health Research (NIHR) Evaluation, Trials and Studies Coordinating Centre, Health Technology Assessment (HTA) programme (the funder of the PERSEPHONE trial). The independent TSC consisted of an independent chairperson (an oncologist) and two other independent members (a breast surgeon and a statistician) and members of the TMG. The NIHR HTA programme and the independent TSC monitored the trial conduct and progress through regular reports, face-to-face meetings and teleconferences about the trial.

The independent monitoring of the trial was undertaken by the independent DSMC, which was established to advise the independent TSC if there was evidence that or a reason why the trial should be amended or terminated based on recruitment rates or safety and efficacy. Reports containing recruitment, protocol compliance, safety data and interim analyses of outcomes (which were confidential and not shared with the investigators) were presented regularly for review along with results from other relevant trials. At each review, the independent DSMC considered whether or not the trial should be stopped prematurely for ethical or safety reasons, including unexpected frequency or severity of toxicity, early indication of an inferior outcome in the experimental arm, unsatisfactory futility analyses or the publication of new data.

At the design stage of the PERSEPHONE trial, it was agreed with the PHARE TMG that one member of the independent DSMC would serve on both trials’ independent DSMCs, with their roles and responsibilities clearly defined for each committee. Details of the membership of the independent TSC, TMG and independent DSMC can be found in the Acknowledgements.

Trial site set-up

A total of 158 hospitals from NHS trusts and health boards in England, Scotland and Wales participated in PERSEPHONE and 152 sites recruited patients (96%); three were screening sites only and three were opened for the purposes of patient follow-up only. A list of all participating sites can be found in the Acknowledgements. Before a site was activated to recruitment, a trial initiation meeting was held, in person or via teleconference, to provide study-specific training to all staff members working on the trial. Continued support was offered to both existing and new staff at participating sites to ensure that they remained fully aware of the trial procedures and requirements.

Monitoring

Several levels of monitoring were applied through the trial.

Remote monitoring of sites was carried out on several occasions and especially after amendments to ensure that the investigator site file (containing the instructional materials and documentation required for the conduct of the trial) was well maintained.

Central monitoring of all data was carried out throughout the trial. Data were routinely cleaned, and queries were sent to sites if needed after (1) automatic validation checks during data entry, (2) manual checks (discrepancies between forms) or (3) annual data freeze to generate interim data-cleaning reports.

To ensure that sites were competent, triggered on-site monitoring was carried out occasionally if serious breaches and/or safety issues were reported.

Since 2010, the sponsor’s regulatory team has been monitoring Addenbrooke’s Hospital on an ongoing 6-monthly basis. Addenbrooke’s has been the highest recruiter throughout the trial. Monitoring involves source data verification for consent, eligibility criteria, LVEF and SAEs. It also includes reviewing the investigator site and pharmacy files. Concurrently, the trial master file held by the Cambridge co-ordinators is reviewed, with a particular focus given to any SAEs and non-compliances reported by all participating sites.

The Warwick trial management file of the PERSEPHONE trial was audited by the MHRA in 2012 and the Cambridge trial management file was reviewed by an independent auditor on behalf of the sponsor in 2015. No critical findings were identified following these reviews.

Route

Trastuzumab was given intravenously to all patients until 2013, when the subcutaneous formulation became available in NHS hospitals.

Following the approval of amendment 11 by the MHRA at the end of 2013, the PERSEPHONE trial allowed participating sites to use the subcutaneous formulation. Switching between intravenous and subcutaneous administration was at the discretion of the treating clinician.

Setting

Trastuzumab was prescribed neoadjuvantly (pre surgery) or adjuvantly (post surgery).

Trastuzumab was prescribed with chemotherapy (concurrently) or post chemotherapy (sequentially). The PERSEPHONE trial recruited patients in both settings.

Dose

Patients were randomised to:

• research arm – to receive nine cycles of trastuzumab (6-month arm)

• standard arm – to receive 18 cycles of trastuzumab (12-month arm).

Each cycle had a 3 week duration with a loading dose given on day 1. Treatment was expected to be administered as per standard practice.

Chemotherapy: commonly used regimens

Details of the chemotherapy regimens received by all patients (as per local institutional protocols) and whether trastuzumab was given concurrently with or sequentially to their chemotherapy were recorded in full, along with reasons for any early cessation of chemotherapy. Chemotherapy regimens were based on local protocols, which were informed by both licensing and NICE guidance.

Endocrine therapy

For women with ER-positive disease, systemic hormonal therapy is advised following completion of chemotherapy and definitive surgery. Concurrent hormone therapies could be administered with trastuzumab, although not with chemotherapy. The PERSEPHONE protocol stated that all endocrine therapy was at the discretion of the responsible clinician in accordance with standard local therapy protocols. However, the following guidelines were suggested: for women who remained pre-menopausal after completing chemotherapy, hormonal therapy options included ovarian suppression and tamoxifen. Entry into the Breast International Group EORTC trials SOFT (Suppression of Ovarian Function With Either Tamoxifen or Exemestane Compared With Tamoxifen Alone in Treating Premenopausal Women With Hormone-Responsive Breast Cancer Trial) or TEXT (Triptorelin With Either Exemestane or Tamoxifen in Treating Premenopausal Women With Hormone-Responsive Breast Cancer) was suggested. For postmenopausal women, tamoxifen or aromatase inhibitors could be used for a minimum of 5 years (or tamoxifen for 2–3 years, switching to an aromatase inhibitor after 2–3 years). Adjuvant hormonal treatment received by the patient was recorded.

Surgery

The PERSEPHONE protocol did not stipulate surgery as part of the trial. Surgery was carried out in accordance with standard practice at each site. Primary surgery was carried out in all patients receiving adjuvant chemotherapy and trastuzumab, mostly consisting of either wide local excision and axillary surgery (sentinel lymph node biopsy with or without axillary nodal dissection) or mastectomy and axillary surgery. Some patients in the trial received neoadjuvant treatment, and in these patients definitive surgery with wide local excision or mastectomy and axillary surgery took place after chemotherapy was completed.

Radiotherapy

The PERSEPHONE protocol stipulated that radiotherapy should be given after definitive surgery according to local protocols. Radiotherapy could be given concurrently with trastuzumab. All radiotherapy treatment received by the patient was recorded.

Inclusion criteria

Patients with the following characteristics were eligible to enter the trial:

• A histological diagnosis of invasive breast cancer.

• No evidence of metastatic disease.

• Known hormone receptor status.

• Overexpression of HER2 receptor. Bilateral breast cancers were eligible provided that one of the tumours overexpressed the HER2 receptor.

• Clear indication for neoadjuvant or adjuvant chemotherapy based on clinical and histopathological features.

• Fit to receive neoadjuvant or adjuvant chemotherapy and trastuzumab in the opinion of the responsible physician.

• No previous diagnosis of malignancy unless:

  • managed by surgical treatment only, and disease free for 10 years

  • previous basal cell carcinoma, cervical carcinoma in situ or ductal carcinoma in situ of the breast.

• Not pregnant and not lactating, with no intention of becoming pregnant during chemotherapy, and agreed to adopt adequate contraceptive measures if pre-menopausal and sexually active.

• No concurrent medical or psychiatric problems that might prevent completion of treatment or follow-up.

• Aged ≥ 18 years.

• Gave written informed consent for the study at any time before the 10th cycle of trastuzumab.

Exclusion criteria

Patients with the following characteristics were ineligible to enter the trial:

• significant concurrent cardiac disease or significant concurrent comorbidity in the opinion of the responsible physician adding to the risks associated with trastuzumab or cytotoxic chemotherapy

• unable to comply with protocol requirements

• received more than nine cycles of trastuzumab

• any other condition that, in the local investigator’s opinion, would make them unsuitable to participate in the trial.

Schedule of assessments

Table 2 shows the schedule of assessments for patients considered for entry into the PERSEPHONE trial prior to randomisation, as well as for patients during their treatment and follow-up. Most of the assessments for PERSEPHONE were mapped on to standard practice.

Toxicity

Based on the available data on the frequency of toxicities experienced with trastuzumab treatment, the 4000 patients recruited to PERSEPHONE were expected to adequately power the analysis of toxicity to allow detection of any clinically relevant differences between the treatment arms, if they existed.

A change in the eligibility criteria for PERSEPHONE was implemented on 11 September 2009 (after 316 patients had been randomised) to allow patients to be randomised before receiving their 10th cycle of trastuzumab. In the case of patients randomised into PERSEPHONE after the start of their trastuzumab treatment, we did not collect information on the toxicities or SAEs experienced during any treatment administered prior to randomisation. In the case of patients randomised before starting trastuzumab, we planned to collect all toxicities and SAEs experienced during their entire treatment.

Cardiac toxicity

Patients were assessed for symptoms or signs of congestive heart failure and information on new or altered cardiac medication was recorded during follow-up visits while on trastuzumab. Cardiac function was assessed using LVEF, either by ECHO or by MUGA scan depending on standard practice at each hospital. As per the recommendations from the trastuzumab SmPC, we requested scans for each patient at baseline and every 3 months for 12 months from the treatment start date. This was revised to standard practice (minimum of 4-monthly) in PERSEPHONE protocol version 4.0 on 31 October 2013.

Blood collection for pharmacogenetics/genetics study (Trans-PERSEPHONE-SNPs)

Trans-PERSEPHONE-SNPs aimed to collect two tubes of whole blood (2 × 9 ml) from patients, to be collected on one occasion at any time before, during or after treatment.

The purpose of collecting this was to build a bank of germline deoxyribonucleic acid (DNA) from breast cancer patients with HER2-positive disease and to carry out germline genome-wide sequencing for pharmacogenetics studies and prognostic and predictive candidate germline mutations. Analysis is planned of (1) germline single nucleotide polymorphisms (SNPs) predisposing to cardiac and other toxicity from trastuzumab, (2) germline SNPs predisposing to HER2-positive breast cancer, and (3) germline SNPs linked to outcomes from adjuvant trastuzumab and any interaction with duration of therapy.

Archival tissue blocks collection (Trans-PERSEPHONE)

Trans-PERSEPHONE is a collection of archival tissue (formalin-fixed, paraffin-embedded tissue) from PERSEPHONE patients. As per standard of care, following surgery, several formalin-fixed, paraffin-embedded blocks per patient were made at sites for diagnosis purposes. One or two representative block(s) were requested from the site’s pathology department. This bank of HER2-positive tissue is to be used to investigate prognostic and predictive molecular signatures of HER2-positive breast cancer and, in particular, to explore this with respect to treatment duration (comparing 6 and 12 months of treatment). Analysis of tumour and normal tissue will involve (1) tissue microarrays for IHC of protein gene products and in situ hybridisation analysis and (2) whole-genome profiling using expression and DNA microarrays.

First 100 patients receiving concomitant trastuzumab and chemotherapy

PERSEPHONE had a safety stage built into the trial to confirm that concurrent administration of chemotherapy and trastuzumab was safe within the trial. Data ‘in real time’ on SAEs and treatment delays for the first 100 patients of this cohort were collected and analysed before being discussed by the independent DSMC (see outcome of the safety phase in Chapter 3, First 100 patients receiving concomitant trastuzumab and chemotherapy).

Serious adverse events/reactions

The definition of a SAE for the purpose of the PERSEPHONE trial was any untoward medical occurrence that at any dose resulted in:

• death

• a life-threatening experience

• initial or prolongation of existing hospitalisation (Hospitalisation was defined as an inpatient admission, regardless of length of stay, even if the hospitalisation was a precautionary measure, for continued observation. Hospitalisation for a preexisting condition, including an elective procedure, that had not worsened did not constitute a SAE)

• persistent or significant disability/incapacity

• a congenital anomaly/birth defect

• a new primary malignancy.

• Any other event that was judged by the responsible investigator to warrant particular attention and for the purposes of the trial included symptomatic LVEF reduction.

Details of all SAEs were documented from the point of randomisation, and therefore not necessarily the start of the trastuzumab treatment, in the trial until 30 days from the last administration of trastuzumab.

SAEs occurring after a patient’s 30-day follow-up assessment were reported only if the investigator believed that the study drug or a protocol procedure may have caused the event.

Each principal investigator was advised to report SAEs using a specific SAE form within 24 hours of becoming aware of the event.

Assessments of all SAEs for expectedness and relatedness were made promptly by the chief investigator. A SAE was deemed to be a serious adverse reaction (SAR) when it was assessed as possibly, probably or likely to be related to trastuzumab.

Listing of SARs was submitted to the MHRA in accordance with national requirements on a yearly basis through development safety update reports/annual safety reports.

Suspected unexpected serious adverse reactions

When the nature and severity of the SAR was not consistent with the trial reference safety information, the event was unexpected and therefore classified as a suspected unexpected serious adverse reaction (SUSAR).

The reference safety information for the PERSEPHONE trial was the trastuzumab SmPC.

The PERSEPHONE Cambridge office reported SUSARs to the MHRA and REC within 7–15 days as per UK regulations.

Serious breaches

To ensure that appropriate action was taken to protect patients, maintain trial integrity and comply with legal requirements and any applicable regulatory guidance, protocol and GCP non-compliances were collected by the trial office before being assessed by the chief investigator and the sponsor.

Any departure from the protocol or regulatory requirements that was likely to affect to a significant degree either the safety of a trial patient or the scientific value of the trial was classified as a serious breach, queried, followed up and reported to the authorities (MHRA and REC) as per UK legislation.

Discontinuation of trial treatment

Patients’ trial treatment was discontinued in the following circumstances, which were reported on the withdrawal CRF:

• The patient opted to discontinue their randomised treatment arm or chose not to comply with the trial procedures.

• The patient was found to be ineligible (i.e. they had been randomised inadvertently without meeting the eligibility criteria).

• The investigator decided that the patient’s trial treatment should be discontinued because of toxicity.

• The patient did not recover from treatment-related toxicity to an extent that would have allowed further trastuzumab treatment.

• The patient had disease progression (radiologically confirmed) while on trial treatment.

• The patient became pregnant while receiving trial treatment and decided to continue her pregnancy.

• During trial treatment the patient relocated to a site that was not participating in the trial.

Follow-up data were collected for all patients who discontinued treatment. Treatment data were collected for patients who had withdrawn but continued to receive trastuzumab up to 12 months, unless the patient had withdrawn because of relapse.

Protocol amendment February 2009

In February 2009, version 2.0 of the protocol was submitted to the authorities. The alterations were to remove limitations of and offer more flexibility in the chemotherapy criteria and options available to patients, such as using the company Healthcare at Home Ltd to administer trastuzumab at patients’ domiciles.

Protocol amendment July 2009

In July 2009, a major amendment was made to the protocol to change randomisation to any point up to and including cycle 9 of the trastuzumab (6 months). Recruitment had been lower than expected and, after extensive consultation with the TMG, TSC, the PHARE group (French trial) and the funder, it was agreed that this was the best option to increase recruitment. As a result of this, changes to the health economic data collection were also made, streamlining it to the patient questionnaires. Detailed health resource use data were collected for the first 300–500 patients only, rather than for all 4000 patients. Mapping the trial on to standard practice also significantly reduced data collection for sites.

Protocol amendment July 2010

A new patient information sheet was designed and submitted to the authorities in July 2010 to improve patient acceptability. Not only was accrual less than expected, but < 60% of the patients consented for the translational substudies, which was lower than in other studies we have conducted. Both the TSC and the patient groups who were contacted for advice strongly believed that (1) the original patient information sheet was too long and complex and (2) when patients agreed to take part in the trial, the substudies should be incorporated into the study as a whole. The TMG, together with patient advisors, therefore developed a more patient-friendly patient information sheet that incorporated the two translational substudies.

Protocol amendment October 2013

The protocol was amended again in October 2013, mainly to reflect the changes to the environment of the trial:

• A full update on recently published data was added to the introduction of version 4.0 of the protocol.

• The long-term follow-up data of the HERA trial12 confirmed the significant benefit for 12 months’ trastuzumab compared with no trastuzumab control at median follow-up of 96 months (8 years) but also demonstrated that 24 months’ trastuzumab brought no additional benefit to 12 months’. Interestingly, had the trial been analysed and reported after only 36 months, a benefit of 24 months’ trastuzumab would have been found. However, with longer follow-up, both the 12- and the 24-month DFS and OS curves are identical.

• Despite a relatively short follow-up, the preliminary results of PHARE were published and showed that the trial had not proven non-inferiority after a median follow-up of 42 months. The PHARE results23 did show that, in a prospectively stratified analysis of subgroups, only patients with ER-negative disease receiving sequential chemotherapy and trastuzumab appeared to be significantly disadvantaged by receiving 6 months of trastuzumab. An examination of the mature DFS and OS curves from the HERA trial shows a clear separation of the 24-month and 12-month curves after 36 months, which resolves with more prolonged follow-up. This time point was the same as that at which PHARE had been analysed.

