Preventing kidney transplant failure by screening for antibodies against human leucocyte antigens followed by optimised immunosuppression: OuTSMART RCT

Article history

The research reported in this issue of the journal was funded by the EME programme as project number 11/100/34. The contractual start date was in April 2013. The final report began editorial review in September 2022 and was accepted for publication in December 2022. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The EME editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Copyright statement

Copyright © 2023 Stringer et al. This work was produced by Stringer et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2023 Stringer et al.

Scientific background

Kidney transplants do not last for the natural lifespan of most recipients, 30–40% of patients have their transplant for < 10 years1 and around 3% of prevalent kidney transplants fail annually. 2 This places a large burden on healthcare services. The single biggest cause of transplant failure is immune-mediated injury, the target of which are mismatched donor human leucocyte antigens (HLA).

A validated prognostic biomarker of kidney transplant failure is the appearance of circulating antibodies (Ab) against HLA. 3–8 Patients with HLA Ab have a three-fold greater risk of graft failure compared to those without7,8 and if these are specific for the kidney donor HLA [donor-specific antibodies (DSA)] there is an even higher risk of graft loss compared to those Ab that are not donor-specific (non-DSA). Inappropriately low levels of immunosuppression, either physician-led or due to patient non-adherence, is an important factor allowing the immune-mediated damage to begin and promoting the appearance of the HLA Ab. 9

Rationale for the study

The mechanisms driving graft dysfunction leading to graft failure are most likely complex and although the HLA Ab themselves might be damaging,10 other components including T- and B-lymphocytes may also play a role. 11 Various novel therapies have been tested in small scale, often uncontrolled human studies with some promising results. 12,13 Two randomised controlled trials concluded that B cell depletion with Rituximab was ineffective at preventing graft dysfunction in patients with biopsy-proven chronic antibody-mediated rejection (CAMR)14,15 as did a smaller trial of the anti-IL-6 receptor Ab tocilizumab. 16 However, two small RCTs of the anti-IL-6 monoclonal Ab Clazakizumab have shown benefit17,18 A larger RCT of clazakizumab, with a planned recruitment of 350 patients is underway (https://clinicaltrials.gov/ct2/show/NCT03744910). Several other studies have suggested that optimised oral treatment with tacrolimus (Tac) and mycophenolate mofetil (MMF) can stabilise graft function in small numbers of patients with existing graft dysfunction. 15,19–26 However, to date there have been no large-scale trials testing this strategy by intervening in patients who develop HLA Ab prior to developing graft dysfunction, and none that have assessed if graft failure can be prevented.

Hypotheses

In the OuTSMART study, we tested the hypothesis that routine surveillance for the development of HLA Ab, combined with optimised ‘treatment’ in those who became HLA Ab+, would prevent the premature failure of transplanted kidney allografts. 27,28

Design

OuTSMART was a prospective, open labelled, randomised marker-based strategy (hybrid) trial design, with two arms stratified by biomarker (HLA Ab) status. Recruits were followed up with regular structured visits for at least 32 months (maximum 64 months) and primary endpoint assessed by remote evaluation after approximately 43 months post-randomisation (or post ‘clock reset’ – see below) was achieved by all. The trial design is represented in Figure 1. All eligible patients were screened for HLA Ab before being randomised 1 : 1 into either double-blinded standard care (SC) or unblinded biomarker-led care (BLC). Biomarker stratification generated three groups of recruits in each arm (DSA+, non-DSA+ and HLA Ab-neg). Patients in the blinded SC arm (groups A1, A2 and C in Figure 1) were blind to their biomarker status and remained on baseline immunotherapy throughout, whereas patients in the unblinded (groups B1, B2 and D in Figure 1) knew their HLA Ab status; in this arm, those with HLA Ab were offered ‘intervention’, whereas HLA Ab-negative patients remained on their existing immunotherapy.

FIGURE 1.

Trial design/flow of patients. Legend to figure 1: Randomisation took part in the HLA laboratory, after the first HLA Ab screening round to allow biomarker stratification within the respective arms. All participants assigned to the HLA Ab negative groups in each arm underwent rescreening for HLA Ab every 8 months till month 32. In both arms, recruits changing from HLA Ab negative to positive, at any time after enrolment were asked to complete a further 32 months follow-up (=‘clock reset’), so that the maximum time under structured follow-up for any recruit was 64 months. To maintain blinding, the randomisation system was programmed at each screening round to choose a small random group of HLA Ab-negative recruits from the SC group to complete a further 32 months follow-up. Therefore, a subset of HLA Ab-negative patients had ‘clock reset’ in the blinded SC arm, whereas only HLA Ab+ participants in unblinded BLC had ‘clock reset’.

Both groups of HLA Ab-negative recruits had regular Ab status monitoring for the first 32 months. Those patients who became positive during subsequent screening rounds were moved to the appropriate HLA Ab positive groups (DSA+ or non-DSA+) for final data analysis. All patients in groups C and D found to be positive on second or subsequent rounds were intensively followed up for an additional 32 months from the time they become positive (=‘clock reset’): those in the BLC arm were also offered the same ‘intervention’ as those patients who were positive in the first screening round. To maintain blinding in the SC arm, the randomisation system was programmed at each screening round to choose a small random group of HLA Ab-negative recruits from this arm to complete a further 32 months follow-up. Thus the maximum amount of time any single patient remained in intensive follow-up was 64 months. At the end of intensive follow-up, most data collection related to secondary endpoints ceased, but data related to graft failure or death were recorded for all participants up and until the end of the trial (see below), irrespective of the length of follow-up.

Setting

Thirteen UK Kidney transplant outpatient clinics.

Participants

Renal transplant recipients > 1 year post-transplantation.

Identification and recruitment

The local transplant clinic database of their prevalent population was used to identify patients meeting the baseline inclusion/exclusion criteria. At the start of the trial, the entire population of transplant clinic attendees who met the eligibility criteria were potentially eligible for recruitment. On subsequent screening rounds, patients who reached 12 months post-transplantation after the start of the trial became eligible. Potentially eligible patients were approached at a routine clinic appointment by the PI or research nurses and given printed and verbal information about the trial. They had the opportunity to return for a second consultation within a few days to give informed consent for recruitment into the study or to do this on their next routine appointment. Alternatively, some eligible patients were sent information about the study through the post, for discussion and consent at their next routine appointment. Following consent, full eligibility criteria were reviewed. This included testing for chronic viral disease (if no such test within last 5 years) or pregnancy (if history suggests possibility of pregnancy).

Randomisation procedure

Prior to randomisation but after consent, site staff registered all recruits online and each assigned a MACRO PIN. Samples from all recruits were sent to the relevant HLA laboratory, along with this PIN and a sample request form containing the other information required for randomisation. HLA laboratory staff performed a screen for HLA Ab and performed single antigen bead testing on positive screening samples to check for the presence of DSA. Once this information was known, the laboratory staff accessed the randomisation system and randomised the patient, using the HLA Ab results and information on the sample form to stratify.

Allocation to SC (blinded) or BLC (unblinded) arms was assigned (1 : 1) by stratified block randomisation with randomly varying block sizes, using a web-based randomisation service provided by the King’s Clinical Trials Unit. Randomisation was stratified by (1) HLA Ab status, to generate three groups within each arm (DSA+, non-DSA+ or HLA Ab-negative), (2) current immunosuppression (to ensure balanced numbers already on Tac or MMF) and (3) site (N = 13). The randomisation allocation was initiated by staff within the five HLA (tissue-typing) laboratories involved in the trial.

Blinding

There was no blinding of arm allocation. In all sites, immediately after randomisation, the PIs and nurses were automatically emailed with information about whether the patient was in the blinded or unblinded groups. If in the unblinded group, the email to the PI contained information about the HLA Ab status. The system told trial staff to enter HLA Ab-negative patients into the subsequent 8 monthly screening rounds.

In blinded patients, HLA Ab status was not fed back to the PIs or trial staff in the emails. All blinded patients had samples taken 8 monthly for HLA Ab screening, though upon sample receipt, HLA laboratory staff used their knowledge of the HLA status to determine those from HLA Ab-negative patients which underwent screening and samples from HLA Ab-positive patients were discarded.

