Clinical results

Between August 2014 and November 2015, 290 patients were assessed for eligibility and 232 randomly assigned to receive intravitreal aflibercept (n = 116) or PRP (n = 116).

The trial was conducted in 22 NHS trusts. Recruitment was completed ahead of schedule in 17 months. Table 5 shows the total number of participants recruited from each site and recruitment by calendar month.

The actual recruitment period was on target to the planned original target of 18 months. Figure 6 shows the actual monthly recruitment compared with the original target.

FIGURE 6 Cumulative monthly accrual of patients into the CLARITY trial. (Figure not reproduced here; see the original article.)

Trial milestones

Study start date was August 2014 and was planned to involve 17 recruiting sites around the UK. The sample size was 220 participants. The eligible patients were identified from the medical retina clinics. New patients were also referred in from the diabetic retinopathy screening programmes. As these patients had to be seen in retinal clinics within 2 weeks of referral and treated within 4 weeks, it was necessary for the trial team to communicate well with the referral team to ensure that these patients were identified before they were treated with PRP in the clinic.

Success in trial recruitment and management

The National Institute for Health Research (NIHR) recognises the need for good trial management and recommends that a clinical trials unit collaborate in the conduct of the study. A multidisciplinary team at the KCTU was funded for the study. As a collaborator on the CLARITY study, the unit ensured the appointment of a dedicated trial manager – a crucial step for the success of this study. The KCTU also has a trial methodologist, data manager, operations manager and statisticians who met with the chief investigator, co-lead and the trial manager for initial brainstorming sessions and then on a regular basis to ensure the smooth running of the study. This group approach and shared ownership were very useful in the successful delivery of the study. During the early part of the set-up phase of the study, the team met often to develop the protocol, obtain all the regulatory approvals and prepare the IMP delivery system. The study source documents and case report forms were user-friendly and we paid significant attention to detail in order to capture all of the required outcome measures. Following the approvals, meetings focused on targeting and solving barriers to site set-up, identifying and managing recruitment hurdles and monitoring recruitment. During these meetings, we evaluated the reasons for initial delay in recruitment.

The start-up delays at some sites were associated with the following:

• delay in completion of training and certification of the staff for the study
• IT system delay
• lack of approved devices to do the study assessments
• delay in site approval process
• patients being treated with PRP at first attendance itself to maintain national time to treatment target
• patients referred from diabetic retinopathy screening programmes not being targeted
• other retinal consultants preferred standard of care to intervention
• research team and clinical team were not at the same area and so patients were being missed.

Recruitment strategies used

The target recruitment rate for the study was one to two patients per month per centre, based on the original 17 centres recruiting and a target recruitment figure of 220. We decided to open five further recruiting centres to counteract initial slow recruitment. We made no changes to the inclusion criteria and the recruitment time was not extended.

The trial management meetings enabled us to pay attention to the trial pathway and ensured that all barriers were tackled in time as we progressed. The trial manager and the chief investigator were always available to provide guidance and support to the clinical sites. The sites were chosen based on previous clinical trial experience or by the estimated volume of potentially eligible patients. Interested sites completed a site feasibility questionnaire. Some sites required support from the UK Clinical Research Network to be able to recruit patients for the study.

Recruitment was monitored closely. We sent out a recruitment league table in a newsletter to all sites every month initially so that the sites were aware of the progress of the study. The chief investigator also e-mailed or telephoned the principal investigators of clinical sites that were recruiting below target. The trial manager had to travel to all sites to monitor the study at the sites.

Screening and eligibility

A total of 290 patients were assessed for eligibility. Of these, 58 were excluded either because they did not meet the inclusion/exclusion criteria (n =  51) or because they were not eligible or not keen (n =  7). The remaining eligible patients (n = 232) were randomised into the trial.

Randomisation

A summary of the recruitment and randomisation across the 22 sites has already been detailed in Table 5. Randomisation was balanced across the treatment groups and hospital sites, and within strata.

