Re‐adjudication of the Trial Evaluating Cardiovascular Outcomes with Sitagliptin (TECOS) with study‐level meta‐analysis of hospitalization for heart failure from cardiovascular outcomes trials with dipeptidyl peptidase‐4 (DPP‐4) inhibitors

Abstract Background Trial Evaluating Cardiovascular Outcomes with Sitagliptin (TECOS) assessed the cardiovascular (CV) safety of sitagliptin versus placebo on CV outcomes in patients with type 2 diabetes and CV disease and found sitagliptin noninferior to placebo. Subsequently, based on feedback from FDA, the Sponsor of the trial, Merck & Co., Inc., engaged a separate academic research organization, the TIMI Study Group, to re‐adjudicate a prespecified set of originally adjudicated events. Methods TIMI adjudicated in a blinded fashion all potential hospitalization for heart failure (HHF) events, all potential MACE+ events previously adjudicated as not an endpoint event, and a random subset (~10%) of MACE+ events previously adjudicated as an endpoint event. An updated study‐level meta‐analysis of four randomized, placebo‐controlled, CV outcomes trials with dipeptidyl peptidase 4 (DPP‐4) inhibitors was then performed. Results After re‐adjudication of potential HHF events in the intent‐to‐treat population, there were 224 patients with a confirmed event in the sitagliptin arm (1.05/100 person‐years) and 239 patients in the placebo arm (1.13/100 person‐years), corresponding to a hazard ratio (HR) of 0.94 (95% confidence interval [95% CI]: 0.78–1.13, p = .49). Concordance between the outcome of the original adjudication and the re‐adjudication for HHF events was 82.7%. The meta‐analysis of CV outcomes trials with DPP‐4 inhibitors with placebo and involving 43 522 patients yielded an HR of 1.07 (95% CI: 0.83–1.39), with moderate heterogeneity (p = .45, I 2 = 62.07%). Conclusion The results of this independent re‐adjudication process and analyses of CV outcomes from TECOS were consistent with the original adjudication results and overall study findings. An updated study‐level meta‐analysis showed no overall significant risk for HHF with DPP‐4 inhibitors, but with statistical heterogeneity.


| BACKGROUND
Sitagliptin, a dipeptidyl peptidase-4 (DPP-4) inhibitor, inhibits the metabolism and inactivation of the incretin hormones glucagon-like peptide-1 (GLP-1) and glucose-dependent insulinotropic polypeptide (GIP), and is approved as an adjunct to diet and exercise to improve glycemic control in adults with type 2 diabetes mellitus. The Trial Evaluating Cardiovascular Outcomes with Sitagliptin (TECOS) was a postapproval, randomized, placebo-controlled, double-blind, event-driven, multinational clinical trial conducted to assess the safety and efficacy of sitagliptin on cardiovascular (CV) outcomes. 1,2 The primary outcome, Clinical Research Institute (DCRI), whose members were blinded to treatment assignment, adjudicated all potential CV events including death, myocardial infarction, stroke, hospitalization for unstable angina, and HHF before database lock.
Due to the association between some DPP-4 inhibitors and heart failure (HF), [4][5][6] all members in this drug class marketed in the United States currently carry warning language regarding HF. Supplemental New Drug Applications (sNDAs) for sitagliptin products were submitted to the US FDA to support addition of the results of TECOS to the products' prescribing information. Based on feedback from the FDA, the trial's Sponsor, Merck & Co., Inc., Kenilworth, NJ, USA, engaged a separate academic research organization also highly experienced in CV event adjudication, the TIMI Study Group (TIMI), to re-adjudicate all potential HHF events, all potential MACE+ events previously adjudicated as not an endpoint event, and a random (approximately 10%) sample of MACE + events previously adjudicated as an endpoint event.
In this paper, we describe the results of re-adjudication on the CV outcomes of TECOS, as requested by FDA, and provide a studylevel meta-analysis of four published, placebo-controlled, CV outcomes trials with DPP-4 inhibitors, updated with the readjudication results from TECOS.

| METHODS
The methods and results of TECOS have been reported previously. 1,2 For the re-adjudication project, events for the main and supplementary analyses were identified via three sources (Figure 1). For the main analyses, two sources were used. The first source was a Sponsorgenerated listing (Master Event List [MEL]) of events previously identified by the original DCRI triggering program for potential endpoint events. Of these events, 100% of previously adjudicated HHF events, 100% of previously negatively adjudicated (i.e., not confirmed as an endpoint event) MACE+ events, and approximately 10% of randomly selected, previously positively adjudicated (i.e., confirmed as an endpoint event) MACE+ events were re-adjudicated by the TIMI Clinical Events Committee (CEC). The second source was via identification of previously unreported potential endpoint events by TIMI CEC Adjudicators or Medical Reviewers during their review of a submitted event. TIMI was able to query sites for additional supporting source documents. To support the main analyses, sensitivity analyses used a third source of previously unreported potential endpoint events that were submitted to TIMI as part of a Sponsor re-monitoring program of select study sites that participated in TECOS. All events adjudicated by TIMI during this readjudication project are referred to as "TIMI-adjudicated" events. Both the original TECOS study and the re-adjudication process were approved by the relevant ethics committees. All enrolled subjects in the TECOS study provided written informed consent.
The re-adjudication project and database review of events conducted by TIMI were performed in a blinded manner without knowledge of treatment assignment or previous adjudication outcome of an event.