On the second and subsequent HLA Ab screening rounds, the laboratory staff updated the randomisation system within 52 days from the date the rescreen sample was taken. Only the results from patients in the unblinded groups were forwarded to the PI and lab staff, via email. This indicated whether status had changed and triggered the initiation of the treatment protocol in those that had changed from HLA Ab negative to positive. It also indicated that patients who had become HLA Ab+ needed ‘clock reset’ to extend the period of intensive follow-up for a further 32 months.

In the blinded arm, PIs were emailed with a list of patients who required ‘clock reset’ so they got follow-up for a further 32 months. This list contained all the recruits who had become HLA Ab+ on rescreening, but also an equivalent number of recruits who had stayed HLA Ab-; these recruits were randomly chosen by the randomisation system as a mechanism to maintain physician and patient blinding to HLA Ab status within this arm.

There were no blinded study medications in the trial so no emergency code break was required. There were no requests for recruits’ HLA Ab status to be unblinded.

Inclusion criteria

1. sufficient grasp of English to enable written and witnessed informed consent to participate;

2. aged 18–75 years;

3. estimated glomerular filtration rate (eGFR by four variable MDRD) of ≥ 30 ml/min (within the previous 6 months of signing consent or taken at screening if not done in the previous 6 months).

Exclusion criteria29

1. recipient requiring HLA desensitisation to remove Ab for a positive cross match (XM) transplant;

2. recipient known already to have HLA Ab who has received specific intervention for that Ab or for CAMR/chronic rejection;

3. recipient of additional solid organ transplants (e.g. pancreas, heart, etc.);

4. history of malignancy in previous 5 years (excluding non-melanomatous tumours limited to skin);

5. HBsAg+, HepC immunoglobulin G (IgG+) or human immunodeficiency virus (HIV+) recipient (on test performed within previous 5 years);

6. history of acute rejection requiring escalation of immunosuppression in the 6 months prior to screening;

7. patient enrolled in any other studies involving administration of another IMP at time of recruitment;

8. known hypersensitivity to any of the IMPs;

9. known hereditary disorders of carbohydrate metabolism;

10. pregnancy or breastfeeding females (based on verbal history of recipient);

11. pre-menopausal females who refuse to consent to using suitable methods of contraception throughout the trial.

Participant withdrawal

Individual recruits were free to withdraw at any time and the PIs also had the right to withdraw patients from the study drug in the event of inter-current illness, adverse events (AEs), serious adverse events (SAEs), suspected unexpected serious adverse reactions, protocol violations, cure, administrative or other reasons. After every withdrawal from ‘treatment’, efforts were made to obtain permission to continue to collect study-specific data before patients were completely withdrawn from the study. Failure to tolerate one or more components of the ‘treatment’ was not seen as a reason to withdraw an individual participant from the trial.

Significant amendments to study protocol

The complete list of changes over the course of the trial is included in Appendix 1. The following are those judged to have altered the conduct of the trial. All changes were discussed and approved by the Trial Steering Committee or Chairman and, where appropriate, by the Data Monitoring Committee.

In Version 4 (13/5/2013), we clarified that the eGFR measurement on which eligibility was be assessed had to be within 1 month of signing consent, and also clarified the definition of a positive HLA Ab test, which was confusing in the previous protocol versions.

In Version 5 (9/7/2013), the definition of diabetes mellitus was updated to incorporate the WHO definition (use of HbA_1c) and the reporting of AEs in this type A trial to the sponsor was clarified.

In Version 7 (7/4/2014), we removed the exclusion criteria ‘history of ongoing or previous infection that would prevent optimisation’ which was being interpreted differently within and across sites. In addition, the gap for the testing of eGFR from within 1 month of signing consent was increased to within the previous 6 months of signing the consent. Finally, the timing of the optimisation process was changed from within 3 months of HLA Ab positivity to ideally within 3 months after positive screening for HLA Ab and allocation to the unblinded treatment arm or as soon as possible thereafter BUT within 8 months of positive screening. This coincided with the realisation that some patients were proving difficult to contact to arrange optimisation and the change was felt to enhance the optimisation process without affecting the outcome of the trial.

In Version 8 (1/7/2014), the time that tissue typing laboratories had to perform the randomisation of patients, was increased from 28 to 56 days post-consent. This was to enhance batching of patient serum for testing, reducing the number of experimental controls and HLA screening beads needed, and therefore the cost of screening.

In Version 10 (11/08/2015), the upper limit for eligibility into the study was increased from 70 to 75 years.

In Version 11 (26/11/2015), the primary objective and endpoint were changed from 3-year graft failure rates in HLA Ab+ patients in the SC versus BLC arms27 to ‘time to graft failure with variable follow-up (with a minimum of 43 months post-randomisation)’. The new primary endpoint was to be assessed remotely from patient notes once 43 months post-randomisation had been achieved by all. This change was required because, after 16 months recruitment, an audit of HLA Ab screening results revealed that the expected 9% prevalence and 3% incidence rates of DSA were actually 5.8% and 1.6%, respectively. 28 All patients already recruited were reconsented to allow this change. This change allowed for a reduction in the number of DSA patients to be recruited, and a significant shortening in the expected study duration while maintaining the power of the study.

Because there was no additional funding for these changes, existing workload was reduced by changing the timing of follow-up visits from 4-monthly to 8-monthly, the end visit for each participant changed from 36 to 32 months, along with the timing of the secondary endpoint assessments. Finally, there was a major reduction in the requirement for SAE reporting to the sponsor.

In Version 12 (1/12/16), we stopped collection of research blood samples and removed the secondary experimental/exploratory ‘scientific’ endpoints. This was required by the funder, who requested that the salary costs associated with the exploratory aspects of the trial be reallocated towards supporting the primary endpoint data collection.

Finally, in Version 14 (08/07/2020), we changed the timing of the collection of the primary endpoint, as a result of the COVID-19 pandemic, in addition to the proposal to included additional sensitivity analyses for the primary endpoint and extension of the study end date.

Statistics methodology

Outcome assumptions and data collection periods

The following treatment effect contrasts for the primary and secondary outcomes were estimated:

1. 1a. DSA+ BLC versus DSA+ SC participants (both at randomisation and rescreening);

2. 1b. non-DSA+ BLC versus non-DSA+ SC participants (both at randomisation and rescreening);

3. 2. all randomised BLC versus SC participants.

For the primary outcome, contrasts 1a and 1b were tested for superiority and contrast 2 was tested for non-inferiority, with non-inferiority concluded if the upper bound of the 95% CI for the HR was less than 1.4. For all secondary outcomes, all contrasts were tested for superiority.

Different comparisons/outcomes used different observation periods. This is outlined in Figure 2.

FIGURE 2.

Diagram to help explain how endpoints were assessed. Figure depicts three recruitment scenarios and explains how different outcomes were assessed, either in relation to recruitment/randomisation or, in the case of recruits turning from HLA Ab-negative to +, in relation to rescreening.

The purpose of these different observation periods for the different comparisons, is that the within DSA+ and within non-DSA+ comparisons aim to estimate the treatment effect of unblinding + optimisation in HLA+ve participants and so include participants at risk from the time they were found to be HLA+ve. The overall unblinded (BLC) versus blinded (SC) comparison aims to estimate the overall effect of the blinding strategy, and so participants time at risk is the time of blinding/unblinding to the HLA result (which is randomisation for all participants).

For the DSA+ and within non-DSA+ comparisons, time at risk started at randomisation for those HLA positive at randomisation and at time of rescreening for those HLA-negative participants who became HLA positive later at rescreening rounds. For the primary outcome, patients’ follow-up time was used up until the pre-COVID-19 collection period. For the overall comparison (statistical hypothesis 2), time at risk started at randomisation for all participants up until the pre-COVID-19 collection period. This was also true for the secondary outcome of death.

For the other secondary outcomes, these were only collected in the intensive data collection period which was from randomisation to 32 months post-randomisation, or in the case of participants who became HLA positive at rescreening rounds, 32 months post-rescreening. Therefore for these secondary outcomes, for the within DSA+ and within non-DSA+ comparisons, time at risk starts at randomisation for those HLA positive at randomisation and at time of rescreening for those HLA-negative participants who became HLA positive later and ends 32 months later. However, for the overall unblinded (BLC) versus blinded (SC) comparison for these other secondary outcomes, time at risk starts at randomisation for all participants and ends 32 months post-randomisation (ignoring any additional follow-up for rescreening HLA-positive participants).

