Systematic review of the clinical effectiveness and cost-effectiveness, and economic evaluation, of denosumab for the treatment of bone metastases from solid tumours

Breast cancer

Bone metastases and their consequences depend on the type of primary tumour. Breast cancer is the commonest cancer in women. In the UK, approximately 124 women per 100,000 are diagnosed with breast cancer each year. 2 Approximately 0.5% of women have bone metastases at diagnosis, with 4.7% developing bone metastases in 5 years. 3 Bone metastases are associated with reduced median survival of approximately 24 months and 5-year survival of 20%. 4 However, survival is more heavily dependent on the presence of visceral organ metastases. Breast cancer commonly spreads to bone, liver, lung and brain. It has been estimated that breast cancer patients with metastatic disease only to bone survive 6 months longer than those with bone metastases and metastases outside a bone (1.6 years compared with 2.1 years). 5

Breast cancer most commonly originates from cells lining ducts or lobules (namely ductal carcinoma or lobular carcinoma). The natural history of the tumour is dependent on a range of different variables which, in turn, contribute to classification. Tumour–node–metastasis (TNM) is the most important prognostic classification and refers to the size of the tumour (T), spread to lymph nodes (N) and presence of metastases (M). Low-grade or precancerous cells are referred to as in situ carcinoma and do not cause metastases, unless the tumour progresses to an invasive carcinoma. Tumour aggressiveness can be predicted by the degree to which tumour cells are differentiated; poorly differentiated cells tend to be more aggressive, whereas well-differentiated cells are less so. Treatment and prognosis depend on receptors expressed by tumour cells. The three most important are oestrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER-2). Generally tumours that are receptor negative are less responsive to treatment and have a worse prognosis.

Prostate cancer

In men the most common cancer is prostate cancer. Approximately 98 men per 100,000 are diagnosed with prostate cancer in the UK each year. Almost 24 men per 100,000 each year die because of prostate cancer. 2 Prostate cancer often progresses to involve bone. At diagnosis 22% of patients have stage IV disease and a further 25% will develop clinically detectable metastases over the course of the disease. 6 One study found that 90% of patients with prostate cancer had some evidence of bone involvement at death. 7 Survival is reduced considerably in the presence of bone metastases, and 5-year survival drops from 56% in patients without bone metastases to 3% in patients with bone metastases. 8 However, this does not imply that bone metastases cause death per se, but rather, they occur in more aggressive cancers.

Prostate cancer originates in glandular cells and is therefore categorised as an adenocarcinoma. Similar to breast cancer, the TNM classification is the most important prognostic indicator. A worse prognosis is associated with the presence of disease in lymph nodes, or beyond. The grade of tumour cells is measured using the Gleason score. A high Gleason score suggests a poorly differentiated tumour and therefore poorer prognosis. Prostate-specific antigen (PSA) is a protein released by the prostate and can be a marker for cancer. However, there has been much debate around PSA testing. High levels of PSA can be found in patients without cancer and normal levels can be found in patients with cancer. 9 Prostate tumours are dependent on androgens to progress. Therefore, antiandrogen treatment can delay progression by either chemical or surgical castration. When tumours respond to castration therapy they are classified as castration-sensitive prostate cancer (CSPC), and when tumours no longer respond to castration treatment they are classified as castration-resistant prostate cancer (CRPC). Hormone-sensitive and hormone-refractory nomenclature has been used. However, some tumours remain dependent on androgens (and amenable to further androgen deprivation)10 to progress irrespective of castration therapy; here, the term castration resistant is more accurate.

Lung cancer

Lung cancer is the second commonest cancer, after breast (in women) and prostate (in men), and has an incidence of 48 per 100,000 per year. Lung cancer prognosis is very poor. More people die from lung cancer each year than from any other cancer (40 patients per 100,000). 2 One-year survival is 25% (in men) and 26% (in women). Five-year survival is only 7.8% (in men) and 8.7% (in women) and reflects cancers that are detected early, at a surgically resectable stage. 11 Spread of tumour to bone is common in lung cancer. Up to 36% of patients with lung cancer have evidence of bone metastases at death. 12 Other organs to which lung cancer often metastasises include the adrenal glands and the brain.

Classification of lung cancer is histological. Non-small cell lung cancer (NSCLC) and small cell lung cancer (SCLC) constitute more than 95% of all lung cancers. NSCLC includes squamous cell carcinoma, adenocarcinoma and large cell carcinoma. SCLC carries a worse prognosis and metastases are usually present at diagnosis. Both SCLC and NSCLC are staged using the TNM classification, or categorised as stage IA (better prognosis) to IV (worse prognosis).

Other solid tumours

Almost any cancer can metastasise to bone. At autopsy, 35–42% of thyroid, renal and bladder tumours have evidence of bone metastases. 13 Colorectal cancer mainly spreads to the liver, but in 6–10% of cases metastasises to bone. 14,15 Since colorectal cancer is the third commonest cancer, after breast (in women), prostate (in men) and lung, the actual number of patients with bone involvement is considerable. Each cancer has different subclassifications, each with its own pathophysiology, treatment and prognosis. For example, papillary thyroid cancer has a very good prognosis compared with anaplastic thyroid cancer. Bladder tumours may be superficial, requiring only local ablation therapy, or may be muscle invasive, requiring surgical resection or radical radiotherapy to the bladder. Therefore, the pathway to bone metastases in each cancer type varies according to primary site, cell type, classification and antineoplastic treatment.

Investigations for bone metastases and skeletal-related events

Bone metastases and SREs can be measured in several different ways. 25 At the time of cancer diagnosis clinicians may screen for metastases. The decision to screen depends on stage of tumour and patients' symptoms. Skeletal scintigraphy (bone scan) uses injected radioactive material, which is then scanned with a gamma camera. Areas of increased bone metabolism are shown. This test shows the whole skeleton and is advantageous for a broad examination of the skeleton in asymptomatic patients. Plain radiographs (X-rays) are used for investigation of specific bones where metastases are suspected. Other investigations can then be used to investigate bone lesions, such as computerised tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET) and single-photon emission CT (SPECT).

Bone markers, measured in blood or urine, have been used to monitor bone turnover in clinical trials. Patients with bone metastases and elevated bone markers are at increased risk of SREs. 26 It has been suggested that bone markers could be used to stratify risk of SRE in individuals with bone metastases, assisting in the choice of bone-modifying agents and monitoring treatment response. 27,28 There are several different bone markers, including BSAP, osteocalcin and N-terminal type 1 procollagen peptides (PINPs) markers for monitoring bone formation, and urinary or serum collagen type 1 cross-linked C telopeptide (CTX) and urinary collagen type 1 cross-linked N-telopeptide (NTX) for monitoring bone resorption. Denosumab trials have included measures of NTX and BSAP as secondary outcomes. 29–31 NTX increases in response to osteoclast-mediated bone resorption and can be measured in the blood or urine. During BP treatment, normalised levels of NTX appear to be associated with a reduced risk of SREs. 32,33 BSAP reflects osteoblastic activity by measuring bone formation. BP and denosumab treatment have been found to reduce BSAP. Conversely, persistent elevation of BSAP despite BP treatment is associated with increased SREs. 32 American Society of Clinical Oncology (ASCO) guidelines do not recommend the use of bone markers outside the trial setting. 34

In routine clinical practice acute uncomplicated pathological fractures are generally investigated by plain radiography. In the trial setting, regular skeletal surveys have been used to screen and diagnose pathological fractures. A skeletal survey is performed by taking plain radiographs of the skull, chest, spine, pelvis and long bones of the arms and legs. Therefore, both asymptomatic (lesions demonstrated radiologically but the patient does not complain of any symptoms) and symptomatic fractures will be observed. For pathological fractures of the spine, plain radiographs may not be sufficient. There may be uncertainty about the presence of a fracture and plain radiographs do not assess the integrity of the spinal canal. In this scenario, imaging with a MRI or CT scan may be necessary. In the case of suspected SCC, MRI is the investigation of choice.

Hypercalcaemia often presents with non-specific symptoms and is easily diagnosed on blood test. Signs and symptoms worsen as serum calcium increases. A serum calcium of more than 2.6 mmol/l is suggestive of hypercalcaemia.

Measuring skeletal-related events

There are several ways of recording SRE data in clinical trials:

• time to first SRE

• time to first and subsequent SREs (multiple event analysis)

• SRE incidence

• proportion of patients with at least one on-study SRE

• skeletal morbidity rate (SMR) – number of events per year

• skeletal morbidity period rate (SMPR) – the number of 12-week periods with new SREs divided by the total observational time.

It is important to note that SRE as a composite end point includes both complications of bone metastases (pathological fracture and SCC) and therapeutic or preventative measures (radiotherapy and surgery). Caution is needed because radiotherapy and surgery would be considered best supportive care (BSC). 35,36 Therefore, measures of radiotherapy and surgery contribute to both the treatment and the outcome measure.

Trinkaus and colleagues37 compared observational SRE frequency in ‘real life’ with SRE frequency in the intravenous BP trials. They found that the rate of SREs was higher in the trial setting than in ‘real life’. This may reflect the fact that bone scans are undertaken fequently in trials.

The various methods of assessing SRE data have evolved to overcome specific problems.

Some outcomes, such as proportion of patients with at least one on-study SRE or SMR, fail to consider time delays in SREs. For example, an individual who suffers SCC on day 1 of a trial is considered equivalent to an individual who suffers SCC after a year. To overcome this issue, time to first SRE can be measured. This outcome does not distinguish the number or timing of subsequent SREs. Consequently, the multiple-event analysis was developed. 38 The Andersen–Gill system is the commonest method used for multiple-event analysis. It includes a measure of both time and number of events. This method has been criticised because it fails to differentiate between individuals who have died and individuals who have left the trial for another reason. 39 Other methods have been described that also attempt to take mortality into account. 40,41

The choice of SRE measure depends on what is considered the most important outcome. To measure SRE prevention, the proportion of patients experiencing a SRE would be more suitable. To measure a reduction in rate, SMR/SMPR would be most appropriate. However, to measure delay, time to first or time to first and subsequent SRE would be more appropriate.

The situation is made more complex because more than one SRE may occur in relation to a single event and therefore the second SRE is dependent on the first. For example, an individual may suffer a pathological fracture, which is treated by radiotherapy or surgery (two SREs). In the pivotal denosumab and BP trials, a subsequent SRE is counted only after a 21-day period. This is not the case for SMR, which assumes independence for each event and can therefore lead to multiple counting of events. In an attempt to address this issue the SMPR outcome has been used.

The incidence of SREs is generally not considered appropriate because of underestimation of time variability within the data (similar criticism could be made of SMR). 42 A patient who suffers several SREs within the first 6 months is considered equivalent to a patient who suffers the same number of events over several years. The former patient is likely to have a reduced quality of life compared with the latter.

Trials have consistently used SRE as a composite outcome. This undoubtedly increases efficiency and power, but some caution is needed. However, the impact on health-care resources and a patient's quality of life is vastly different for SCC compared with an asymptomatic rib fracture. Nor does this SRE composite outcome directly measure factors that are important to patients such as mobility or pain (these are measured indirectly through need for radiotherapy or surgery). 43

Bisphosphonates

Bisphosphonates reduce bone resorption by inhibiting osteoclasts. 54 Clinical effectiveness starts after 6–12 months of treatment. 55 There are first-, second- and third-generation BPs. Early non-aminobisphosphonates include clodronate and etidronate. The addition of a nitrogen group to the BP structure was found to increase potency by inhibition of the 5-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) reductase pathway. These aminobisphosphonates include ibandronic acid, disodium pamidronate and zoledronic acid.

During the early studies of oral nitrogen-containing BPs, an association with oesophagitis was frequently reported. 56 Therefore, zoledronic acid and disodium pamidronate are available only as intravenous preparations. Ibandronic acid is available as an oral or intravenous preparation. Intravenous BPs are excreted rapidly from the kidneys and are typically associated with a higher incidence of hypocalcaemia and renal impairment than oral BPs. 57 Administration time varies from 15 minutes for zoledronic acid to 120 minutes for disodium pamidronate.