In addition, following scrutiny of the published PHARE data, the PERSEPHONE Independent Data Monitoring and Safety Committee confidentially examined the data in the PERSEPHONE trial. As a result of this, the committee advised the TMG and Trial Steering Group that there were no adverse signals in the PERSEPHONE trial data and that the trial be continued.

• The equivalence of the subcutaneous formulation of trastuzumab and the intravenous formulation was reported38 and therefore version 4.0 of the protocol stated that the use of the subcutaneous formulation was allowed.

• This low-risk trial was made flexible, with the aim of making it as pragmatic as possible (i.e. following standard practice to facilitate recruitment). In particular, cardiac function, which was initially assessed every 3 months for 12 months, was subsequently evaluated at intervals up to 4-monthly in accordance with a change in standard practice guidelines. 36

Protocol amendment December 2018

Protocol version 5.0 included changes to account for the European Union General Data Protection Regulation, revised trial timelines and permission to follow up patients using sites’ standard practice, which could include telephone and e-mail follow-up.

Non-inferiority design

The non-inferiority design of clinical trials was introduced to evaluate new treatment approaches, drugs, devices, biologics and other medical treatments to demonstrate whether or not a new treatment is a ‘good substitute’ (i.e. has similar efficacy to that of an established treatment). 39 Increasing in use by a factor of 6 over a 10-year period, this design is especially useful when evaluating whether a new treatment offers greater safety, reduced toxicity and reduced cost, together with confirmation that efficacy is not effectively compromised. It is the favoured design for trials in oncology that seek to de-escalate treatments by reducing either the duration or the intensity and therefore reduce toxicity, improve safety and improve patient experience. Cancer trials that have used this design include the IDEA meta-analysis,40 which established that 3 months rather than 6 months of an adjuvant capecitabine/oxaliplatin regimen for colorectal cancer is non-inferior and significantly less toxic. The non-inferiority of outcomes for head and neck cancer was demonstrated for positron emission tomographic scanning surveillance followed by radical dissection if demonstrated positron emission tomographic scanning positive or equivocal results, rather than routine radical neck dissection. 41 The TAILORx study used a non-inferiority design to compare adjuvant chemotherapy and endocrine therapy with endocrine therapy alone in women with ER-positive, HER2-negative, node-negative breast cancer and a moderate recurrence risk (score of 11–25) on a genomic test (OncotypeDx; Genomic Health, Inc., Redwood City, CA, USA). The trial demonstrated that endocrine therapy alone was non-inferior to adjuvant chemotherapy with endocrine therapy. 42 All of the trials examining reduced durations of trastuzumab in HER2-positive breast cancer have used non-inferiority designs. 23–26

Sample size and non-inferiority limit/margin

The PERSEPHONE trial was designed to allow the non-inferiority of the experimental arm (6 months’ trastuzumab) compared with the control arm (12 months’ trastuzumab) to be demonstrated in terms of the primary end point of DFS. The power calculations assumed that the DFS of the standard treatment of 12 months’ trastuzumab would be 80% at 4 years as this was the result available from the HERA trial. 10 The margin for non-inferiority was set as a 3% level, implying that the 4-year DFS of the experimental arm should not be below 77%, a difference equivalent to a HR of 1.17. On this basis, with 5% one-sided significance and 85% power, a trial randomising 4000 patients in total (2000 to each arm) would be able to prove the non-inferiority of the experimental arm.

The trial was expected to recruit for 4 years, with an additional follow-up period of 5 years, and allowed for loss to follow-up of up to 4%. The sample size was calculated by simulation assuming unadjusted analysis with a Cox’s proportional hazards model. The sample size also allowed for the 4-year DFS rate of the control arm to vary between 77% and 83%. The independent DSMC monitored the assumptions underlying the sample size calculation throughout the study.

Statistical methods

All randomised patients were included in all analyses where possible. Patients were analysed according to the treatment group to which they had been randomised on an intention-to-treat basis. Analyses were guided by the PERSEPHONE statistical analysis plan, which was prepared before data were available, and subsequently agreed by the PERSEPHONE independent DSMC.

Patient and tumour characteristics were presented to evaluate the comparability of the treatment arms and also the generalisability of the results to clinical settings. Categorical variables were presented using counts and percentages, and continuous variables were presented using either mean [standard deviation (SD)] or median [interquartile range (IQR)] depending on normality, and all were tabulated by treatment arm. A Consolidated Standards of Reporting Trials (CONSORT) flow diagram43 was also presented (see Chapter 3).

Primary end point: disease-free survival

The primary end point of DFS was measured for every patient from the date of diagnosis to date of first relapse (local or distant) or death. Patients who were disease free and either on follow-up or lost to follow-up were censored at the latest date at which they were known to be alive and disease free. Survival curves were plotted using the Kaplan–Meier method. Non-inferiority was defined as no worse than 3% below the control arm’s 4-year DFS rate. To test the non-inferiority of the experimental arm (i.e. 6 months’ trastuzumab), the HR was estimated using a Cox’s proportional hazards model containing only the trial treatment effect. If the 95th percentile of the estimated HR was less than the critical value, then the experimental arm (6 months’ trastuzumab) was regarded as non-inferior. Critical values were calculated for different scenarios with regard to the observed 4-year DFS on the control arm, as follows:

• With a 4-year DFS on the control arm of 80%, the critical value to be used was 1.1712853.

• With a 4-year DFS on the control arm of 85%, the critical value to be used was 1.2210943.

• With a 4-year DFS on the control arm of 88%, the critical value to be used was 1.2713341.

• With a 4-year DFS on the control arm of 90%, the critical value to be used was 1.3217671.

Proportionality of hazards was checked using an assessment of log–log plots.

A secondary analysis adjusting for stratification and baseline prognostic factors was planned. The treatment effect on DFS was also presented for stratification variables (ER status, chemotherapy type, chemotherapy timing and trastuzumab timing) using HR plots with interaction statistics using methods described by the Early Breast Cancer Trialists’ Collaborative Group in 1990. 44

Timing of primary analysis of primary end point

In 2015, the TMG and independent DSMC assessed the latest 4-year DFS rate for trial patients randomised to the control arm. This was found to be approximately 88%. The primary analysis of the primary end point for the 4000 patients was therefore planned to be undertaken when 500 DFS events had been observed, using a conditional power calculation. 45

Pre-planned early stopping guidelines

Initially, the PERSEPHONE protocol stated that, to control the overall alpha level, interim analyses of non-inferiority of the experimental arm with regard to the primary outcome would be reported to the independent DSMC using conservative tests, with significance determined by a p-value of 0.001. In 2015, the TMG and independent DSMC concluded that there was an additional requirement for prospective formal futility stopping rules.

The timings of the interim analyses for futility were configured on the 500 DFS events required for the primary analysis of DFS; these are shown in Table 3, including a recommended harm analysis46 at 25% of the total number of events.

Using a p-value of 0.01 (one sided), the HRs in the final column show the limits above which, with the corresponding number of events observed, the trial would have been stopped on the grounds of futility (i.e. concluding that non-inferiority could not reasonably be expected to be achieved).

The three interim analysis times specified above were reached in October 2012, June 2014 and March 2016. The independent DSMC considered the report presented to them and concluded that the data were immature and that there were no signals in the data that caused concern or would warrant a change in the study plan. The results found at these three time points did not cross the boundaries specified in Table 3.

The first of these time points (i.e. October 2012) coincided with the presentation of the PHARE47 and HERA 2-year data. 48

As the PERSEPHONE independent DSMC was confidentially examining the data in the PERSEPHONE trial, it advised the PERSEPHONE Trial Management Group and Trial Steering Committee that there were no adverse signals in these data, and recommended that the trial continue. On the strength of its advice, the NIHR HTA programme agreed a funding grant extension to complete recruitment of 4000 patients and to allow appropriate follow-up.

Overall survival

Overall survival (all-cause mortality) and a sensitivity analysis of cause-specific mortality were assessed using Kaplan–Meier curves and a Cox’s proportional hazards model containing only the trial treatment effect. OS across treatment arms was also assessed by stratification variables (ER status, chemotherapy type, chemotherapy timing and trastuzumab timing) using HR plots with interaction statistics. A landmark analysis of OS was also undertaken.

Trastuzumab treatment delivery

The total number of trastuzumab cycles administered per patient is shown graphically, separated by treatment arm. Counts and proportions for the reasons for early discontinuation of trastuzumab treatment as per protocol were documented. The total duration of trastuzumab treatment was calculated from the date of first infusion until 21 days after the date of the last cycle. Counts and proportions of the number of cycle delays/holds are presented, split by treatment arm, along with the stated reasons for them. Following an amendment to the protocol (see Protocol amendment October 2013), trastuzumab was allowed in the trial in its subcutaneous formulation at a fixed dose of 600 mg for each cycle without a loading dose. Method of administration for each trastuzumab cycle was recorded and is presented in tabular form (see Chapter 3).

Adherence to the PERSEPHONE protocol with respect to trastuzumab treatment was also assessed using a subgroup analysis including only cycles of trastuzumab administered to patients after they were randomised into the trial. Therefore, this analysis excluded any variation in trastuzumab treatment given before randomisation that adhered to local practice but was not in accordance with the PERSEPHONE protocol.

Toxicities

Initially, an analysis of all of the toxicity data received from all randomised patients was undertaken. For each toxicity in turn, each patient’s worst reported Common Terminology Criteria for Adverse Events (CTCAE) grade was identified and the frequencies were tabulated.

Following this, an analysis was undertaken on the population who were randomised upfront before any trastuzumab treatment to compare the randomisation arms in terms of the frequency and severity of toxicity through their first 6 months of treatment and, for patients randomised to 12 months’ trastuzumab, through their second 6 months of treatment. The frequency with which patients experienced each CTCAE grade was tabulated.

Serious adverse events

Frequencies of SAEs, SARs and SUSARs are presented by treatment arm. Frequencies of reasons for reporting, severity, causality and outcome are tabulated by treatment arm; a full breakdown of reported primary event categories of the SAEs and SARs is also given.

Patient review of trial information and trial conduct

At the start of PERSEPHONE, Maggie, who was the consumer representative in the National Cancer Research Institute Breast Cancer Clinical Studies Group, was co-opted to help with reviewing the patient materials and facilitating the patient focus groups during which the study was explained. These focus groups were instrumental in our revising the patient materials and streamlining the number of consent forms that the ethics committee had mandated (five forms reduced to two) owing to the translational substudies and quality-of-life data collected. Initially, patients were being lost to recruitment as the consent process was too long and the research team did not have time to approach patients. Maggie informally asked various consultants why they were not recruiting patients to this trial, and feedback highlighted the time needed for consent and the burden on the research team and patients when using the five consent forms. Maggie sent a letter to the ethics committee requesting that the consent forms be reduced from five to two and once this was agreed by the committee it greatly increased the numbers of patients approached and recruited.

Throughout the trial, the study team engaged with the charity Independent Cancer Patients’ Voice; Maggie is president of this charity, and it has reviewed patient materials including the patient leaflet and clinic posters. Independent Cancer Patients’ Voice was formed in 2009 by a group of cancer patients with a keen interest in research; it is a patient-led organisation that aims to bring the views and experiences of cancer patients and their families and carers into the cancer research community. Independent Cancer Patients’ Voice believes that clinical research is enhanced when patients are in partnership with health-care professionals, instead of passively receiving health care. 49 As a result of Maggie’s involvement with PERSEPHONE, Independent Cancer Patients’ Voice was invited to work with researchers at the University of Warwick and University of Cambridge; this has been invaluable at all stages of the trial.

Other local groups such as the consumer group at Guilford and the Camazons at Cambridge have contributed to focus groups on the barriers to recruitment. The patients endorsed the trial and the continued importance of the question being addressed.

Patient participation in dissemination

The PERSEPHONE results were first presented at the American Society of Clinical Oncology in 2018, which was followed by a Lancet publication that was e-mailed to all of the recruiting centres. Patients were asked as part of the consent process if they wanted to receive a copy of the results once these were available. A patient-friendly summary of the key findings was sent alongside the main paper. Maggie reviewed the one-page summary and simplified the content for easier understanding. Including patients in the dissemination of patient-approved materials is essential for wider dissemination. Finally, Maggie volunteered to be interviewed by the press and provided a quotation for the PERSEPHONE press release.

The final phase of the patient involvement will be a survey of patients to see what they think about the impact of these results and to endorse a recommended change in the guidelines.

Screening

Routine screening logs were requested from sites every 6 months for 6 years and then annually for the last 2 years of recruitment. Screening log returns were low; however, analysing the logs that were returned, 1848 out of the 7975 (23%) patients known to have been screened were recruited. The most common reasons that screened patients were not recruited were that they preferred to receive standard treatment (26%), they were ineligible (24.5%) or they declined to take part in the trial without giving a reason (16.5%) (see Appendix 1, Table 26).

Recruitment

Between 4 October 2007 and 31 July 2015, 4089 patients were randomised from 152 sites: 2045 to 12 months’ trastuzumab and 2044 to 6 months’ trastuzumab. Cumulative recruitment over time, compared with predicted accrual revised in October 2009, is shown in Figure 1. There was some slowing of recruitment in the second half of 2012, and it then remained steady at a reduced level. This was when the PHARE trial reported its early findings at the European Society for Medical Oncology,47 a result that failed to demonstrate non-inferiority. The presented data that were subsequently published23 were reviewed in detail by the independent DSMC alongside a confidential report of the PERSEPHONE data. The independent DSMC noted that the results from the PHARE trial had been reported with a median follow-up of 42.5 months and a primary end point of DFS at 2 years, and it interpreted the results as inconclusive. Following a confidential review of our trial data, the independent DSMC reported to the TMG and TSC that, taking into account the new data from PHARE, together with our own safety and efficacy data, there was no reason to stop the trial early, and it encouraged continuing to full recruitment of 4000 patients. All active sites were informed of the independent DSMC review and the advice to continue recruitment.

FIGURE 1.

Recruitment over time.

Consolidated Standards of Reporting Trials flow diagram

In total, 2045 patients were randomised to 12 months’ trastuzumab and 2044 patients were randomised to 6 months’ trastuzumab (Figure 2). One patient was double randomised and followed randomisation to 12 months, and this duplicate is the only randomised ‘patient’ excluded from the intention-to-treat analysis. Full treatment details are available for 1942 out of 2045 (95%) 12-month patients and for 1979 out of 2043 (97%) 6-month patients and therefore these are the patients included in the treatment analysis. The numbers of patients in each category of treatment are shown in the CONSORT flow diagram (see Figure 2). Notably, 82% of 12-month patients and 89% of 6-month patients received the number of cycles to which they had been randomised. In the 6-month arm, 5% (102/1979) received more than nine cycles of trastuzumab. Seventeen per cent (336/1942) of 12-month patients received fewer than the randomised 18 cycles, the most common reason being cardiovascular toxicity (146/336; 43%), followed by patient request (94/336; 28%) and relapse or death (26/336; 8%). Among 6-month patients, 6% (113/1979) received fewer than nine cycles, the most common reason being cardiovascular toxicity (60/113; 53%), followed by patient request (26/113; 23%) and relapse or death (12/113; 11%).

FIGURE 2.

The CONSORT flow diagram.

Recruitment by site and across treatment arms

The full recruitment by site is shown in Appendix 1, Table 27, with randomisation to 6 or 12 months, and the percentage of the total recruitment. This demonstrates the importance of recruiting from this large number of cancer units (n = 128) and cancer centres (n = 24) to achieve the high recruitment target. The majority of recruitment was from cancer units, with the research personnel supporting the trial provided initially by the National Cancer Research Network and then by the integrated newly developed Clinical Research Network in England; the NHS Research Networks in Scotland (funded by the Chief Scientist’s Office); and the Health and Care Research Office in Wales.

Withdrawals from trial treatment

Patients were withdrawn from trial treatment for reasons specified in the protocol, and this occurred for 346 (17%) 12-month patients and 223 (11%) 6-month patients. Among 12-month patients, the most common reason for withdrawal was toxicity (8% of patients randomised), followed by patient request (5%) and disease progression (2%) (Table 4). Among 6-month patients, the most common reason was patient request (6%, with 100 patients requesting a change to standard treatment of 12 months), followed by toxicity (3%), disease progression (< 1%) and ineligibility (< 1%). Follow-up data were collected from all withdrawn patients, except 106 patients (12-month patients, n = 65; 6-month patients, n = 41) who also withdrew their consent for further data collection.

Ineligible patients

Nineteen patients were deemed ineligible after randomisation (12-month patients, n = 7; 6-month patients, n = 12), principally for previous cancers or ductal carcinoma in situ treated with radiotherapy as well as surgery:

• 12-month patients –

  • four with previous cancer/ductal carcinoma in situ treated with surgery and radiotherapy

  • two who were HER2 negative

  • one with primary cancer confined to the axilla.