For the primary outcome, the proportional hazards assumption was checked by testing for an interaction between treatment and time or more precisely, testing for a non-zero slope in a generalised linear regression of the scaled Schoenfeld residuals on functions of time (which is equivalent to testing the interaction).

Economic evaluation

Adherence to drug therapy and perceptions of risk to the health of the transplant

Health surveys, consisting of validated psychological measures adapted for this specific health context, were performed at baseline and 12 and 24 months post screening for HLA Ab+, and included the medication adherence report scale (MARS) questionnaire, which consisted of measuring six items on a five-point Likert scale with higher scores representing greater adherence. Most items assessed intentional medication non-adherence; one item measured unintentional non-adherence. MARS was completed for each medication patients received. MARS correlates well with relatively objective measures of adherence in a range of illness contexts, including electronic measures of inhaled corticosteroids for asthma and blood pressure control for hypertension. 34,35 It has also been shown to have good levels of internal consistency, test-retest reliability and construct validity. 35 For Tac, 12-hour trough levels were also compared against the target trough levels (4–8 ng/ml) and a composite adherence scale based on combining MARS scores with trough levels was developed. Concern about the risk of transplant failure was measured using the Brief Illness Perceptions Questionnaire. 36

Analysis of questionnaires was performed separately to the main trial data by the team at University College London. Analyses were based on imputed data: where values were missing for a given survey item, the mean score for that item across all participants was used, providing that a given case had at least 80% complete data for other items on that scale. Mann–Whitney U or chi-squared tests were used to compare scores or proportions across patients in the BLC DSA+ compared to SC DSA+ groups, and BLC non-DSA compared to SC non-DSA groups.

Anti-HLA Ab determination

Serum prepared from 10 mL of blood was used in the commercially available LABScreen tests (One Lambda, Canoga Park, CA through VH Bio, Gateshead, UK), analysed on Luminex equipment (Luminex Corp, Austin, Texas) in the five original sites (Guy’s, Birmingham, Manchester, Leeds and Royal London). All worked to the same standard operating procedure, agreed pre-trial (see Appendix 2). Serum was first analysed using mixed HLA Class I and Class II Ab screening beads, with a positive or negative result assigned based on batch-specific cut-offs designated using validated protocols at the Guys laboratory site. In those patients with positive results, serum was further analysed using single antigen-coated Class I or Class II beads. A positive result was defined as giving a mean fluorescence intensity (MFI) of binding ≥ 2000. Laboratory staff then compared assigned DSA/non-DSA status depending on whether HLA Ab were directed against a mismatched donor HLA antigen. HLA Ab in which it was difficult to label as DSA+ or non-DSA+ (because of insufficient data on donor mismatches, for instance), were categorised as being non-DSA+. Samples with a positive reaction on screening but lacking reactivity with the single antigen beads were considered negative. Screening of all HLA Ab– patients was undertaken every 8 months.

Trial oversight

Patient and public involvement

Kidney transplant patients were involved in the grant application process, first, via their involvement in the local review of projects at GSTT via the TRU Project Board Steering Committee, and second, through a specific meeting with members of the GSTT Kidney Patients Association during the grant application process. At this stage patient involvement led to three significant changes in trial design, including dropping the inclusion of protocol kidney transplant biopsies, reducing the maximum dose of prednisolone used, and because of concerns about the communication of risk associated with the biomarker, recruitment of Professor Rob Horne into the team. The KPA helped Prof Horne with the design of the patient information sheets for the trial and also provided a representative to sit on the TSC. Throughout the conduct of the trial, the KPA were kept updated about how the trial was progressing via their representative on the TSC and through regular updates via the MRC Centre for Transplantation newsletter, annual Clinical Trials Day literature and CI contributions to their quarterly newsletter.

Participants, recruitment and flow

Recruitment was from 13 UK transplant centres and took place between 11 September 2013 and 27 October 2016. During that time, 5519 renal transplant recipients (see Figures 1 and 3) were assessed for eligibility of which 2094 were enrolled after consent. Fifty-seven patients were found to be ineligible after post-consent checks. 2037 were randomised after HLA Ab screening into two arms, each containing three groups based on the HLA Ab screening results; blinded SC (A1, A2 or C), and unblinded BLC (B1, B2 and D). Randomisation broken down by site and year is shown in Table 2. Screening of the HLA Ab-negative groups for HLA Ab finished in June 2017, at which time a further 63 with DSA (28 blinded, 35 unblinded) and 263 non-DSA (116 vs. 147) were identified, leaving 1019 remaining HLA Ab-negative through the course of screening (524 vs. 495). The end of the intensive follow-up period (last person, last visit) occurred 32 months later in March 2020, with the remote primary endpoint collection at a minimum of 43 months post-randomisation, originally scheduled for June 2020 moved to March 2020 because of the pandemic (as described above).

FIGURE 3.

Consort diagram for OuTSMART. ¹ Two patients were randomised in error to blinded HLA Ab-negative SC group. These were included from the intention-to-treat analysis. Refer to Supplementary Material for further information.

Of the 90 patients ‘lost’ during the trial, 29 withdrew consent (group A1 n = 1: A2 n = 7: B1 n = 1: B2 n = 6: C n = 8: D n = 6), 16 became uncontactable (group A1 n = 0: A2 n = 3: B1 n = 1: B2 n = 3: C n = 3: D n = 6), 1 was withdrawn for an AE (group B2) and the remaining 44 were withdrawn for reasons listed as ‘other’, but 43 of these were because patients had transferred care to another, non-trial transplant unit and were therefore out of touch (group A1 n = 5: A2 n = 9: B1 n = 1: B2 n = 6: C n = 13: D n = 10).

Protocol violations/randomisation errors

There were two randomisation errors, both participants were randomised to the blinded (SC) arm and were HLA Ab-negative at baseline. One participant had graft failure prior to randomisation and one participant was randomised but was found to have died shortly before randomisation. Randomisation was carried out by lab staff following HLA screening and in error, it was not communicated to the lab staff that these events had occurred prior to randomisation. These participants are excluded from all analyses.

Further, for the primary analysis, one rescreened randomised participant is not included in the DSA group as they were found to have graft failure prior to being rescreened and becoming HLA Ab+ DSA and so were not at risk for the purpose of this analysis. This participant is included in any group/sensitivity analyses where time at risk starts at randomisation for all.

There were several other errors in recording of HLA status in the randomisation system. As per the intention-to-treat principle, these were analysed in the original groups as recorded in the randomisation data and not in the corrected group (as the HLA status as per randomisation data was communicated to the PI if in the unblinded arm, and treatment strategy would have been based on this data). These errors were the following:

• One participant randomised to the BLC arm and entered as HLA Ab+ DSA at baseline in randomisation system was actually HLA Ab+ definite non-DSA at baseline according to the lab data.

• One participant randomised to the SC arm and entered as HLA Ab+ DSA at baseline was actually HLA Ab+ definite non-DSA at baseline according to the lab data.

• One participant in the BLC arm was moved from the HLA Ab-negative to the HLA Ab+ DSA group at rescreening (month 16). However, according to their lab data, their Ab at that time indicated ‘unknown whether DSA’ and should have been allocated to the non-DSA group as per the protocol.

• One participant randomised to the BLC arm and entered as DSA at baseline actually had Ab at that time indicating ‘unknown whether DSA’ and should have been randomised to the non-DSA group.

• Two participants randomised to the SC arm who were HLA Ab-negative at baseline were rescreened according to lab data at month 16 and became HLA Ab+ with unknown DSA. However, this was erroneously not entered into the randomisation system (and are considered HLA Ab-negative for the purpose of ITT analysis).

Baseline data

Baseline and change in IS during the study

Eighteen per cent of all participants were taking ciclosporin at randomisation, and 15% taking azathioprine. Interestingly, the proportions on ciclosporin or azathioprine were highest in those with DSA compared to those with non-DSA and those who were HLA Ab-negative (see Table 3). The majority of patients were taking Tac (73%) or MMF (67%) at randomisation, though fewer were taking maintenance prednisolone (55%). The proportions on Tac or MMF were lowest in those with DSA, compared to those with non-DSA and those who were HLA Ab-negative (see Table 3). Finally, 27% overall were taking all three drugs. The proportion taking all three drugs was lowest in those with DSA compared to those with non-DSA and those who were HLA Ab-negative (see Table 3). These differences were maintained when considering patients who developed new HLA Ab during the rescreening process. Therefore, the proportions on ciclosporin or azathioprine were still highest in those with DSA after rescreening compared to those with non-DSA and those who were HLA Ab-negative (see Table 4). The proportions on Tac or MMF were still lowest in those with DSA after rescreening, compared to those with non-DSA and those who were HLA Ab-negative (see Table 4). Finally, the proportion taking all three drugs was still lowest in those with DSA after rescreen, compared to those with non-DSA and those who were HLA Ab-negative (see Table 4).

