CardioExchange Archive

In a recent Viewpoint in JACC two members of the FDA’s Cardiovascular and Renal Drugs Advisory Committee, Mori Krantz and Sanjay Kaul, write about the problem of missing data in the ATLAS ACS 2-TIMI 51 Trial and consider some of the larger implications of missing data in clinical trials. One of the authors, Sanjay Kaul, agreed to answers questions about the topic from CardioExchange’s Harlan Krumholz.

HK: What is the issue with the missing data in ATLAS and why is it important?

SK: The amount of missing data, especially lack of vital status information, is inordinately large in that it challenges the interpretation of the data. Furthermore, there is potential for “informative censoring,” i.e., patients who are dropping out of the study because of a bleeding side effect from rivaroxaban might also be at risk for experiencing the thrombotic events (primary endpoint). This differential dropout together with missing information regarding thrombotic events from these patients might amplify treatment differences in favor of rivaroxaban.

HK: How do you think the missingness problem in ATLAS should affect our view of the drug?

SK: Missingness contributes to lack of confidence in the data being reliable. However, missingness is not the only issue that should affect our view of the drug. The data are not coherent in the sense that the 2 doses of the drug have a different effect on the components of the primary endpoint (2.5 mg dose reducing CV death but not MI; 5 mg dose reducing MI but not CV death). In addition, there is little external support for incremental benefit with the use of novel anticoagulant agents on top of dual antiplatelet therapy. Finally, even if missingness was not a major concern, the results are not ‘statistically persuasive’ to justify an ACS indication on the basis of a single trial (the FDA Guidance requires 2 successful trials or one trial with persuasive evidence plus confirmatory evidence).

HK: Should ATLAS results be used in meta-analyses – or do you have concerns about whether we can rely on them?

SK: Although missingness challenges the interpretation of the ATLAS results, it does not necessarily invalidate them. If the ATLAS data end up being used in a meta-analysis, I would like to see a sensitivity analysis addressing the impact of the ATLAS results on the overall pooled estimates. I would question the strength of evidence if the overall evidence is driven by the ATLAS results.