TIMI adjudicated events based on the Clinical Events Classification
Committee Charter for TECOS dated January 23, 2014, 1 with the following exceptions: (1) suspected event-used in instances where the available documentation supported a diagnosis of MI, stroke, or HHF, but was insufficient to meet the charter definition; and (2) urgent HF visit-an HF event that meets all of the following: the patient has an urgent, unscheduled office/practice or emergency department visit for a primary diagnosis of HF, but not meeting the criteria for an HHF event, the patient exhibits new or worsening symptoms of HF on presentation, the patient has objective evidence of new or worsening HF, the patient receives initiation or intensification of treatment specifically for HF, with the exception that changes to oral diuretic therapy do not qualify as initiation or intensification of treatment. 7 The TECOS trial utilized a combination of a manual and algorithm-driven adjudication process for HF events which are summarized in Supporting Information: Table S1. For the TIMI adjudications, each event is assigned to two CEC Adjudicators who independently review the event and enter his or her adjudication. If Concordance, defined as the agreement between TIMI and DCRI (i.e., the proportion of the total number of events in agreement with respect to adjudication outcome among the total number of events adjudicated by both TIMI and DCRI), was calculated for HHF for all cases that were adjudicated by both TIMI and DCRI, as this is the only endpoint where every potential DCRI case was re-adjudicated by TIMI. We also report the percent of events downgraded and upgraded, with DCRI adjudication being the base case.
Similar to the original TECOS Statistical Analysis Plan (SAP), 1 the TIMI re-adjudication project prespecified that the following hypotheses be tested in a sequential manner that began with testing for noninferiority for the primary composite CV outcome (MACE+) and key secondary composite CV outcome (MACE) in the PP population, with supporting analyses performed in the ITT population. Other secondary outcomes including HHF were evaluated for superiority in the ITT population. A total event analysis to account for multiple HHF events was performed under a counting process assumption based on the method described by Anderson-Gill with robust variance estimator.
A study-level meta-analyses of four placebo-controlled, CV outcomes trials of DPP-4 inhibitors using random-effects models on trial-level data were performed for HHF and the composite of HHF or CV death from the following trials: SAVOR-TIMI 53, 4,5 EXAMINE, 6,8 CARMELINA, 9,10 and TECOS using updated results based on re-adjudication. Supporting Information: Table S2 summarizes the similar HHF definitions across studies. The HRs from each trial were combined by random effects model with a restricted maximum likelihood approach and Hartung-Knapp adjustment to yield a point estimate of the pooled effect and 95% CI. The pooled effect describes a weighted average of the DPP-4 inhibitor treatment effect on the outcomes of interest listed above, giving more weight to a trial with smaller variance as well as accounting for in-between trial variability. Statistical heterogeneity across the two trials was assessed by Cochran's Q statistic and Inconsistency index (I 2 ) 11 . The meta-analyses were conducted using R version 3.6.1 (R Core Team) and the R package metaphor. 12 All analyses were performed by TIMI who had access to the raw data files from the original TECOS trial as well as the updated re-adjudication results.

| DISCUSSION
The results of this independent re-adjudication process and analyses of TECOS were entirely consistent with the original adjudication results and study findings presented by the TECOS investigators in 2015. Specifically, the HRs and confidence intervals were nearly identical for the main analyses, supporting the noninferiority of sitagliptin compared to placebo for MACE+ and MACE. In addition, after systematically reviewing all potential HF events identified during the trial and in the re-adjudication process, there was no difference between sitagliptin and placebo with respect to risk of first episode of HHF, the total number of HHF events, or the composite of CV death or HHF.
Using the same endpoint definition for HHF, the degree of concordance between the outcomes of the TIMI and the previous adjudications was high at 82.7%. We believe the small amount of discordance is to be expected considering these events were adjudicated more than 5 years apart, and in many cases with the benefit of additional clinical data being available to the TIMI CEC. Any Analyses of HHF were adjusted for a history of heart failure at baseline. discordance between the adjudication processes in determining which cases did or did not meet endpoint definitions, or through the identification of previously unreported events during this process, would not be expected to introduce any bias into the treatment effect of sitagliptin versus placebo as both adjudication processes were blinded to treatment assignment. TIMI, which did not participate in the original adjudication process, did not receive any unblinded data or treatment code until all adjudications were F I G U R E 3 Study-level meta-analysis of the risk for HHF in patients with (left) and without (right) a history of heart failure before hospitalization from the published outcomes trials of dipeptidyl peptidase-4 (DPP-4) inhibitors including the updated TECOS results from re-adjudication. HHF, hospitalization for heart failure; TECOS, Trial Evaluating Cardiovascular Outcomes with Sitagliptin.
the re-adjudication of the Rosiglitazone evaluated for CV outcomes in oral agent combination therapy for type 2 diabetes (RECORD) trial, re-adjudication led to a disagreement in 21% of cases. 13 There is no method to determine which adjudication in the discordant cases is  19 These findings from uncontrolled observational data run counter to the observations in carefully conducted randomized, placebo-controlled trials, raising the concern that the residual confounding or bias in the observational studies related prescribing patterns or patient risk could not be completely addressed, even with rigorous statistical methodology.
In summary, after a prospective, comprehensive, and systematic re-adjudication of all HHF events in TECOS, we confirmed that there was no difference between sitagliptin and placebo with respect to the risk of HHF.

DATA AVAILABILITY STATEMENT
The trial database cannot be shared, but parties interested in collaborating should contact the corresponding author.