Five hundred twelve of the five hundred thirty-two (97%) HLA Ab+ patients in the BLC arm had an optimisation interview and 33% of the DSA group, and 24% of the non-DSA group underwent steroid boost (see Table 7). The proportion taking all three IMPs increased from 23% immediately post-screening to 54% immediately post-optimisation in the DSA+ BLC group, and from 25% to 44% in the non-DSA+ BLC group. These changes were sustained to the last visit and were statistically significant (see Table 7). There were no discernible differences in demographic characteristics between those optimised according to the full protocol and those not optimised to the full protocol (see Table 8). However, there was significant site variation in the implementation of the full optimisation protocol, with only 6 of 13 sites optimising more than 50% of their BLC HLA Ab+ patients on all three IMPs (see Table 8).

Completion of follow-up visits

The majority of participants completed all four of the formal intensive study follow-up visits at months 8, 16, 24 and 32 months. 1.2% were withdrawn, died or reached the primary endpoint prior to the month 8 visit, and 2.5% of the remainder missed this visit. The corresponding figures for month 16 are 3.6% and 1.1%: month 24, 6.4% and 2.8% and month 32, 9.4% and 1.1%.

Primary analysis – time to graft failure in HLA Ab± groups (hypotheses 1a and 1b)

There were 34 graft failures in the blinded SC HLA Ab+ groups (12 DSA+, 22 non-DSA+) compared to 42 in the unblinded BLC HLA Ab+ groups (19 DSA+, 23 non-DSA+), with no evidence that the unblinded BLC strategy is superior to the SC strategy. 95% CIs included the null HR, in both the HLA Ab DSA+ group [HR1.54 (95% CI 0.72 to 3.30)] or non-DSA+ group [HR 0.97 (0.54 to 1.74)] (see Figures 4 and 5, Table 9).

FIGURE 4.

Kaplan–Meier Curves – DSA+ groups. Graph compares time to graft failure in the DSA+ groups. Blue (unbroken) line = patients in unblinded, BLC arm. Black (broken) line = patients in blinded SC arm. The number at risk of graft failure at each time point is shown beneath the graph, followed by (in brackets) the number of graft failures. NB: One HLA-Ab-negative participant in the blinded (SC) group who developed DSA on rescreening was not included in this analysis as the graft failed prior to rescreening, so they were not at risk for the purpose of this analysis.

FIGURE 5.

Kaplan–Meier Curves – non-DSA+ groups. Graph compares time to graft failure in the non-DSA+ groups. Blue (unbroken) line = patients in unblinded, BLC arm. Black (broken) line = patients in blinded SC arm. The number at risk of graft failure at each time point is shown beneath the graph, followed by (in brackets) the number of graft failures.

Post-COVID, there were 39 graft failures in the blinded SC HLA Ab+ groups (15 DSA+, 24 non-DSA+) compared to 49 in the unblinded BLC groups (21 DSA+, 28 non-DSA+). Nevertheless, the sensitivity analysis showed no appreciable difference from the primary analysis (see Table 10). In the BLC HLA Ab+ groups, there were 18 graft failures in those who underwent full optimisation according to the protocol, but only 16 in those not optimised according to the protocol and when the former were used in the sensitivity analysis, there was no appreciable difference in the primary outcome. The same is true for all the other planned sensitivity analyses (see Table 10). Post-hoc sensitivity analyses adjusting additionally for factors unbalanced at baseline (sex and time since transplant) or only looking at unblinded BLC recruits that underwent ‘best’ optimisation had no impact on the effect estimates had no impact on the effect estimates (see Table 10).

Secondary outcome analysis

Time to graft failure in all unblinded BLC versus all blinded SC (hypothesis 2)

Overall there were 62 graft failures in the blinded care arm (including 28 HLA Ab-negatives) compared to 64 in the unblinded care arm (including 22 in the HLA Ab-negative groups), providing insufficient evidence for non-inferiority of the unblinded BLC strategy with the upper 95% confidence limit for the HR exceeded the pre-specified threshold of 1.4 (HR 1.02, 95% CI 0.72 to 1.44) (see Figure 6). Time to graft failure in the HLA Ab-negative groups only is shown in Figure 7.

FIGURE 6.

Kaplan–Meier Curves – BLC versus SC. Graph compares time to graft failure in the entire unblinded BLC arm versus blinded SC arm. Blue (unbroken) line = patients in unblinded, BLC arm. Black (broken) line = patients in blinded SC arm. The number at risk of graft failure at each time point is shown beneath the graph, followed by (in brackets) the number of graft failures.

FIGURE 7.

Kaplan–Meier Curves – HLA Ab negatives. Graph compares time to graft failure in the HLA Ab-negative groups. Blue (unbroken) line = patients in unblinded, BLC arm. Black (broken) line = patients in blinded SC arm. The number at risk of graft failure at each time point is shown beneath the graph, followed by (in brackets) the number of graft failures.

Health economic analysis

Complete data (i.e. baseline and 16-month costs and EQ-5D-5L) were available for 173 blinded and 189 unblinded cases. The number and percentage of respondents using specific services is shown in Table 15. At baseline, the most commonly used services were renal outpatient, other outpatient, and GP care. Relatively more of the blinded group used other inpatient care. Other services were used by fewer respondents and there were no key differences between arms. At follow-up, similar patterns were evident and again there were few notable differences between the arms.

TABLE 15 - Number (percentage) of respondents using services at baseline and 16-month follow-up by trial arm

	Baseline		Follow-up
Service	Blinded (n = 173)	Unblinded (n = 189)	Blinded (n = 173)	Unblinded (n = 189)
Residential care	3 (1.8)	3 (1.6)	1 (0.6)	1 (0.5)
Renal inpatient	33 (20.6)	37 (20.0)	36 (20.9)	32 (17.0)
Intensive care	9 (5.6)	5 (2.7)	2 (1.2)	3 (1.6)
Other inpatient	30 (18.6)	20 (11.2)	26 (15.3)	31 (16.7)
Renal outpatient	146 (85.4)	158 (86.3)	136 (80.0)	154 (81.9)
Other outpatient	71 (42.8)	92 (51.1)	77 (45.3)	87 (47.3)
Day hospital	4 (2.5)	5 (2.9)	4 (2.4)	7 (3.9)
A&E	33 (20.3)	36 (20.1)	38 (22.8)	36 (20.3)
GP	150 (87.2)	163 (87.6)	139 (80.8)	136 (72.3)
Physiotherapist	27 (16.2)	27 (15.1)	18 (10.5)	35 (19.1)
OT	4 (2.4)	6 (3.3)	3 (1.8)	6 (3.3)
Speech therapist	1 (0.6)	1 (0.6)	0 (0)	2 (1.1)
Dietitian	30 (17.9)	18 (9.9)	23 (13.5)	17 (9.3)
Nutritionist	4 (2.4)	5 (2.8)	3 (1.8)	1 (0.6)
Social worker	6 (3.6)	3 (1.7)	1 (0.6)	4 (2.2)
Homecare worker	2 (1.2)	4 (2.2)	1 (0.6)	3 (1.7)
Psychologist	9 (5.5)	3 (1.7)	3 (1.8)	9 (4.9)
Complementary healthcare	13 (7.9)	10 (5.7)	5 (2.9)	6 (3.3)
District nurse	6 (3.6)	5 (2.8)	5 (2.9)	7 (3.8)
Psychiatrist	4 (2.4)	2 (1.1)	0 (0)	3 (1.7)
Counsellor	7 (4.2)	5 (2.8)	11 (6.5)	12 (6.6)

The mean number of contacts (or days for residential care and inpatient care) is shown in Table 16. This is for the whole sample and so includes those with zero use. While most services do not differ much between arms, other inpatient care stands out. At baseline this is higher for the blinded group and this reflects the greater proportion of blinded respondents using it. At follow-up there were similar proportions using other inpatient care (as shown in Table 15) but the unblinded group had patients with longest lengths of stay.