Withdrawals

Table 6 shows the numbers of participants who did not complete week 52 in both arms. In total, 21 out of 232 (9.1%) did not complete week 52: 9 from the aflibercept arm and 12 from the PRP arm. Table 7 shows the time to withdrawal and reason for withdrawal, for all patients. The withdrawals were requested by the participants themselves. The withdrawals were balanced between arms.

Baseline characteristics

Baseline characteristics were well balanced between treatment groups, as shown in Table 8. A total of 123 (53%) treatment-naive and 109 (47%) non-naive patients were recruited. Mean baseline BCVA was 81.4 (SD 8.1) ETDRS letters. The proportion of patients with baseline BCVA 54–69 and ≥ 70 ETDRS letters was 9% and 91%, respectively. Table 9 shows the baseline BCVA.

The trial flow

The CONSORT flow diagram (Figure 7) shows the participant flow through the study period.

FIGURE 7 Participant flow through the study period.66 a, ITT population included in the sensitivity analysis. Reprinted from The Lancet, Vol. 389, Sivaprasad S et al.,66 Clinical efficacy of intravitreal aflibercept versus panretinal photocoagulation for best corrected visual acuity in patients with proliferative diabetic retinopathy at 52 weeks (CLARITY): a multicentre, single-blinded, randomised, controlled, phase 2b, non-inferiority trial, pp. 2193–203, © 2017, with permission from Elsevier. (Figure not reproduced here; see the original article.)

Derivation of the intention-to-treat model and per-protocol populations

Patients included in the prespecified ITT LME model were derived as follows:

1. The BCVA data were available for 211 patients of 232 randomly assigned patients (n = 107 in the aflibercept arm and n = 104 in the PRP arm) at 52 weeks and for 214 patients at 12 weeks (n = 109 in the aflibercept arm and n = 105 in the PRP arm).
2. A total of four patients in the PRP arm at 12 weeks and two patients in the aflibercept arm at 52 weeks were excluded because of the presence of vitreous haemorrhage within 3 months of BCVA recordings and BCVA was more than 3 SD below the mean at that time point (including all measurements).
3. There were 198 patients with BCVA available at both 12 and 52 weeks. A total of 11 patients had BCVA recorded at 52 weeks and not 12 weeks (n = 8 in the PRP arm and n = 3 in the aflibercept arm). In addition, there were 12 patients who had BCVA recorded at 12 weeks but not at 52 weeks (n = 5 in the PRP arm and n = 7 in the aflibercept arm).
4. There were therefore 221 patients who contributed to the analysis in the LME model for the ITT strategy (n = 109 in the PRP arm and n = 112 in the aflibercept arm).
5. A total of 18 patients did not meet the PP definition and were not included in the PP population (n = 214). This included 11 (9.5%) patients in the aflibercept arm and 7 (6.0%) in the PRP arm, with 4 patients in the aflibercept arm and 4 in the PRP arm not being compliant with the eligibility criteria and a further 7 patients in the aflibercept arm and 3 in the PRP arm not receiving initial mandatory treatment requirements. There were therefore 210 patients who contributed to the PP analysis in the LME model (n = 106 in the PRP arm and n = 104 in the aflibercept arm).

Primary outcome

The primary outcome at 52 weeks showed that aflibercept was both non-inferior and superior to PRP in terms of BCVA in both ITT and PP populations (Table 10 and Figure 8). The 95% CI for the adjusted difference between arms fell both above the prespecified acceptable non-inferiority margin of –5 letters and above the superiority margin of 0 letters, at both 12 and 52 weeks.

FIGURE 8 Primary outcome at 52 weeks (ITT and PP). (Figure not reproduced here; see the original article.)

