Significant Increases in Bleeding
On the safety side, rivaroxaban significantly increased the risk of all bleeding events, compared with placebo.
Across all strata, the 2.5-mg dose significantly increased the risk of the following TIMI classifications of bleeding: major, major or minor, life-threatening, intracranial hemorrhage, minor, clinically significant, and warranting medical attention. In stratum 2, the 2.5-mg dose significantly increased all TIMI major or minor bleeding, major bleeding, life-threatening bleeding, clinically significant bleeding, and medical attention bleeding. The 2.5-mg dose did not significantly increase the risk of TIMI major fatal bleeding or fatal intracranial hemorrhage.
In stratum 2, the 2.5-mg dose reduced the rate of the primary end point by 15% while increasing the rate of bleeding not related to coronary artery bypass graft surgery, Dr. Hicks said. "There was a twofold increase in TIMI major fatal bleeding and threefold increases in TIMI major bleeding, intracranial hemorrhage, hemorrhagic stroke, TIMI life-threatening bleeding, and TIMI major or minor bleeding. There was also an 18% increase in the risk of fatal stroke. Although the hazard ratios were increased, the absolute incidence of these events was low."
Subjects aged 75 years and older, and individuals weighing less than 60 kg, had an increased risk of bleeding events.
ACS subjects appeared to have a more pronounced tendency for liver injury than did populations studied in the atrial fibrillation and deep vein thrombosis prophylaxis studies. "Therefore, rivaroxaban, even in lower doses than what is recommended for other uses, appears possibly to cause mild liver injury in some patients. This finding likely reflects some increased susceptibility to drug-induced liver injury in patients with ACS," Dr. Hicks said.
Data Quality as a Counterargument to Approval
Dr. Hicks said that in arriving at her approval recommendation, she considered the argument that the amount of missing data from the ATLAS study weighed against approval.
Across all strata, 2,402 subjects (15.5%) discontinued the study prematurely, including 1,294 subjects (8%) who withdrew consent. "There were [more than] 1,000 subjects at the end of the trial with unknown vital status," Dr. Hicks said. Additionally, there were incomplete follow-up and uncounted deaths. "The quantity of missing data in ATLAS could affect the overall interpretability of the trial."
Nevertheless, Dr. Hicks suggested she was satisfied with the sponsors’ attempt to obtain vital status information on missing subjects and its responses to the FDA’s information requests on the issue. She also pointed to the broader problem of missing data in the setting of CV outcomes studies.
Shades of Ticagrelor
However, Dr. Thomas Marciniak, the review division’s medical team leader on the rivaroxaban application, takes a tougher stand on the issue of missing data and its ramifications for approval. In an April 26 memo, Dr. Marciniak suggests that the data quality may not support the favorable statistical results reported with the 2.5-mg dose.
The percentage of patients with incomplete follow-up in the study was high, averaging about 12%, he noted. "This percentage is far higher than the differences between the placebo and rivaroxaban arms in end point rates, which typically are about 1%-1.5%. The difference between placebo and rivaroxaban for bad follow-up rates matches these differences in end point rates. By [Johnson & Johnson’s] patient status statistics, there appears to be plenty of opportunity for incomplete data to obscure or magnify any differences in end points."
Furthermore, the rates of patients with unknown vital status exceed the reported differences in mortality rates, Dr. Marciniak said, asserting, "We cannot have confidence that the claimed mortality benefits are real."
He also challenged the sponsors’ approach to censoring patients and their classification of deaths that occurred after withdrawal of consent, saying this may have led to an overstatement of rivaroxaban’s efficacy.
"Because of the extent of missing follow-up in ATLAS, we cannot have confidence in either the calculated mortality or CV endpoint benefits," he said.
Marciniak’s criticisms about ATLAS are reminiscent of his attacks on the pivotal efficacy study for AstraZeneca PLC’s ticagrelor (Brilinta). During a July 2010 advisory committee review of the antiplatelet agent, Dr. Marciniak criticized the sponsor for inadequate end point and vital status follow-up in the 18,624-patient PLATO trial.
Dr. Marciniak’s criticisms were even more strenuous in his written reviews, in which he cited inadequate data collection and analyses, poor follow-up, faulty adverse event reporting, failure to adjudicate potential end points, an inappropriate primary end point, and censoring problems. The review division approved ticagrelor in July 2011 despite Dr. Marciniak’s recommendation against approval.
This coverage is provided courtesy of "The Pink Sheet." This news organization and "The Pink Sheet" are owned by Elsevier.