TABLE 16 - Mean number of service contacts/days at baseline and 16-month follow-up by trial arm

	Baseline		Follow-up
Service	Blinded (n = 173)	Unblinded (n = 189)	Blinded (n = 173)	Unblinded (n = 189)
Residential care	0.14	0.16	0.03	0.02
Renal inpatient	3.36	1.31	1.12	1.48
Intensive care	0.18	0.07	0.02	0.02
Other inpatient	1.34	0.69	0.45	2.65
Renal outpatient	4.75	3.94	3.19	3.34
Other outpatient	1.72	1.97	1.71	1.72
Day hospital	0.04	0.03	0.04	0.05
A&E	0.29	0.35	0.29	0.33
GP	2.53	3.26	2.56	2.54
Physiotherapist	0.64	0.54	0.43	1.12
OT	0.07	0.06	0.03	0.10
Speech therapist	0.02	0.02	0.00	0.04
Dietitian	0.33	0.28	0.19	0.19
Nutritionist	0.03	0.04	0.03	0.01
Social worker	0.08	0.13	0.01	0.04
Homecare worker	0.61	2.91	0.03	0.63
Psychologist	0.50	0.32	0.08	0.35
Complementary healthcare	0.43	1.12	0.24	0.76
District nurse	0.41	0.16	0.09	0.31
Psychiatrist	0.05	0.03	0.00	0.08
Counsellor	0.14	0.44	0.27	0.66
Total	0.14	0.16	0.03	0.02

Service costs were highest for inpatient care and outpatient contacts (see Table 17). At baseline, renal and other inpatient costs were substantially higher in the blinded arm. By follow-up, this had switched with the unblinded group having far higher inpatient costs. GP costs were relatively low even though most of the sample used them. Other costs were low and did not differ substantially between arms.

TABLE 17 - Mean service costs at baseline and 16-month follow-up by trial arm

	Baseline		Follow-up
Service	Blinded (n = 173)	Unblinded (n = 189)	Blinded (n = 173)	Unblinded (n = 189)
Residential care	14	16	3	2
Renal inpatient	1678	654	558	739
Intensive care	236	88	24	29
Other inpatient	671	346	224	1324
Renal outpatient	641	532	430	451
Other outpatient	232	266	230	233
Day hospital	4	3	4	5
A&E	54	64	53	61
GP	86	111	87	87
Physiotherapist	41	35	28	72
OT	6	5	3	9
Speech therapist	3	2	0	5
Dietitian	30	26	18	17
Nutritionist	3	4	3	1
Social worker	4	7	1	2
Homecare worker	17	81	1	18
Psychologist	43	28	7	31
Complementary healthcare	25	65	14	44
District nurse	18	7	4	14
Psychiatrist	7	5	0	10
Counsellor	8	25	16	38

The overall costs are reported in Table 18 along with the main cost-effectiveness results. At baseline the costs were higher for the blinded group. At follow-up, the costs were higher for the unblinded arm. The regression model showed that the unblinded group had follow-up costs that were £1523 higher than for the blinded group but this was not statistically significant (95% CI, –£49 to £4074).

TABLE 18 - Total costs, QALYs and incremental cost-effectiveness ratio

Difference between groups and confidence interval produced from a seemingly unrelated regression model with group and baseline cost entered as independent variables.

Difference between groups and confidence interval produced from the same seemingly unrelated regression model with group and baseline utility entered as independent variables.

	Blinded	Unblinded
Baseline cost (£s)	3600	2287
Follow-up cost (£s)	1672	3137
Adjusted follow-up cost difference (£s)	1523 (95% CI, −49 to 4074)a
Baseline utility	0.7959	0.8091
Follow-up utility	0.7828	0.7950
QALYs	1.0525	1.0694
Adjusted follow-up QALY difference	0.0009 (95% CI, −0.0195 to 0.0237)b
Incremental cost-effectiveness ratio	£1,692,222 per QALY

Baseline and follow-up EQ-5D-5L utility scores were similar between groups and with little change over time. The unblinded group had 0.0009 more QALYs than the blinded group and this was not statistically significant (95% CI, −0.019 to 0.024). The ICER showed that for unblinded care to produce one extra QALY, a cost of £1.7 million would be incurred. Uncertainty around the results is shown in Figure 8. The ICER of £1.7 million is far in excess of the threshold used by NICE (£20,000–30,000) and so there is little likelihood of the unblinded option being more cost-effective than blinded care (see Figure 9).

FIGURE 8.

Cost-effectiveness plane. Scatter plot showing that unblinded care is likely to be more expensive, but outcomes are symmetrically distributed around the horizontal axis indicating that unblinded care can result in higher costs with either worse or better outcomes.

FIGURE 9.

Cost-effectiveness acceptability curve. Graph shows low probability that unblinded care is cost-effective, despite high costs.

Adverse events

8189 AEs (670 SAEs) were reported, and 1570 patients (77%) experienced at least one AE (see Table 20). Significant differences were observed for five outcomes/codes with HLA Ab+ participants in the unblinded BLC arm being more likely to experience cardiovascular, respiratory, gastrointestinal and GU/renal AEs than HLA Ab+ participants in the blinded SC arm (see Figure 10). These comparisons are not adjusted for multiple testing however, and any AEs of concern are covered by existing secondary outcomes.

FIGURE 10.

Adverse events recorded in the HLA Ab+ groups. Left panel compares the proportion of patients suffering AEs grouped by body system in blinded (SC: blue circles) and unblinded (BLC: light blue triangles) HLA Ab+ groups. Right panel shows relative risk (95% CI) of developing an AE in the BLC patients, ordered by size of relative risk.

Changes in HLA antibodies

By the end of intensive follow-up, fewer than 2% of the HLA Ab-negative groups became Ab+, more than 50% of the DSA+ participants became HLA Ab-negative, 16–23% lost their DSA but retained non-DSA HLA Ab, and 60–70% of the non-DSA+ participants became Ab-negative. 5.1% of the blinded SC non-DSA+ recruits had developed DSA or possible DSA, compared to 1.6% of the unblinded non-DSA+ participants (see Table 5).

Within the blinded SC group, the same proportion (2/21 [9.5%]) of those with persisting DSA had graft failure as those who became DSA-negative (4/48 [8.3%]). Although in the BLC group, only 2/50 (4%) of the recruits who lost DSA suffered graft failure, compared to 6/28 (21.4%) with persisting DSA, a formal post-hoc analysis of the interaction between persisting DSA and time to graft failure in the main analysis revealed non-significant differences (p = 0.316) in revised HRs. Further analysis of within-group interactions was not undertaken as numbers were small.

Equality, diversity and inclusion statement

We have reported the sex and ethnicity of recruits in several tables because these have a well-documented impact on kidney transplant survival and performance. The centres from which we recruited contained very broad and mixed populations drawn from all ethnic backgrounds. The requirement that recruits had a sufficient grasp of English was not felt to disproportionately bias recruitment of any particular group of individuals. All the recruiting centres have well-defined and published equality, diversity and inclusion policies meaning that our research teams comprised of individuals from diverse backgrounds.

Discussion of results

The OuTSMART trial tested the hypothesis that regular screening of kidney transplant recipients for HLA Ab, a validated biomarker for graft failure, followed by an intervention to improve adherence and optimise immunosuppression, would prolong the life of the organ allografts. Using an open-labelled randomised marker-based strategy (hybrid) design, in which all patients were screened, but only half were unblinded to their results and from which only biomarker-positive patients received the intervention, we recruited more than 2000 patients between 2013 and 2016 and followed them until 2020.

As the largest double-blinded study in transplantation to test a stratified medicine approach to post-transplant care, based on HLA Ab status, and the largest RCT to use graft failure as the primary endpoint, there is no ambiguity about how to interpret the results: OuTSMART further validates the prognostic value of DSA, but finds no evidence to support our hypothesis that intervening can prevent graft failure, with little separation in the Kaplan–Meier curves by group and confirmatory 95% CIs for HRs that included the null value. Further, there were no clear signals in favour of biomarker-led care in the HLA Ab+ groups from any of the secondary outcomes, despite indications of improved adherence, especially in DSA+ patients. That said, fewer DSA+ and non-DSA patients in the BLC arm (4 out of 106, and 5 out of 427, respectively) had biopsy-proven rejection than in the SC arm (9 out of 92 and 8 out 391, respectively), though neither of these differences failed to reach statistical significance. Interestingly, fewer HLA Ab-negative patients in the BLC had rejection (5 out of 495) compared to those in the SC arm (12 out of 525), such that there was a statistically significant reduction in biopsy-proven rejection in the whole BLC arm. These data are difficult to explain but they support our signal of improved adherence and potentially suggest that patient awareness of positive or negative risk associated with a prognostic biomarker impacts on behaviour.