Three sensitivity analyses on the population with completed follow-up at 52 weeks were carried out, adjusting for sites, outliers and missing data. No patients were offered anti-VEGF treatment for macular oedema in the PRP arm so sensitivity analysis for concomitant treatments was not required. When sites were considered, the adjusted difference in BCVA between arms remained significant at 4.1 letters (95% CI 2.4 to 5.7 letters; p < 0.0001) and 4.1 letters (95% CI 2.4 to 5.7 letters; p < 0.0001) in the ITT and PP populations, respectively. A total of 207 and 198 patients remained after outliers in the ITT and PP populations, defined as less than or more than 4 SDs, were removed. This sensitivity analysis showed the adjusted difference in BCVA between arms as significant at 4.0 letters (95% CI 2.7 to 5.4 letters; p < 0.0001) in the ITT and 4.1 letters (95% CI 2.7 to 5.5 letters; p < 0.0001) in the PP population. The sensitivity analysis for missing data also confirmed a superiority effect in both ITT (n = 232) and PP populations (n = 214) for three prespecified alternative scenarios.

The primary outcome was also analysed at 12 weeks and the superiority of aflibercept was noted as early as 12 weeks in both the ITT and PP population (Figure 9).

FIGURE 9 Visual outcome at 12 weeks in the (a) ITT population; and (b) PP population. (Figure not reproduced here; see the original article.)

Subgroup analysis of BCVA in both the ITT and PP populations at 52 weeks was carried out based on the baseline PDR status and HbA_1c categories (Table 11).

Table 12 shows the change in BCVA in patients categorised into diastolic BP > 90 mmHg versus those ≤ 90 mmHg. There was no significant difference in BCVA outcome between patients who presented with BCVA 54–69 letters than those who presented with 70 letters or better.

Secondary outcomes

The proportion of patients with a ≥ 10-letter improvement and able to do so with a baseline BCVA of ≤ 90 was 5% (5/101) in the aflibercept arm compared with 2% (2/95) in the PRP arm (difference between arms was 2.8%, 95% CI –3.1% to 9.1%; p = 0.45). The proportion of patients with a ≥ 10-letter worsening was 5% (5/107) in the aflibercept arm compared with 15% (16/104) in the PRP arm (difference between arms was 10.7%, 95% CI 2.6% to 19.3%; p = 0.009). There were 5% (5/107) of patients with a ≥ 15-letter worsening in the aflibercept arm and 6% (6/104) in the laser arm (difference between arms was 1.1%, 95% CI –5.5% to 7.9%; p = 0.72) (Figure 10).

FIGURE 10 Best corrected visual acuity (≥ 10 letters) improvement and worsening at week 52. (Figure not reproduced here; see the original article.)

Binocular Esterman scores showed significant worsening in the PRP arm. This was also reflected in lower binocular visual acuity scores in the PRP arm. Other visual function tests did not vary between arms. Table 13 shows changes in visual function between baseline and week 52.

The NEI-VFQ-25 scores did not show significant differences between arms (Table 14).

The RetTSQ scores showed that patient satisfaction scores were significantly better in the aflibercept arm and the adjusted mean difference was 3.0 (95% CI 0.4 to 5.5; p = 0.022). There are no thresholds for clinically meaningful change for the RetTSQ.67 However, in the validation of the RetTSQ a three-point difference was observed between those with very good visual acuity and those in the category below (five-category visual acuity variable).38 The quality-of-life change assessed using RetDQoL did not show any difference between arms (Table 15).

Anatomical outcomes

Macular thickness and volume (Table 16) significantly increased in the PRP arm compared with the aflibercept arm. The proportion of patients with new-onset centre-involving macular oedema also increased significantly in the PRP arm.

The proportion of patients with macular oedema at 52 weeks assessed by the clinical investigator was significantly higher in the PRP arm than in the aflibercept arm (Table 17).

Treating investigators determined regression and reactivation patterns of retinal new vessels to decide retreatment based on predefined criteria. Table 18 shows the proportion of patients with each regression pattern in each arm.

A significant proportion of eyes showed total regression of retinal new vessels in the aflibercept arm compared with the PRP arm. The difference in proportions of total regression favouring the aflibercept arm was 30% (95% CI 16% to 42%; p < 0.0001) at 52 weeks.