• 6-month patients –

  • seven with previous cancer/ductal carcinoma in situ treated with surgery and radiotherapy

  • two with metastatic disease

  • one who was HER2 negative

  • one who had received > 9 cycles of trastuzumab at time of randomisation

  • one who had been double randomised (excluded from analysis).

Follow-up data continued to be collected for all patients unless they withdrew from the study and requested not to be followed up. The remainder of this report comprises 4088 randomised patients: 2045 randomised to 12 months’ trastuzumab and 2043 randomised to 6 months’ trastuzumab.

Data return

Return rates of trial CRFs from sites to the Clinical Trials Unit were very high, with > 97% of all expected baseline and treatment CRFs returned and > 90% of all expected annual follow-up forms returned (see Appendix 1, Table 28).

Patient and tumour characteristics, split by randomised treatment arm

Table 5 shows patient and tumour characteristics split by whether patients were randomised to 12 months’ (n = 2045 patients) or 6 months’ (n = 2043 patients) trastuzumab. Patient and tumour characteristics were balanced across the two randomised treatment arms in terms of both minimisation variables and other prognostic factors. Forty-one per cent of patients received anthracycline-based chemotherapy, 10% received taxane-based chemotherapy without anthracyclines and 49% received anthracycline and taxane-based chemotherapy; therefore, 90% received anthracycline-containing chemotherapy. Sixty-nine per cent of patients had ER-positive tumours and 31% had ER-negative tumours. Eighty-five per cent of patients received adjuvant chemotherapy and 15% received neoadjuvant treatment. Of those patients receiving adjuvant chemotherapy and trastuzumab, 58% were node negative and 41% were node positive; 47% had tumours ≤ 2 cm and 50% had tumours > 2 cm. The profile of standard prognostic factors of ER status, nodal status and tumour size for the population included in the trial was considerably better than that for patients in the licensing trials (HERA, NSABP B-31 and NCCTG N9831, and BCIRG-006), and similar to that for patients in other trastuzumab duration trials (see Appendix 1, Table 29).

Patient and tumour characteristics split by adjuvant and neoadjuvant patients

Appendix 1, Table 30, shows the patient and tumour characteristics of all patients, split by whether they received adjuvant (n = 3462 patients) or neoadjuvant (n = 626 patients) chemotherapy. As can be seen, more neoadjuvant chemotherapy patients than adjuvant chemotherapy patients were ER negative (39% vs. 30%: p < 0.0001), received concurrent chemotherapy and trastuzumab (79% vs. 41%: p < 0.0001), and received anthracycline and taxane chemotherapy (89% vs. 41%). In addition, fewer neoadjuvant chemotherapy patients than adjuvant chemotherapy patients received anthracycline-based chemotherapy (9% vs. 48%) or taxane-based chemotherapy (2% vs. 11%: p < 0.0001 for differences in types of chemotherapy). The age profile for neoadjuvant chemotherapy patients was younger than that for adjuvant chemotherapy patients (p < 0.0001), and more neoadjuvant chemotherapy patients were pre-menopausal (39% vs. 26%; p < 0.0001). Follow-up was shorter for neoadjuvant chemotherapy patients (5.4 years, IQR 4.1–7.2 years) than for adjuvant chemotherapy patients (6.2 years, IQR 4.6–7.7 years).

Patient and tumour characteristics split by whether patients received concurrent or sequential trastuzumab and chemotherapy

Appendix 1, Table 31, shows the patient and tumour characteristics of all patients, split by whether they received chemotherapy and trastuzumab concurrently (n = 1900 patients) or sequentially (n = 2188 patients). As had been expected, 79% of concurrent patients received anthracycline and taxane-based chemotherapy compared with 22% of sequential patients, 3% of concurrent patients received anthracycline-based chemotherapy (E-CMF) compared with 75% of sequential patients, and 18% of concurrent patients received taxane-based chemotherapy compared with 3% of sequential patients (p < 0.0001). Patients given trastuzumab and chemotherapy concurrently compared with sequentially were more often node positive (53% and 32%, respectively; p < 0.0001), were more likely to have larger tumours (> 2 cm: 55% vs. 47%, respectively; p < 0.0001), were more likely to receive neoadjuvant treatment (26% vs. 6%, respectively; p < 0.0001) and, in addition, had shorter median follow-up (concurrent 5.3 years, IQR 4.2–6.6 years, vs. sequential 6.7 years, IQR 4.2–6.6 years).

Change in characteristics and standard practice over time throughout the trial

The characteristics of the patients being entered into PERSEPHONE and standard practice by the clinicians over the 8-year recruitment period from 4 October 2007 to 31 July 2015 were seen to change (Figure 3). The rate of ER-positive tumours increased from 62% of all patients randomised in 2008 to 74% of all patients randomised in 2015. The use of concurrent trastuzumab with chemotherapy increased from 24% in 2008 to 70% in 2015. The type of chemotherapy also changed. Anthracycline-based chemotherapy decreased from 63% in 2008 to 25% in 2015, taxane-based chemotherapy without anthracyclines increased from 7% to 18%, and anthracycline and taxane chemotherapy increased from 30% to 57%. In 2008 only 13% of patients received neoadjuvant treatment, but by 2015 this had increased to 20%. The PHARE trial presented early results in October 2012 and published in June 2013, including a subgroup analysis, which supported the use of concurrent chemotherapy particularly in ER-negative patients. The trends continued in the trial for more anthracycline and taxane-based chemotherapy to be used, for more concurrent rather than sequential treatment, and for more ER-positive patients to be included. Among the adjuvant patients the proportion who were node-negative increased from 53% at the start of the trial to 60% in the final year of recruitment, and tumour sizes decreased, with 46% having tumours of ≤ 2 cm at the start of the trial and this increasing to 58% in the final year of the study. The prognosis of patients entered into the study was good at the start and had improved further by the end of the study.

FIGURE 3.

Changing characteristics and treatments of patients randomised into PERSEPHONE: (a) stratification variables for all patients; and (b) prognostic variables for patients receiving adjuvant chemotherapy. CT, chemotherapy.

Length of time on treatment

The protocol-stated total duration of trastuzumab treatment is:

• 378 days for 12 months of trastuzumab (18 cycles each 21 days apart)

• 189 days for 6 months of trastuzumab (nine cycles each 21 days apart).

The median (IQR) time in days on treatment was:

• 383 days (378–392) for 12-month patients

• 190 days (189–197) for 6-month patients.

The good compliance with the protocol reflects the tolerability of each of the treatment arms.

Frequency and reasons for delays/holds of trastuzumab treatment

Of the 50,856 trastuzumab cycles received in total, 34,557 were administered during the first 6 months of treatment and 16,299 were administered during the second 6 months of treatment.

• In the first 6 months of treatment, delays/holds were reported in 2685 (8%) cycles, by 1769 (45% of the 3918) patients:

  • 1420 (8%) delays/holds reported by 920 (47%) 12-month patients

  • 1265 (7%) delays/holds reported by 849 (43%) 6-month patients.

• In the second 6 months of treatment, delays/holds were reported in 1203 (7%) cycles, by 802 (42% of the 1911) patients:

  • 1132 (7%) delays/holds reported by 753 12-month patients (42% of the 1809 who had ≥ 10 cycles)

  • 71 (6%) delays/holds reported by 49 6-month patients (48% of the 102 who had ≥ 10 cycles).

The 3968 reasons for the 3888 reported delays/holds throughout treatment are shown in Appendix 1, Table 34. Interestingly, delays/holds were much more likely to be a result of patients’ requests and logistic reasons than because of toxicity. Reasons recorded in the 12-month/6-month groups were, respectively, 23%/22% for patient holidays, 15%/15% for logistical and administrative reasons (rescheduling for bank holidays, change of clinic days, waiting list, poor intravenous access), 7%/5% by patient request for personal reasons and 5%/5% for booked surgery/radiotherapy/procedures (permanent line insertion). Only 5%/6% were due to cardiotoxicity from trastuzumab; 4%/6% were due to sepsis (probably chemotherapy related); 4%/3% were due to unrelated medical problems; and 2%/2% were due to awaiting cardiac function tests.

However, 7798 (15%) of the 50,856 trastuzumab cycles received were given prior to randomisation into PERSEPHONE. Of the 43,058 trastuzumab cycles received under the PERSEPHONE protocol by 3864 patients:

• delays/holds were reported in 3396 (8%) cycles by 1915 (50%) patients –

  • 2289 (8%) by 1170 (60%) 12-month patients

  • 1107 (8%) by 745 (39%) 6-month patients.

First 100 patients receiving concomitant trastuzumab and chemotherapy

The first 100 patients in PERSEPHONE who received concomitant chemotherapy and trastuzumab provided data ‘in real time’ on SAEs and treatment delays. These data were collected and analysed by the PERSEPHONE trials office and were reported to and discussed by the Independent Data and Safety Monitoring Committee on 15 December 2010. These were confirmed to be within the acceptable safety limits expected for concomitant trastuzumab and chemotherapy.

Serious adverse events

During the trial, 475 SAEs were reported by 384 patients:

• 291 by 227 12-month patients

• 184 by 157 6-month patients.

The most common reason for reporting a SAE was inpatient hospitalisation or prolongation of hospital stay (83%) (see Appendix 1, Table 36). There were nine patients for whom death was either the reason for reporting or the outcome of the SAE, and none of these was judged related to trastuzumab. One sudden death occurred 6 months after completion of trastuzumab, and autopsy revealed a pulmonary embolus with no evidence of metastatic disease. Eight deaths occurred during trastuzumab treatment and causes were as follows: one pulmonary embolus on chemotherapy; one pneumonia with metastatic liver and mediastinal disease; one subarachnoid haemorrhage; two pneumonitis and pulmonary infection related to taxane; one suicide by drug overdose; one death on chemotherapy (autopsy report stated that no cause found); and one patient second primary diagnosis glioblastoma multiforme.

In the 12-month arm the five most common SAEs reported were infection (118, 41% of SAEs reported), cardiac (38, 13%), pain (19, 7%), gastrointestinal (10, 3%) and neurology (14, 5%). In the 6-month arm, the five most common SAEs reported were infection (80, 43%), cardiac (15, 8%), gastrointestinal (12, 7%), pain (11, 6%) and neurology or pulmonary/upper respiratory (both 8, 4%) (see Appendix 1, Table 36). Only 103 (22%) of the SAEs were SARs (see Serious adverse reactions), as the cause of many SAEs was chemotherapy. The severity of the reported SAEs illustrate that only the minority were mild (19% of 12- and 6-month patients), and 5% of SAEs in the 12-month arm and 6% in the 6-month arm were fatal or life-threatening.

Serious adverse reactions

A total of 103 serious adverse reactions (SAEs that were deemed by the PERSEPHONE CI as having a causality either possibly, probably or definitely related to trastuzumab) were reported during the trial. Twice as many 12-month as 6-month patients reported a serious adverse reaction:

• 67 by 63 12-month patients

• 36 by 31 6-month patients.

The primary event categories reported for the 103 SARs are shown in Appendix 1, Table 37. There were no deaths related to trastuzumab. The five most common reasons for reporting a SAR were cardiac [30 (45%) in the 12-month arm, 15 (42%) in the 6-month arm], infection [12 (18%) and 9 (25%), respectively], pulmonary/upper respiratory [8 (12%) and 1 (3%), respectively], cardiac arrhythmia [5 (7%) and 3 (8%), respectively] and allergy/immunology [5 (7%) and 2 (6%), respectively], and these were all more common in 12-month patients.

Suspected unexpected serious adverse reactions

In total, since the trial opened, five SUSARs were reported among the 4088 patients recruited, although two were subsequently downgraded to a SAE:

• TNO (trial number) 209 – 39-year-old patient had a miscarriage. The patient became pregnant approximately during cycle 14–15 of trastuzumab and subsequently had a miscarriage around her 10th week of pregnancy. The miscarriage was reviewed as possibly related to trastuzumab.

• TNO 254 – 61-year-old patient presented with idiopathic pulmonary hypertension. The patient was still receiving treatment for this condition in 2018. The event was reviewed as possibly related to trastuzumab.

• TNO 662 – 61-year-old patient died suddenly at home following cycle 2 of trastuzumab/docetaxel and cyclophosphamide. The postmortem (macroscopic) showed no obvious cause of death. Pneumonia, pulmonary embolism and myocardial infarction were excluded. The toxicology analyses showed no drugs detected in the blood. The death was classified as unlikely to be related to trastuzumab and the event was downgraded to a SAE.

• TNO 1506 – 48-year-old patient was hospitalised with blistering of mouth and fever 10 days following the fifth cycle of trastuzumab. The event was reviewed as probably related to trastuzumab.

• TNO 3657 – mild visual symptoms not interfering with function. The event was classified as unlikely to be related to trastuzumab and the event was downgraded to a SAE.

Serious breaches

Six serious breaches were reported to the authorities during the trial. Three were directly linked to clinical incidents and the three others were related to misconduct by teams involved in participation or co-ordination. All breaches were escalated to the sponsor and subsequently reported to the MHRA and the REC. Each breach was investigated and followed up with a corrective and preventative action plan.

1. Patient TNO 858 was administered two cycles of trastuzumab despite having a LVEF measurement of 35%. The patient was admitted to hospital with chest pains. The error was picked up by the trial office 10 months later only after receiving a SAE form for the hospital admission. This error occurred as the principal investigator at the site had not been aware of the LVEF result until two further trastuzumab cycles had been administered, at which point treatment was held. Corrective and preventative actions were discussed with the site and a reminder was sent to all sites regarding LVEF measurements and trastuzumab treatment.

2. Owing to poor weather conditions, patient TNO 1970’s trastuzumab cycle was delayed by 4 days. This dose was manufactured by the company Healthcare at Home on the day of intended administration and had an expiry date of 24 hours later. During administration 4 days later, a research nurse noticed the expiry date and stopped treatment, by which point around 30–40 mg had been given to the patient. The PERSEPHONE trial team was alerted to this mistake only 15 days later. The principal investigator at the site reported no ill effects of this administration. Corrective and preventative actions were discussed with the company, and also with the site, with regard to communication.

3. Patient TNO 2968 had been confused with another patient taking part in another trial. The PERSEPHONE patient had potentially been administered pertuzumab/placebo. The site underwent rigorous monitoring by the PERSEPHONE team and corrective and preventative actions were discussed with staff at the site.

4. An audit review at a participating site uncovered a number of issues related to patient consent. Six patients randomised to the trial were consented by a research nurse who had not been authorised to undertake this responsibility as per protocol. Five additional patient consent forms did not adhere to GCP guidelines (i.e. the dates had been changed and the consent forms had been posted to the patient’s address to be signed). Some issues were also found with co-investigators who had taken consent from patients but were not GCP trained or authorised to take consent as per the site signature and delegation log. Following the event, the site received a monitoring visit to ensure that the corrective and preventative actions were implemented.

5. During the development safety update report 2016 development, the sponsor identified that several updates to the trastuzumab SmPC had been made during the trial. Although important treatment alert letters were sent to all clinical investigators at the time of the changes, these were not amended in the PERSEPHONE protocol and patient information sheet. Corrective and preventative actions were discussed between the sponsor and the trial office. By this time the trial had completed recruitment (completed July 2015), and after discussion it was not felt that an amended patient information sheet would be helpful. All sites who had remaining patients in the trial were contacted with reminders that all patients who were pre-menopausal and could become pregnant were required to take effective contraceptive precautions during trastuzumab treatment and up to and including 7 months after completion. In addition, all sites were asked to identify any other patients who had become pregnant.

6. After her tissue was re-tested during a routine audit of the hospital laboratory, patient TNO 4077 was found to be HER2 negative. The patient was withdrawn from the trial at this point and followed up as per local practice. A protocol non-compliance was not generated after notification of the event and the sponsor was not informed as it was felt that the initial laboratory error occurred outside the trial protocol. The sponsor was made aware of the event when the patient sought legal action against the hospital. By that time, the hospital had already conducted a root cause analysis, and corrective and preventative actions had been taken.

During the trial, several updates were made to the trastuzumab SmPC, but these were not implemented in the PERSEPHONE protocol and patient information sheet. Specifically, the two following major patient safety updates were made:

• continued use of contraception for 6 months post treatment – SmPC updated August 2010 (up to 7 months in 2016)

• monitoring of long-term cardiac effects of trastuzumab toxicity – SmPC updated May 2015.

Although the PERSEPHONE trial protocol and patient information sheet were not amended to include these issues, the team was aware of the changes and sent out important treatment alert letters to all clinical investigators at the time of the changes in both 2010 and 2015.