Our health economic analyses confirm that renal transplant patients in both arms of the trial had relatively high levels of hospital use at baseline and follow-up. Patterns of care did not differ substantially over time. However, at follow-up, the unblinded BLC group were making more use of inpatient care, with consequently higher costs. QALYs were almost the same for the groups. The incremental cost per QALY for BLC over SC care was substantially greater than the threshold used by NICE. If the costs of screening were included, then this ratio would be even higher indicating that screening is unlikely to be cost-effective. The screening costs were not available at the time of the analysis. However, subsequently these have been reported as approximately £140 per test. It will be clear that this will make no meaningful difference to the findings on cost-effectiveness. The costs of immunotherapy were not included but these amount to only around £100 per patient (with the widespread availability of generics) and so again would not change the findings reported here.

All these data suggest that routine screening for HLA Ab is currently difficult to justify in the absence of any treatment that impacts of transplant survival. These results will impact significantly on how transplant centres around the world organise their post-transplant monitoring and should encourage a global effort to find novel approaches and treatments to prolong allograft survival in the face of DSA.

The validity of HLA Ab as a prognostic biomarker for kidney transplant failure was first demonstrated in retrospective case–control studies showing a higher prevalence of IgG Ab against donor HLA in failed compared to working transplants. 3,4 Later prospective studies reported a higher graft failure rate in those with HLA Ab compared to patients without. 5,6 Lachmann et al.,7 in a cohort of > 1000 patients reported 5-year graft failure rates of 51% for patients with DSA, 30% for patients with non-DSA and 17% for patients with no HLA Ab. This study also established the importance of repeat testing for HLA Ab and demonstrated that the majority of HLA Ab+ patients who were biopsied showed changes consistent with chronic immune-mediated injury. These findings were corroborated by a second study from the Netherlands,8 in which the risk of graft failure with HLA Ab was also shown to be independent of graft dysfunction and proteinuria. These studies provided one of the foundations for OuTSMART.

At the time OuTSMART was conceived, there were no tested strategies for how to intervene in patients with HLA Ab. Multiple trials, reporting since OuTSMART started, have tested agents targeting B cells with rituximab (± IVIg)14,15 or plasma cells with bortezomib and these have failed to show any impact. 37 Agents targeting IL-6 or IL-6 receptor have shown early promise in early phase studies,17 but larger studies assessing their impact are awaited. Other innovative treatments are at earlier stages of assessment. 38

Our hypothesis was that targeting the cells of the immune system rather than the HLA Ab, might prevent graft failure. There were three aspects to this. First, the knowledge that activated T cells associated with development of HLA Ab. 9,11,39–41 Second, the well-described link between immunosuppression reduction, including from non-adherence and development of DSA and graft dysfunction,9,42,43 meant it was logical to try and enhance immunosuppression in this group. Finally, optimised oral immunosuppression to target cell-mediated responses is known to prevent the development of HLA Ab20,21,44 and graft dysfunction,23–25 but has also been shown to stabilise deteriorating function in those with established immune-mediated dysfunction. 15,19,26,45–47

In keeping with previous work, OuTSMART showed that 15–20% of grafts in patients with DSA failed within the period of follow-up (after DSA detection) compared with 7% in the population who stayed consistently HLA Ab-negative. Although in line with recent observations from the Collaborative Transplant Survey (CTS Newsletter 2 : 2020 1st May), this is much lower than we had expected to see based on Lachmann et al. 7 Another important observation from OuTSMART was that patients who developed a non-DSA had a similar time to graft failure as patients without HLA Ab, in contrast to Lachmann’s data, which suggested a graft survival disadvantage associated with non-DSA HLA Ab. 7 A likely explanation for both differences is the different population demographics, most prominently baseline maintenance immunosuppression. For example, the proportion of patients in OuTSMART taking either Tac (73%) or MMF (67%) were double that in Lachmann’s cohort (35% and 33%), reflecting shifts in practice over the last 20 years.

Having screened > 5000 patients, only 37% were randomised, but 25% failed to provide consent, for various reasons, and 34% failed to meet eligibility, which were designed to ensure safe running of the trial and unambiguous interpretation of the results. We are therefore confident our results have generalisability. Several elements of the trial design need further explanation. First, we chose not to exclude patients who were known to be DSA+ (but XM–) at the time of transplantation, and these accounted for ~23% of recruits. Second, the majority of HLA Ab+ patients in all groups had either DSA (~66%) or non-DSA (~68%) at the point of randomisation, and although most of these were de novo Ab and had developed post-transplantation, only a relative minority developed de novo Ab during our rescreening process. Both these were practical compromises, as we calculated that recruiting sufficient numbers of HLA Ab-negative patients to collect enough DSA+ patients from rescreening alone was not feasible. Since patients with DSA that persist > 12 months post-transplantation are known to be at high risk of chronic rejection and graft failure,48,49 and at least one study has shown a similar prognostic significance for persistent non-DSA,48 both these decisions do not compromise the validity of the data. Third, we changed the primary endpoint during the study from graft failure rate over three years, to time to graft failure with minimum follow-up of 43 months. This was because the prevalence and incidence rates of DSA were lower than anticipated with consequent implications for the number of patients needed. 27,28 This change preserved the power of the trial, without affecting the protocol or general modelling strategy. Although the minimum follow-up period was shortened due to the unplanned COVID-19 pandemic, our sensitivity analyses suggested this did not impact on our conclusions. Fourth, in the original design, development of HLA Ab triggered a transplant biopsy to correlate with graft pathology even in the absence of graft dysfunction. This design aspect was removed after a Patient Public Involvement session at which patients raised serious concerns. Fifth, after allocation into HLA Ab+ groups, no further screening was done until the final visit, at which point we were able to retest 70–80%, revealing that only 50% remained DSA+. While we are confident that our testing regimen, which involved a screening test followed by single antigen testing was identifying genuine DSA, these data might indicate heterogeneity within the DSA+ groups not accounted for in our design. However, a formal analysis of interaction between DSA persistence and our primary endpoint indicated non-significant differences on the HRs. Finally, we designed the trial as a ‘real world’ effectiveness study, such that optimisation was tailored to individual patients, according to compliance, tolerance and achievement of target levels (for Tac). This meant that failure to tolerate one or more of the components of the protocol (or refusal to take any of the agents) was not used as a reason for withdrawal from the study. This aspect of the study was regarded as highly relevant and important by patients and PIs, but had the following consequences: all groups had average Tac levels within our target range, only 50% of the unblinded DSA+ group received the ‘steroid boost’ and many in the blinded groups were on immunosuppression that resembled our optimised regimen.

Nevertheless, more than 95% of the unblinded Ab+ group had the intervention interview, the proportion at the end of the trial on Tac, MMF and prednisolone in the unblinded DSA+ group rose from 64% to 82%, 59% to 73%, and 59% to 76%, respectively, whereas proportions in the blinded DSA+ group stayed constant (~60%), and the proportion taking all three IMPs at the end of the trial rose from 23% to 48% in unblinded DSA+ but stayed constant (~22%) in the blinded DSA+ group. In addition, at the end of the trial, the unblinded DSA+ group had the highest average Tac levels and were on the highest average dose of MMF. Moreover, our formal assessments of adherence revealed evidence of significant differences for each of the IMPs, at least at 12 months after the intervention. All these indicate measurable differences related to our intervention. Moreover, these all probably contributed to the lower rates of biopsy-proven rejection in the BLC patients.

There are several methodological limitations that need highlighting. Firstly, within DSA and non-DSA groups, for rescreened participants we defined the start of the time at risk (time zero) for graft failure when they became HLA Ab Positive rather than at time of randomisation. This assumes there is no effect of the blinding/unblinding to the HLA biomarker in the absence of optimised immunosuppression up until that point. However, the overall comparison would seem to support this assumption, given we found no evidence of a treatment effect for the overall unblinding strategy. We also did carry out a sensitivity analysis within the DSA and non-DSA comparisons using time at risk as randomisation for all participants which gave similar results to the main analysis.