Oral BPs are absorbed by passive diffusion in the gastrointestinal tract. As a result, less than 6% of the active compound is absorbed, and this is further reduced with the presence of food. In addition, oral BPs increase the risk of oesophageal erosions, inflammation and neoplasm. 58 It is therefore recommended that patients remain upright for 30–60 minutes after ingestion. Consequently, oral BPs become burdensome for patients. 59 Location of treatment is important to patients. One study found that patients prefer administration at home, but this is not often possible with intravenous treatments. 60

Bisphosphonates are considered to be relatively safe drugs. Possible adverse reactions include renal failure, osteonecrosis of the jaw (ONJ), hypocalcaemia and acute-phase reaction. To avoid renal impairment, renal function is checked before administration, dose is adjusted if necessary and the intravenous infusion is given slowly. McDermott and colleagues61 assessed predictors of renal impairment in patients given zoledronic acid. The following predictive factors were found on multivariate analysis: age, myeloma or renal cell cancer, number of doses, concomitant non-steroidal anti-inflammatory drug therapy and current or prior treatment with cisplatin. ONJ has only recently been associated with BPs;62 ONJ leads to oral or periodontal lesions, which are usually associated with previous dental procedures. Hypocalcaemia can be rectified with oral calcium. Acute-phase reaction usually presents with transient pyrexia following first administration.

Four BPs are currently licensed in the UK for bone metastases:

1. Zoledronic acid (Zometa™, Novartis, Basel, Switzerland) is licensed for the reduction of bone damage in advanced malignancies involving bone. It is administered by intravenous infusion over at least 15 minutes at a dose of 4 mg every 3–4 weeks.

2. Disodium pamidronate (Aredia^®, Novartis) is licensed for osteolytic lesions and bone pain in bone metastases associated with breast cancer or multiple myeloma. It is administered by slow intravenous infusion (over at least 2 hours) at a dose of 90 mg every 4 weeks.

3. Sodium clodronate (Bonefos™, Bayer Schering, Berlin, Germany; Clasteon™, Beacon, Tunbridge Wells, UK; Loron 520™, Roche, Basel, Switzerland) is licensed for osteolytic lesions, hypercalcaemia and bone pain associated with skeletal metastases in patients with breast cancer or multiple myeloma. It is administered by mouth at a dose of 1.6–3.2 g daily.

4. Ibandronic acid (Bondronate™, Roche) is licensed for the reduction of bone damage in bone metastases in breast cancer. It is administered either by mouth (50 mg daily) or by intravenous infusion (6 mg every 3–4 weeks).

Therefore, zoledronic acid is the only drug licensed for cancer involving bone, other than breast or multiple myeloma. Zoledronic acid has been the most studied BP and, according to expert opinion, is the most widely used BP. The patent for zoledronic acid is expected to expire in 2013. There are currently no firm criteria to advise when BPs should be stopped.

Best supportive care

Best supportive care varies between each primary cancer type.

In patients with breast cancer and bone metastases, BSC encompasses the use of BPs to prevent SRE and reduce pain. However, for the purpose of this report, the definition of BSC does not include BPs. Pain is also managed by the use of both simple and opioid analgesics, corticosteroids and non-steroidal anti-inflammatory agents. External beam radiotherapy is used to control pain at specific sites and, less commonly now, systemic radiopharmaceuticals may be used to alleviate widespread pain at multiple sites not controlled by other means. All patients with metastases in a long bone should be assessed for the risk of pathological fracture and referred to an orthopaedic surgeon for consideration of prophylactic fixation. Not all patients will require treatment with all modalities discussed above. The NICE guidelines currently recommend that all patients with bone metastases receive a BP, while ASCO guidelines recommend the use of a bone-modifying agent in patients with bone metastases and evidence of bone destruction. There is variation in the use of the other interventions mentioned, dependent on local practice and patient factors.

In patients with bone metastases, current BSC encompasses the use of systemic anticancer therapies including chemotherapy and further hormone therapies. Palliative external beam radiotherapy and systemic radionucleotides, such as strontium-89, are widely used and may be used on multiple occasions to treat metastatic bone pain. Despite these measures, pain may continue to be burdensome, and analgesics, often requiring specialist pain services, are frequently required. Attitudes to systemic anticancer therapies used in this context vary across the UK; in particular, there remains widespread controversy about the optimal timing of docetaxel-based chemotherapy, some clinicians opting to use it to prevent symptoms such as bone pain, whereas others save it until symptoms become burdensome. Two new drugs, cabazitaxel and abiraterone acetate, which are licensed for this indication, may change BSC patterns in this population, but neither drug has been the subject of published NICE review and access outside of clinical trials remains limited in the UK. The treatment of SRE is similar to that of other solid tumours (OSTs). Pathological fractures can be treated or prevented with surgery, radiotherapy or analgesics. Current practice is that BPs are not given to prevent complications of bone metastases, such as pathological fractures and SCC. However, BPs are used to treat pain when first-line analgesics have not alleviated pain.

In patients with lung cancer with bone metastases, BSC may include chemotherapy, palliative radiotherapy, antibiotics, steroids, surgery, analgesics and antiemetics. 63 Certain treatments are aimed at slowing disease progress (chemotherapy), while others are aimed at alleviating (analgesics and antiemetics) or preventing (surgery to prevent pathological fracture) symptoms. BSC may vary according to the location or primary tumour and presence of distal metastases. BPs are generally not used to prevent SREs. However, clinicians may consider BPs as a second-line analgesic option for painful bone metastases. BSC for pathological fracture and SCC in lung cancer is similar to that for OSTs.

Current treatments of skeletal-related events

Treatment of pathological fractures depends on the severity of injury, the bones involved and the degree of destruction. Management options include analgesics, immobilisation, surgical fixation, radiotherapy or a combination of the above. The impact of pathological fractures varies widely; some may be unnoticed and asymptomatic while more severe fractures may be associated with SCC and paraplegia.

Management of metastatic SCC has been described. 47 The guidelines highlight the need for early diagnosis and imaging with MRI. Acute treatment recommendations include good nursing care, corticosteroids and appropriate case selection for surgery or radiotherapy. Moreover, the guidelines make recommendations for long-term care, including management of pressure ulcers, bladder or bowel incontinence, postural hypotension and lung secretions, prevention of thromboprophylaxis and planning for rehabilitation or long-term care.

Hypercalcaemia of malignancy can present with various different signs and symptoms. If untreated, HCM can lead to confusion, drowsiness or coma. Rehydration and BP treatment are the cornerstone of management. Loop diuretics and steroids can also be used. Older agents such as plicamycin, calcitonin and gallium nitrate are not commonly administered.

Variation in service

There is variation among oncologists in the choice of BPs and more so in breast cancer, for which four BPs are licensed. With no clear guidelines about which BP to use, the decision is often made by the individual clinician. Based on expert opinion, zoledronic acid is the most widely used BP.

Bisphosphonates are used consistently in breast cancer; however, the use of BPs in other cancers varies. Among patients with metastatic tumours other than breast cancer, some clinicians use BPs routinely, wherease others reserve BPs only for uncontrolled pain and still others rarely use BPs. With the imminent patent expiry of zoledronic acid and the anticipated reduction in price, patterns of use may change significantly in the near future.

Fallowfield and colleagues64 conducted a UK survey to evaluate BP prescribing habits among oncologists. They found that 53% of oncologists gave intravenous and oral drugs, 40% gave only intravenous drugs and 7% gave only oral drugs. Zoledronic acid (56–85%) and disodium pamidronate (23–42%) were the commonest intravenous drugs, and ibandronic acid (66%) was the commonest oral BP used. Reasons reported for using oral preparations included ‘health authority/primary care trust only funds oral preparation’, ‘local guidelines dictate which patients receive oral/intravenous’ and ‘intravenous preparations are not listed on the local formulary’.

Variation in BSC exists between treatment centres. Local policy, available resources and clinician prescribing habits all affect the likelihood of patients being offered certain BPs, analgesics or antineoplastic medications.

Bisphosphonates

Bisphosphonates considered as a comparator included:

• sodium clodronate

• disodium pamidronate

• ibandronic acid

• zoledronic acid.

Etidronate was initially considered as an unlicensed (for this purpose) comparator, because of its much lower cost. However, clinical advice suggests that it should be used infrequently because it may cause gastrointestinal toxicity.

Currently, zoledronic acid has UK marketing authorisation for the reduction of bone damage in all advanced malignancies involving bone. Disodium pamidronate and sodium clodronate are licensed for breast cancer and multiple myeloma, and ibandronic acid is licensed only for breast cancer. However, we also considered inclusion of trials of these BPs when used outside their licensed indications.

Clinical experts and NICE guidelines were consulted to determine the place of BPs in the care pathway. For patient groups in which BPs are considered the current standard of care, denosumab was compared with BPs only.

A BP class effect was not assumed. As data allowed, all BPs would be included within a NMA.

Best supportive care (excluding bisphosphonates)

Best supportive care was considered a comparator where BPs were not recommended. This varied depending on the type of cancer. The relevant NICE CGs are CG81 for advanced breast cancer,45 CG58 for prostate cancer,46 CG121 for lung cancer48 and CG75 for metastatic SCC. 47 All of these guidelines recommend radiotherapy and analgesics within BSC. Other recommended supportive care for bone metastasis includes surgical fixation in breast cancer and multiple myeloma, strontium-89 in prostate cancer and nerve blocks in lung cancer.

Breast cancer

NICE CG81 on breast cancer recommends offering BPs to patients with newly diagnosed bone metastases to prevent SREs and to reduce pain. 45 Therefore, BSC was not used as a comparator in patients with advanced breast cancer and bone metastases. The planned NMA is shown in Figure 1.

FIGURE 1.

Network meta-analysis for those with bone metastases from breast cancer.

Prostate cancer

The NICE guidance, CG58, on prostate cancer recommends that ‘the use of BPs to prevent or reduce the complications of bone metastases in men with hormone-refractory prostate cancer is not recommended. Bisphosphonates for pain relief may be considered for men with hormone-refractory prostate cancer when other treatments (including analgesics and palliative radiotherapy) have failed’. 46 Therefore, in prostate cancer denosumab is compared with both BPs and BSC.

Lung cancer

No guideline recommendation for the use of BPs exists for bone metastases from lung cancer. NICE CG121 suggested that there was insufficient evidence to recommend BPs as a first-line treatment in bone metastases from lung cancer. 66 However, the standard treatments such as analgesics, or single-fraction radiotherapy, are recommended for the relief of symptoms from bone metastasis.

As the NICE guidelines for prostate and lung cancer recommend BSC before giving a BP, for these patient groups we plan to include BSC as a comparator, where data exist. The planned NMA for prostate cancer, lung cancer and OSTs is shown in Figure 2.

FIGURE 2.

Network meta-analysis for those with bone metastases from prostate cancer, lung cancer or OST.

Other solid tumours

In the protocol we stated that if we obtained enough data on OSTs for which no relevant NICE guidelines existed, we would seek expert opinion as to the place of BPs in the clinical pathway.

Expert opinion suggested that BPs, mainly zoledronic acid, were used in OSTs. Therefore, the network diagram will be as in Figure 2 and denosumab is compared with both BPs and BSC.

Overall

A flow diagram outlining the screening process for the overall review of clinical effectiveness is shown in Figure 3.

FIGURE 3.

Flow diagram of the searches and screening process.

The searches identified 989 records, of which 585 were unique studies (after removing duplicates). Following screening of titles and abstracts, the full text of 352 articles was obtained for further assessment. With the addition of four reports received from the manufacturer, this resulted in 39 studies (74 reports) meeting the inclusion criteria for the review of clinical effectiveness (see Appendix 4). However, of these 39 studies, 31 were not able to contribute data to the AG's NMA and none reported denosumab, and therefore these studies were not reported further in the results chapters. The reasons why they were not able to contribute data to the NMA included:

1. studies did not report uniform definition of SREs

2. studies did not report standardised SRE rates

3. studies did not report outcomes separately for different cancer types

4. studies included patient groups where some patients were not diagnosed with bone metastases.

