Typically, an error that affects so much of a paper would undermine the results (and possibly lead to a retraction). But in this case, the revised dose calculations actually strengthened the findings, according to the first author.
The NEJM study aimed to clarify whether patients with a neuromuscular disease called myasthenia gravis benefit from a surgical procedure to remove the thymus. About half of the patients received surgery plus the steroid prednisone, while the rest only received the steroid. The researchers found patients who received the surgery fared better.
Shortly after the paper was published in August 2016, the authors discovered an error in the calculation of the average prednisone dose. According to Gil Wolfe, the first author of the paper, when the researchers corrected the error:
The recalculation didn’t reverse any of the findings; it actually strengthened them, making thymectomy look even better.
In other words, people who had surgery needed less prednisone overall, and the authors found the doses went down faster in the surgery group than they initially thought.
Here’s the correction notice for “Randomized Trial of Thymectomy in Myasthenia Gravis,” which explains the calculation error:
Randomized Trial of Thymectomy in Myasthenia Gravis (Original Article, N Engl J Med 2016;375:511-522). An error in calculating the time-weighted average of the doses of prednisone led to incorrect point estimates for the primary and secondary outcomes. For patients whose prednisone dose was tapered and then stopped before month 36 and who remained off prednisone, the time-weighted dose calculations did not account for the months during which the patients were not receiving the medication. As a result, the time-weighted prednisone doses for such patients were overestimated. The errors affected several parts of the article: the Abstract, the Results section, the Discussion, Figure 1B, and Table 2. The Supplementary Appendix was also affected. The direction of change of the primary and secondary outcomes and the overall conclusions of the trial are not altered from the original published version. The article is correct at NEJM.org.
Wolfe, professor and chair of the Department of Neurology at the University at Buffalo School of Medicine and Biomedical Sciences, the State University of New York, told us:
We wanted prednisone data for the full 36 months patients were followed. But for patients who went down to 0 mg of the drug during the study, the time-weighted dose calculation stopped before 36 months.
Wolfe explained that when the authors recalculated the drug dose over the three year study period, the revised numbers made the surgery option look even better than the original numbers had.
Wolfe said that the recalculation only affected the prednisone dose data, not the data on disease severity, patient quality of life, or complications from surgery or prednisone.
A spokesperson for NEJM told us the journal never considered a retraction:
In the process of looking at the computer program that did the analysis for this trial, the study statisticians found that that the amount of prednisone was miscounted. They reviewed the published article and found that they needed to make changes in dose reported for each group.
Like Retraction Watch? Consider making a tax-deductible contribution to support our growth. You can also follow us on Twitter, like us on Facebook, add us to your RSS reader, sign up on our homepage for an email every time there’s a new post, or subscribe to our new daily digest. Click here to review our Comments Policy. For a sneak peek at what we’re working on, click here.