Deaths

At database lock on 6 March 2019, 389 deaths (10% of the 4088 patients) had been reported [182 (9%) 12-month patients and 207 (10%) 6-month patients]. Median follow-up of the 3699 alive patients was 6.1 years (IQR 4.5–7.6 years), with 97% of alive patients followed up for at least 2 years.

The reported causes of death (multiple for some patients) illustrate that 77% of deaths were from breast cancer (Table 9). Among the 88 patients who died without breast cancer reported as a cause, two had reported a local relapse and eight had reported a distant relapse.

Relapses and second primaries

In total, 482 (12%) patients reported a relapse (12-month patients, n = 229; 6-month patients, n = 253) (Table 10). One hundred and sixty-four (4%) patients reported a local relapse [12-month patients, n = 81 (4%); 6-month patients, n = 83 (4%)]. This was most commonly reported in the ipsilateral breast/chest wall (n = 92), ipsilateral axillary nodes (n = 28) and ipsilateral supraclavicular nodes (n = 19) (see Appendix 1, Table 38). Three hundred and ninety-nine (10%) patients reported a distant relapse [12-month patients, n = 192 (9%); 6-month patients, n = 207 (10%)]. This occurred most commonly in the liver [86/192 patients (45%) in the 12-month arm and 91/207 patients (44%) in the 6-month arm], bone [63 (33%) in the 12-month arm and 89 (43%) in the 6-month arm], lung/pleura [74 (39%) in the 12-month arm and 78 (38%) in the 6-month arm] and brain [40 (21%) in the 12-month arm and 43 (21%) in the 6-month arm]. One hundred and thirty-four (3%) patients reported a second primary [12-month patients, n = 67 (3%); 6-month patients, n = 67 (3%)]. This was most commonly reported in the contralateral breast [n = 36 (27%): 12-month patients, n = 23 (34%); 6-month patients, n = 13 (19%)], lung [n = 17 (13%): 12-month patients, n = 9 (13.5%); 6-month patients, n = 8 (12%)] and bowel/colon [n = 15 (11%): 12-month patients, n = 4 (6%); 6-month patients, n = 11 (16.5%)].

Disease-free survival

At a median follow-up of 6.1 years, 566 (14%) patients either had experienced a relapse or had died. The 4-year DFS rate was 90.3% (95% CI 88.9% to 91.5%) in the 12-month group and 89.5% (95% CI 88.1% to 90.8%) in the 6-month group (Figure 6). Thus, with the non-inferiority margin of 3%, the non-inferiority limit for the HR was set at 1.33. The HR for relapse or death with 6 months’ compared with 12 months’ trastuzumab was 1.10 (90% CI 0.96 to 1.26). This outcome met the prespecified definition of non-inferiority (non-inferiority p = 0.01). The two-sided p-value for the difference between treatments was 0.26. Adjustment for all stratification factors gave the same results, with a HR of 1.10 (90% CI 0.96 to 1.26; non-inferiority p = 0.01).

FIGURE 6.

Disease-free survival according to randomised group: (a) Kaplan–Meier plot; and (b) HR plot.

With a difference of 0.008 in absolute risk of a DFS event at 4 years between patients randomised to 6 months and those randomised to 12 months of trastuzumab, the number needed to treat is 125 for the overall population. That is, on average, 125 patients would need to receive 12 months’ trastuzumab for one of them to benefit in terms of a DFS event, compared with the same 125 patients receiving only 6 months of trastuzumab.

Overall survival

Analysis of the secondary end point of OS also met the prespecified definition of non-inferiority. The 4-year OS rate was 94.9% (95% CI 93.9% to 95.8%) in the 12-month group and 94.2% (95% CI 93.0% to 95.1%) in the 6-month group (Figure 7). With the same non-inferiority margin of 3%, the non-inferiority limit for the HR was set at 1.61. The HR for death with 6 months’ trastuzumab compared with 12 months’ trastuzumab was 1.14 (90% CI 0.96 to 1.34), meeting the prespecified definition of non-inferiority (non-inferiority p = 0.0003). The two-sided p-value for the difference between treatments was 0.21, and adjusting for stratification factors gave similar results (HR 1.13, 90% CI 0.95 to 1.33; non-inferiority p = 0.0002).

FIGURE 7.

Overall survival according to randomised group: (a) Kaplan–Meier plot; and (b) HR plot.

With the difference of 0.007 in absolute risk of an OS event at 4 years between patients randomised to 6 months and those randomised to 12 months of trastuzumab, the number needed to treat is 143 for the overall population. That is, on average, 143 patients would need to receive 12 months’ trastuzumab for one of them to benefit in terms of an OS event, compared with the same 143 patients receiving only 6 months’ trastuzumab.

Landmark analysis of disease-free survival and overall survival

The exploratory landmark analysis, including only patients who were alive and disease free at least 6 months after the start of their trastuzumab treatment, was undertaken on 4008 patients (12-month patients, n = 2007; 6-month patients, n = 2001). The cohort was balanced across randomised arms in terms of demographics and baseline disease characteristics (see Appendix 1, Table 39) and, with a median follow-up of 5.1 years, 365 (9%) deaths and 537 (13%) DFS events have been recorded. DFS and OS were recalculated from the landmark time point (6 months after each patient’s start of trastuzumab treatment).

The landmark 4-year DFS rate was 88.7% (95% CI 87.2% to 90.1%) in the 12-month group and 88.4% (95% CI 86.8% to 89.7%) in the 6-month group, and hence the non-inferiority limit for the HR was set at 1.29 (Figure 8). The calculated HR for relapse or death with 6 months’ trastuzumab compared with 12 months’ trastuzumab was 1.09 (90% CI 0.95 to 1.26), meeting the prespecified definition of non-inferiority (non-inferiority p = 0.03). The two-sided p-value for difference between treatments was 0.30.

FIGURE 8.

Landmark DFS according to randomised group: (a) Kaplan–Meier plot; and (b) HR plot.

The landmark 4-year OS rate was 93.2% (95% CI 91.9% to 94.3%) in the 12-month group and 92.6% (95% CI 91.3% to 93.7%) in the 6-month group, and hence the non-inferiority limit for the HR was set to 1.46 (Figure 9). The calculated HR for death with 6 months’ trastuzumab compared with 12 months’ trastuzumab was 1.12 (90% CI 0.95 to 1.34), meeting the prespecified definition of non-inferiority (non-inferiority p = 0.006). The two-sided p-value for difference between treatments was 0.27.

FIGURE 9.

Landmark OS according to randomised group: (a) Kaplan–Meier plot; and (b) HR plot.

Additional analyses

To ensure the robustness of findings from the PERSEPHONE trial, and to facilitate comparisons across various other studies, additional analyses of invasive DFS [including invasive contralateral breast cancers and second primary invasive cancers (non-breast) according to the STEEP (standardised definitions for efficacy end points in adjuvant breast cancer trials) system], distant DFS, and breast cancer-specific survival were undertaken. Survival curves of these end points were comparable with those of the protocol-specified primary and secondary end points (see Appendix 2, Figures 25–27).

All patients

Forest plots for DFS including all patients (Figure 10) showed heterogeneity in the treatment effect of chemotherapy type (p = 0.01), predominantly driven by the small number of events in the taxane-only group, in which 56% of patients received docetaxel with cyclophosphamide. The timing of trastuzumab relative to chemotherapy (concurrent or sequential) showed heterogeneity in the treatment effect (p < 0.001) favouring 12-month patients receiving concurrent trastuzumab. In terms of DFS, no heterogeneity in the treatment effect was observed for ER status, chemotherapy timing (neoadjuvant or adjuvant), age, grade, menopausal status or IHC status (IHC3+ or IHC2+ and FISH positive). In terms of OS, similar results were found with heterogeneity for concurrent/sequential chemotherapy and trastuzumab, although the heterogeneity in treatment effect was not observed for chemotherapy type (p = 0.11) (see Appendix 2, Figure 28). Forest plots for landmark DFS show heterogeneity for chemotherapy type (p = 0.01), chemotherapy timing (adjuvant/neoadjuvant, p = 0.04) and concurrent/sequential treatment (p = 0.003) (see Appendix 2, Figure 29). Landmark OS showed heterogeneity only for concurrent/sequential trastuzumab and chemotherapy (p = 0.02) (see Appendix 2, Figure 30).

FIGURE 10.

Forest plot of DFS for all patients. CT, chemotherapy; O–E, observed–expected.

Adjuvant patients

For the subset of patients receiving adjuvant chemotherapy, exploratory forest plots of the randomised treatment effect on DFS (Figure 11) showed heterogeneity for chemotherapy type (p = 0.006) and trastuzumab timing (p = 0.003) and on OS (see Appendix 2, Figure 31) heterogeneity for trastuzumab timing (p = 0.02). There was no additional heterogeneity in the treatment effect for nodal status at surgery, tumour size, or combined ER and nodal status.

FIGURE 11.

Forest plot of DFS for adjuvant patients only. CT, chemotherapy; O–E, observed–expected.

Neoadjuvant patients

For the subset of patients receiving neoadjuvant chemotherapy, exploratory forest plots of the randomised treatment effect on DFS (Figure 12) and OS (see Appendix 2, Figure 32) found no heterogeneity in the treatment effect within the stratification variables. In addition, there was no heterogeneity in the treatment effect for response to neoadjuvant chemotherapy and trastuzumab (pathological complete response vs. non-pathological complete response). Thirty per cent (186/614) of patients achieved pathological complete response (i.e. they had no cancer in breast or axillary lymph nodes) after neoadjuvant chemotherapy and trastuzumab, and to date DFS events have been seen at follow-up in only 6.4% of these patients (12/186), compared with 26% (113/428) of patients who failed to achieve pathological complete response. This is the expected result, as pathological complete response has been shown to predict better outcomes. The numbers in this group are too small to indicate any interaction with treatment duration.

FIGURE 12.

Forest plot of DFS for neoadjuvant patients only. CT, chemotherapy; O–E, observed–expected; pCR, pathological complete response.

PERSEPHONE exclusion criteria for cardiac problems

The protocol detailed that patients should have no significant cardiac disease or comorbidity that, in the opinion of the principal investigator, added to the cardiac risks associated with trastuzumab and chemotherapy. PERSEPHONE was a trial designed to ‘map on to standard practice’ and therefore did not exclude any patients who, in the opinion of the principal investigators at recruiting sites, would otherwise be considered fit enough to receive chemotherapy and the standard of 12 months’ adjuvant trastuzumab.

Definition of clinical cardiac dysfunction

‘Clinical cardiac dysfunction’ was defined as any or all of the following: symptoms of cardiac disease or signs of CHF or new/altered cardiac medication prescribed during the 12 months after starting trastuzumab. Following randomisation, side effects specific to trastuzumab were reported in the CRFs with the aid of patient-recorded diary sheets. Every 3 months, symptoms or signs of CHF and new/altered cardiac medication were reported on the CRFs to the trials office.

Left ventricular ejection fraction

Treatment modifications

If LVEF fell below 50%, the protocol advised ‘holding’ trastuzumab and repeating LVEF monitoring at 6-week intervals until LVEF returned to ≥ 50%. After delays for a total of 3 months on account of cardiotoxicity, the protocol advised to discontinue trastuzumab permanently. In the case of NYHA class III/IV heart failure symptoms (breathlessness at rest or on minimal exertion), trastuzumab was permanently discontinued even after resolution of symptoms or normalisation of LVEF with treatment. 65,66 In the later part of the study, standard practice for patients on trastuzumab changed in many sites following the publication of new guidelines. 36 Patients on trastuzumab who had an asymptomatic reduction in LVEF to below the lower limit of normal and ≥ 40% could be treated with cardiac medication while continuing trastuzumab and having their LVEF monitored. Changes to the protocol were not made and therefore these instances were reported as protocol non-compliances.

Chemotherapy, timing of randomisation and prior cardiac medication

Factors relevant to cardiac risk appeared balanced across randomised treatment arms; 90% of patients (12-month patients, 90%; 6-month patients, 90%) received an anthracycline as part of either anthracycline-based (12-month patients, 41%; 6-month patients, 41%) or anthracycline plus taxane-based (12-month patients, 49%; 6-month patients, 49%) chemotherapy (see Table 5). Fifty-four per cent of patients received trastuzumab sequentially after chemotherapy (12-month patients, 54%; 6-month patients, 53%), and 46% received trastuzumab concurrently with the non-anthracycline component of chemotherapy (12-month patients, 46%; 6-month patients, 47%). Only 2% of patients reported being on cardiac medication prior to starting trastuzumab (12-month patients, 2%; 6-month patients, 3%). Randomisation was before the start of trastuzumab or at any time up to and including the ninth cycle. Forty-four per cent (1782/4088) of patients were randomised upfront and 81% (3320/4088) were randomised prior to cycle 5, and the timing of randomisation was balanced between the 6- and 12-month arms (see Table 6).

Cardiac events

Full clinical cardiac data were available for 3995 patients (98% of 4088 randomised): 1987 (97% of 2045 randomised) 12-month patients and 2008 (98% of 2043 randomised) 6-month patients. Clinical cardiac dysfunction was reported in 384 (10% of 3995) patients (Table 11), with 188 (5% of 3995) reporting symptoms of cardiac disease, 65 (2% of 3995) having signs of CHF, and 274 (7%) reporting new or altered cardiac medication (see Appendix 1, Table 40). Clinical cardiac dysfunction was more common in 12-month than in 6-month patients [228/1987 (11%) vs. 156/2008 (8%), respectively; p < 0.0001]. Some differences in rates were apparent in the first 6 months of trastuzumab [168/1987 (8%) 12-month patients vs. 127/2008 (6%) 6-month patients; p = 0.01], which we would assume a chance finding as we would expect these to be similar at this point. A more significant difference between the two arms emerged during the 7- to 12-month period [158/1953 (8%) vs. 97/1914 (5%), respectively; p = 0.0002], as expected.

In total, 19,458 LVEF measurements (ECHO, 73%; MUGA, 17%; unknown, 10%) were reported in 4080 patients: 10,193 in 2042 12-month patients and 9265 in 2038 6-month patients. The use of each modality was balanced across the randomised treatment arms (12-month patients: ECHO, 73%, MUGA, 18%, unknown, 9%; 6-month patients: ECHO, 73%, MUGA, 17%, unknown, 10%). MUGA scans appear to have a higher rate of low LVEF readings (< 50%, or percentage unknown but classified on report as abnormal): 10% of MUGA scans compared with 3% of ECHO (see Appendix 1, Table 41). During the first 6 months of treatment, the proportion of patients with low LVEF was 7% in both groups, 149 out of 2042 (7%) in the 12-month arm and 145 out of 2038 (7%) in the 6-month arm (p = 0.87). However, during months 7–12, this proportion increased for 12-month patients (150/1942; 8%) but fell for 6-month patients (84/1758; 5%) (p = 0.0003 for difference between groups). Patients with substantial decreases in LVEF, defined as absolute decrease in LVEF of ≥ 10% from baseline to < 50%, were seen in months 1–6 of treatment in similar numbers: 99 out of 1955 (5%) patients in the 12-month arm and 101 out of 1957 (5%) in the 6-month arm. However, during months 7–12, there were more patients with substantial decreases in the 12-month arm than in the 6-month arm [102/1880 (5%) and 61/1701 (4%), respectively; p = 0.01]. Substantial falls in LVEF, defined as decreasing to < 50% after a baseline of ≥ 59%, occurred in the first 6 months in similar numbers of patients: in 64 out of 1955 (3%) 12-month patients and in 70 out of 1957 (4%) 6-month patients. However, in months 7–12 these figures were 71 out of 1880 (4%) and 33 out of 1701 (2%), respectively (p = 0.0015).

A landmark analysis was carried out on 3415 patients who had not reported any cardiac dysfunction (defined as clinical cardiac dysfunction or low LVEF) within their first 6 months of treatment and for whom data were available in months 7–12. Significantly more 12-month patients (97/1714; 6%) than 6-month patients (39/1701; 2%) reported cardiac dysfunction in the second 6-month period (p < 0.0001).