Of these 31 studies, 6 reported on bone metastases from breast cancer,71–76 13 reported on bone metastases from prostate cancer77–89 and 12 reported on bone metastases from OSTs. 90–101

Of the remaining eight studies that did contribute data to the network meta-analyses, four reported breast cancer31,102–104 [18 reports22,31,102–116; and Amgen Ltd. Multiple Technology Appraisal: Denosumab for the treatment of bone metastases from solid tumours (unpublished report). London: National Institute for Health and Care Evidence; 2011], two reported prostate cancer29,117 (15 reports19,29,117–129) and two reported OSTs, excluding breast and prostate cancer30,130 (seven reports30,130–135). Therefore, across the review of clinical effectiveness, eight studies (40 reports) contributed data to the NMAs.

All of the included studies were RCTs. No systematic reviews were identified that exactly met our inclusion criteria. The ASCO clinical practice guideline update on the role of bone-modifying agents in metastatic breast cancer was the most relevant systematic review identified. This review included denosumab, disodium pamidronate and zoledronic acid but did not include ibandronate or clodronate (because they are not licensed for this indication in the USA), which, therefore, were not considered further. 34

A search of safety-related articles identified 28 additional studies. 61,62,136–161

Breast cancer

The primary comparator for denosumab was considered to be BPs (zoledronic acid, disodium pamidronate, ibandronic acid or sodium clodronate) as recommended in NICE guideline CG81 for all patients with advanced breast cancer and newly diagnosed bone metastases. 45

One RCT (10 reports,31,105,106,110–114,116 including CSR 20050136) was identified comparing denosumab with zoledronic acid, with the primary published report considered to be that by Stopeck and colleagues. 31 An additional three studies contributed data to the NMA. One study, by Kohno and colleagues,102 compared zoledronic acid with placebo. One study (four reports22,103,107,115) compared disodium pamidronate with placebo, with the primary published report considered to be that by Lipton and colleagues. 103 One study (three reports104,108,109) compared zoledronic acid with disodium pamidronate, with the primary published report considered to be the 2003 paper by Rosen and colleagues. 104

Overall

All 31 studies that were excluded from the NMA included comparisons of BPs with placebo or another BP, and some compared BSC with placebo or another BSC. Table 2 provides a summary of the interventions and comparators included in the trials and a list of studies included or excluded from the NMA. Studies meeting the inclusion criteria but not contributing data to the NMA were not reported on in the chapters on clinical effectiveness because none provided direct evidence on denosumab compared with BPs, placebo or BSC. However, the results from these studies have been presented in appendices; see Appendix 6 for the characteristics of the participants and description of the interventions/comparators with the reasons for exclusions from the NMA and Appendix 7 for the results of these studies. Appendix 8 shows the characteristics of the included studies.

Breast cancer

Table 3 shows summary information for the four studies that provided direct evidence for denosumab or were included in the NMA. The study by Kohno and colleagues102 was undertaken between May 2000 and May 2003 and enrolled adults with at least one osteolytic bone metastasis from breast cancer from 51 centres in Japan. Patients received 4 mg zoledronic acid or placebo every 4 weeks for 12 months. The primary outcome was the ratio of the SRE rate (defined as the total number of SREs divided by the total years on study) for patients treated with zoledronic acid divided by the SRE rate for the placebo group. Follow-up was 52 weeks. The study was funded by Novartis.

The study by Lipton and colleagues103 reports results of two similarly conducted RCTs. 22,115 The studies were undertaken between 1990 and 1996 and enrolled women with stage IV breast cancer and at least one predominantly lytic metastatic bone lesion measuring ≥ 1 cm from 106 centres in the USA, Canada, Australia and New Zealand. Patients received 90 mg disodium pamidronate every 3–4 weeks or placebo every 4 weeks for 24 cycles. The primary outcome was the SMR, defined as the ratio of the number of skeletal complications experienced by a patient divided by the time on the trial for that patient (expressed as the number of events per year). Follow-up was 24 months. The study was funded by Novartis.

The study by Rosen and colleagues104 was undertaken between October 1998 and January 2000 and enrolled women with at least one bone metastasis (osteolytic, osteoblastic, or mixed) secondary to stage IV breast cancer. The primary analysis of this study included advanced multiple myeloma, but a subgroup of those patients with breast cancer is presented separately. 110 The study was described as multicentre and international. Patients received 4 mg zoledronic acid or 90 mg disodium pamidronate every 3–4 weeks for 24 months. Zoledronic acid was initially infused over 5 minutes in 50 ml of hydration solution. However, because of concerns over renal safety a protocol amendment in June 1999 changed the infusion time to 15 minutes and increased the volume of the infusion to 100 ml. The primary outcome was the proportion of patients who experienced at least one SRE during the study period. Follow-up was 25 months. The study was funded by Novartis.

The study by Stopeck and colleagues31 was undertaken between April 2006 and December 2007 and enrolled women with confirmed breast cancer and at least one bone metastasis from 322 centres in Europe, North America, South America, Japan, Australia, India and South Africa. However, few (academic-in-confidence information has been removed) of patients were from the UK (MS). Patients with creatinine clearance < 30 ml/minute, prior intravenous BP treatment, current or prior oral BPs for the treatment of bone metastases, non-healed dental/oral surgery and prior malignancy within 3 years before random assignment were excluded. Patients received a subcutaneous injection of 120 mg denosumab and an intravenous infusion of placebo or an intravenous infusion of 4 mg zoledronic acid and a subcutaneous injection of placebo every 4 weeks. The study was powered to detect both non-inferiority and superiority with respect to time to first on-study SRE (primary outcome), and risk of first and subsequent on-study SREs. Follow-up was around 34 months. The study was funded by Amgen and Daiichi Sankyo.

Skeletal-related events by type

In the denosumab RCT Stopeck and colleagues31 did not report SRE by type. The MS reported that denosumab reduced the risk for time to radiation in bone by (academic-in-confidence information has been removed) compared with zoledronic acid. Table 6 shows the distribution of first on-study SRE by type of SRE in the denosumab study. The distribution of type of SRE was similar across the treatment groups, with radiation to bone and pathological fracture being the most commonly occurring.

In the study by Lipton and colleagues,103 the median time to first pathological fracture was significantly longer in the disodium pamidronate group compared with the placebo group (by almost 12 months). The time before requiring bone radiation was not reached in the disodium pamidronate group compared with a median of 16 months in the placebo group (p < 0.001). 103

History of skeletal-related events

The MS reported time to first on-study SRE by history of SRE for the denosumab study 136 (Table 7). This showed that for those without a prior SRE (academic-in-confidence information has been removed). Covariate analysis showed that patients with a prior SRE history had an increased risk (academic-in-confidence information has been removed) compared with those without a SRE history.

The study by Rosen and colleagues,104 comparing zoledronic acid with disodium pamidronate, reported time to first on-study SRE by lytic and non-lytic subgroup. There was no significant difference between the non-lytic treatment groups. For those lytic cases, the time to first SRE was much longer in the zoledronic acid (∼ 10.3 months) group compared with the disodium pamidronate group (∼ 5.8 months).

Skeletal-related events by type

None of the studies reported risk of first and subsequent SREs by individual SRE type.

The MS reported the distribution of first and subsequent on-study SRE by type of SRE in the denosumab RCT (study 136) (Table 9). As for first on-study SRE by type, the distribution of type of SRE was similar across the treatment groups, with radiation to bone and pathological fracture again the most commonly occurring.

Prior history of skeletal-related events

The MS reported risk of first and subsequent on-study SRE by history of SRE for study 136 (Table 10). (Academic-in confidence information has been removed.) Covariate analysis as presented in the manufacturer's table showed that patients with a history of SRE had an increased risk (academic-in confidence information has been removed) compared with those without a SRE history.

Skeletal-related events by type

The MS did not report SMR by type of SRE.

The study by Lipton and colleagues103 comparing disodium pamidronate with placebo reported SMR for different types of SREs including radiation to bone, radiation to bone for pain relief, pathological fracture, surgery to bone, SCC and hypercalcaemia. A statistically significant difference was reported between disodium pamidronate and placebo for all types of SRE other than SCC. Among all the SREs, the highest rate (events per year) was reported for pathological fracture (1.6 vs 2.2) and the lowest rate was reported for SCC (0.07 vs 0.37).

Prior history of skeletal-related events

The MS did not report SMR by prior history of SREs.

In the study by Kohno and colleagues102 the SRE rate reduction for zoledronic acid was more than 30% higher in patients without a prior fracture (unadjusted SRE rate ratio 0.43) than in patients with a prior fracture (unadjusted SRE rate ratio 0.81).

In the subgroup analysis of patients with lytic lesions, Rosen and colleagues104 reported SRE rates in the zoledronic acid arm (1.16 events per year) that were almost half of those in the disodium pamidronate arm (2.36 events per year; p = 0.008). In those with non-lytic lesions, the difference between the treatment groups for SRE rate was reported to be non-significant (0.81 vs 0.97; p = 0.904).

Skeletal-related events by type

The MS did not report this outcome.

The studies by Kohno and colleagues102 and Lipton and colleagues103 reported the proportions of patients experiencing types of SRE at 1 year and 2 years, respectively. For each type of SRE reported (other than for SCC in the study by Lipton and colleagues103), the BP group experienced lower rates compared with placebo. In the study by Lipton and colleagues, 103 the difference between the treatment groups for each type of SRE was statistically significant other than for SCC. In both studies the most frequently occurring type of SRE was fractures (25.4% vs 39.8% at 1 year in the study by Kohno and colleagues102 and 40% vs 52% at 2 years in the study by Lipton and colleagues103), followed by radiation to the bone.

In a subgroup analysis comparing patients with lytic and non-lytic lesions, Rosen and colleagues104 reported a non-significant difference for the proportion experiencing a SRE between zoledronic acid and disodium pamidronate in each subgroup at 13 months.

Prior history of skeletal-related events

None of the studies reported incidence of SRE by prior history of SREs.

Prior history of skeletal-related events

None of the studies reported overall survival by prior history of SREs.

Functional Assessment of Cancer Therapy

The Functional Assessment of Cancer Therapy – Breast (FACT-B) questionnaire consists of the Functional Assessment of Cancer Therapy -General (FACT-G) questionnaire plus additional questions specific to breast cancer. For each component of the FACT-B [FACT-G total score, FACT-B total score, physical well-being domain, functional well-being domain and trial outcome index (TOI): a composite of the functional well-being domain, physical well-being domain, and the prostate cancer subscale], a higher score indicates better HRQoL.

Stopeck and colleagues31 reported quality of life using the FACT-G questionnaire completed by patients at baseline, day 8, and before each monthly visit through to the end of the study (73 weeks). At 73 weeks 30% of patients had discontinued the study (academic-in-confidence information has been removed) (CSR 136).

Patients were divided into two subgroups at baseline: no/mild pain or moderate/severe pain, based on BPI. For those with no/mild pain at baseline, an average of 4.1% more patients (range −0.6% to 9.3%) treated with denosumab had a ≥ 5-point increase in the FACT-G score and an average of 2.4% fewer patients (range −4.4% to 6.3%) had a ≥ 5-point decrease in the FACT-G score at 18 months compared with those patients treated with zoledronic acid. For those with moderate/severe pain at baseline, a similar proportion of patients treated with denosumab had either a ≥ 5-point increase (average 3% more; range −1.7% to 7.9%) or a ≥ 5-point decrease (average 3.5% fewer; range −1.1% to 11.5%) in the FACT-G score at 18 months compared with those treated with zoledronic acid. 106 An average of 3.2% (range 1% to 7%) more patients in the denosumab group experienced a clinically meaningful improvement in quality of life (≥ 5-point increase in FACT-G total score) from week 5 through to week 73. 105

European Quality of Life-5 Dimensions

For both components of EQ-5D [the health index and the visual analogue scale (VAS)], a higher score indicates a more preferred health status. For the health index questions of the EQ-5D, a three-level response was used to assess quality of life (academic-in-confidence information has been removed) (CSR 136).

(Academic-in-confidence information has been removed) (CSR 136).

Lipton and colleagues,103 comparing disodium pamidronate with placebo, reported mean change in the quality-of-life scores from baseline to 24 months and to the last visit. Quality of life was evaluated using the Spitzer quality-of-life index. From baseline to the last visit quality of life worsened in both the disodium pamidronate group (−1.80; SD 2.81) and the placebo group (−2.13; SD 2.63) (p = 0.088).

