PLATO DEATHS LISTS
Lately our Task Force gained access to the detailed dataset of 938 PLATO deaths reported to the FDA. We matched those records with local patient-level data from sites controlled by the sponsor, and found that the existence, precise dates and proper causes of some deaths in PLATO were inaccurately reported. 7 Several clopidogrel deaths were reported earlier than actual, while their causes were switched from “non-vascular” or “unknown” to “vascular”. In contrast, some ticagrelor deaths were reported later while other vascular deaths were incorrectly entered in the FDA list as “non-vascular” or “unknown”.7 We explore here any differences between the FDA-issued death list (n=938) and the shorter PLATO Investigators dataset (n=905). The overall numbers of fatalities and some primary causes for both lists are presented in the Table.
With regard to the discrepancy in total PLATO deaths, it is our understanding that the dataset we have in possession includes deaths after the analysis period that are not counted in the primary efficacy analysis, but the absolute majority of deaths (over 900) should match precisely (Figure). Indeed, we identified several events when the patient never received a single drug dose, serious enrollment mistakes, and inclusion/exclusion errors, but the most common restriction was the length of follow-up. In fact, the PLATO death count was restricted to no more than 12 months follow-up duration. Any death within the analysis period should be counted, although PLATO also excluded deaths after “withdrawal of consent” or “volunteer discontinuations mostly for ticagrelor patients.3,8 Importantly, there were some PLATO patients who were still on drug and died 12-16 months after enrollment while the trial was still active in the second half of 2008, but these patients were excluded as being beyond the analysis timeframe. Finally, we are skeptical that the reductions in the FDA-issued list removed more ticagrelor (n=21) than clopidogrel (n=12) events for the final count of 905 deaths used in all PLATO publications and affiliated presentations.
Furthermore, we presumed that the precise death characteristics and single primary causes of all remaining PLATO deceased patients should be identical between the FDA and trial investigators. All 905 deceased patients in the “official” trial dataset should be included in the broader list submitted to the FDA. Since some of the 938 deaths were reported inaccurately to the FDA, we started looking at which of the erroneous records were transferred into the smaller revised PLATO pool used in the original1 and over 90 peer-reviewed secondary publications. Examining details of 2 high-quality secondary PLATO publications 9,10 focused on mortality revealed several fundamental problems outlined in the Table.
The well-structured and detailed data from the PLATO Investigators, clearly indicate that Tables 5 (vascular) and 6 (non-vascular) reported primary causes of deaths.9 The FDA list also identify a single primary cause of death, perhaps “unknown” or “other” but just one cause. However, looking at the PLATO Investigators’ statements reveals obvious discrepancies. With regard to reported numbers of vascular deaths the significant differences were observed between FDA- and trial investigators lists. There were more sudden deaths in the shorter list than in the FDA dataset (161 vs.138; p<0.03), post-AMI (373 vs.178; p<0.001) but fewer heart failure deaths (73 vs.109; p=0.02). Stroke numbers match well (39 vs. 37; p=NS) with only 2 ticagrelor cases removed. Among non-vascular causes of deaths cancer matched well (32 vs.31; p=NS), and sepsis cases were identical (30 vs. 30; P=NS). However, suicides deserve special attention. Two extra cases9, in the shorter list are impossible to comprehend. Not only suicide (PLATO code 12-5) is a valid primary death cause, but both “missing” patients are from the clopidogrel arm. We directly approached PLATO Investigators with the detailed specific inquiry regarding these 2 suicides no reported to the FDA but got no response or any explanation regarding the 2 extra deaths in the Investigators’ list. Not solving such a simple issue raises questions regarding not only vascular deaths prevention but also the claimed all-cause mortality benefit of ticagrelor.