NetwORK UK, masked to treatment allocation, graded ETDRS diabetic retinopathy severity scores from colour fundus photographs obtained at baseline, 12 weeks and 52 weeks.14 Of patients with gradable photographs (n = 227), 175 (77%) were graded low-risk PDR (levels 61 and 65) and 52 (23%) high-risk PDR (levels 71 and 75). Three eyes were graded below level 61. Improvement from diabetic retinopathy severity score is difficult to assess in lasered eyes and so the improvement of the level of remaining retinopathy was graded. A significantly higher proportion of patients in the PRP arm remained at PDR (level 61 or above) than in the aflibercept arm at both 12 and 52 weeks.

Treatment outcomes

The proportion of patients who received treatment in accordance with protocol was 94% (109/116) in the aflibercept arm and 97% (113/116) in the PRP arm. The treatment allocation guess form, which measures the success of masking of primary assessors to treatment allocation, was reported for 210 participants. Assessors guessed correctly for 15% (32/210), incorrectly for 10% (20/210), and were unable to tell for 75% (158/210) of participants.

By 52 weeks, aflibercept arm patients received a mean (SD) of 4.4 (1.7) injections (95% CI 4.1 to 4.7) [median (IQR) 4.0 (3.0 to 5.0)] including the three mandated loading doses. The mean number of aflibercept injections in treatment-naive patients was 4.6 (1.6) [median (IQR) 4 (3.0 to 6.0)] while non-naive patients received a mean number of injections of 4.1 (1.8), [median (IQR) 4.0 (3.0 to 4.8)]. A total of two (1.6%) patients required supplemental PRP in the aflibercept arm.

In the PRP arm, 78 (69%) received multispot laser and the remaining received single-spot laser. The type of laser delivery was not recorded for three patients. From week 12, 75 patients (65%) in the PRP arm required supplemental PRP.

Safety outcomes

A breakdown of the ocular adverse events reported is shown in Table 19. Trials such as CLARITY are not designed to detect reasonably small effects in secondary safety outcomes because adverse events are typically sparsely categorically distributed and sample size is set to provide power instead for the primary outcome. The ratio of false-positive significant findings to genuine significant findings is consequently larger than for primary outcomes. We therefore use p-values as a guide only, with caution, and we use 95% CIs whose width demonstrate the extent of the lack of information around the effects between arms in adverse event rates. Twice as many patients in the PRP arm (18%) developed vitreous haemorrhage as patients in the aflibercept arm (9%; p = 0.034). This 9% higher PRP rate had a wide 95% CI from 1% to 18%. The percentage of patients requiring vitrectomy was 6% in the PRP arm compared with 1% in the aflibercept arm (p = 0.066). This 5% higher PRP rate had a 95% CI from –0% to 11%. Later phase trials may be helpful to validate these results and to add to the body of safety evidence around these and other preliminary safety event results. There were no cases of NVI, NVA or NVG in either eye observed in a study of this size, where absence of evidence (low rates) means that there is low precision in estimated rate differences; this would be ameliorated for study treatments taken forwards in larger and longer later phase trials. There were no cases of endophthalmitis in the study. Table 20 shows the frequency of non-ocular adverse events in both groups. Twice as many patients in the aflibercept arm (n = 8) had APTC-defined events than in the PRP arm (n = 4), representing a 3% excess (Table 21). This was not statistically significant (p = 0.25) and the 95% CI for the between-arm difference was –3% to 10%. Three patients died during the trial (two in the aflibercept arm and one in the PRP arm).

Pregnancy

Two trial participants in the aflibercept arm and three in the PRP arm became pregnant despite being advised on double contraception. One patient’s partner became pregnant in the aflibercept arm (Table 22). Aflibercept was stopped in both the patients who were randomised to aflibercept, but none of them required supplemental PRP during the trial. The patient whose partner became pregnant did not need further treatment as per retreatment guidelines. All pregnancies were followed up after the trial and there were no adverse events reported among the mothers or babies.