Hypocalcaemia

The MS reported that hypocalcaemia events were mainly non-serious and transient and resolved either spontaneously or with calcium supplementation (MS). More hypocalcaemia adverse events occurred in the denosumab group than in the zoledronic acid group [5.5% (56/1020) vs 3.4% (34/1013) respectively].

Kohno and colleagues102 reported that 39% of the zoledronic acid group and 7% of the placebo group experienced grade 1 hypocalcaemia. There were no grade 2 or 3 hypocalcaemia events in the zoledronic acid group, while one patient in each group experienced grade 4 hypocalcaemia. 102 Lipton and colleagues,103 comparing disodium pamidronate with placebo, reported that one patient (1/367) discontinued disodium pamidronate after a symptomatic hypocalcaemia episode. Rosen and colleagues104 did not report this outcome in their study comparing zoledronic acid with disodium pamidronate.

An observational study165 reported on 177 patients receiving BPs over 13 months. They found the incidence of hypocalcaemia to be 15.8% in patients treated with zoledronic acid over this period. However, this study included all grades of hypocalcaemia.

Osteonecrosis of the jaw

The rates of ONJ in the denosumab RCT were low and similar between the denosumab group and the zoledronic acid group [2.0% (20/1020) vs 1.4% (14/1013); p = 0.39]. 31 The cumulative incidence of ONJ in the denosumab and zoledronic acid groups, respectively, was 0.8% and 0.5% at 1 year, 1.9% and 1.2% at 2 years, and 2.0% and 1.4% at 3 years. 31 Stopeck and colleagues31 reported that, as of February 2010, 10 (50%) denosumab-treated patients and six (43%) zoledronic acid-treated patients had resolution of the ONJ event; 10 (50%) denosumab-treated patients and nine (64%) zoledronic acid-treated patients reported local infection; and seven patients in each group (35%, denosumab; 50%, zoledronic acid) reported undergoing limited surgical procedures such as debridement and sequestrectomy.

None of the other RCTs or observational studies reported ONJ.

Renal toxicity

In the denosumab RCT, a statistically significant lower rate of adverse events potentially associated with renal impairment occurred in the denosumab group compared with the zoledronic acid group [4.9% (50/1020) vs 8.5% (86/1013), respectively; p = 0.001]. 31 Stopeck and colleagues31 also reported that the rates of severe and serious adverse events (SAEs) associated with renal impairment were also lower for denosumab than for zoledronic acid (0.4% vs 2.2%, and 0.2% vs 1.5%, respectively). The incidence of renal adverse events among patients with baseline renal clearance ≤ 60 ml/minute was also lower in the denosumab group (5.9%) than in the zoledronic acid group (20.0%), and a greater proportion of patients had decreases in their baseline creatinine clearance from ≥ 60 ml/minute to < 60 ml/minute with zoledronic acid (16.1%) compared with denosumab (12.7%). 31

It should be noted that, as zoledronic acid is contraindicated in patients with poor renal function, such patients were excluded from the denosumab study. The manufacturer stated that the incidence of renal toxicity observed in the denosumab group represented a background rate for patients with advanced cancer, as such patients were predisposed to renal dysfunction, for example through the use of nephrotoxic drugs (MS).

Rosen and colleagues130 reported that there was no significant difference in renal safety profiles between the 4 mg zoledronic acid group and the 90 mg disodium pamidronate group. After 25 months, a change in the creatinine level of more than 0.5 mg/dl from baseline had occurred in 7.7% of patients in the zoledronic acid group and 6.0% of patients in the disodium pamidronate group. 130

Kohno and colleagues102 stated that there was no evidence of decreased renal function among patients in either group. In the zoledronic acid group, mean serum creatinine was 0.79 mg/dl at baseline and 0.78 mg/dl at the end of study while in the placebo group it was 0.79 mg/dl at baseline and 0.85 mg/dl at the end. In one patient in the zoledronic acid group, serum creatinine increased notably from a baseline of 1.3 mg/dl to 2.0 mg/dl, compared with seven patients in the placebo group. No patient in the zoledronic acid group developed a Common Terminology Criteria for Adverse Events (CTCAE) grade 3 or 4 serum creatinine increase, while one patient in the placebo experienced such an event. 102

Acute-phase reactions

Acute-phase reactions encompass flu-like symptoms including pyrexia, chills, flushing, bone pain, arthralgias and myalgias. 31 Stopeck and colleagues31 reported that acute-phase reactions in the first 3 days after treatment were 2.7 times more common in the zoledronic acid group than in the denosumab group [27.3% (277/1013) vs 10.4% (106/1020), respectively]. In the MS, SAEs of acute-phase reactions within 3 days of first dose were reported. (Academic-in-confidence information has been removed.)

Other adverse events

Table 14 shows, for the denosumab RCT, rates of a number of selected other adverse events, including those leading to treatment discontinuation, CTCAE grade 3 or 4 events, serious and fatal adverse events. The rates for both groups were broadly similar.

For details of all other adverse events extracted from the RCTs meeting the review's inclusion criteria and also adverse events extracted from a number of observational studies identified, see Appendix 10.

Time to first on-study skeletal-related event

The results from the AG's and MS's NMAs are shown below in Table 17.

In both the AG's NMA and the MS's NMA, time to first SRE favoured denosumab compared with zoledronic acid, disodium pamidronate and placebo. In the AG's NMA, the difference was statistically significant for denosumab versus zoledronic acid and denosumab versus placebo (academic-in-confidence information has been removed). The AG did not compare denosumab with ibandronic acid because they considered the studies too heterogeneous to provide meaningful results. (Academic-in-confidence information has been removed.) Risk of first and subsequent on-study SREs (academic-in-confidence information has been removed).

The results for risk of developing first and subsequent on-study SREs are provided below in Table 18.

Risk of first and subsequent SREs favoured denosumab compared with zoledronic acid, disodium pamidronate or placebo in both the AG's NMA and the MS's NMA. In the AG's NMA the difference was statistically significant. (Academic-in-confidence information has been removed.) SMR and SMPR (academic-in-confidence information has been removed).

The AG did not have access to SMPR for denosumab compared with zoledronic acid and were therefore unable to perform this comparison (Table 19).

The SMRs in both the AG's NMA and the MS's NMA favour denosumab. There was a statistically significant difference for denosumab compared with placebo (AG's NMA), zoledronic acid compared with placebo (AG's NMA). (Academic-in-confidence information has been removed.) Proportion of patients with on-study SRE (academic-in-confidence information has been removed).

The AG undertook a NMA comparing the proportion of patients with an on-study SRE (Table 20). This is a less informative outcome as it does not differentiate between lengths of study. However, the AG judged the study lengths to be similar enough to be included within the NMA. It also provided an opportunity to compare interventions by individual SRE.

Compared with zoledronic acid denosumab non-significantly reduced the risk of any SRE, pathological fracture and radiation to bone. There was a non-significant increase in SCC compared with zoledronic acid. Compared with placebo both denosumab and zoledronic acid non-significantly reduced the risk of each individual SRE. It should be noted that none of the above results was statistically significant and the NMA is not sufficiently powered to detect differences. Individual SREs should not be compared with each other, for example comparing the effectiveness of an intervention to prevent pathological fractures compared with SCC, because of the low numbers of events.

Skeletal-related event by type

Neither study reported the time to first SRE for individual SREs.

Table 23 shows the distribution of first on-study SRE by type of SRE in the study by Fizazi and colleagues. 29 The distribution of type of SRE was similar across the treatment groups, with radiation to bone and pathological fracture being the most commonly occurring.

Saad and colleagues117 did not report this outcome.

Prior history of skeletal-related events

The MS reported time to first on-study SRE by prior history of SREs for study 103 (Table 24). This showed a statistically significant difference in favour of denosumab for those patients with no prior SRE (academic-in-confidence information has been removed). Covariate analysis showed that patients with a prior SRE history had an increased risk (academic-in-confidence information has been removed) compared with those without a SRE history.

Saad and colleagues117 reported that the median time to first on-study SRE for those with a previous SRE (n = 144) was 361 days for the zoledronic acid group compared with 258 days for the placebo group (p = 0.066), whereas for those with no previous SRE (n = 277) it was 499 days for the zoledronic acid group and 337 days for the placebo group (p = 0.065). 119

Skeletal-related event by type

Neither study reported risk of first and subsequent on-study SRE by type of SRE.

The MS reported the distribution of first and subsequent on-study SREs by type of SRE in the denosumab RCT (study 103) (Table 25). As for first on-study SRE by type, the distribution of type of SRE was similar across the treatment groups, with radiation to bone and pathological fracture again the most commonly occurring.

Prior history of skeletal-related events

The MS reported risk of developing first and subsequent on-study SREs by history of SRE for study 103 (Table 26). (Academic-in-confidence information has been removed.) Covariate analysis as presented in the manufacturer's table showed that patients with a history of SRE had an increased risk (academic-in confidence information has been removed) compared with those without a SRE history [although in the text the manufacturer reported the covariate effect (academic-in confidence information has been removed) and increased risk (academic-in confidence information has been removed)].

Saad and colleagues reported that among the 144 patients with a SRE before study entry, zoledronic acid significantly reduced the risk of SREs by 40% compared with placebo (RR 0.60; p = 0.028), and among the 277 patients without a SRE before study entry, zoledronic acid significantly reduced the overall risk of SREs by 33% compared with placebo (RR 0.67; p = 0.027). 119

Skeletal-related event by type

The MS did not report SMR by type of SRE.

Table 27 shows the SMR by type of SRE for the study by Saad and colleagues. 117

Prior history of skeletal-related event

SMR by history of SRE was not reported for the denosumab RCT.

Saad and colleagues reported that the mean on-study SRE per year, for those patients with a previous SRE (n = 144) was 0.8 for zoledronic acid compared with 2.3 for placebo (p = 0.036), whereas for those with no previous SRE it was 0.77 for zoledronic acid and 0.98 for placebo (p = 0.06). 119

Skeletal-related event by type

Incidence of SREs by type of SRE was not reported for the denosumab RCT.

Table 29 shows the proportions of patients with different types of SRE for the study by Saad and colleagues. 117 More SREs occurred in the placebo group overall. The most frequently occurring SRE in both groups was radiation therapy to bone, followed by pathological fractures.

Prior history of skeletal-related events

Neither study reported incidence of SRE by history of SREs. However, Saad and colleagues117 reported that for those with a previous SRE (n = 144), the proportion of patients with one or more SRE while on study was 41% (27/66) for zoledronic acid compared with 51% (40/78) for placebo (p = 0.215), whereas for those with no previous SRE (n = 277) this was 37% (54/147) for zoledronic acid compared with 47% (61/130) for placebo (p = 0.087). 119

Prior history of skeletal-related events

Neither study reported overall survival by history of SREs.

Functional Assessment of Cancer Therapy

The Functional Assessment of Cancer Therapy – Prostate (FACT-P) questionnaire consists of the FACT-G questionnaire plus additional questions specific to prostate cancer. For each component of the FACT-P (FACT-G total score, FACT-P total score, physical well-being domain, functional well-being domain, and TOI: a composite of the functional well-being domain, physical well-being domain, and the prostate cancer subscale), a higher score indicates better HRQoL.

Table 30 shows the change in FACT scores from baseline to week 73. (Academic-in-confidence information has been removed) (CSR 103).

Saad and colleagues117 reported that the total FACT-G score decreased from baseline to the last measurement, with no statistically significant differences between the zoledronic acid and placebo groups.

European Quality of Life-5 Dimensions

For both components of EQ-5D (the health index and the VAS), a higher score indicates a more preferred health status. For the health index questions of the EQ-5D, a three-level response was used to assess quality of life. (Academic-in-confidence information has been removed) (CSR 103).

Saad and colleagues117 reported that the EQ-5D scores decreased from baseline to the last measurement, with no statistically significant differences between the zoledronic acid and placebo groups.