We have made certain assumptions as to missing data, the primary analysis assumes a missing at random mechanism (with data censored at loss to follow-up or death). The effect of the intercurrent event of death was assessed with a sensitivity analysis (again which showed similar results). Other than death, the percentage lost to follow-up was very low (4%) and we consider that if the missingness mechanism was ‘missing not at random’, the impact on the results would be quite small. Secondary analyses also assumed a missing at random mechanism. Measures of the secondary outcomes (other than death) are additionally affected by graft failure and death as intercurrent events, for which the data will be subsequently missing. We haven’t directly assessed the impact of these intercurrent events (other than in the competing analysis of graft failure and death) for the secondary outcomes, and so the estimand for the secondary outcomes should be considered as following a treatment policy strategy (i.e. it is the estimated treatment effect in the absence of death or graft failure). Given we did not find evidence of an effect of the intervention on death or graft failure, this approach would seem reasonable. We have not adjusted reported p-values for multiplicity as we have defined a clear single primary outcome,50 and consequently the results on the secondary outcomes should be considered subsidiary and exploratory rather than confirmatory.

Future research

We believe there are several potential areas of future research suggested by these data. The first is understanding of why some people with DSA deteriorate more quickly than others. If we are correct in asserting that improvements in transplant immunosuppression are responsible for why graft failure rates associated with DSA have improved over time, the implication is that better control of adaptive alloimmunity plays a part in preserving graft function. Recent insights into how subpopulations of regulatory T and B cells associate with phenotypes associated with DSA are consistent with this,51,52 and are an exciting area for future research. Understanding whether these insights relate to whether DSA persist or disappear is another area worthy of further research. Work in both these areas might help elucidate the precise pathophysiological contributions that different immune effector mechanisms, beside DSA, make to graft failure. This should all lead to a more rational design of specific therapies that prevent or halt these processes and help preserve graft function. With this in mind, it is very important that promising potential therapies, such as biological agents targeting IL-6 or its receptor, are properly evaluated in well-designed and appropriately powered RCTs.

Contributions of authors

Mr Dominic Stringer (https://orcid.org/0000-0001-5624-1733) (Statistician, Health Informatics) was part of the team of trial statisticians.

Dr Leanne Gardner (https://orcid.org/0000-0002-1233-9613) (Trial Manager, King’s CTU) was the trial manager.

Dr Olivia Shaw (https://orcid.org/0000-0003-4560-0619) (HLA Bioscientist, Guy’s) was part of the team of HLA laboratory principal investigators.

Dr Brendan Clarke (https://orcid.org/0000-0001-7243-1916) (HLA Bioscientist, Leeds) was part of the team of HLA laboratory principal investigators.

Dr David Briggs (https://orcid.org/0000-0002-6796-7086) (HLA Bioscientist, Brimingham) was part of the team of HLA laboratory principal investigators.

Dr Janet Worthington (https://orcid.org/0000-0002-1241-3975) (HLA Bioscientist, Manchester) was part of the team of HLA laboratory principal investigators.

Dr Matthew Buckland (https://orcid.org/0000-0002-5646-4707) (HLA Bioscientist, Barts) was part of the team of HLA laboratory principal investigators.

Dr Rachel Hilton (https://orcid.org/0000-0001-5118-2017) (Consultant Nephrologist, Guy’s) was part of the team of principal investigators at the original recruiting sites.

Dr Michael Picton (https://orcid.org/0000-0002-3066-0638) (Consultant Nephrologist, Manchester) was part of the team of principal investigators at the original recruiting sites.

Dr Raj Thuraisingham (https://orcid.org/0000-0003-2073-9569) (Consultant Nephrologist, Royal London) was part of the team of principal investigators at the original recruiting sites.

Dr Richard Borrows (https://orcid.org/0000-0003-2218-3023) (Consultant Nephrologist, Birmingham) was part of the team of principal investigators at the original recruiting sites.

Dr Richard Baker (https://orcid.org/0000-0002-9260-7664) (Consultant Nephrologist, Leeds) was part of the team of principal investigators at the original recruiting sites.

Ms Rose Tinch-Taylor (https://orcid.org/0000-0003-3861-6364) (Statistician, Health Informatics) was part of the team of trial statisticians.

Professor Robert Horne (https://orcid.org/0000-0002-3068-8438) (Professor of Behavioural Medicine, UCL) led the health beliefs and compliance team.

Professor Paul McCrone (https://orcid.org/0000-0001-7001-4502) (Health Economist, University of Greenwich) was the trial health economist.

Dr Joanna Kelly (https://orcid.org/0000-0002-4389-5284) (Clinial Trials Operations, King’s CTU) was responsible for database design and helped with trial design.

Dr Caroline Murphy (https://orcid.org/0000-0001-7547-8998) (Director of Operations, King’s CTU) was responsible for database design and helped with trial design.

Professor Janet Peacock (https://orcid.org/0000-0002-0310-2518) (Senior Statistician, Health Informatics) was part of the team of trial statisticians.

Professor Anthony Dorling (https://orcid.org/0000-0003-3102-2600) (Professor of Transplant Inflammation and Repair, KCL) was the chief investigator and designed the trial, held the EME grant and wrote the paper.

List of investigators and UK recruiting centres

The following PIs, based in 13 UK centres were: (1) Anthony Dorling (CI), Rachel Hilton (PI) and Olivia Shaw (Lab PI), Guy’s Hospital, Guy’s and St Thomas’ NHS Foundation Trust; (2) Sapna Shah (PI), King’s College Hospital, King’s College Hospital NHS Foundation Trust; (3) Richard Baker (PI) and Brendan Clarke (Lab PI), St James’s University Hospital, Leeds Teaching Hospitals NHS Trust; (4) Michael Picton (PI) and Judith Worthington (Lab PI), Manchester Royal Infirmary, Manchester University NHS Foundation Trust; (5) Raj Thuraisingham (PI) and Matthew Buckland (Lab PI), Royal London Hospital, Bart’s Health NHS Trust; (6) Richard Borrows (PI) and David Briggs (Lab PI), Queen Elizabeth Hospital, University Hospitals Birmingham NHS Foundation Trust; (7) Keith McCullough (PI), York Hospital, York and Scarborough Teaching Hospitals NHS Foundation Trust; (8) Waqar Ayub (PI), University Hospital Coventry, University Hospitals Coventry and Warwickshire NHS Trust; (9) Aimun Ahmed (PI), Royal Preston Hospital, Lancashire Teaching Hospitals NHS Foundation Trust; (10) Janet Hegarty (PI), Salford Royal Hospital, Northern Care Alliance NHS Foundation Trust; (11) John Stoves (PI), Bradford Royal Infirmary, Bradford Teaching Hospitals NHS Foundation Trust; (12) Kin Yee Shiu and Stephen B Walsh (PIs), Royal Free Hospital, Royal Free NHS Foundation Trust; (13) Mysore Phanish (PI), St Helier Hospital, Epsom and St Helier University Hospitals NHS Foundation Trust.

Ethics statement

The initial study protocol was approved on 14 January 2013 and given the reference number 12/LO/1759 by the National Research Ethics Service (NRES Committee London–Hampstead, Skipton House, Ground Floor, NRES/HRA, 80 London Road, London SE1 6LH), who approved all subsequent amendments.

Data-sharing statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Disclaimers

This report presents independent research. The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, the MRC, the EME programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, the EME programme or the Department of Health and Social Care.

Summary of changes to protocol approved by the ethics committee

All changes were discussed and approved by the Trial Steering Committee or Chairman and, where appropriate, by the Data Monitoring Committee.

The changes made in Version 11 of the protocol reflect the major changes in the design and endpoints that are incorporated into the final version.

Version 2 07/11/12

• Change to section 3.1 to reflect that MMF was being used outside its marketing authorisation.

Version 3 29/1/2013

• Changes to sections 2.2.1 and 6.2 relating to assessment of adherence and risk perception. Rather than collecting prescription redemption data (version 1 and 2), we proposed tablet counts on randomly selected patients in addition to the use of iPads or similar tablets to collect the data, prior to electronic transfer to a secure server hosted by University College London. Thirdly, we proposed to pilot the questionnaires and perform quantitative interviews on a small number of participants initially recruited to Guy’s, to inform whether existing standardised questionnaires required change to suit this population.

