COBALT: A Confirmatory Trial of Obeticholic Acid in Primary Biliary Cholangitis With Placebo and External Controls

Abstract

Introduction: Obeticholic acid (OCA) treatment for primary biliary cholangitis (PBC) was conditionally approved in the phase 3 POISE trial. The COBALT confirmatory trial assessed whether clinical outcomes in patients with PBC improve with OCA therapy.

Methods: Patients randomized to OCA (5-10 mg) were compared with placebo (randomized controlled trial [RCT]) or external control (EC). The primary composite endpoint was time to death, liver transplant, model for end-stage liver disease score ≥15, uncontrolled ascites, or hospitalization for hepatic decompensation. A prespecified propensity score-weighted EC group was derived from a US healthcare claims database.

Results: In the RCT, the primary endpoint occurred in 28.6% of OCA (n = 168) and 28.9% of placebo patients (n = 166; intent-to-treat analysis hazard ratio [HR] = 1.01, 95% confidence interval = 0.68-1.51), but functional unblinding and crossover to commercial therapy occurred, especially in the placebo arm. Correcting for these using inverse probability of censoring weighting and as-treated analyses shifted the HR to favor OCA. In the EC (n = 1,051), the weighted primary endpoint occurred in 10.1% of OCA and 21.5% of non-OCA patients (HR = 0.39; 95% confidence interval = 0.22-0.69; P = 0.001). No new safety signals were identified in the RCT.

Discussion: Functional unblinding and treatment crossover, particularly in the placebo arm, confounded the intent-to-treat estimate of outcomes associated with OCA in the RCT. Comparison with the real-world EC showed that OCA treatment significantly reduced the risk of negative clinical outcomes. These analyses demonstrate the value of EC data in confirmatory trials and suggest that treatment with OCA improves clinical outcomes in patients with PBC.

Trial registration: ClinicalTrials.gov NCT02308111.

Graphical abstract

Figure 1.

Kaplan–Meier survival curves for the COBALT primary composite endpoint. This figure illustrates the survival probabilities and number of OCA and non-OCA patients with primary biliary cholangitis who are at risk of death, liver transplant, or hepatic decompensation over time. (a) Intent-to-treat analysis of OCA vs placebo in the COBALT RCT. (b) COBALT OCA arm vs external control from the Komodo Healthcare Map database. The endpoint of this analysis did not include model for end-stage liver disease score ≥15 because not all component laboratory data were available in the Komodo database. Censoring is indicated by + symbols. HR, hazard ratio; OCA, obeticholic acid; RCT, randomized controlled trial.

Figure 2.

ALP levels vs time after randomization for placebo and OCA patients who discontinued study visits or started commercial therapy before an endpoint event. Each data point represents a patient's ALP level at the last study visit before an endpoint event. ALP, alkaline phosphatase; OCA, obeticholic acid; ULN, upper limit of normal.

Figure 3.

Evidence of confounding in the ITT analysis of the COBALT RCT. Panels (a) and (b) illustrate the number of patients who discontinued IP before an endpoint event (bars) over time vs the proportion of patients who initiated commercial PBC treatments (curves) over time in the placebo arm (a) and OCA treatment arm (b). Commercial PBC treatments included commercial OCA, fibrates, or UDCA (if the patients were not receiving UDCA at baseline). The dotted lines show the time since randomization for 25% of patients to initiate commercial treatment in the placebo and OCA cohorts. Panel (c) shows mean ALP levels before and after initiating commercial treatment by treatment arm. ALP, alkaline phosphatase; IP, investigational product; ITT, intent-to-treat; OCA, obeticholic acid; PBC, primary biliary cholangitis; RCT, randomized controlled trial; UDCA, ursodeoxycholic acid.

Figure 4.

Patients at risk of a primary composite endpoint event during the COBALT RCT. The proportion of placebo patients who had regularly scheduled study visits (solid lines) steadily decreased >12 months after randomization and was less than that in OCA patients during the remainder of the study. Patients who did not have follow-up study visits were followed by telephone calls and medical record review (i.e., proportion = 1 − proportion who had study visits), as described in the main text. The proportion of patients who discontinued IP (dashed lines) and initiated second-line therapy (dotted lines) steadily increased >12 months since randomization and was higher in placebo vs OCA patients during the remainder of the study. IP, investigational product; OCA, obeticholic acid; RCT, randomized controlled trial.

Figure 5.

Hazard analyses of the composite endpoint in COBALT data sets. Forest plot comparing hazard ratios for the composite endpoint among patients treated with OCA vs placebo in the ITT analysis of the RCT, IPCW analyses to adjust for potential sources of bias in the RCT, the as-treated analysis of the RCT, and the real-world EC analysis. 2L, second line; EC, external control; HR, hazard ratio; IP, investigational product; IPCW, inverse probability of censoring weights; ITT, intent-to-treat; OCA, obeticholic acid; RCT, randomized controlled trial; UDCA, ursodeoxycholic acid. ^aDefined as initiation of commercial OCA, UDCA for patients who were not receiving UDCA at baseline, or fibrates for patients who were not administered fibrates at baseline. ^bRatio of the hazard of OCA treatment to the hazard of placebo treatment.