Hypocalcaemia

The MS reported that hypocalcaemia events were mainly non-serious and transient and either resolved spontaneously or with calcium supplementation (MS). More hypocalcaemia adverse events occurred in the denosumab group than in the zoledronic acid group [13% (121/943) vs 6% (55/945), respectively], a statistically significant difference (p < 0.0001). 29 Calcium decreases of grade 3 or higher occurred in 48 patients (5%) receiving denosumab and 13 patients (1%) receiving zoledronic acid.

In the study by Saad and colleagues,117 1.9% (4/214) of patients in the zoledronic acid group experienced grades 3 or 4 hypocalcaemia compared with none in the placebo group.

Osteonecrosis of the jaw

In the denosumab RCT, more patients in the denosumab group than in the zoledronic acid group experienced ONJ events [2% (22/943) vs 1% (12/945)], although the difference was not statistically significant (p < 0.09). 29 Of those, 17 (77%) on denosumab and 10 (83%) on zoledronic acid had a history of tooth extraction, poor oral hygiene, or use of a dental appliance. Fizazi and colleagues29 reported that, by April 2010, 10 patients (45%) on denosumab had received limited surgical treatment for ONJ (debridement, sequestrectomy, or curettage) and two (9%) had undergone bone resection, whereas three patients (25%) on zoledronic acid had received limited surgery and one (8%) had undergone bone resection. They also reported that, overall, resolution of ONJ, as defined by mucosal coverage, was recorded in four patients (18%) on denosumab and one patient (8%) on zoledronic acid.

Saad and colleagues117 did not report ONJ.

The proportion of patients experiencing ONJ was slightly lower than in observational studies (see Appendix 11). Walter and colleagues160 retrospectively studied patients prescribed BPs and found that 18.6% of patients experienced ONJ (time at risk not reported). However, three other observational studies reported a cumulative incidence of 2.2–6.5% over 12–15 months. 62,137,144

Renal toxicity

In the denosumab RCT, a similar rate of adverse events potentially associated with renal impairment occurred in the denosumab and zoledronic acid groups [15% (139/943) vs 16% (153/945), respectively]. 29 The rates of SAEs associated with renal impairment were also similar [5.9% (56/943) vs 5.6% (53/945) respectively] (MS). It should be noted that, as zoledronic acid is contraindicated in patients with poor renal function, such patients were excluded from the trial. The manufacturer stated that the incidence of renal toxicity observed in the denosumab group represented a background rate for patients with advanced cancer, as such patients were predisposed to renal dysfunction, for example owing to the use of nephrotoxic drugs (MS).

Saad and colleagues117 reported that renal function deterioration occurred in 15.2% of patients who received zoledronic acid and 11.5% of those receiving placebo. They stated that Kaplan–Meier estimates of time to first renal function deterioration indicated a comparable RR between the groups, so that compared with the placebo group the zoledronic acid group had a RR of 1.07 (95% CI 0.46 to 2.47; p = 0.882). 117

Observational studies of zoledronic acid reported a higher incidence of renal toxicity. Oh and colleagues152 found that 23.8% of patients experienced renal toxicity over 10 months while Bonomi and colleagues137 reported a figure of 6.5%. However, these studies had a broader definition of renal toxicity than the RCTs.

Acute-phase reactions

In the denosumab RCT, during the first 3 days of treatment, fewer patients in the denosumab group than in the zoledronic acid group experienced symptoms associated with acute-phase reactions [8% (79/943) vs 18% (168/945), respectively]. 29

Saad and colleagues117 did not report this outcome.

Other adverse events

Table 31 shows, for the denosumab RCT, rates of a number of selected other adverse events, including those leading to treatment discontinuation, CTCAE grade 3 or 4 events, serious and fatal adverse events. The rates for both groups were broadly similar.

Saad and colleagues117 reported that similar proportions of patients who received zoledronic acid (9.8%) and placebo (10.1%) discontinued the study drug because of a SAE.

In the denosumab group, 337 patients (36%) developed anaemia compared with 341 (36%) in the zoledronic acid arm. In the study by Saad and colleagues, a higher proportion of patients in the zoledronic acid group than in the placebo group experienced anaemia (26.6% vs 17.8%). 117 The clinical significance of this is unclear.

For details of all other adverse events extracted from the RCTs meeting the review's inclusion criteria and also adverse events extracted from a number of observational studies identified, see Appendix 11.

Time to first skeletal-related event

Results from the NMAs for time to first on-study SRE are shown in Table 33.

The NMA results from both the AG and the MS show that time to first SRE favoured denosumab compared with zoledronic acid or placebo. The AG's NMA found these differences to be statistically significant in favour of denosumab, (academic-in-confidence information has been removed).

Risk of first and subsequent skeletal-related events

The NMA results for risk of developing first and subsequent on-study SREs are shown in Table 34.

The NMA results show the risk of developing first and subsequent SREs favoured denosumab compared with zoledronic acid or placebo. The AG's NMA found these differences to be statistically significant in favour of denosumab, (academic-in-confidence information has been removed).

Skeletal morbidity rate

The NMA results for SMR are shown in Table 35.

The AG's NMA found a non-significant difference in favour of denosumab compared with zoledronic acid and a significant difference in favour of denosumab compared with placebo, whereas there was a statistically significant difference in favour of zoledronic acid compared with placebo. (Academic-in-confidence information has been removed.)

Proportion of patients with on-study skeletal-related events

The AG compared the proportion of patients with an on-study SRE for individual SREs and with a subgroup with a SRE history. This outcome does not differentiate between time on study and, therefore, the results should be interpreted with caution. However, it provides an opportunity to indirectly compare SRE types and SRE history.

Denosumab non-significantly favoured zoledronic acid and placebo throughout. Owing to the small numbers, however, these results should not be used to compare the relative effectiveness of denosumab for preventing individual SRE types.

Time to first on-study skeletal-related event

The results for time to first on-study SRE are shown in Table 40. The NMA results favoured denosumab compared with zoledronic acid or placebo for time to first on-study SRE but were not statistically significant.

Risk of first and subsequent on-study skeletal-related events

The results for the risk of developing first and subsequent on-study SREs are presented below in Table 41. The NMA results favoured denosumab compared with zoledronic acid or placebo for risk of developing first and subsequent SREs, although only the result compared with placebo was statistically significant.

Proportion of patients with on-study skeletal-related events

Results for the proportion of patients with an on-study SRE are shown below in Table 42. The NMA results favoured denosumab compared with zoledronic acid or placebo for the proportion of patients with an on-study SRE but were not statistically significant. These results should be interpreted with additional caution because this outcome does not differentiate between lengths of study, thereby adding to the uncertainty.

All patients

An ad hoc analysis by Henry and colleagues30 reported a non-significant difference in overall survival between the denosumab and zoledronic acid groups (HR 1.08; 95% CI 0.90 to 1.30).

The study by Rosen and colleagues130 did not report overall survival for those with OSTs excluding NSCLC.

Prior history of skeletal-related events

Neither study reported overall survival by history of SREs for those with OSTs excluding NSCLC.

Time to first on-study skeletal-related event

The results for time to first on-study SRE are shown in Table 43. There was a statistically significant difference in favour of denosumab compared with zoledronic acid or placebo for this outcome.

Risk of first and subsequent on-study skeletal-related events

The NMA results for risk of developing first and subsequent SREs are presented in Table 44. There was a statistically significant difference in favour of denosumab compared with placebo for this outcome.

Proportion of patients with an on-study skeletal-related event

The results for the proportion of patients with an on-study SRE are shown in Table 45. The results for denosumab compared with zoledronic acid or placebo were not statistically significant although the direction of effect favoured denosumab. These results should be interpreted with additional caution because this outcome does not differentiate between lengths of study, thereby adding to the uncertainty.

Skeletal-related events by type

The time to radiation to the bone was reported in the post-hoc analysis of study 244 (excluding multiple myeloma). The median time to radiation to the bone in the zoledronic group and in the denosumab group (academic-in-confidence information has been removed), and the risk reduction for denosumab (academic-in-confidence information has been removed) (MS) were reported.

In the study by Henry and colleagues30 (including multiple myeloma), denosumab reduced the risk of having radiation to bone by 22% compared with zoledronic acid (HR 0.78; 95% CI 0.63 to 0.97; p = 0.03). 134

(Academic-in-confidence information has been removed) (CSR 244).

Table 47 shows the distribution of first on-study SRE by type of SRE as reported in the MS (post-hoc analysis of CSR 244, excluding multiple myeloma). The distribution of type of SRE was similar across the treatment groups, with radiation to bone and pathological fracture being the most commonly occurring.

The study by Rosen and colleagues130 reported that the median time was not reached for individual SRE except for median time to first pathological fracture, which was longer in the zoledronic acid group (238 days) compared with the placebo group (161 days) (p = 0.031). Rosen and colleagues131 further reported that the time to first vertebral fracture and time to first radiation therapy were significantly longer in the zoledronic acid group (p = 0.05).

Prior history of skeletal-related events

The MS reported time to first on-study SRE by prior history of SREs for post hoc study 244 (excluding myeloma) (Table 48). (Academic-in-confidence information has been removed.)

The published study by Henry and colleagues30 did not report time to first on-study SRE by previous history of SRE. (Academic-in-confidence information has been removed) (CSR 244).

Rosen and colleagues130 did not report time to first on-study SRE by previous history of SRE.

Skeletal-related events by type

Neither study reported multiple event analysis for SRE by type.

In the MS (post-hoc analysis CSR 244), there was no difference reported between denosumab and zoledronic acid for the proportion of patients with each type of SRE. The distribution of each type of SRE is shown in Table 50. Radiation to bone and pathological fracture were the most commonly occurring SREs, whereas surgery to bone and SCC were reported for only a small proportion of patients.

The published studies by Henry and colleagues30 and Rosen and colleagues130 did not report on risk of first and subsequent on-study SREs by type of SRE.

Prior history of skeletal-related events

The MS reported risk of first and subsequent on-study SREs by history of SRE for post hoc study 244 (excluding multiple myeloma) (Table 51). (Academic-in-confidence information has been removed) (MS).

(Academic-in-confidence information has been removed) (CSR 244).

The studies by Henry and colleagues30 and Rosen and colleagues130 did not report risk of first and subsequent on-study SREs by prior history of SRE.

Skeletal-related events by type

The SMR by type of SRE was not reported for the denosumab RCT. 30

Rosen and colleagues130 reported that the SMR for each type of SRE was lower in the zoledronic acid treatment groups than in the placebo group except for surgery to bone and SCC; however, no data were reported.

Prior history of skeletal-related events

Neither study reported SMR by history of SREs.

Skeletal-related events by type

Incidence of SREs by SRE type was not reported for the denosumab RCT.

Rosen and colleagues130 reported the distribution of SRE type in zoledronic acid compared with placebo as shown in Table 53. For each individual SRE, a lower proportion of patients receiving zoledronic acid experienced a SRE than those receiving placebo. Radiation to bone and pathological fracture were the most frequently occurring SREs while SCC occurred least.

Prior history of skeletal-related events

Neither study reported incidence of SRE by history of SREs.

Prior history of skeletal-related events

Neither study reported overall survival by history of SREs.

Functional Assessment of Cancer Therapy – General

(Academic-in-confidence information has been removed) (Table 54).

(Academic-in-confidence information has been removed) (CSR 244).

The study by Rosen and colleagues130 stated that there were no statistically significant differences between zoledronic acid and placebo with respect to any of these global quality-of-life outcomes and that changes in FACT-G scores were also comparable between treatment groups; however, no data were reported.

European Quality of Life-5 Dimensions

(Academic-in-confidence information has been removed) (CSR 244).

Hypocalcaemia

Henry and colleagues30 reported that 10.8% of denosumab-treated patients had hypocalcaemia compared with 5.8% of zoledronic acid-treated patients. The statistical difference between the groups was not reported. Grade 3 or 4 decreases in albumin-adjusted calcium values were reported in nine patients (1.0%) receiving zoledronic acid and 20 patients (2.3%) receiving denosumab. Although the number of patients reporting hypocalcaemia is small the total number of events is higher for denosumab compared with zoledronic acid (academic-in-confidence information has been removed) (CSR 244).

The study by Rosen and colleagues130 did not report hypocalcaemia.

