Of the four novel oral anticoagulants (NOACs) billed as alternatives to warfarin, Eliquis (apixaban) is the only one that can claim it reduces deaths in people with atrial fibrillation.
But is that claim real?
An FDA reviewer doesn't think so.
In the clinical trial of Eliquis (apixaban), patients getting the drug were 11% less likely to die from any cause than those getting warfarin, which for decades had been the only anticoagulant available to prevent strokes in people with atrial fibrillation.
But because vital data -- primarily involving deaths -- was missing from the trial, the FDA reviewer concluded that the benefit very well could be pushed over the threshold of statistical significance, according to a memo he wrote.
Although statistical significance standards can vary, most researchers -- and the FDA -- agree that a P-value of less than 0.05 to indicates a finding that is not the play of chance. A P-value equal to 0.05 indicates a trend toward significance.
The P-value for Eliquis and its death benefit stood at .0465 -- just enough to be deemed significant -- when reviewer Thomas Marciniak, MD, wrote his critical memo .
He said if there was one more death among Eliquis patients or one less warfarin death, then the Eliquis mortality benefit would become insignificant because it would climb above the .05 threshold.
Marciniak noted there were more than 300 Eliquis patients with missing data.
He said the missing data and other data quality issues also reduced confidence in the drug's main benefit -- preventing strokes -- though he said those results were pointing "reasonably" in the right direction.
"I, like most FDA reviewers, would like to conclude that apixaban (Eliquis) is effective in atrial fibrillation -- we would like to have alternatives to warfarin," .
But the data problems "destroy our confidence" that the drug reduces deaths, he concluded.
Marciniak's review also took issue with other drugs approved by the FDA.
He said the Eliquis trial was just one of many other trials -- he did not name the others -- that had substantial problems with data quality.
"Some of the responsibility for the data quality problems rests with us, the FDA: We have approved drugs ignoring similar data quality issues, granting superiority claims and not discussing in the labels the data quality issues," he said. "We must stop doing this."
Marciniak was so concerned about the Eliquis missing data, and how it might wipe out the drug's superiority claim about reducing deaths, he said the data problems should be described at length on the drug's label.
Ultimately, he was overruled by his fellow reviewers at the FDA.
In a response memo, two other FDA officials discounted Marciniak's comments and wrote they were "somewhat unrealistic."
They said the missing data were not "especially large" and it was unclear if the "missingness" favored Eliquis, actually making the death benefit more significant.
"Perhaps some (patient data) were lost to follow-up because they had a stroke or died, which potentially biases the study results," they wrote in response. "However, there is no reason to believe that this was more likely in the apixaban (Eliquis) arm than in the warfarin arm. We cannot directly address whether the missing data are biased in one direction."
In addition, they said it was not appropriate to have to provide a "cushion" of fewer deaths to account for "what ifs" when it comes to establishing statistical significance.
In an email, representatives of Pfizer and Bristol-Meyers Squibb, which co-market the drug, said data compiled in its trial came from more than 1,000 sites in 40 countries.
The data were "thorough and complete, despite inherent challenges with following such a large and diverse population," they said. "As the FDA ultimately decided, the small fraction of patients lost to follow-up did not adversely alter the interpretation of superior results versus warfarin."
However, , an expert in clinical trial design and assistant professor of medicine at Yale University Medical School, said Marciniak makes a valid point.
"I would tend to agree that touting this as a potential (mortality) benefit of the drug is likely very premature," said Wilson who reviewed the FDA documents for this story.
By contrast, the two top-selling anticoagulants failed to show a statistically significant death benefit.
The clinical trial of Pradaxa (dabigatran), showed that deaths were slightly lower in those getting the drug, but the P-value was .051, not enough to be considered statistically significant. Pradaxa's label does not mention any death benefit.
The other top-seller, Xarelto (rivaroxaban), also failed to show a statistically significant mortality benefit, showing a P-value of .15.
The other anticoagulant alternative to warfarin, Savaysa (edoxaban), also did not a have a statistically significant benefit in reducing deaths from any cause.
The P-value for the recommended dose approved in the U.S. for atrial fibrillation was .08.
John Fauber is a reporter with the Milwaukee Journal Sentinel. Coulter Jones is a reporter with 鶹ý. This story was reported as a joint project of the Milwaukee Journal Sentinel and 鶹ý.