Statistical analyses and sample size

The DGF rate in DCD transplanted kidneys was established previously at approximately 50% [3] and this was confirmed with contemporary registry data held by the UK National Health Service Blood and Transplant. Further details are provided in S5 Supplementary Material in S1 File.

A Bayesian sequential analysis of efficacy data was planned to allow for the possibility of stopping early for success or failure in this single-arm study. The statistical impact of various sample sizes on type I error rates was explored for a background DGF rate of 50%, along with the power for detecting a targeted 35% DGF rate with the GSK1070806 experimental therapy. This represents a 30% relative reduction in DGF from the 50% background rate (S5 Table in S1 File), is deemed to be clinically meaningful and predictive of a significant impact on clinical outcomes in any subsequent larger, late-stage clinical trials. Given the desire to minimize the number of patients in this proof-of-concept study, it was planned with a maximum number of 30 patients, which was considered a feasible sample size that could be recruited at a small number of clinical sites over a reasonable time period.

Although a maximum cohort size of 30 was selected for this study, the actual number was determined based on patients’ sequential DGF outcomes. The sample size and decision criteria were selected to be adequately powered to detect the prespecified treatment effect and so there was a reasonable type 1 error rate in the event treatment was ineffective. The design yielded the probability of an erroneous “Go” decision of 0.139 when the GSK1070806 DGF rate is 50% (i.e., null hypothesis), and a probability of an appropriate “Go” decision of 0.659 if a DGF rate of 35% occurs (Fig 2A; S5 Table in S1 File).

(A) Probability of a Go decision by DGF rate* (B) Sequential Go/No Go/Continue rule^†. *At a maximum of 30 patients, this design yields the probability of a Go decision of 0.139 when the GSK1070806 DGF rate is 50% (i.e., null hypothesis) and 0.69 at what has been a clinically impactful GSK1070806 DGF rate of 35%. ^†The number in the first column indicates the number of patients who have completed study treatment. A sequential Go/No Go/Continue rule is based on the predictive probability of success. A high predictive probability (PP) of success means that GSK1070806 is likely to be efficacious by the end of the study given the observed data, whereas a low PP suggests that the treatment may not have sufficient activity. If the PP value <2% (red region) the alternative hypothesis is rejected. If the PP is >92% (green), the conclusion may be made that GSK1070806 has better efficacy than the standard of care. If the PP is 2–92% (white region), the trial will continue to the next interim or until reaching 30 completed patients. The sequential path observed in the study is represented by the orange line. Although the pathway ends in the white region, the decision to terminate the study was made as only 1 of the 7 patients completing study treatment was not on dialysis and had creatine <400 μmol/L; this suggested that it was unlikely that GSK1070806 3 mg/kg reduced the risk of DGF. DGF, delayed graft function.

“Success” at full enrolment was defined as 11 or fewer DGF events in 30 patients (<37% DGF rate) with the outcome of the study depending on the DGF rate in the presence of GSK1070806 administration. With patient DGF status evaluated sequentially as each patient reached 7 days post transplantation, early success on the primary endpoint would be declared if the predictive probability of success was >0.92 (i.e., 92% probability the eventual number of DGF events would be 11 or fewer if the study continued to full enrolment) and a failure would be declared on the primary endpoint if the predictive probability of success was <0.02. Fig 2B illustrates which events may trigger an early Go (success) decision as the predictive probability of success is >0.92 (green box), and a No Go (failure) decision as the predictive probability of success is <0.02 (red box).

A summary of statistical analyses for the study endpoints can be found in S6 Supplementary Material in S1 File.