Observational studies reported a higher incidence of hypocalcaemia compared with the RCTs. However, the observational studies are likely to have broader criteria for hypocalcaemia. Chennuru and colleagues138 reported an incidence of 8.3% over 2 years in patients prescribed zoledronic acid. Zuradelli and colleagues161 reported an incidence of 4.6% in patients prescribed zoledronic acid (time at risk not reported).

Osteonecrosis of the jaw

Henry and colleagues30 reported that rates of ONJ were similar in the denosumab (1.3%) and zoledronic acid (1.1%) groups (p = 1.00). The cumulative incidence rates of ONJ at years 1 and 3 was reported to be slightly higher in the zoledronic acid group compared with the denosumab group, which was 0.6% versus 0.5% at year 1 and 1.3% versus 1.1% at year 3 (p = 1.0). At year 2, ONJ events were slightly higher in the denosumab group (1.1%) compared with the zoledronic acid group (0.9%).

The study by Rosen and colleagues130 did not report ONJ.

Two large observational studies were found. Hoff and colleagues147 reported an incidence of 0.7% (29/3994) over 21.2 months in patients taking zoledronic acid or disodium pamidronate. Vahtsevanos and colleagues159 reported an incidence of 4.9% (80/1621) over 20.4 months in patients taking any BP.

Renal toxicity

Henry and colleagues30 reported that renal adverse events occurred more often in the zoledronic acid group (10.9%) than in the denosumab group (8.3%). In both treatment groups, renal failure was reported to be similar. The MS reported a higher number of patients in the zoledronic acid group compared with the denosumab group with serious renal adverse events (34 patients compared with 24 patients). (Academic-in-confidence information has been removed) (CSR 244). The small discrepancy in these results is unclear.

Rosen and colleagues130 reported that the proportion of patients with decreased renal function was higher in the zoledronic acid group than in the placebo group. When zoledronic acid was given as a 5-minute infusion, the proportion of patients with decreased renal function was much higher in the zoledronic acid group (16.4%) than in the placebo group (5.6%). After the implementation of a 15-minute infusion of the given dose, 10.9% in the zoledronic acid group and 6.7% in the placebo group experienced decreased renal function.

The largest observational study155 (n = 966) evaluated renal impairment in patients taking any BP and found an incidence of 2.9% over 9.6 months.

Acute-phase reactions

Henry and colleagues30 reported that acute-phase reactions occurred more often in the zoledronic acid group (14.5%) than in the denosumab group (6.9%). In the MS, SAEs of acute-phase reaction occurred within 3 days of first dose. (Academic-in-confidence information has been removed.)

Rosen and colleagues130 did not report this outcome.

Other adverse events

In the study by Henry and colleagues,30 SAEs were reported in 66% of those treated with zoledronic acid and in 63% of those treated with denosumab (p = 0.16). Pyrexia and anaemia were reported to be significantly higher in the zoledronic acid group than in the denosumab group. Other adverse events were similar in both groups.

In the study by Rosen and colleagues,130 a higher proportion in the zoledronic acid group than in the placebo group was reported to have nausea (46% vs 34%), vomiting (36% vs 29%) and dyspnoea (33% vs 26%). The incidence of bone pain was reported to be higher in the placebo group (59%) than in the zoledronic acid group (51%).

There were no other adverse events of note from the observational studies assessed. Anaemia was similar between all groups.

For details of all other adverse events extracted from the RCTs meeting the review's inclusion criteria and also adverse events extracted from a number of observational studies identified, see Appendix 12.

Time to first on-study skeletal-related event

The results for time to first on-study SRE are shown in Table 55. The AG's NMA results were statistically significant in favour of denosumab compared with zoledronic acid or placebo. (Academic-in-confidence information has been removed.)

Risk of first and subsequent on-study skeletal-related events

The results for risk of developing first and subsequent on-study SREs are presented in Table 56. The AG's NMA results were statistically significant in favour of denosumab compared with placebo, whereas the result for the comparison with zoledronic acid was not statistically significant, although the direction of effect favoured denosumab. (Academic-in-confidence information has been removed.)

Proportion of patients with on-study skeletal-related event

The results for the proportion of patients with an on-study SRE are shown in Table 57.

In the AG's NMA, the differences between denosumab and zoledronic acid or placebo were not statistically significant, although the direction of effect favoured denosumab. This outcome does not account for differences in length of study, thereby adding to the uncertainty, and thus these results should be interpreted with caution.

Full papers

Dranitsaris and Hsu168 estimate the cost-effectiveness of disodium pamidronate compared with BSC over a 12-month trial among breast cancer patients with bone metastases. This drew on the findings of Hortobagyi and colleagues,22 who report the clinical effectiveness of the then only relevant disodium pamidronate trial. Over a mean duration of therapy of 10 months, disodium pamidronate and BSC were associated, respectively, with the following events:

• non-vertebral fractures: 20% vs 30%

• radiation to the bone: 19% vs 33%

• surgery to the bone: 4 % vs 10%

• any SRE: 46% vs 62%

• any SRE excluding hypercalcaemia: 43% vs 56%.

Costs per health state were estimated by chart review, with unit costs being drawn from the Princess Margaret Hospital and the Centenary Hospital of Ontario, Canada.

The main aspects of the paper that are of interest are the utility data, which are drawn from a time trade-off (TTO) exercise among 25 women from the Canadian general public and 25 female health workers. There is a lack of detail within the paper, and it seems likely that the health state descriptors include elements of both the treatment aspects and the clinical effectiveness for each arm. With this noted, the TTO exercise yields the following estimates (Table 58).

The source of the anticipated benefit from disodium pamidronate over placebo when no SRE is experienced is unclear and is not specified within the paper. Health worker responses are reasonably consistent, with a consistent reduction in quality of life from a SRE of around 50% for both the disodium pamidronate and the placebo health states. Results are more mixed within the public responses, with SREs causing a similar, approximate 50%, reduction in quality of life in the placebo group, but only a 30% reduction in the disodium pamidronate group.

Dranitsaris and Hsu168 estimate that disodium pamidronate results in an additional cost of £1758 (C$2800). Based on the SRE rates including hypercalcaemia of 46% and 62%, this results in an estimated gain from disodium pamidronate of 0.15 quality-adjusted life-years (QALYs), with an associated cost-effectiveness of £11,740 (C$18,700) per QALY based on public preferences and £10,359 (C$16,500) per QALY based on health-care worker preferences. Results are sensitive to the costs of surgery to the bone.

Hillner and colleagues169 estimate the cost-effectiveness of disodium pamidronate compared with BSC for breast cancer patients over a 2-year time horizon in the USA. The utility values are taken from expert opinion, with fractures at 0.8, radiation at 0.6, surgery at 0.4 and both hypercalcaemia and SCC at 0.2. The duration applied to these is not clear from the paper, but it may be 1 month. Disodium pamidronate is estimated to result in an additional 1.13 months SRE free with a net cost increase of £3593 (US$3968) for chemotherapy patients, resulting in a cost-effectiveness of £97,973 (US$108,200) per QALY. For hormone-treated patients the correspoding amounts are 0.82 additional months free of SRE at a cost of £6958 (US$7685) to yield a cost-effectiveness of £276,444 (US$305,300) per QALY.

Ross and colleagues,55 in the 2004 Health Technology Assessment (HTA) monograph reviewing the role of BPs in metastatic disease, model the cost per SRE avoided for breast cancer patients with bone metastases. This uses a cost-effectiveness Markov model with a monthly cycle. This simulates rates of SREs, with the health states also including hypercalcaemia and pain reduction, this latter being distinct from palliative radiotherapy. Note that SCC is not considered. The RRs for SREs and hypercalcaemia in the model for BPs compared with BSC are not differentiated by BP, but are differentiated by event type:

• 0.90 for vertebral fracture

• 0.79 for non-vertebral fracture

• 0.71 for palliative radiotherapy

• 0.59 for surgery to the bone

• 0.51 for hypercalcaemia.

Direct drug and administration costs are based on the cost of disodium pamidronate plus an oncology outpatient appointment. The cost per fracture is taken as the average of the relevant inpatient health-care resource groups (HRGs) within NHS reference costs £2786 (£2017), with surgery to the bone being costed at £2813 (£2036), while radiotherapy is based on three radiotherapy sessions in an outpatient setting to yield a cost of £978 (£708). Ross and colleagues55 undertook their own bottom-up costing for hypercalcaemia to estimate an average cost of £4840 (£3503). Note that this study was undertaken when discount rates were differentiated between costs at 6% and benefits at 1%.

The model estimates a 4-year survival of 16%, with patients being treated monthly with disodium pamidronate until death or to the end of the fourth year. This results in an average 1.45 SREs being averted compared with BSC: 0.54 non-vertebral fractures, 0.16 vertebral fractures, 0.64 courses of palliative radiotherapy and 0.12 episodes of surgery to the bone. An additional 0.34 episodes of hypercalcaemia are modelled as being prevented together with an average 3.2 months bone pain reduction. The total cost of therapy is estimated to be £7235 (£5237), but cost offsets reduce this to £613 (£444). Excluding hypercalcaemia, this results in a cost per SRE avoided of £423 (£306). With the application of a 0.33 QALY loss per SRE drawn from Dranitsaris and Hsu169 as reviewed above but adjusted for an increased SRE duration of 22 months, this translates into a cost-effectiveness estimate of £1851 (£1340) per QALY gained.

Reed and colleagues170 (supported by Novartis) compare the cost-effectiveness of zoledronic acid with BSC for prostate cancer patients with bone metastases, mainly within the context of the USA and Medicare. This analyses within-trial SRE rates and resource utilisation data over 15 months to estimate the cost per SRE avoided. An additional cost–utility analysis is conducted based on the EQ-5D VAS scores. The average number of SREs within the zoledronic acid group is 0.78 compared with 1.24 in the BSC group, resulting in incremental cost-effectiveness ratios (ICERs) of £11,137 ($12,300) per SRE avoided and £105,976 (US$159,200) per QALY.

De Cock and colleagues171 model the cost-effectiveness of oral ibandronate compared with zoledronic acid and disodium pamidronate among UK breast cancer patients receiving hormonal therapy. Treatment with oral ibandronate is estimated to result in a direct utility gain of 0.02 compared with intravenous administration. Discontinuation rates are also assumed to be lower, it being estimated that 96.9% of ibandronate patients are treated for an average of 7.2 months out of a total survival of 14.3 months. This compares with 71% for zoledronic acid and 73% for disodium pamidronate, although 12% of these patients switch to oral ibandronate. Oral ibandronate is estimated to be as effective as zoledronic acid for those on therapy in preventing SREs, and both are slightly superior to disodium pamidronate. Given this, ibandronate is estimated to yield an additional 0.02 QALYs over both zoledronic acid and disodium pamidronate, while saving £390 (£307) and £201 (£158), respectively.

In a parallel paper, De Cock and colleagues172 model the cost-effectiveness of oral ibandronate compared with zoledronic acid and disodium pamidronate among UK breast cancer patients receiving chemotherapy. This applies the same SRE rates and RRs for those on therapy as those applied in De Cock and colleagues,172 with the same discontinuation rates and percentages switching to oral ibandronate. There is also the same anticipated average survival of 14.3 months and the same quality-of-life values. There is the same average gain from ibandronate of 0.02 QALYs compared with zoledronic acid and disodium pamidronate, but the costs savings differ marginally: £490 (£386) compared with zoledronic acid and £285 (£224) compared with disodium pamidronate.

Guest and colleagues173 (supported by Mayne Pharma) undertake a cost minimisation analysis of disodium pamidronate compared with zoledronic acid for breast cancer patients in the UK, with a 1-year time horizon. This draws clinical effectiveness estimates from the literature, distinguishing between those on chemotherapy and those on hormonal therapy. Disodium pamidronate is estimated to be marginally superior in preventing any SRE among the chemotherapy group, and slightly inferior to zoledronic acid in preventing any SRE among the hormonal therapy group. These rates are then qualified by rates of individual SREs, with disodium pamidronate typically resulting in slightly more of all SREs among those experiencing a SRE, with the exception of fractures among those receiving hormonal therapy. Disodium pamidronate has a higher discontinuation rate, particularly among those being treated with hormonal therapy. For chemotherapy treated patients, this results in an average 3.77 SREs for disodium pamidronate compared with 2.79 for zoledronic acid. For hormone-treated patients, this resulted in an average 3.44 SREs for disodium pamidronate compared with 2.93 for zoledronic acid. The authors conclude that there is little clinical difference, and that as a consequence cost minimisation is appropriate.

