Efficacy and Safety of RBX2660 in PUNCH CD3, a Phase III, Randomized, Double-Blind, Placebo-Controlled Trial with a Bayesian Primary Analysis for the Prevention of Recurrent Clostridioides difficile Infection

Trial Design

PUNCH CD3 was a randomized, double-blind, placebo-controlled, phase III trial (Fig. 1 of the Electronic Supplementary Material [ESM]). This trial was conducted in the USA and Canada according to the ethical principles of the Declaration of Helsinki, Good Clinical Practice guidelines, principles of informed consent, and requirements of publicly registered clinical trials. The protocol received institutional review board approval before its commencement and was conducted under an FDA Investigational New Drug application. Authors had full access to the data and vouch for the accuracy and completeness of the data and for compliance with the protocol and statistical analysis plan.

The development program was originally planned to include two randomized, placebo-controlled, pivotal phase III trials. However, the widespread availability and use of FMT under enforcement discretion made it increasingly difficult to enroll patients into a placebo-controlled trial within this orphan indication. The FDA acknowledged the increasing recruitment difficulties and recommended that innovative designs, such as formal borrowing of data in a Bayesian framework, could be pursued. Therefore, PUNCH CD3 was analyzed using a Bayesian hierarchical model borrowing data from the previous phase IIb trial (PUNCH CD2). In addition, two interim analyses were added for early stopping for futility or efficacy.

Participants

Participants were adults (aged ≥ 18 years) with documentation of rCDI (defined as one or more recurrences after a primary episode) who had completed one or more rounds of standard-of-care antibiotic therapy or had two or more episodes of severe CDI resulting in hospitalization within the past year. Within 30 days before enrollment, participants were required to have a positive stool test for the presence of C. difficile with the capability to produce toxins assessed by polymerase chain reaction (PCR), enzyme immunoassay (EIA), or other assays. Participants must have been taking or just been prescribed antibiotics to control rCDI symptoms. The antibiotic, dose, and regimen for the qualifying event were at the discretion of the treating physician; therefore, taper or taper/pulse regimens were allowed. Participants agreed not to ingest over-the-counter/prescription probiotics for 8 weeks following study treatment. Participants were excluded if they had a known history of refractory CDI, inflammatory bowel disease, irritable bowel syndrome, chronic diarrhea, celiac disease, colostomy, active colitis, continued diarrhea despite antibiotic therapy, required antibiotic therapy for another condition, or had a previous FMT. All key inclusion/exclusion criteria are listed in the Appendix of the ESM, page 5.

Randomization, Interventions, and Masking

Enrolled participants were randomly assigned (2:1) to receive RBX2660 (microbiota suspension) or placebo (normal saline). Randomization was stratified by antibiotics used for the qualifying CDI event (vancomycin alone, vancomycin in combination with another antibiotic, fidaxomicin alone, or other); randomization was not stratified by site.

Both RBX2660 and placebo were administered rectally following the instructions for use and standard site procedures after a full course of antibiotic treatment for rCDI and following a washout period of 24–72 h, and within 14 calendar days of randomization. Enema administration was performed by a qualified and trained healthcare professional not involved in other procedures or assessments during the trial, and product-specific instructions for use were provided to ensure consistency in the procedure. No bowel preparation was required. Study blinding was maintained during administration by covering the infusion bag and tubing in an opaque sleeve. In the event of treatment failure within the first 8 weeks of blinded treatment, participants were offered the opportunity to receive an open-label treatment of RBX2660.

Study Outcomes of PUNCH CD3

The primary endpoint was treatment success, defined as the absence of CDI diarrhea within 8 weeks of study treatment. Sustained clinical response was a secondary endpoint, defined as treatment success of the presenting CDI recurrence and no new CDI episodes for greater than 8 weeks through 6 months after completing a study treatment. Evaluation of safety and tolerability included the incidence and severity of AEs and serious AEs (SAEs), which were collected during follow-up visits, telephone assessments, and diaries through the 6-month follow-up. Although AE collection started at the time of consent, the follow-up for treatment-emergent AEs started at the time of study drug administration.

Statistical Analysis

During PUNCH CD3 enrollment, an agreement was reached with the FDA to implement an adaptive design, with two interim analyses to evaluate the primary endpoint for possible early declaration of study success or futility. Two prespecified sponsor-blinded interim analyses were planned when 160 and 220 participants had completed the week 8 visit; a Pocock α-spending approach was used to determine the stopping criteria at the interim analyses (see ESM, page 8).

Given the recruitment challenges noted above, the FDA agreed with analysis of the primary endpoint using a Bayesian hierarchical model that dynamically borrowed information about the treatment effect from the previous phase IIb trial, PUNCH CD2, taking into account differences in response rates between the two trials. This analysis was considered appropriate, given the similarity of the PUNCH CD2 and CD3 study designs (Table 1 of the ESM). This model incorporated data from the PUNCH CD2 study from the one-dose RBX2660 group and placebo control group (not the two-dose RBX2660 group). The advantage of modeling the data jointly in this manner is that if the treatment effect is similar in both studies, a combined analysis can reduce the amount of uncertainty in the estimate. The Bayesian hierarchical model provides estimates of the treatment success rates for each treatment group in PUNCH CD3 as well as the estimated treatment effect and the associated posterior probability of superiority. The Bayesian hierarchical model (see ESM, pages 7–11) was included in the statistical analysis plan before enrollment completion and unblinding for any interim or final analyses.

This study included two superiority thresholds: (1) posterior probability of superiority > 0.999 selected to control the nominal type I error rate without borrowing at one-sided 0.00125; and (2) posterior probability of superiority > 0.975 selected to control the nominal type I error rate without borrowing at one-sided 0.025. The higher threshold therefore corresponds to a statistically very persuasive finding. The lower threshold provides evidence of a statistically significant phase III trial (see ESM, pages 7–11).

PUNCH CD3 analysis populations included the modified intent-to-treat (mITT), ITT, per-protocol (PP), and safety populations (defined in ESM, page 12). The efficacy results in the mITT population were considered the primary outcome, while the results in the ITT and PP populations were considered supportive. The original intention was to incorporate PUNCH CD2 data from the ITT population only as this was the primary analysis population in that study. During the Biologics License Application review, the FDA recommended increasing exchangeability between the two studies by applying the PUNCH CD3 analysis population definitions to the PUNCH CD2 data and matching the analysis populations when borrowing (i.e., PUNCH CD2 mITT data were borrowed for the PUNCH CD3 mITT primary analysis and PUNCH CD2 ITT data were borrowed for the PUNCH CD3 ITT sensitivity analysis). For completeness, results based on both approaches for borrowing are presented for the primary endpoint. Results for prespecified subgroups are presented based on the FDA-recommended approach only.

Sustained clinical response through 6 months was assessed with frequentist analyses using Pearson’s chi-squared test, with a two-sided α = 0.05. Two-sided 95% confidence intervals for the differences in proportions between treatment arms were calculated. Statistical analyses were performed using SAS version 9.2 (SAS Institute Inc., Houston, TX, USA), R version 3.4.3 (R Core Team 2017), and Stan version 2.17.2 (Stan Development Team 2018).

Data Monitoring

To ensure the safety and well-being of participants throughout the study, an independent data and safety monitoring board reviewed safety data for trends and stopping rules. An endpoint adjudication committee provided independent blinded adjudications of treatment success or failure. An independent statistical analysis committee conducted both interim analyses and the final primary efficacy analysis independent of the sponsor, data and safety monitoring board, or endpoint adjudication committee.

留言 (0)

沒有登入
gif