Eleven deaths were reported to have had a ‘cardiac’ cause, either first cause or contributory cause (see Appendix 1, Table 42). None occurred during the first 12 months after starting trastuzumab treatment. Nine patients died with no metastatic disease and two had metastatic disease. In all cases, the TMG and the cardiologist who reviewed the cases judged trastuzumab as unrelated or unlikely to have been related to cardiac problems. Four deaths were caused by ischaemic heart disease, which has no known association with trastuzumab. One patient developed a decrease in LVEF with CHF; this was very unlikely to have been related to trastuzumab, as the patient received only four cycles, treatment was stopped because of reduced LVEF, and left ventricular function then recovered > 3 months after treatment had stopped. The patient was then diagnosed with chronic obstructive pulmonary disease, which accounted for her breathlessness, and died of chronic obstructive pulmonary disease 78 months after breast cancer diagnosis. There was one case of cardiac amyloid and one of cardiac sarcoid; neither was judged to be related to trastuzumab. One patient was reported to have right ventricular arrhythmic cardiomyopathy, which has a known genetic basis (mostly desmosomal genes) with physiological triggers (e.g. extreme endurance sports), and therefore this was judged unlikely to be related to trastuzumab. One patient had a low LVEF of 48% recorded after three cycles of trastuzumab; the trastuzumab was stopped, and the patient was diagnosed with symptomatic CHF and started cardiac medication. Her LVEF normalised, and she remained on preventative cardiac medication. She had received 300 mg/m² epirubicin. Her death was due to metastatic breast cancer. One patient had a reduced LVEF during trastuzumab treatment, which was discontinued after four cycles when LVEF was 30%. The patient was started on cardiac medication and further investigations revealed that she had coronary artery disease. After she stopped trastuzumab, her LVEF recovered to normal, at 50%. The cause of her death 6 years later was recorded as metastatic breast cancer, with a secondary cause of death of cardiomyopathy. However, no supporting evidence was provided for cardiomyopathy (no autopsy). Presence of cardiomyopathy was felt unlikely to be related to trastuzumab and most likely to be related to documented coronary artery disease or otherwise anthracyclines (300 mg/m² of epirubicin).

Trastuzumab modifications

Complete trastuzumab treatment data are available for 3921 (96%) of the 4088 patients (see Figure 2). Delays in trastuzumab cycles were caused by cardiotoxicity in 5% of the 12-month patients (fifth most common reason) and 6% of the 6-month patients (third most common reason) (see Appendix 1, Table 34). However, cardiotoxicity was the most common reason reported for trastuzumab treatment being stopped early [206 (46%) out of 449 patients who received fewer cycles than randomised to receive: (146; 43%) of 336 12-month patients stopping early and 60 (53%) of 113 6-month patients stopping early] (see Figure 4). Most commonly, patients’ trastuzumab treatment was stopped early because of cardiotoxicity (i.e. cardiotoxicity was the reason in over half of cases during that cycle) in cycles 4–8 for both 6- and 12-month patients, and also in cycles 10 and 13 for 12-month patients (Figure 14). Cardiac monitoring with LVEF commenced after four cycles of trastuzumab, and this corresponded to the time point at which we observed an increase in the number of patients stopping trastuzumab early because of cardiac toxicity.

FIGURE 14.

Total number of trastuzumab cycles received, indicating the numbers of patients stopping because of cardiotoxicity: (a) 12-month patients; and (b) 6-month patients. The light blue bar indicates patients who received the exact number of cycles as randomised.

We were interested to see whether cardiac problems occurring within the first 3 months of trastuzumab (early) correlated with early cessation more than those occurring at 4–12 months (late). Clinical cardiac dysfunction occurred early in 138 12-month patients, of whom 82 (59.2%) completed 18 cycles, and late in 217 patients, of whom 105 (48%) completed 18 cycles (see Appendix 1, Table 43). Clinical cardiac dysfunction occurred early in 149 6-month patients, of whom 106 (71%) completed nine cycles, and late in 169 patients, of whom 119 (70%) completed nine cycles. Low LVEF occurred early in 74 12-month patients, of whom 31 (42%) completed 18 cycles, and late in 205 patients, of whom 78 (38%) completed 18 cycles (see Appendix 1, Table 44). Low LVEF occurred early in 81 6-month patients, of whom 53 (65.5%) completed nine cycles, and late in 145 patients, of whom 96 (66%) completed nine cycles. Therefore, patients reporting cardiac dysfunction soon after being randomised to their treatment arm were as likely as those reporting it late to complete their randomised trastuzumab treatment. There was a steady increase in the cumulative number of patients stopping trastuzumab early as treatment continued.

Relative change of left ventricular ejection fraction at 6 and 12 months

A total of 3925 patients had their LVEF percentage reported both at baseline and at least one other time point: 1964 12-month and 1961 6-month patients. The two treatment arms were well matched on baseline LVEF percentages [median (IQR) 63% (59–67%) for 12-month patients and 63% (59–67%) for 6-month patients; p = 0.43]. There were statistically significant reductions in LVEF at the 6- and 12-month time points in both treatment arms (all p < 0.0001). Despite similar relative changes from baseline to 6 months across treatment arms [median (IQR) 0.97 (0.90–1.02) for 12-month patients and 0.97 (0.90–1.02) for 6-month patients; p = 0.50], 12-month patients showed significantly greater reductions in LVEF by 12 months [median (IQR) being 0.97 (0.90–1.02) for 12-month patients and 0.98 (0.92–1.03) for 6-month patients; p = 0.0002].

Random-effects modelling

A quadratic curve was found to fit the LVEF data over all patients, demonstrating that cardiac function recovers post treatment (Figure 15). This was seen for the entire trial population (p < 0.0001) and for each randomised arm separately (both p < 0.0001). There was a significant difference between treatments in terms of change over time (p = 0.016); quadratic modelling predicts an earlier recovery of cardiac function after treatment completion in 6-month patients. No significant differences were found in trastuzumab timing (concurrent or sequential; p = 0.77) in terms of LVEF changes over time (see Appendix 2, Figure 33). Quadratic modelling predicted some differences between chemotherapy types in terms of LVEF changes over time, with anthracycline-based chemotherapy resulting in a lower nadir and a slower recovery of LVEF than anthracycline/taxane chemotherapy (p = 0.04) (see Appendix 2, Figure 34). In addition, in terms of LVEF change over time, a significant difference was found depending on the number of anthracycline cycles patients received, whether three or fewer or more than three (p = 0.04) (see Appendix 2, Figure 35).

FIGURE 15.

Random-effects modelling predicted lines and 95% CIs, by randomised treatment arm.

Predictive factors of low left ventricular ejection fraction

Chemotherapy type was found to be predictive of low LVEF (p < 0.0001), with taxane and anthracycline chemotherapy demonstrating an increased risk of low LVEF compared with taxane only (OR 2.30, 95% CI 1.18 to 4.47) in the 12-month group but not in the 6-month group (OR 1.31, 95% CI 0.68 to 2.51) (Table 12). Anthracycline-based chemotherapy without taxanes also demonstrated an increased risk of low LVEF compared with the taxane-only group (OR 2.79, 95% CI 1.43 to 5.44 for 12-month patients, and OR 2.13, 95% CI 1.12 to 4.06 for 6-month patients).

The number of anthracycline cycles received was also found predictive of low LVEF across all patients (p < 0.0001). Receiving more than three cycles of anthracycline was associated with a significant increase in the odds of developing low LVEF in both the 12-month arm (OR 1.69, 95% CI 1.27 to 2.25; p = 0.0003) and the 6-month arm (OR 1.76, 95% CI 1.27 to 2.43; p = 0.0007). Baseline LVEF was also found predictive of low LVEF for all patients (reference ≥ 65%: 55% to < 65%, OR 2.15, 95% CI 1.66 to 2.78; < 55%, OR 15.6, 95% CI 10.8 to 22.5; p < 0.0001). This was the case for both 12-month (reference ≥ 65%: 55 to < 65%, OR 2.30, 95% CI 1.63 to 3.26; < 55%, OR 18.2, 95% CI 10.8 to 30.6; p < 0.0001) and 6-month patients (reference ≥ 65%: 55% to < 65%, OR 1.94, 95% CI 1.33 to 2.84; < 55%, OR 13.7, 95% CI 8.1 to 23.0; p < 0.0001). Left-sided radiotherapy (p = 0.38), body mass index (p = 0.93) and ethnicity (p = 0.15) were not found to influence the incidence of low LVEF. Among the small number of patients who were taking cardiac medication before starting trastuzumab (44 in the 12-month arm and 55 in the 6-month arm), those in both arms were at significantly increased risk (over fourfold) of developing cardiac dysfunction (both p < 0.0001). Age at randomisation predicted for low LVEF in the 6-month arm (p = 0.0009) but not in the 12-month arm (p = 0.33), with 6-month patients aged > 60 years having increased risk compared with the reference age of < 50 years (60 to < 70 years; OR 1.96, 95% CI 1.27 to 3.03, and ≥ 70 years, OR 2.70, 95% CI 1.59 to 4.58).

Within-trial analysis

Individual patient data from the PERSEPHONE trial (database lock 6 March 2019) were analysed to determine the costs and QALYs associated with the two trastuzumab treatment durations (6 months and 12 months). To boost recruitment, a protocol change was implemented to allow the individuals to be randomised into the trial at any point up to 6 months into their trastuzumab treatment. A landmark analysis from 6 months into trastuzumab treatment was therefore appropriate for the economic analysis as this represented the point at which the treatment pathways diverged between the arms. The analysis followed the intention-to-treat principle to calculate the outcomes and costs incurred by patients in the trial arms during the follow-up period of 18 months. QALYs were estimated using the patient-completed EuroQol-5 Dimensions, three-level version (EQ-5D-3L), questionnaire77 in combination with survival data. Information on resource use and treatments received was recorded in structured CRFs and patient-reported questionnaires, allowing costs to be estimated. In July 2009, a change to the health economics data collection was made (amended in the protocol). This change affected the number of data collected at baseline. These data were poorly completed up to that date and, following the change to the randomisation point, they were no longer needed for the health economic analysis and so were removed. Full details of the changes can be found in Appendix 1, Table 45. Cost-effectiveness is summarised as the incremental cost-effectiveness ratio (ICER) and incremental net health benefit (INB). All analyses were conducted using the statistical package R (The R Foundation for Statistical Computing, Vienna, Austria).

Lifetime decision model analysis

Some of the long-term health benefits and costs associated with the two trastuzumab treatment durations occur beyond the PERSEPHONE trial follow-up period, such as the costs associated with the treatment of recurrence and secondary primary cancers. A de novo decision-analytic model was therefore developed to extrapolate the costs and health benefits of the two treatment durations over a lifetime horizon. In line with the within-trial analysis, the base case adopts an NHS and social care perspective. Costs and QALYs beyond the first year were discounted at an annual rate of 3.5%. The statistical software R, version 3.6 (The R Foundation for Statistical Computing, Vienna, Austria), was used to build the model.

Modelling approach and structure

A Markov model was built to estimate the total QALYs and costs per patient in each treatment arm. Patients with HER2-positive breast cancer are at long-term risk of relapsing or dying and therefore the model captures the time to these events and the associated quality of life and costs.

The structure of the model was developed in discussion with clinical experts and health economists, and it was adapted from the structure of a previously published model. 91 A 3-month model cycle was selected in line with this model. Patients enter the model 6 months into their trastuzumab treatment. Cost and QALY data were elicited directly from patients and from CRFs for the subsequent 18 months, and these data were used to inform the first six cycles of the model (summarised in the within-trial analysis). During this period, the costs associated with symptomatic or asymptomatic reversible cardiac toxicity were captured. As each state in a Markov model has no memory of ’time in state’, there is also the assumption that each cycle in a recurrence state incurs identical transition probabilities. Where necessary, this was overcome by building in tunnel states that allow transition probabilities, costs and utility parameters to change depending on the number of cycles in that state.

After the treatment period, individuals could move to either a disease-free (follow-up) state or a recurrence state. As data from the PERSEPHONE trial are the primary source of data for the model parameters, the recurrence states and their definitions align with those used in the trial:

• locoregional (ipsilateral breast/chest wall, axillary and ipsilateral supraclavicular nodes)

• distant (excluding ipsilateral supraclavicular nodes)

• second primary (including contralateral malignant breast disease).

There is evidence that the cost of locoregional recurrence is high in the first year after relapse and then falls to minimal levels. 92 For this reason, we incorporated a 1-year tunnel state for non-distant relapses, after which patients move into a ‘disease free after locoregional recurrence state’, provided that no distant relapse has occurred. This also allowed us to build in an ongoing higher risk of developing a distant recurrence for those who had experienced a locoregional recurrence. Those who had a second primary cancer stayed in that state but could move to the locoregional, distant or death state. If individuals developed a distant recurrence, they could stay in the same state or die, but they could not transition to any of the other states. Individuals in the death state could not transition to any of the other states, as death is an absorbing state. In addition, individuals in all states were at risk of dying as a result of breast cancer, chronic heart failure or unrelated causes. The model structure is depicted in Figure 16.

FIGURE 16.

Model structure. BC, breast cancer.

The model was run for the lifetime of the cohort (assuming that no one survived beyond 100 years old), with a start age of 56 years (the median age at the landmark date in the trial). A half-cycle correction was implemented in the model.

Cost-effectiveness analysis

The relative cost-effectiveness of using trastuzumab for 6 compared with 12 months was assessed using standard decision rules, estimating the pairwise ICER. 106 The ICER examines the additional cost that the use of trastuzumab for 6 months incurs over its use for 12 months and compares this with the additional benefits. The ICER estimate represents the additional cost required to generate one additional unit of health outcome (QALY), which provides the basis for establishing whether or not the new approach appears to provide good value for money to the NHS. Guidance from NICE suggests that an incremental cost per additional QALY of around £20,000–30,000 is considered to represent an appropriate threshold to establish value for money in the NICE Guide to the Methods of Technology Appraisal 2013. 78

To provide monetary value for the effectiveness, net monetary benefits were estimated by assigning a monetary value for the threshold λ, where λ = £20,000 was used:

The analysis followed the intention-to-treat principle, so that patient data were analysed based on the treatment group to which patients were randomised. As the two trial arms were expected to differ after 6 months of trastuzumab treatment onwards, the economic analysis was performed on data reported from that point, consistent with the landmark analysis. The baseline EQ-5D-3L values were obtained from the patient questionnaires returned at 6 months of trastuzumab treatment.

The point estimate of the ICER was obtained from the ratio of the difference between the adjusted mean QALYs in the two trial arms to the difference in the mean costs.

Validation

Internal validation was conducted for both the within-trial analysis and the decision-analytic modelling. Two health economists worked on different parts of the R code necessary to conduct the analyses and verified the equations and parameters used in the analyses relative to their sources. Furthermore, each one explained their part of the code to the other to look for possible computational errors. In the end, all sections of the code were verified by both health economists.

In the discussion, we also compare the results of the present analysis with those of other relevant published studies.

Subgroup analyses

Across both the within-trial analysis and the lifetime decision model, subgroup analyses were conducted for the following groups, in line with the statistical analysis of the main trial:

• ER status – negative versus positive

• chemotherapy type – (1) anthracycline-based (no taxanes), (2) taxanes and anthracyclines or (3) taxane-based (no anthracyclines)

• chemotherapy timing – adjuvant versus neoadjuvant

• trastuzumab timing – concurrent versus sequential (with respect to chemotherapy).

For the within-trial subgroup analyses, each imputed iteration of the trial data was filtered to those in each subgroup, and a probabilistic sensitivity analysis via bootstrap simulation with replacement (10,000 simulations) was conducted.

For the lifetime model subgroup analyses, the ‘msm’ package in R97 was used to calculate transition probability matrices for each subgroup. Subgroup-specific DFS curves were then estimated and the proportion of local and distant recurrences was calculated using the same method described for the main model analysis (see Transition probabilities).

Within-trial results

Cost-effectiveness

The within-trial cost-effectiveness results are presented in Table 18 (deterministic results) and Table 19 (probabilistic results). During the 18-month follow-up period from the landmark point (6 months), 6 months of trastuzumab was associated with an estimated incremental cost saving of £9537 (95% CI £9183 to £9890) and an incremental effect of +0.003 QALYs (95% CI −0.015 to 0.021 QALYs) compared with 12 months of trastuzumab. Six months of trastuzumab is, therefore, cost-effective and dominates the longer duration of trastuzumab treatment (INB 0.48, 95% CI 0.45 to 0.51).

TABLE 18 - Within-trial analysis base-case deterministic cost-effectiveness results (NHS perspective and imputed QALYs)

Treatment duration (months)	Total cost (£)	Total QALY	Life-years	Incremental cost (£)	Incremental QALY	ICER (£)	Net benefit (QALYs)	Incremental net benefit (QALYs)
12	15,298	1.14	1.472	–	–	–	0.38	–
6	5762	1.15	1.475	–9536	0.003	Dominant	0.86	0.48

TABLE 19 - Within-trial probabilistic sensitivity analysis cost-effectiveness results (NHS perspective and imputed QALYs)

Treatment duration (months)	Total cost, £ (95% CI)	Total QALY (95% CI)	Life-years (95% CI)	Incremental cost, £ (95% CI)	Incremental QALY (95% CI)	ICER (£)	Net benefit (95% CI)	Incremental net benefit (95% CI)
12	15,298 (15,023 to 15,579)	1.14 (1.13 to 1.16)	1.471 (1.464 to 1.478)	–	–	–	0.38 (0.36 to 0.40)	–
6	5762 (5548 to 5981)	1.15 (1.13 to 1.16)	1.475 (1.468 to 1.481)	–9537 (–9890 to –9183)	0.003 (–0.015 to 0.021)	Dominant	0.86 (0.84 to 0.88)	0.48 (0.45 to 0.51)

Figure 17 shows the within-trial probabilistic cost-effectiveness analysis on the cost-effectiveness plane. Each point (light blue) is one of the 10,000 simulated results and the triangle is the average of these points. The spread around this point illustrates the estimated uncertainty around the average cost-effectiveness result. The diagonal line represents a willingness-to-pay threshold of £20,000 per QALY. All of the points fall below this line, indicating 100% probability that 6 months of trastuzumab is more cost-effective than 12 months. There is 100% probability that 6 months of trastuzumab is cost saving compared with 12 months; however, the probability that 6 months is more effective in terms of QALYs is 63%.