Drug administration times for the base case are estimated as 184 to 214 minutes for disodium pamidronate compared with 204 to 232 minutes for zoledronic acid, though this latter includes patients waiting 90 minutes for test results. It is unclear quite how this has been costed. Expert opinion supplies much of the resource-use estimates (Table 59).

In the light of the above, disodium pamidronate is estimated to be cost-saving compared with zoledronic acid: £1130 (£936) for chemotherapy patients and £776 (£643) for hormone-treated patients.

Reed and colleagues170 (supported by Novartis) compare the costs and consequences of zoledronic acid with disodium pamidronate among breast cancer patients with bone metastases, again mainly within the context of the USA and Medicare. This analyses within-trial SRE rates and resource utilisation data, with a mean patient follow-up of 10 months. Zoledronic acid is estimated to have a RR of a SRE of 0.80 compared with disodium pamidronate. Costs in the zoledronic acid group are estimated to be marginally higher: £14,218 (US$15,703) compared with £14,198 (US$15,680) for disodium pamidronate. This was not taken through to a cost-effectiveness estimate specific to breast cancer patients.

Botteman and colleagues174 (authorship includes an employee of Novartis) compare the cost-effectiveness of zoledronic acid, oral ibandronate, intravenous ibandronate, disodium pamidronate, oral clodronate and BSC for breast cancer patients with bone metastases. This uses a cost–utility model from a UK NHS perspective, with a monthly cycle over a 10-year time horizon. Patients can discontinue active therapy due to non-compliance, which might be because of an adverse event. Fifty per cent of those discontinuing move on to another active therapy: oral if previously on intravenous and intravenous if previously on oral. Disease progression is also assumed to lead to therapy being stopped.

A baseline annual rate of 3.05 SREs is assumed for BSC, with this being multiplied by the relevant HR to arrive at the treatment-specific SRE rates: 0.56 for zoledronic acid, 0.62 for oral ibandronate, 0.71 for intravenous ibandronate and 0.70 for disodium pamidronate.

Quality-of-life values for without a SRE and with a SRE are drawn from Dranitsaris and Hsu168 on the grounds that it was the only published source available. There is some arbitrariness in the estimation of benefits, with the oral ibandronate being assumed to be postponed to the 12th week, while oral clodronate was assumed to have half the benefits of the other therapies. Survival was unaffected by treatment, with a mean survival of 20 months.

Zoledronic acid is estimated to require 11 minutes of physician time, 11 minutes of pharmacy technician time and 44 minutes of nurse time, in contrast to 8, 12 and 152 minutes for disodium pamidronate and 10, 11 and 98 minutes for intravenous ibandronic acid. This results in staff administration costs of £42.17 (£37.42) for zoledronic acid, £88.23 (£78.29) for disodium pamidronate and £65.20 (£57.85) for intravenous ibandronic acid.

The SRE costs are averaged across the SREs, with an average inpatient cost of £2272 (£2016) plus an additional average of £1826 (£1620) outpatient and care in the community costs. These are stated as being based on the Ross and colleagues55 BPs review HTA monograph.

The base-case results are an average 6.11 SREs for BSC, with this being reduced to 3.71 SREs for zoledronic acid; 4.41 SREs for disodium pamidronate; 4.46 for intravenous ibandronate; and 4.06 for oral ibandronate with this last SRE possibly being the result of the high discontinuation rate and second-line intravenous therapy. Given the figure for BSC and the average survival of 2 years, it is not obvious how progression was included in the modelling.

Average QALY estimates are surprisingly similar between the BPs – 1.18 QALYs to 1.20 QALYs – and BSC – 0.99 QALYs. Total costs are £21,032 (£18,662) for BSC, with disodium pamidronate and intravenous ibandronate exceeding this by £127 (£113) and £516 (£458), respectively, to yield cost-effectiveness estimates relative to BSC of £658 (£584) per QALY and £2671 (£2370) per QALY. Zoledronic acid and oral ibandronate are estimated to save £2554 (£2267) and £2382 (£2114) compared with BSC, and so dominate it, with zoledronic acid further dominating oral ibandronate. Across the therapies, zoledronic acid is estimated to be the preferred treatment at all values of willingness to pay.

Joshi and colleagues (who include Botteman and a Novartis employee) estimate the cost-effectiveness of zoledronic acid compared with BSC for NSCLC patients across five European countries in what appears to be an update of the Botteman 2009 abstracts, as summarised below. 175–181 This is based on the NSCLC subset of the Phase III trial populations, within which the median survivals were not statistically different between zoledronic acid, 201 days, and BSC, 157 days. As a consequence, a Weibull distribution is fitted to the zoledronic acid arm to yield an estimated average survival of 272 days. This is then multiplied by each arm's SRE-specific SMR to derive the number of SREs: 1.38 for zoledronic acid and 2.17 for BSC, though the latter includes some episodes of hypercalcaemia.

The SREs are assumed to be associated with only 1 month loss of quality of life, the baseline NSCLC HRQoL of 0.63 being reduced by 6.8% by vertebral fracture, 20% by non-vertebral fracture, 40% by radiation therapy, 60% by surgery to the bone and 80% by both SCC and hypercalcaemia, as drawn from Hillner and colleagues. 169 This results in zoledronic acid being estimated to yield 0.44 QALYs compared with 0.42 QALYs for BSC.

For the UK, in common with the approach of the 2004 Ross HTA monograph,55 the costs per SRE were derived mainly from averaging a range of HRG costs. This yields costs of £138 (€187) for vertebral fracture; £4520 (€6105) for non-vertebral fracture; £745 (€1007) for radiation to the bone; £2456 (€3318) for surgery to the bone; £3714 (€5017) for SCC; and £3822 (€5163) for hypercalcaemia. Administration costs and supplies for zoledronic acid are based on the micro-costing of DesHarnais and colleagues182 with 11-minute physician time, 11-minute pharmacist time and 44-minute nurse time to yield a total administration cost of £38.82 (€52.43). Total UK costs are reported as £3062 (€4136) for zoledronic acid compared with £3086 (€4168), which suggests a small net saving from zoledronic acid of £22 (€32), though the paper reports this as a saving of £155 (€209). The 0.79 fewer SREs are estimated to provide cost offsets of £1217 (€787), and zoledronic acid is estimated to dominate BSC for NSCLC patients with bone metastases.

Carter and colleagues183 (a similar authorship list to Joshi and colleagues' 2011 NSCLC paper,181 and with the support of Novartis) model the cost-effectiveness of zoledronic acid versus BSC for prostate cancer patients in France, Germany, Portugal and the Netherlands. 183 Quality-of-life data are drawn from the Reed and colleagues184 paper through a back calculation using the ICER and the estimated additional costs. This suggests an average gain from zoledronic acid over placebo of 0.034 QALYs. Rates of individual SREs are estimated solely to inform the drug and SRE costing exercise attached to this estimate of QALY gains. The base-case results are that 0.759 SREs are avoided on average, generating savings of between £2094 (€2396) and £3162 (€3617) per patient. The direct drug and administration costs of zoledronic acid are less geographically variable at between £3012 (€3446) and £3269 (€3704), with the resulting increase in costs leading to cost-effectiveness estimates ranging from a low of £2124 (€2430) in the Netherlands, to a high of £31,476 (€36,007) in France.

Xie and colleagues185 (supported by Novartis) estimate the cost-effectiveness of denosumab compared with zoledronic acid for patients with hormone refractory prostate cancer with bone metastases. This uses a 1-year Markov model with a 13-week cycle. The justification for using a 1-year time horizon rather than a 3-year time horizon is the anticipation of zoledronic acid being available in generic form from March 2013. But the analysis is from a US perspective, and the costs are not particularly relevant. The paper is of interest in part because in addition to modelling rates of SREs, the probability of a SRE is dependent on whether the patient is progression free or with progression. The likelihood of progression is not differentiated by treatment arm, but progression increases the rate of SREs by 2.14 compared with the without-progression SRE rate, as drawn from Tchekmedyian and colleagues. 186 Among those without progression denosumab was estimated to have a RR of first on-study SRE of 0.83 and a HR of 0.82 for subsequent SREs, with these estimates probably being carried over to the with-progression patients (Table 60).

These cost-effectiveness results are summarised below (Table 61), within which unless otherwise stated the cost-effectiveness estimates are the cost per QALY for the more effective treatment over the less effective treatment.

Available only as abstracts

A number of other papers available only as abstracts were identified by the literature review. Few details are provided within the abstracts and the results for zoledronic acid compared with BSC, or for denosumab versus zoledronic acid, are summarised below for completeness. Note that all these studies are supported by Novartis. The AG has also been in contact with John Carter of Pharmerit with a view to accessing the full texts of the two cost–utility studies of denosumab versus zoledronic acid. Apparently these are ready for full publication and will be made available, but are yet to be received by the AG.

Note that some of the abstracts that were identified, for example Stephens for lung,187 simply report the results available in other abstracts, in this case Botteman188 for lung, and are therefore not repeated in Table 62.

Full papers

Patient characteristics

Baseline patient characteristics are drawn from the relevant denosumab trials (Table 70).

Survival data

On the basis of the Akaike information criterion, the survival analysis of the data pooled across the arms of the denosumab trials suggests modelling breast cancer survival using a Weibull, prostate cancer using a gamma and OSTs using a log-logistic functional form (Table 71).

The key assumption in the above is that there is no overall survival difference between denosumab and zoledronic acid, with this assumption of no survival differences also being carried over to the other comparators where applicable. In other words, survival is not affected by rates of SREs. Any frailty distribution around multiple SREs in the same patient similarly is assumed to not affect survival. The survival curves are, for reasons that are not entirely clear, augmented with the age-specific non-solid tumour mortality rates as drawn from UK life tables. This results in the following survival percentages within the modelling (Table 72).

Balance between types of skeletal-related events

The balance between the different SREs is taken from the denosumab trials, with the data being pooled between the arms (Table 73). The balance between the SRE types is time invariant, with the exception that once a SRE-naive patient has experienced a first SRE the balance between SREs is that for subsequent SREs as applied to SRE-experienced patients.

Rates of skeletal-related events for zoledronic acid

Zoledronic acid is taken as the numéraire against which the other treatments' HRs and RRs are measured. The rates of first SREs and subsequent SREs for the comparator treatments are derived through the application of the relevant HRs and RRs. The rates of SREs for zoledronic acid are split into:

• the time to first on-study SRE for SRE-naive patients

• the SRE rate per cycle for patients who are SRE experienced at baseline.

Times to first skeletal-related event among skeletal-related event-naive patients

A reasonably standard set of time to event functional forms are fitted to the time to first on-study SRE among SRE-naive patients for the zoledronic acid arm of the denosumab trials. This results in the log-normal form being assessed as best by the Akaike information criterion for prostate cancer and OSTs.

But the gamma function is estimated as being superior for breast cancer patients with an Akaike information criterion of 3327 compared with 3330 for the log-normal, which is the next best fit. The manufacturer justifies the adoption of a common log-normal form on the basis of the probabilistic model often simulating a shape parameter for the gamma distribution of less than 0.08, which is apparently problematic. But even if this is the case, it would seem desirable to have applied the fitted gamma function within the deterministic modelling to test any sensitivity to this assumption. Unfortunately, the submission does not outline the parameterised form of the gamma distribution for breast cancer. If the central estimate for this postpones the first SRE beyond that suggested by the fitted log-normal distribution this may have tended to bias the analysis in favour of denosumab. The parameter estimates are shown in Table 74.

Rates of subsequent skeletal-related events among skeletal-related event experienced patients

The SRE cycle rate is calculated as the total number of SREs divided by the patient-years of exposure, and adjusted to the 28-day cycle length. The base case applies the 21-day window definition of a SRE, which results in the following cycle rates. The manufacturer assumes a cycle lasts 4/52nds of 1 year within this calculation. This is marginally longer than the true 28/365ths and serves to slightly increase the rate of SREs within the zoledronic arm, but this is unlikely to have much, if any, material effect on results (Table 75).