• Change to section 7.3 relating to a change in the pharmacovigilance policy of the sponsor to ensure Important Medical Events are recorded as SAEs.

Version 4 13/5/2013

• Changes to sections 2.2 and 2.3 to clarify that randomisation will be stratified by site.

• Change to section 4.1 to clarify that the eGFR measurement on which eligibility will be assessed has to be within 1 month of signing consent.

• Change to section 5.2.1 to clarify the definition of a positive HLA Ab test, which was confusing in the previous protocol versions.

• Changes to mention of recruitment targets, shifting emphasis away from precise predictions towards a more pragmatic approach that highlights recruitment will stop once minimum numbers required for statistical power have been recruited to each of the individual groups.

Version 5 9/7/2013

• Change to lab PI at the Royal London Hospital.

• Change to section 3.1 clarifying the dosing of one the IMPs, Prednisolone.

• Change to reflect updated WHO definition of DM (addition of HbA_1c testing).

• Change to section 7.3.1, reflecting the fact that certain AEs in this type A trial may not require reporting to the sponsor, but may still require recording in the eCRF.

Version 6 6/12/2013

• Change to abandon the requirement that recruits be tested for hepatitis B core Ab, as it was hindering collection of samples for scientific analysis. Since core Ab positivity was not a contraindication to optimisation, and testing was not required by King’s College London, as the infectivity of samples from core Ab positive, surface antigen-negative samples is very low, this change was felt not to compromise the trial in any way but would enhance the number of scientific samples obtained at recruitment.

• Change to allow urine as well as blood testing to rule out pregnancy.

• Changes relating to recruitment of live donors, to ensure they were tested for HIV and Hepatitis B and C if not tested within the previous 5 years, to increase volume of blood taken to 80 ml and to allow consent to be obtained by non-clinicians be allowed to consent these patients.

Version 7 7/4/2014

1. Changes to maximise and standardise recruitment across sites:

   • Removal of exclusion criteria ‘history of ongoing or previous infection that would prevent optimisation’. This criterion was vague (i.e. did not define which infections were important) and was being interpreted differently within and across sites. As the optimisation for each participant was optimised to that particular individual, immunosuppression could be tailored according to their medical history.

   • Increase in the gap for the testing of eGFR from within 1 month of signing consent to within the previous 6 months of signing the consent. Participants had to have an eGRF≥30 to be eligible for the study, and by increasing the time-period to within the previous 6 months, screening for potential participants can be more efficient as measurements of eGFR from previous renal clinic appointments could be used.

2. Removal of need to perform total immunoglobulin testing:

   • This measurement proved to be a difficult and expensive test to perform and was not routinely performed by all hospital laboratories. This test had originally been incorporated into the study to ensure participants were not developing MMF-induced hypogammaglobulinemia. Fortunately, this could still be detected by maintaining requirement to test for IgG, IgM and IgA.

   • Testing and recording of IgG, IgM and IgA moved to every year instead of every four months. Testing for the levels of these immunoglobulins every 12 months was felt to be sufficient by the TSC sufficient for monitoring MMF-induced hypogammaglobulinemia.

3. Clarification of and changes to follow-up procedures:

   • Clarification that participants would see a research nurse for all trial-related procedures at follow-up appointments, which will be held at the same time a participant is in routine clinic.

   • Details regarding the fact that only medications being taken at the time of the follow-up would be recorded.

   • Clarification of the timings for questionnaire completion in table 2.2.1.

   • Patients could be given an appointment slip containing a telephone number and/or an email address to contact research nurses if their routine appointments are rescheduled.

   • Clarification of the time windows for follow-up appointments that were allowed without deviation to the protocol.

4. Clarification of the optimisation process for participants allocated to the unblinded HLA Ab positive arm:

   • Change of optimisation timing from within 3 months of HLA Ab positivity to ideally within 3 months after positive screening for HLA Ab and allocation to the unblinded treatment arm or as soon as possible thereafter BUT within 8 months of positive screening. This coincided with the realisation that some patients were proving difficult to contact to arrange optimisation and the change was felt to enhance the optimisation process without affecting the outcome of the trial.

   • Clarification of the way that patients will be informed about the group to which they have been allocated. Participants with HLA Ab allocated to the unblinded arm will be told the results of this allocation as soon as possible and invited to undergo optimisation. This can be performed over the phone. Those participants in the blinded groups or in the unblinded HLA Ab negative group will be told the result of their randomisation at their next clinic visit.

   • Clarification that recording details of the optimisation process will be in an Optimisation Log at each site and not in the eCRF.

5. Addition of three new sites.

Version 8 1/7/2014

• Extension to the time that tissue typing laboratories had to perform the randomisation of patients, from 28 to 56 days post consent. This was to optimise batching of patient serum for testing, reducing the number of experimental controls and HLA screening beads needed, and therefore the cost of screening.

• Clarification, in section 7.1, of when testing of HbA_1c should occur.

• Clarification about tests to be performed to monitor for MMF-induced hypogammaglobulinemia.

• Changes to allow the use of results from routine clinic blood tests taken up to a week prior to consent, to minimise duplication of tests in sites where it was routine for patients to attend for blood tests prior to their clinic visit.

• Changes to allow study information to be collected via telephone to minimise time spent with each patient during busy routine clinics.

Version 9 15/10/2014

• Clarification around timing and need for collection of experimental research samples (laboratory) at point of consent.

• Further clarification of the time windows for follow-up appointments that were allowed without deviation to the protocol. Changes made to try to ensure collection of three study assessments per year.

• Change to reduce nurse paperwork: as a Type A trial with a large recruitment target, missing data around sample collection was to be coded in the eCRF but not recorded as a protocol deviation.

• Change to the PI at one of the sites.

Version 10 11/08/2015

• Changes to two coinvestigators in the Tissue Typing Laboratories and addition of three new sites.

• Change to eligibility age range, from 18–70 years to 18–75 years. Originally, we believed that non-transplant-related mortality may be higher in 71–75-year-old age group, however this was reconsidered and felt not to be an issue.

• Clarification of the inclusion criteria relating to timing of the eGFR used when considering eligibility.

Version 11 26/11/2015

1. Change in primary endpoint

   • The primary endpoint of the trial was changed from ‘graft failure rates over three years’ to ‘time to graft failure with variable follow-up (with a minimum of 43 months post-randomization)’. This change was required to account for the low numbers of DSA-positive participants being recruited to the trial. This change will allow for a reduction in the number of DSA patients to be recruited, and a significant shortening in the expected study duration while maintaining the power of the study. Section 8 on sample size and statistical analyses were changed accordingly. The new primary endpoint was to be assessed remotely from patient notes once 43 months post-randomisation was achieved by all. All patients already recruited were to be reconsented to allow this change.

2. Changes to reduce costs associated with extension of the trial

   • Follow-up visits changed from 4-monthly to 8-monthly to reduce nurse workload.

   • End visit for each participant changed from 36 to 32 months.

   • Secondary endpoint assessment changed from 36 to 32 months, except for health economics which was moved from 36 to 16 months.

   • Major reduction of SAE reporting to sponsor incorporated into sections 7.2 and 7.3.

3. Change in trial statistician, change in site PIs and addition of a new site.

Version 12 1/12/16

• Cessation of collection of research blood samples, associated with removal of the secondary experimental ‘scientific’ endpoints. This was required by the funder, who requested that the salary costs associated with the experimental aspects of the trial be reallocated towards supporting the primary endpoint data collection.

Version 13 21/11/18

• Change to reflect inclusion of albumin:creatinine ratio as a measure of proteinuria in addition to protein:creatinine ratio and clarification that one or the other (not both) are required as one assessment of graft dysfunction.

• Change to the way that change in eGFR was to be compared between arms.

• Inclusion of proposed details for how the primary outcome data was to be collected during the period March 2020–June 2020.

• Inclusion of collection of a final sample for HLA Ab screening from all participants at their final research clinic visit at month 32.

• Clarification that results from the trial will be presented as estimates and 95% CIs.

• Clarification that baseline covariates were to be included in the statistical model for the primary outcome.

Version 14 08/07/2020

• Change to the timing of the collection of the primary endpoint as a result of the COVID-19 pandemic, in addition to the proposal to include additional sensitivity analyses for the primary endpoint and extension of the study end date.

• Inclusion of a thank you card for all participants who have taken part in the OuTSMART trial.

SOP relating to HLA Ab determination