FIGURE 17.

Within-trial probabilistic analysis on the cost-effectiveness plane.

Lifetime model results

Cost-effectiveness results

The results of the lifetime model base-case analysis are presented in Table 22. Compared with 12 months of trastuzumab, reducing treatment duration to 6 months is estimated to produce a lifetime cost saving of £9316 and a slight reduction of 0.01 QALYs per individual. Reducing the length of treatment of trastuzumab therefore has a negligible negative impact on QALYs and is significantly cost saving, with an expected INB of 0.46 QALYs. There is notable uncertainty around this result, however, with a wide 95% CI around this mean value (95% CI –2.21 to 1.98 QALYs).

TABLE 22 - Lifetime decision model probabilistic sensitivity analysis of cost-effectiveness results (NHS perspective and imputed QALYs)

Treatment duration (months)	Total cost (95% CI)	Total QALY (95% CI)	Life-years (95% CI)	Incremental cost (95% CI)	Incremental QALY (95% CI)	ICER	Net benefit (95% CI)	Incremental net benefit (95% CI)
12	25,340 (21,660 to 30,702)	11.13 (10.35 to 11.85)	14.018 (13.23 to 14.66)	–	–	–	9.86 (8.91 to 10.70)	–
6	16,024 (10,371 to 28,569)	11.12 (9.09 to 12.27)	14.017 (11.57 to 15.14)	–9316 (–16,485 to 2907)	–0.008 (–2.09 to 1.19)	9016	10.32 (7.69 to 11.69)	0.46 (–2.21 to 1.98)

The lifetime model base-case analysis results are presented in Figure 18 on the cost-effectiveness plane. The triangle point is the mean incremental cost and QALYs across all of the simulations. Each of the points on the graph represent one of the 10,000 simulated model results. The spread of these points therefore illustrates the uncertainty around the average result. The diagonal line represents the NICE willingness-to-pay threshold of £20,000 per QALY. Points that fall beneath this line are deemed cost-effective; the spread of points above and below the line shows the uncertainty around the lifetime cost-effectiveness of reducing the duration of trastuzumab to 6 months. This uncertainty is driven primarily by uncertainty in the estimated effectiveness of 6 months’ trastuzumab: the probability that reduced treatment duration is cost saving is 95%, whereas the probability that 6 months’ trastuzumab is as effective as or more effective than 12 months’ is 58%.

FIGURE 18.

Lifetime decision model probabilistic sensitivity analysis on the cost-effectiveness plane.

The cost-effectiveness acceptability curve (Figure 19) shows the probability that each treatment duration is the most cost-effective alternative (i.e. has the highest expected net benefit) across varying willingness to pay per QALY thresholds. At a threshold of £20,000, 6 months of trastuzumab has a 73% probability of being cost-effective compared with 12 months of trastuzumab, dropping to 66% at a £50,000 willingness-to-pay threshold and remaining above 61% at a £150,000 threshold.

FIGURE 19.

Lifetime decision model cost-effectiveness acceptability curve.

Subgroup analyses

The results of the lifetime decision model subgroup analyses are reported in Table 23 and shown in Figure 20. For some subgroups (concurrent trastuzumab timing, taxane-based chemotherapy and neoadjuvant chemotherapy timing), the incremental net benefits are negative. For these subgroups, the total QALYs for 12 months of trastuzumab treatment compared with 6 months are notably higher. For example, individuals who received taxane-based chemotherapy and 6 months of trastuzumab had an estimated incremental QALY loss of –0.73 (95% CI –1.52 to 0.33) compared with 12 months of trastuzumab. Six months of trastuzumab remained cost-saving, although the cost savings were slightly lower than those observed in the main trial analysis (–£8318 vs. –£9316). The cost-effectiveness results for the remaining subgroups were roughly in line with the main trial analysis results, except for those receiving trastuzumab sequentially to chemotherapy. The incremental QALY gain for the sequential subgroup was 0.32 (95% CI –0.10 to 0.75), with estimated cost savings of £10,871 (95% CI –£13,770 to –£8412). However, the CIs indicate that there is some uncertainty associated around this result.

TABLE 23 - Lifetime decision model probabilistic sensitivity analysis cost-effectiveness results: subgroup analyses

Subgroup	Treatment duration	Total cost, £ (95% CI)	Total QALY (95% CI)	Incremental cost, £ (95% CI)	Incremental QALY (95% CI)	ICER (£)	Net benefit (95% CI)	Incremental net benefit (95% CI)
ER status negative	12 months	27,596 (23,612 to 33,069)	10.67 (10.06 to 11.25)	–	–	–	9.29 (8.57 to 9.97)	–
	6 months	19,712 (15,284 to 25,747)	10.38 (9.74 to 10.98)	–7884 (–11,351 to –4108)	–0.29 (–0.88 to 0.28)	26,851	9.39 (8.61 to 10.11)	0.10 (–0.65 to 0.84)
ER status positive	12 months	25,177 (22,385 to 28,980)	11.24 (10.68 to 11.78)	–	–	–	9.98 (9.36 to 10.57)	–
	6 months	15,726 (12,837 to 19,699)	11.24 (10.69 to 11.78)	–9450 (–11,749 to –7109)	–0.002 (–0.39 to 0.38)	3,924,828	10.45 (9.82 to 11.04)	0.47 (–0.03 to 0.96)
Anthracycline-based chemotherapy	12 months	24,299 (21,522 to 27,935)	11.16 (10.60 to 11.69)	–	–	–	9.94 (9.33 to 10.52)	–
	6 months	14,767 (12,164 to 18,330)	11.33 (10.79 to 11.85)	–9532 (–11,701 to –7392)	0.17 (–0.18 to 0.52)	Dominant	10.59 (9.99 to 11.16)	0.65 (0.19 to 1.09)
Concurrent trastuzumab	12 months	26,421 (23,685 to 30,084)	11.30 (10.76 to 11.82)	–	–	–	9.98 (9.38 to 10.55)	–
	6 months	19,569 (15,766 to 24,875)	10.75 (10.15 to 11.32)	–6852 (–9521 to –3472)	–0.55 (–1.01 to –0.10)	12,564	9.78 (9.05 to 10.44)	–0.20 (–0.81 to 0.38)
Sequential trastuzumab	12 months	26,705 (23,120 to 31,566)	10.79 (10.23 to 11.34)	–	–	–	9.45 (8.79 to10.09)	–
	6 months	15,835 (12,789 to 19,975)	11.11 (10.55 to 11.64)	–10,871 (–13,770 to –8412)	0.32 (–0.10 to 0.75)	Dominant	10.32 (9.68 to 10.91)	0.86 (0.33 to 1.42)
Adjuvant chemotherapy	12 months	25,210 (22,189 to 29,350)	11.12 (10.57 to 11.66)	–	–	–	9.86 (9.24 to 10.46)	–
	6 months	15,235 (12,344 to 19,235)	11.18 (10.62 to 11.72)	–9975 (–12,440 to –7560)	0.05 (–0.34 to 0.45)	Dominant	10.41 (9.78 to 11.01)	0.55 (0.05 to 1.06)
Neoadjuvant chemotherapy	12 months	29,321 (25,377 to 34,656)	10.73 (10.12 to 11.33)	–	–	–	9.26 (8.51 to 9.97)	–
	6 months	25,948 (19,699 to 34,464)	9.83 (9.09 to 10.52)	–3373 (–7726 to 2340)	–0.91 (–1.64 to –0.22)	3727	8.53 (7.56 to 9.41)	–0.74 (–1.71 to 0.17)
Taxane-based chemotherapy	12 months	27,323 (23,167 to 35,160)	11.01 (9.95 to 11.73)	–	–	–	9.64 (8.28 to 10.48)	–
	6 months	19,005 (14,524 to 25,428)	10.27 (9.56 to 10.92)	–8318 (–14,911 to –3011)	–0.73 (–1.52 to –0.33)	11,344	9.32 (8.43 – 10.09)	–0.32 (–1.35 to 1.07)
Anthracycline- and taxane-based chemotherapy	12 months	28,146 (24,319 to 33,300)	10.90 (10.30 to 11.48)	–	–	–	9.49 (8.79 to 10.16)	–
	6 months	18,995 (15,008 to 24,503)	10.75 (10.12 to 11.36)	–9151 (–12,479 to –5617)	–0.15 (–0.69 to 0.39)	63,049	9.80 (9.05 to 10.51)	0.31 (–0.41 to 1.01)

FIGURE 20.

Lifetime decision model on the cost-effectiveness plane: subgroup analyses. CT, chemotherapy.

Twelve months of trastuzumab is therefore the preferred option in the three contextual subgroups: neoadjuvant, concurrent and taxane-only chemotherapy.

Statistical methods

To assess the quantification of patients’ health, the frequencies of the different responses to the general health question were assessed, and patients’ EuroQol-5 Dimensions (EQ-5D) visual analogue scale scores over time were assessed graphically using box-and-whisker plots.

The free-text data were entered into the NVivo database and coded by treatment arm and time point of the questionnaire. Two researchers and a patient representative looked at the data and began by performing a simple content analysis. Many themes emerged from these data. This broad initial analysis included all free text across all time points, regardless of treatment arm, to explore the whole experience for patients undergoing trastuzumab.

Which side effects are due to trastuzumab?

The side effects patients reported were sometimes difficult to attribute to one specific treatment, as many of the treatments were given concurrently or in close succession. This could potentially cause problems with treatment management and should be kept in mind when reviewing the data:

Aches and pains especially below the waist slowing me down considerably; not sure if it’s my age, the Herceptin treatment or the drug Arimidex to be taken for next 5 years. Also not sure if itching on my back and top of arms is caused by Herceptin or Arimidex. I know the Arimidex can cause hot flushes etc. so it could be a combination of each.

878

It is difficult to decide whether tiredness and hot flushes are due to Herceptin or the taking of Anastrozole/Arimidex tablets.

1084

It is really difficult to tell which side effects are due to Herceptin and which ones are due to the chemotherapy – especially when they are taken together.

1052

Results from the simple content analysis

From the first, simple content, analysis, 19 themes emerged (Figure 23).

FIGURE 23.

Number of reports of identified theme of treatment/study.

The five most commonly reported aspects were:

1. aches and pains

2. fatigue

3. shortness of breath

4. emotional fragility

5. financial impact.

Aches and pains

Although aches and pains are effects known and reported by health-care organisations (Table 24), they do not appear at the top of lists of side effects and are, arguably, not highlighted, but they appear to have had a huge impact on patients and their daily activities. It is not known whether patients wished to highlight them because of the perceived lack of emphasis elsewhere or because their impact was particularly great but, for many patients, these side effects did appear to be particularly distressing and intractable.

Aches and pains were mentioned in 524 responses. Common descriptions of this included an ‘aching of leg muscles’ (4027), ‘a general feeling of aching in the joints which impedes exercise’ (4001) and ‘joints and muscles hurt most of the time’ (325).

However, this could be extremely severe and debilitating and often seemed to increase over time:

Muscle and bone pain so bad it reduces me to tears.

2489

My body aches more. I feel like a woman of 80 some days.

17

Joint pain has increased as months have passed.

619

Bone pain in the hands and wrists – very painful. Plus occasional bone pain in elbows and ankles. Pain still occurring since finishing Herceptin approximately six weeks ago.

3719

I still suffer with joint pain which I’ve had since I started Herceptin.

481

Achy joints. Shooting pains in arm . . .

860

Fatigue

Fatigue or tiredness was reported in 461 responses. Fatigue is the most common side effect of cancer treatment and is increased with a combination of therapy. 108 Its cause and effects are highly complex and disruptive and have often been, arguably until recently, downplayed or overlooked. 109 Here this was described as ‘always tired’ (379), ‘extreme tiredness’ (489), ‘totally depleted of energy’ (4018) and ‘my energy levels are down’ (14). However, this was again was often experienced as extreme and debilitating:

I do think that people receiving Herceptin should be told that it can cause increasing fatigue because I thought I was going mad until someone told me.

188

Be prepared to be very tired during your treatment, I have found the fatigue hard to deal with. I am usually a very active person and the tiredness has been difficult to deal with. The tiredness is horrid.

3650

Fatigue gets worse as treatment goes on.

3606

The fatigue and anxiety is unreal; I’ve become quite depressed. It’s not easy to explain to people this might happen . . . but more information would be so helpful.

223

Fed up with tiredness & weariness. Sometimes only able to iron or garden for 20 mins then need to sit down.

200

You should put in a question about extreme tiredness which cannot be regarded as an illness but really affects your life.

1977

Shortness of breath

It is not clear whether shortness of breath correlates with the cardiac damage that trastuzumab is known to cause,64 anaemia or infection. However, it is known that shortness of breath is a particularly distressing, uncomfortable and frightening experience that limits both physical and social activity, which may have led to its dominance among the responses:

Following my Herceptin treatment I became breathless very soon after walking (more than 500 yards) or hoovering.

419

I have mitral valve weakening thus I am breathless. I am to visit [a specialist hospital] for this. For these reasons I have to catch a taxi for dentist, doctor, optician etc.

3889

I am still suffering with shortness of breath, extreme tiredness and severe aches in my joints. I had hoped that by now I would have made a far better recovery from my treatment but sadly I feel rather despondent that I am still so low and exhausted.

3626

Shortness of breath may also be significant as it is a reminder of potential cardiac damage, which seemed particularly distressing to some patients:

I was extremely shocked at the damage to my heart as a direct result of Herceptin at such an early stage in treatment. I feel that future patients should be made aware of the risk of heart failure in relation to Herceptin.

1294

Herceptin treatment affected my heart. Although the MRI scan shows heart function is now OK, I get anxiety and fear that the future prognosis is not good due to this . . . and worry constantly that I will have problems in the future. I remain positive on a day to day basis, but have some bleak and despairing moments.

123

I was told about the risks to my heart but I don’t think this was emphasised to the extent it should have been. Because of the affect it has had on my heart I now have to undergo an MRI scan and angiogram. Therefore I have decided to quit the Herceptin programme.

1051

Emotional fragility or depression

‘Emotional fragility’ was a term coined by one participant (1698), who felt that it was a more appropriate descriptor than ‘depression’. Another participant (553) wanted to emphasise that anxiety and depression should never be ‘lumped together’ as she felt that they were significantly different.

Many patients described a ‘black cloud’ (1510) and felt that they had lost all of their self-confidence as a result of the impact of the treatment and side effects. They wanted to warn others that they should be ‘prepared to feel down’ (1) and would ‘cry a lot’ (2627). Although patients seemed to equate this with being in the trial or receiving treatment, arguably it may have been due to the loss and fear that is known to occur after cancer treatment:

As far as the herceptin treatment the side effects are manageable and not as bad as chemo or radiotherapy . . . I find my anxiety and depression increasing now though. I feel a bit lost and my safety nets are being removed one at a time; I find myself dwelling increasingly on the possibility of recurrence of the cancer my moods are often erratic and often unreasonable. I’m snappy and sullen instead of the normal happy and philosophical/positive.

1377

The network & support has been really important e.g. friends/well-being group/family/hospital services/activities etc. Initially for more physical & practical support but towards the end of treatment for mental/emotional support & support in managing long-term recovery which has taken longer than I expected & I don’t feel I’m there yet. Anything you can do to support & help people through this stage I feel is just as important as the medical help & support given at early diagnosis & during treatment.

4061

Not sure whether it is the side effects from the Herceptin but have been feeling really low which is out of character for me. Feeling really tired, depressed and anxious a lot.

381

At times I have been quite depressed for a couple of days after having Herceptin . . . for 2/3 days following treatment I felt quite depressed- on occasions this could be fairly bad: I’d be very tearful, could not face socialising and was pessimistic about the future.

665

I have felt a little stressed and quite tearful over the past weeks.

735

I feel that a little more emphasis should be placed on the psychological effects felt by Cancer patients.

1038

I felt somewhat abandoned as I seemed to fall through the cracks of the support services on offer. I didn’t know or have the emotional strength to call on the breast care nurse service and was unsure whether I was still eligible for such support.