Note that the SRE rate per cycle for SRE-experienced patients excludes the data on SREs subsequent to the first on-study SRE among the SRE naive at baseline patients. The manufacturer justifies this on the grounds that it would break randomisation. This justification is not understood by the AG. It could be argued that applying the SRE rate estimated from patients who were SRE experienced at baseline to the patients who were SRE naive at baseline but have experienced an on-study SRE is a more serious violation of randomisation or stratification within the trials. Note also that the proportions of patients who were SRE naive at baseline were 59% for breast cancer, 74% for prostate cancer and 52% for OSTs.

Hazard ratios and relative risks for skeletal-related events for comparator treatments

The MS applies the hazard ratios for time to first on-study SRE and RRs for time to first and subsequent SRE as estimated from the denosumab trial data for denosumab versus zoledronic acid (table 24 of the MS), and from the NMA for the other comparators (tables 50, 51 and 52 of the MS) with zoledronic acid being the numéraire as outlined above. These are summarised in Table 76.

Note that while the submission suggests that the subgroups of SRE-naive and -experienced patients are analysed separately, the subgroup-specific HRs and RRs for denosumab versus zoledronic acid are not applied. Only pooled results are presented for comparator drugs because owing to a lack of published data neither the AG nor the manufacturer was able to undertake a NMA for SRE-experienced or SRE-naive patients. The modelling submitted by the manufacturer applies the HRs and RRs pooled across all patients, whether modelling SRE-naive patients or SRE-experienced patients. This is likely to have mainly affected the cost-effectiveness results presented for prostate cancer and for the OSTs group.

It would seem sensible to apply the SRE-naive- and -experienced-specific HRs and RRs for denosumab versus zoledronic acid when analysing these subgroups. The SRE-experienced-subgroup-specific central estimates suggest a smaller effect from denosumab compared with the pooled estimates for these patients.

Adverse events and discontinuations

The model includes the following SAEs:

• ONJ

• renal toxicity

• hypercalcaemia

• hypocalcaemia

• skin infections.

For the main comparators of denosumab and zoledronic acid the rates of these are drawn from the denosumab trials. Each of these SAEs is also associated with a treatment-specific discontinuation rate, again drawn from the denosumab trials (Table 77). A further treatment-specific general discontinuation rate is drawn from the denosumab trials, though it is not clear whether or not the definition of this excluded the discontinuations due to SAEs. The key assumption within the handling of adverse events and discontinuations is that their rates are constant over the period of the modelling.

The rates of adverse events for the other BPs are drawn from the literature, and are assumed to apply equally across the three cancer groups being modelled. Discontinuation rates due to SAEs for the other BPs are assumed to be the average across the rates observed for denosumab and zoledronic acid.

Discontinuation rates for the other BPs not due to SAEs are drawn from another three papers within the literature.

Rates of adverse events for BSC are assumed to be zero. This may be unrealistic and may tend to worsen the cost-effectiveness estimates for the active treatments relative to BSC. In the main, adverse event rates do not appear to be key model drivers as there is sufficient differentiation between active treatments and BSC in terms of reducing SRE rates. Sensitivity analyses that compare active treatments with BSC and assume minimal differences between them in terms of SRE rates may not be reliable, as the assumption of zero adverse events in the BSC arm may have come to the fore of the analysis. But given the cost-effectiveness estimates for active treatments versus BSC as outlined below this may not be a particular concern (it is also, at least in part, addressed in the AG modelling through sensitivity analyses that assume zero adverse events for all treatments). Note that those discontinuing denosumab or BP therapy are assumed to immediately assume the BSC RR for SREs. There is no waning protective effect from having received denosumab or BP therapy.

Discontinuations also introduce what may appear to be a perversity within the model structure. The model estimates both denosumab and zoledronic acid to have a very poor cost-effectiveness when compared with BSC. Because of this, a treatment that has a high discontinuation rate sees patients rapidly move off active treatment and on to the more cost-effective BSC. As a consequence, a high discontinuation rate for an active treatment improves the cost-effectiveness estimate for that treatment. This requires some qualification, in that the situation is more complicated if the main sources of discontinuations are SAEs, with their associated HRQoL and cost impacts. But as can be seen from the above, for both denosumab and zoledronic acid the vast majority of discontinuations are not related to SAEs.

Drug and administration costs

The list price of denosumab is £309.86 per vial. The manufacturer cites the BNF as the source of the direct drug costs of the comparators. The BNF used by the manufacturer may predate the current BNF62, which differs slightly from table 72 of the submission, giving the list prices as:

• £174.17 for a 4-mg vial of Zometa^®zoledronic acid

• £165.00 for a 90-mg vial of generic disodium pamidronate.

This compares with the costs applied by the manufacturer of £183.30 and £167.73 respectively. This mainly affects the comparison with zoledronic acid, the manufacturer cost for it being 5% higher than BNF62.

To estimate the administration costs associated with the different administration routes the manufacturer commissioned a micro-costing study, as summarised in the MS. This study was undertaken in the UK among 80 oncology nurses and 20 oncology pharmacists. It is unclear to what extent any of the nursing staff would have had actual experience of denosumab, but they would obviously be fully familiar with subcutaneous injections. The micro-costing study provided estimates of the staff times involved in administering denosumab and BPs, and costed these from a NHS perspective using standard Personal Social Services Research Unit staff costs.

Note that the micro-costing study prompted respondents about the administration times associated with different infusion durations: ‘Question: It is assumed that an infusion of IV X would typically occur over a minimum of X minutes according to the SPC. Is this correct for your centre? If not, please specify the infusion time.’ This wording may have framed responses to the question. It also does not appear to ask whether or not the duration of the intravenous infusion involved any additional nursing time: 15 minutes for zoledronic acid, 15 minutes of intravenous ibandronic acid and 90 minutes for disodium pamidronate. These timings were included in the costing.

For the comparison between denosumab and zoledronic acid the main differences in terms of minutes of staff time reported by the oncology nurses and as outlined in the MS to the nearest minute are given (Table 78).

Owing to the apparently highly skewed nature of replies, the manufacturer has chosen to use the medians rather than the means for costing purposes. The requirement to make this adjustment may suggest that the micro-costing study is not entirely reliable. (Academic-in-confidence information has been removed.)

The manufacturer estimates that denosumab will result in staff time savings compared with zoledronic acid (academic-in-confidence information has been removed) per administration. These arise in part from the preadministration savings (academic-in-confidence information has been removed), but more from drug administration savings (academic-in-confidence information has been removed) within which avoiding the need for infusion saves (academic-in-confidence information has been removed) staff time.

Taking these elements together with the consumables and fixed costs estimated within the micro-costing study yields the total annual direct drug and administration costs (Table 79).

Without the patient access scheme (PAS) the annual denosumab cost of £4467 is around £1102 more expensive than zoledronic acid.

The PAS proposed by the manufacturer has recently been approved. (Commercial-in-confidence information has been removed.)

(Commercial-in-confidence information has been removed.)

The base case assumes 4-weekly dosing for both denosumab and the BPs. The manufacturer also supplies a sensitivity analysis that retains 4-weekly dosing for denosumab, but assumes that a percentage of BP patients receive 3-weekly dosing in line with their chemotherapy regimen.

Within the denosumab trials intravenous therapy could be withheld because of elevated creatine. This affects the average dose received within the zoledronic acid arm. The CSRs provide the subject incidence of intravenous dose withholding, though it is not clear to the AG whether this corresponds to the number of patients having their dose withheld or the number of doses withheld. It appears possible that since exposure to zoledronic acid could be resumed only once creatine levels had returned to acceptable levels, some of these incident patients may have had more than one dose withheld. But on the conservative assumption that the incident patient dose withheld data is equivalent to only one dose being withheld the figures in Table 80 are implied.

The impact of this has not been included within the direct drug and administration costs calculated by the manufacturer.

Skeletal-related event costs

Trim points and manufacturer's costings

For the derivation of the average inpatient cost per event the manufacturer's costings include an allowance for the excess bed-days within the NHS reference costs. The manufacturer calculates a weighted average length of stay across elective inpatients, non-elective long-stay inpatients and non-elective short-stay inpatients for the identified HRGs. This average HRG length of stay is taken as the trim point. If the average length of stay observed within the STARs study exceeds this, the manufacturer costs this excess at the excess bed day rate for the identified HRGs, averaged across elective inpatients and non-elective long-stay inpatients (Table 81).

For instance, the average length of stay across the three HRGs identified for non-vertebral fractures treated as an inpatient is calculated as 7.93 days. Among those treated as inpatients for non-vertebral fracture, the STARs study average length of stay is given (academic-in-confidence information has been removed). The manufacturer calculates the excess bed-days (academic-in-confidence information has been removed) minus 7.93 days: (academic-in-confidence information has been removed) days are costed at £217 per day to yield an excess bed-day cost (academic-in-confidence information has been removed). This is added to the weighted average inpatient cost across the three HRGs (academic-in-confidence information has been removed) to yield an overall total cost for non-vertebral fractures treated on an inpatient basis (academic-in-confidence information has been removed).

But the 2010–11 episode trim points for the three identified HRGs (HD39A, HD39B and HD39C) are 45 days, 21 days and 19 days, respectively. Whereas the average treatment duration within the STARs study will encompass a spread of values, it is questionable whether or not any allowance for excess bed-day costs should have been made by the manufacturer.

These considerations around excess bed-day trim points apply throughout the manufacturer's costings of inpatient stays for the other SREs and AEs.

Radiotherapy to the bone costing

For reasons that are not clear, to cost radiotherapy planning and administration the manufacturer uses 2008–9 reference costs and indexes these for inflation, rather than using the 2009–10 reference costs which are employed for all the other SREs.

For the planning of radiotherapy the manufacturer includes the HRG codes SC01Z through to SC03Z, which seems reasonable. It may be more questionable to have included SC04Z relating to planning multiple phases of complex radiotherapy and SC010Z relating to planning ‘other’ radiotherapy. The weighted average cost across inpatients, day cases, outpatients and ‘other’ settings is applied to all those receiving radiotherapy.

Similarly, for the delivery of radiotherapy the manufacturer includes the HRG codes SC21Z through to SC24Z, all of which relate to delivering a single fraction of radiotherapy. Again, it may be more questionable to have included SC29Z relating to the delivery of ‘other’ radiotherapy, the unit costs of this typically being somewhat higher than that of the HRGs specifically relating to delivering a single fraction of radiotherapy. The weighted average cost across inpatients, day cases, outpatients and ‘other’ settings is multiplied by the average number of fractions (academic-in-confidence information has been removed) drawn from the STARs study.

Base-case skeletal-related event costs

The STARs-based costing results in the following cost estimates (Table 82).

Vertebral fracture is something of an outlier within these costings, with quite significant costs being associated with outpatient visits and outpatient procedures. Possibly because of the questionable reliability of the resource use around vertebral fractures and the numbers observed (academic-in-confidence information has been removed), coupled with expert opinion that vertebral fractures are typically asymptomatic to the extent of not being treated, the manufacturer applies no cost for vertebral fractures in the base case.

Adverse event costs

As already noted, the cost of treating hypercalcaemia, £4579 (£379; 2004), as drawn from the Ross HTA monograph is used for the base case.

For hypocalcaemia the manufacturer assumes that this will require one haematology consultant-led outpatient appointment, one intravenous calcium injection, and two follow-up visits. Each visit is associated with a blood test, to yield a total cost per event of £443.

For the other adverse events the manufacturer assumes that all will be treated as inpatients and simply averages the inpatient cost over a range of HRGs:

• ONJ HRGs: CZ16 minor maxillofacial procedures, CZ17 intermediate maxillofacial procedures, CZ18 major maxillofacial procedures, and CZ19 complex maxillofacial procedures, to arrive at an average cost of £2465

• renal toxicity HRGs: all LA07 acute kidney injury and all LA08 chronic kidney disease but not LA09 general renal disorders, to arrive at an average cost of £1681

• skin infections HRG: only JD04B minor skin disorders category 3 with Intermediate CC at £1440.