1100

. . . you feel well looked after physically, but the emotional side of the treatment journey is not really talked about. I need and I feel I want to be looked after even now, I need more hugs, cuddles, reassurances.

1344

Financial impact

A total of 125 responses mentioned the importance of subsequent patients being made aware of the financial aspects of treatment and involvement. This impact was from increased costs such as travel and car parking or because of an inability to function and, therefore, to work in the same capacity as before. Although patients seemed to be aware that this was unavoidable, they wanted to raise awareness:

Financially it’s crippling. If you’re low paid there is no help, nearly cost me my home as I wasn’t always well enough to work overtime or do my 2nd job. Still in debt now. But I’m alive!!

2840

I have had increased regular outgoings arising solely from the effects of my illness since diagnosis. I live alone and am not eligible for a social services care plan because I am in my 50s. Consequently I have had to buy in the following services from time to time: home help/meal deliveries, laundry services, taxis to appointments, putting the cat in the cattery during periods of incapacity/hospitalisation. Since diagnosis, I would estimate that these additional regular cost amount to several thousands of pounds. This together with the reduction in my income (end of paid sick pay, return to work on reduced hours, ill health retirement) has had a considerable effect on my standard of living.

1581

Although I have not incurred big one off costs I have had to spend between £4–6 every three weeks whilst having Herceptin. This soon adds up over the year I will be having this treatment. Free parking would be of great benefit.

2262

The financial cost of my diagnosis has been huge to my family. We have had no help with this burden as my husband is working.

3843

The main input on my diagnosis has been on my work/employment. I have had to reduce my work hours from 37 per week to 18 1/2 hours. Although I now spend more time keeping fit i.e. swimming, walking, etc. I no longer feel that I have a career – just a job. I can’t see me going back to full time employment which makes me sad. On the other hand I now have a better work/life balance but this would not be the case if I had children or my husband did not have such a good job. I don’t think I would have had the choice about taking a reduction in hours and would have struggles with the ongoing fatigue.

582

People should be aware of the cost involved in coming to hospital every 3 weeks – car parking, petrol, time off work; it all mounts up.

39

My treatment and care has been excellent, but no-one prepares you for life after. I have been 24 years as a Senior Court Usher and loving my work. Since I returned on a phase return, I can’t do my work because of my pain and disability in my legs after the Herceptin, sleepless nights and this has made me very low although I’m grateful to be here after my treatment. You do need ongoing support to help you through, especially if your work is affected. I’m now doing a job I dislike but as a widow at 47 I have to earn a wage.

2106

The Herceptin injections cause the most trouble. They make you ache to the extreme you can’t walk or stand if at all not for long. Due to my work (HGV driver) I can’t afford to not be able to stand. Now I’ve had 6 months I’m stopping further injections.

3683

Other participant experiences

Between 50 and 100 responses cited another raft of aspects that were arguably more ‘visible’ to others and matched more closely with the side effects described by organisations (see Table 24):

• Anxiety (95 responses): ‘I have lost a lot of confidence and get very anxious’ (336).

• Nail changes and soreness (95 responses): ‘Every fingernail was broken/torn (3429)’; ‘I can’t peel an orange or use Sellotape’ (2728); ’very soft nails, sore hands, skin irritation’ (816).

• Hot flushes (88 responses): ‘I never feel very comfortable! I have to wear cotton/light clothing – not nice in cold weather’ [14].

• Sore/’dripping’ nose (68 responses): ‘Constantly dripping nose’ [1424]; Very, very, very dry, scabby nose (internal) when having the Herceptin [1412].

• Weight gain (67 responses): ‘Gradually worsening problem with water retention/swollen feet/legs/body with each of the first 4 Herceptin injections – sometimes gaining over a stone in weight in 3 or 4 days and starting earlier with each cycle. For cycle 5 have swapped to intravenous Herceptin to see if it makes any improvement – watch this space!’ [4081].

• Neuropathy (64 responses): ‘Continuous pins and needles in hands and feet, with finger tips becoming increasingly sore. Painful balls of feet (like walking on marbles!) making walking uncomfortable – have to wear shoes with very squashy, soft soles’ [4081].

• Skin rashes/dryness (52 responses): ‘Dry and cracked skin’ [1016]; ‘Skin looks like porridge oats’ [1185].

Other side effects

Although it is impossible to explore issues further with patients, certain aspects stood out during analysis. These were the cumulative impact of the side effects of Herceptin, the perceived ‘downplaying’ of the effects of Herceptin by health-care staff and the impact of Herceptin treatment on other chronic illnesses:

General health has been good. However, the cumulative effect of treatment – chemotherapy, radio-therapy and Herceptin – has been one of fatigue, exhaustion and general debilitation.

1628

Although Herceptin is said to have few side effects (by comparison with chemotherapy), I think it would be helpful to patients if they were made aware of the full range possible. In my own case, I remain very tired after each treatment for several days so that it would not be possible to resume my employment at present. I also continue to have diarrhoea a day or two after each treatment and also disturbed sleeping patterns. Each effect taken singly is not significant but in combination is most significant and does impact on my ’normal life’. My GP – generally sympathetic – reads the current literature provided by Macmillan and does not necessarily appreciate the debilitating effect of this treatment in my case and so is pressing me to return to work. This isn’t helpful – indeed it is very stressful. So, fuller communication to GPs and patients needed I think.

1581

I have found Herceptin equally as challenging as chemo because it has caused significant fluid retention and joint pain in hands, fingers, knees and ankles that makes normal day to day tasks and walking difficult. Unfortunately I have been discriminated against at work and lost my job – I believe as a result of being absent – but couldn’t afford to take my ex-employer to court. I think people should be very wary of taking time off as the law does not protect cancer patients as well as you expect.

2476

I think Herceptin is very ’chemo’ like. My treatment is 3 weekly and the first 2 weeks are miserable. I am bolstered by the fact that after that I begin to feel myself again. I am looking forward to when the Herceptin is out of my system and I’ll begin to feel my old self again – hopefully!

3943

I was given a great deal of information and support to assist with the effects of chemotherapy and radiotherapy but far less on the possible effects of the Herceptin injections. I had assumed that my health would start to improve after the chemotherapy and radiotherapy and was not expecting that the Herceptin alone would have such an impact.

3941

My condition with other problems have become worse. I had problems doing things because of mobility problems before breast cancer but the pain has got worse.

470

As I have chronic fatigue syndrome and fibromyalgia I have found all the treatment for cancer extra hard and very tiring! I can’t live my life because my body doesn’t work properly.

3517

I suffer with osteoarthritis in hands and knees and have noticed a very considerable increase in the pain and stiffness since the start of Herceptin, particularly in my hands, sometimes making it impossible to unlock my hand joints to do simple tasks.

242

The Diabetes that I had under control [has] become uncontrollable. My blood sugar has been between 10 and 20 since last September; now I am type 1 taking insulin.

1606

Additional themes

Having received such a large number of rich, qualitative data from patients, the researchers felt that it would be interesting to share some emerging themes not necessarily related to side effects of the diagnosis or treatment:

• Taking part in the trial: how does this impact on patients?

• What do patients on the trial think about receiving less than the standard 12 months’ treatment with Herceptin?

• Fear of recurrence.

• Needing more information or support.

• Patients’ relationships with health-care professionals.

• Feeling abandoned after treatment.

• Wanting to move on.

Taking part in the trial: how does this impact on patients?

Reading through the responses from the patients on the trial, it seems that a lot of them felt grateful that they had been given the opportunity to take part. Sometimes this feeling appeared to be attributable to the media attention Herceptin received when it first appeared as a treatment for breast cancer:

I feel very privileged to have this treatment as the press call it the post code lottery.

820

I would like to take the opportunity to say how extremely grateful I am to have been a recipient of Herceptin.

1141

. . . I am alive & very grateful for the treatment & realise I am lucky to have had the opportunity to have it.

1483

I am grateful for the Herceptin & believe it has saved my life.

549

The chance to join the trial was a blessing. Any small inconvenience having the treatment paled at chance to Herceptin . . . and giving hope that my survival could be many more years.

1519

. . . from what I have understood about Herceptin I think it seems a ’wonder drug’. I feel blessed it has been available for my recovery.

1566

For some, being in the trial gave them a sense of being watched more carefully:

I am happy to be a part of your study/clinical trial of Herceptin. It really helps me psychologically because with this treatment I felt safe and secure of not recurring from breast cancer.

1453

I do feel slightly more confident due to the extra care the trial affords.

1377

My family are reassured about my health more because of the extra tests that I have had due to being on the trial.

1016

What do patients on the trial think about having less than the standard 12 months’ treatment with Herceptin?

As the PERSEPHONE trial randomised patients to standard of care (12 months of Herceptin) or a reduced treatment duration, it is interesting to hear how the patients felt about this, especially those who received the shorter duration.

Perhaps surprisingly, not many patients gave negative responses related to risk of receiving fewer Herceptin treatments. This is perhaps a result of careful explanation of the trial given to prospective patients. Just a handful of patients referred to being anxious about being ‘undertreated’:

Not having completed half the treatment has left me somewhat anxious as I feel that somehow I have lowered my chances of the recurrence of cancer. This is having the effect of me feeling vulnerable also.

513

I would sometimes like more re-assurance and information about the 6 months – have I done the right thing??

544

Would have been nice to have a chat at the end of the 9 cycles re the [trial] . . . Reassurance that [trial] was good and results were showing further cycles didn’t not seem to improve survival [rates] would have been nice.

896

Just one participant reported that she withdrew from the trial in order to continue with a full year of Herceptin:

I have decided to stop the trial and continue with 1 years treatment of Herceptin – following discussion with family I feel if the cancer returned within the next five years I would maybe feel that it was partly my fault – by not having the full treatment.

889

Fear of recurrence

Many of the responses indicated a fear of recurrence:

It never leaves you, and I am always thinking it is going to come back . . . I am always thinking on special occasions like Christmas birthdays etc., will I be here for the next one?

470

It’s hard to come to terms with the reality that this ’weight’ will never truly be far from my immediate thoughts.

496

I still think I have cancer or it will come back soon.

1482

Although I do not feel ill. I am finding that my fears of cancer returning are a constant ‘niggle’.

963

This is evidently something that causes a lot of angst for patients with this diagnosis and this manifests in many ways.

Some reported that this fear of recurrence caused them to feel depressed:

Still worry about the cancer coming back which makes me depressed at certain times.

641

Some described it monopolising their life:

I find myself dwelling increasingly on the possibility of recurrence of the cancer. My moods are often erratic and often unreasonable. I’m snappy and sullen instead of the normal happy and philosophical/positive. I am more negative than I have even been.

1377

Many sought more reassurance that their cancer had really gone and was not coming back:

It is frustrating that there is no test to confirm whether the cancer has gone – you are left to see what happens.

1140

I would also like to know when any tests could be performed to see if I’m completely cured or not (always at the back of my mind) . . . You must communicate with sample patients more even if it takes a phone call to put their mind at rest.

608

I finished Herceptin 6 months ago and am coming up to 2 years Post OP. I’m feeling very low and anxious about reoccurrence as I haven’t got an ultrasound or mammogram for another year.

1414

I feel personally that I need to know my cancer has really gone and is not going to return in the other breast/or somewhere else in a few years/months. I have not been given this reassurance yet.

730

Many said that they remained anxious about every symptom and felt the need to be able to ask a health professional about this:

Unfortunately, once having had cancer, one is never free of the worry about it returning, so it is very important to feel able to report any worrying symptoms that may arise.

1166

Some were worried about becoming overly anxious:

The one element that you have to deal with is that of heightened vigilance of any bump, ill feeling. I think it makes you ’ultra-aware’ which inevitably leads to scares/panics about the cancer returning. I am not a hypochondriac – but you begin to feel like one.

1140

Some reported difficulty in remaining objective, with every possible symptom signifying risk of recurrence:

With time, the anxiety over a relapse increases and it becomes difficult to be/remain objective regarding any sensation.

719

Just a few responses suggested feeling safe from the risk of recurrence:

. . . with this treatment I felt safe and secure of not recurring from breast cancer . . . I feel better and worry free now that I finish my dose of Herceptin.

1453

Needing more information or support

A common theme running through the responses is wanting more information. Some wanted more about their own diagnosis and treatment pathway:

It would be good to know how many treatments exactly are involved rather than just a vague 6 or 12 months. I have been asking for 4 months each time I attend for treatment and have been told they don’t know.

740

Please keep patients fully informed about their diagnosis as it is not always easy to know the right questions to ask.

1272

It would have been better if I had known I would not be randomised until after Herceptin (9) – this was not what I was led to expect earlier.

1368

Some felt that the information they had been given was too complicated or difficult to understand:

The information pack I was given was put in medical terms and not lay man’s terms the information should be easy to understand it was far too complicated you needed a medical degree to understand it.

1160]

I think that it would be helpful to the patient if they were told (obviously in layman’s terms) how the Herceptin works in the body. I think that some people are afraid to ask questions.

1243

. . . have treatments explained in layman’s terms.

1524

The time at which information is given is important, as some patients pointed out how difficult it is to absorb information at the point of diagnosis, when they are in shock and struggling to process everything:

After being informed of the diagnosis by one consultant my husband and I were taken to another room and asked if we had any questions. As we were both in a state of shock at the time it was impossible to think clearly.

505

I would appreciate that things be explained more than once as I tend to forget a lot of information.

1470

Some wanted more clarification about the side effects of Herceptin:

The possible side effects of medication should, most definitely, be explained to patients as, given their emotional state of mind after being diagnosed with cancer, could cause some anxiety, the patient believing, as I did, that they could have cancer in their bones!

1038

I wish I had more information about Herceptin and joint pain.

1383

I have peripheral neuropathy in my fingers and toes, and would have liked to be told more about this.

1524

Others wanted more information about Herceptin itself:

I would like to know if they have found out any new information on Herceptin, just interested to find out.

541

Many wanted more information to help them understand the various tests for their heart function as well as the symptoms of heart problems:

I think people should be advised about what symptoms may indicate heart problems. I didn’t know my symptoms of coughing and shortness of breath were due to a lowered E.F. as a result of a reaction to the Herceptin. If I had I would have drawn the Dr’s attention to it sooner.

1355

I think patients should receive more information regarding the effects Herceptin can have on the heart. I feel had I received this I would have declined the treatment.

1051

Several asked to be told the results of the trial, perhaps seeking reassurance that they had made the right decision to take part:

I would like to know the result of trial.

736

If possible, to be kept up to date with findings from trial & outcomes from trial.

928

I would be interested to know what conclusions are reached in the future concerning the Herceptin 6 months vs 1 year trial.

1141

I would be really interested in receiving a copy of the results of the RCT when they are available.

1256

Patients’ relationships with health-care professionals

Many of the responses made reference to the importance of the patient’s relationship with members of the health-care team. A lot of these responses emphasised that a good relationship led to trust and less stress about treatment:

Most of the time I am fine. When I go to the hospital if the doctors and nurses explain things and tell me what is going to happen next I am fine but if I get a letter I am not expecting for the hospital then I get stressed.

582

I have found it useful that I could consult my research nurse as she put my mind at rest over any problems I may be suffering.

807

I also feel as if I don’t want to bother staff with trivial questions, but to me these questions are important.

1041

Where to go what to do if you feel you have a problem between appointments with consultants – leaflet or card . . . Not being treated like an over anxious idiot when asked about ‘new’ lump.

1090

Feeling abandoned after treatment

Some respondents described feeling unsupported or abandoned after completing their treatment. It would seem that the treatment period can be intense and involves a great deal of contact with the health-care team and other patients. Some people have trouble adjusting to the period of follow-up, when they have little contact with others:

I feel any patient who has gone through this experience and treatment need more contact and support after treatment finished. The help + support received during treatment was excellent but it just felt like when treatment was ended I seemed to be abandoned.

1436

After I finished radiotherapy I did tend to feel a bit depressed. Think that was after I had received a great deal of attention attending 5 days over 5 weeks.

1335

The initial flurry of activity for treatment has died down and now I have to deal with the psychological implication of my diagnosis. An issue that I was not expecting.

1100

Wanting to move on

A fairly common theme throughout the responses was patients expressing a wish to ‘move on’ with their lives and to resume some sense of normality or a semblance of ‘pre-diagnosis’ life:

Today is my last treatment and it is with huge relief, as every 3 weeks reminds you, that you are a cancer patient. It will be somehow nice to close this chapter. Emotionally post treatment is harder than the chem + radio time, which was most unexpected.

496

I found it too long a constant reminder of being a patient. I wanted to get on with my life.

533

Had hoped to be selected for 6 month trial . . . want to get on with life and put the cancer into the past.

614

. . . now looking forward to living again.

1502

. . . now it’s time to move on.

644

Some of the respondents felt very positive about this:

I feel the cancer is a thing of the past; I feel positive about my future.

614

Since treatment ceased my feeling of health and wellbeing has improved tremendously and I would say is nearly back to what it was in April 2009 before I knew anything was wrong and that I had breast cancer.

616

My health is doing really well. I am just looking forward to the future.

644

Some suggesting ways of keeping well and staying positive:

I’ve joined a choir which has improved my breathing possibly due to chemo? Being pro-active has definitely been good for me!

1285

Healthy eating and moderate exercise would help.

1324

PHARE: 6 versus 12 months

The PHARE trial was a French national study, funded by the Institut National du Cancer, that was carried out between May 2006 and July 2010; 3380 patients were randomised to either 6 or 12 months’ trastuzumab (Table 25). Prior to the start of the PERSEPHONE trial and during the trial’s recruitment phase, the trials teams met and agreed on common data collection items so that a joint analysis of individual patient data could occur once both trials had been completed and published. The two trials had significant similarities, including mapping on to standard national practice for the treatment of HER2-positive early breast cancer. Differences included the randomisation time point; in PHARE, this was between 3 and 6 months of trastuzumab, which would have excluded all patients who developed cardiac toxicity in the first 3–6 months, whereas at the start of the PERSEPHONE trial randomisation was carried out prior to any trastuzumab treatment. Subsequently, the PHARE TMG advised that entering patients into a 6 versus 12 months trial was in their opinion more acceptable to patients after they had already received some trastuzumab. As recruitment into PERSEPHONE was slow at this point, the TSC and TMG decided to make the randomisation point more flexible, allowing recruitment up to and including the ninth cycle of trastuzumab. The original PHARE statistical plan was to recruit 7000 patients so that non-inferiority could be detected at the 2% level. With a 2-year DFS estimated at 85%, the HR limit or critical value was predetermined to be 1.15. Shortly after the start of the trial, there was a change in the statistical plan that allowed for longer recruitment time and follow-up, which altered the recruitment target to 3400 patients. The PHARE trial published the results of 2-year DFS in 2013,23 which failed to demonstrate non-inferiority, with 12-month patients having a 2-year DFS of 93.8% (95% CI 92.6% to 94.9%) and 6-month patients having a 2-year DFS of 91.1% (95% CI 89.7% to 92.4%) (HR 1.28, 95% CI 1.05 to 1.56; non-inferiority p = 0.29). More recently, the PHARE trial published long-term follow-up data and demonstrated a HR after 7.5 years’ median follow-up very similar to that in PERSEPHONE (PHARE HR 1.08, 95% CI 0.93 to 1.25; p = 0.39). The statistical analysis plan with a critical value of HR 1.15 does not allow a conclusion of non-inferiority, although the similarity of the results from the two trials is striking.

HORG: 6 versus 12 months

This trial from the Hellenic Oncology Research Group reported 3-year DFS in 481 randomised patients (see Table 25); 67% of patients had ER-positive tumours and 79% had malignant involvement of axillary lymph nodes, with 14% having more than four nodes positive, demonstrating a higher risk of relapse profile than patients in the other duration trials. The non-inferiority HR limit of 1.53 was derived from an estimated absolute difference in 3-year DFS of 8%, based on an expected DFS in the 12-month group of 85%. This was calculated before the start of the trial. The 3-year DFS was 95.7% for 12 months’ trastuzumab and 93.3% for 6 months’ trastuzumab, with a HR of 1.57 (95% CI 0.86 to 2.10; non-inferiority p = 0.137), thus failing to demonstrate non-inferiority. Had non-inferiority been demonstrated in this trial, it is debatable whether or not, with a non-inferiority margin of 8%, the result would have been practice-changing.

SOLD: 9 weeks versus 12 months

The SOLD trial has been described in Chapter 1. In a time-driven analysis 2 years after the last patient was randomised, the 5-year DFS of 2174 patients was analysed, with a median follow-up of 5.2 years25 (see Table 25). The non-inferiority limit was derived from the true (observed) standard-arm 5-year DFS of 88.7%, and the HR limit or critical value was 1.385, which was calculated when the statistical design of the trial was changed from superiority to non-inferiority in February 2014. The results showed a 5-year invasive DFS of 90.5% for the 12-month arm and 88.0% for the 9-week arm, with a HR of 1.39 with two-sided 90% CI of 1.12 to 1.72, and therefore the upper CI crossed the non-inferiority boundary, failing to demonstrate non-inferiority. However, distant DFS (94.2% for 12 months and 93.2% for 9 weeks) and OS (95.9% and 94.4%, respectively) did not differ substantially between the groups. The design of the trial was based on the original FinHer experimental arm in which 9 weeks’ trastuzumab was given concurrently with docetaxel and was the first adjuvant treatment, followed by FEC chemotherapy and then completion of 12 months’ trastuzumab in the standard arm. The total dose of trastuzumab in the 9-week arm was 20 mg/kg, in comparison with 56 mg/kg in the 6-month arms of PERSEPHONE, PHARE and HORG, and perhaps this lower dose is insufficient to demonstrate non-inferiority, despite being given upfront and concurrently with docetaxel. An exploratory subgroup analysis with relatively small numbers suggests that docetaxel at a dose of 100 mg/m² rather than 80 mg/m² may be associated with non-inferiority of 9 weeks’ trastuzumab, but this would have to be confirmed in a further randomised trial.

Short-HER: 9 weeks versus 12 months

The Short-HER trial has been described in Chapter 1. There were some differences in the chemotherapy given in the experimental arm and the standard arm, which was required by the government funders. The standard arm used chemotherapy that was in the national guidelines and the experimental arm was the same as in the FinHer and SOLD trials. Chemotherapy in the standard arm consisted of AC or EC with a dose of epirubucin of 90 mg/m² administered every 3 weeks for four cycles followed by 175 mg/m² paclitaxel or 100 mg/m² docetaxel every 3 weeks for four cycles. Trastuzumab was administered every 3 weeks for 18 cycles, starting with the first taxane dose (8 mg/kg loading dose at first cycle, and 6 mg/kg thereafter). Chemotherapy in the experimental arm consisted of 100 mg/m² docetaxel every 3 weeks for three cycles with concurrent trastuzumab weekly for 9 weeks followed by FEC (60 mg/m² of epirubicin) for three cycles. Therefore, the standard arm had higher doses of epirubucin, and four cycles rather than three of each chemotherapy type. Accrual was planned for 2332 patients, but the trial was required to meet timelines from the funders and therefore had to complete recruitment after 1254 patients, which meant that the trial was underpowered, with a power of 56% for the non-inferiority end point (see Table 25). Five-year DFS was 88% in the standard and 85% in the experimental arm, with a HR of 1.13 (90% CI 0.89 to 1.42) and with the upper limit of the CI crossing the non-inferiority limit, which was set at 1.29. Therefore, the trial could not claim non-inferiority. A Bayesian analysis showed that the probability that the 9-week arm was non-inferior to the 12-month arm was 80%. The 5-year OS did not differ substantially between the groups, at 95.2% in the standard arm and 95.0% in the short arm (HR 1.07, 90% CI 0.74 to 1.56).

Trial-level meta-analysis

In a meta-analysis of the latest published data from the PHARE trial51 and the PERSEPHONE trial (this analysis), we compared 6 months and 12 months of trastuzumab treatment. 119 As the PHARE trial randomised patients to 3 or 6 months of trastuzumab treatment and timed DFS from randomisation, we used the landmark analysis timed from 6 months after the start of trastuzumab as the most appropriate end point comparable with that used in the PHARE trial. The weighted 4-year DFS rate was 88.9%. Thus, with the methodology used in our study, and with a non-inferiority limit of 2%, the non-inferiority limit for the HR was calculated as 1.19. There was no detectable difference between the two trials’ results (heterogeneity p = 0.94), and the pooled HR for DFS of 1.08 (90% CI 0.98 to 1.18) met the prespecified definition of 2% non-inferiority (non-inferiority p = 0.04). The data from both PERSEPHONE and this meta-analysis of trial-level evidence demonstrate that 6 months’ trastuzumab is non-inferior to 12 months’ trastuzumab in the population as a whole. In this population there will, therefore, be a significant number of patients in whom reducing trastuzumab to 6 months should be possible.

Strengths of the study

The PERSEPHONE trial was pragmatic and mapped on to standard practice in the UK. Study sites could approach all patients in whom treatment with adjuvant or neoadjuvant chemotherapy and trastuzumab was planned. It therefore examined the effect of reducing trastuzumab to 6 months in a ‘real world’, post-licensing population treated with chemotherapy and trastuzumab. All types of chemotherapy were included in the trial and therefore as standard chemotherapy practice gradually changed over the course of the trial, continued recruitment was facilitated and the recruitment target was successfully reached. This also means that results will be applicable to this population treated in routine clinics. Detailed cardiology and health economic substudies are further strengths and include within-trial cost-effectiveness analysis and modelling the long-term cost-effectiveness of trastuzumab. Collection of germline blood samples and formalin-fixed paraffin-embedded tissue from around 80% of patients provides an invaluable translational resource to explore pharmacogenetics and trastuzumab cardiotoxicity, inherited germline predisposition to HER2-positive breast cancer and molecular markers predicting outcomes for different durations of trastuzumab.

Limitations of the study

There are some limitations because of the wide range of chemotherapy treatments accepted in the trial, and subgroup analysis based on specific regimens is not possible. However, the recruitment of required numbers for sufficient levels of significance and statistical power outweighs this limitation in our view. The selection of chemotherapy and trastuzumab timing according to perceived risk by the investigators (i.e. concurrent preferred in higher-risk patients) is a further limitation. The changes in standard practice during the trial and, therefore, the different lengths of follow-up also bring some limitations. The variable timing of randomisation potentially introduces ascertainment bias. The analyses were performed in accordance with the intention-to-treat principle. We believe that this is the most appropriate approach to use as it ensures unbiased estimates. There are some concerns that it can potentially underestimate differences between treatment arms and thus drive the results towards non-inferiority. However, this concern is valid only when adherence to treatment is low and/or differential loss to follow-up occurs. In the PERSEPHONE trial, adherence to protocol-mandated treatment is high and loss to follow-up is low and balanced across randomised treatment arms, which we believe mitigates this concern.

Trastuzumab and pertuzumab

The APHINITY (Adjuvant Pertuzumab and Trastuzumab in Early HER2-Positive Breast Cancer) trial was published in 2017. 141 This compared adjuvant chemotherapy and 12 months’ trastuzumab with or without 12 months’ pertuzumab for the treatment of HER2-positive breast cancer; 63% of patients were node positive and 36% were ER negative. Invasive DFS was timed from randomisation and included local and distant recurrence, contralateral new breast cancer and death from any cause, but did not include second primaries. The estimates of 3-year invasive DFS rate were 94.1% in the pertuzumab group and 93.2% in the placebo group (HR 0.81, 95% CI 0.66 to 1.00; p = 0.045). Among the cohort of patients with node-positive disease, the 3-year rate of invasive DFS was 92.0% in the pertuzumab group, compared with 90.2% in the placebo group (HR for an invasive-disease event 0.77, 95% CI 0.62 to 0.96; p = 0.02). Among patients with node-negative disease there was no improvement in invasive DFS with added pertuzumab at this early time point. In December 2017146 the US Food and Drug Administration (FDA) licensed adjuvant pertuzumab for the treatment of patients who had a high risk of recurrence. In March 2019, NICE approved pertuzumab for use in the NHS with trastuzumab biosimilars147 in patients with node-positive disease. The landscape of the treatment of HER2-positive early breast cancer is therefore very different from that when the trial started.

Neratinib for 12 months after trastuzumab

Neratinib is a potent, irreversible tyrosine kinase inhibitor of HER1, HER2, HER3 and epidermal growth factor receptor. It inhibits the phosphorylation of these receptors by irreversibly binding to the intracellular signalling domain, thereby inhibiting HER-mediated downstream signalling. The ExteNET trial142,143 examined the use of an additional 12 months of neratinib in patients after they had completed 1 year of adjuvant treatment with chemotherapy and 12 months of trastuzumab. A total of 2840 women were randomly assigned up to 2 years after completing primary treatment (amended to 1 year after, during the study) to 12 months’ neratinib or placebo. The 2-year invasive DFS rate was 93.9% (95% CI 92.4% to 95.2%) in the neratinib group and 91.6% (95% CI 90.0% to 93.0%) in the placebo group, with a stratified HR of 0.67 (95% CI 0.50 to 0.91; p = 0.0091). The most common severe adverse event (grade 3 or above) was diarrhoea, with a 40% incidence. In a prespecified subgroup analysis, patients with ER-positive disease seemed to gain more benefit than those with ER-negative disease [ER positive, HR 0.51, 95% CI 0.33 to 0.77, p = 0.0013; ER negative, HR 0.93, 95% CI 0.60 to 1.43, p = 0.74; p (for interaction) = 0.054]. Neratinib’s activity in ER-positive disease has raised questions about the predominant mechanism of action in these tumours and whether or not it also inhibits ER-related signalling pathways. Although 1 year’s neratinib has received marketing authorisation approval from both the FDA and the European Medicines Agency, uptake has been low in the USA and Europe, thought to be because of neratinib’s modest effect and the high incidence of severe diarrhoea. NICE has now appraised neratinib and has approved it for 1-year extended use in patients with ER-positive, HER2-positive breast cancer but only if they had received 1 year of single agent adjuvant trastuzumab and no pertuzumab. 148 However, the NICE appraisal states that it is unclear which patients would receive neratinib in clinical practice.

HERA and 24 months’ trastuzumab

The HERA trial also tested 24 months’ trastuzumab compared with 12 months’ trastuzumab and a no-treatment control arm. 10 The 11-year follow-up, published in 2017,13 showed that 24 months was no better than 12 months in terms of DFS (HR 1.02, 95% CI 0.89 to 1.17), with 10-year DFS of 69% for 1 year and 69% for 2 years of trastuzumab. Extending trastuzumab treatment beyond 12 months did not improve DFS further.

Neoadjuvant trials in HER2-positive disease

Neoadjuvant trials are an excellent route for accelerated drug approval and have been particularly effectively used in HER2-positive breast cancer to speed the process of new drug approval in the adjuvant setting. 149 The NeoSphere trial looked at neoadjuvant trastuzumab and pertuzumab144,145 in a 417-patient randomised Phase II study with four arms. Group A received standard trastuzumab with docetaxel; group B received pertuzumab and trastuzumab with docetaxel; group C received pertuzumab and trastuzumab; and group D received pertuzumab and docetaxel. The primary end point was pathological complete response. This was highest in group B, at 45.8%, followed by group A (29%) and then group D (24%), with the lowest rate in group C (16.8%). The 5-year follow-up analysis confirmed that progression-free survival mirrored the pathological complete response rates and demonstrated differences in 5-year progression-free survival that, because of the small numbers in the trial, did not reach statistical significance. These were 86% (95% CI 77% to 91%) for group B, 81% (95% CI 71% to 87%) for group A, 73% (95% CI 64% to 81%) for group C, and 73% (95% CI 63% to 81%) for group D (group B vs. group A: HR 0.69, 95% CI 0.34 to 1.40; group C vs. group A: HR 1.25, 95% CI 0.68 to 2.30; group D vs. group B: HR 2.05, 95% CI 1.07 to 3.93). NeoSphere prompted the accelerated approval of neoadjuvant pertuzumab with trastuzumab and chemotherapy from the FDA,149 the EMA150 and NICE. 151 The promising early results of neoadjuvant pertuzumab had already led to the APHINITY trial, which tested the addition of pertuzumab to standard chemotherapy and trastuzumab treatment in the adjuvant setting. 141 This trial had already completed recruitment when the FDA stipulated that if the adjuvant APHINITY trial did not demonstrate benefit from the addition of pertuzumab, then approval for its use in the neoadjuvant setting would be withdrawn.

In the recently published Katherine Trial,152 trastuzumab emtansine (Kadcyla^®; Roche, Switzerland) was tested against trastuzumab after the failure of neoadjuvant chemotherapy and anti-HER2 therapy to produce a complete pathological response. The estimated proportion of patients who were free of invasive disease at 3 years was 88.3% in the trastuzumab emtansine group and 77.0% in the trastuzumab group. Invasive DFS was significantly higher in the trastuzumab emtansine group than in the trastuzumab group (HR for invasive disease or death 0.50, 95% CI 0.39 to 0.64; p < 0.001). This significant improvement in outcome is likely to lead to an appropriate escalation of standard treatment in the post-neoadjuvant setting, at least for patients who do not achieve a pathological complete response with neoadjuvant treatment.

Trials are now being developed of the de-escalation of adjuvant anti-HER2 therapy after a pathological complete response to neoadjuvant treatment has been achieved. This is an important development in the portfolio of trials for HER2-positive early breast cancer.