Primary Submission Category: Applications in Health and Biology
Small sample size but still definitive: secondary analysis of a randomized trial on central neck dissection for papillary thyroid cancer
Authors: Benjamin Cher, Qinyun Lin, Rebecca Sippel, Kenneth Frank, Courtney Balentine,
Presenting Author: Benjamin Cher*
A recent randomized controlled trial (RCT) sought to address a major controversy in endocrine surgery: whether low-risk thyroid cancer should be treated by removing only the thyroid or whether it is also necessary to routinely resect lymph nodes. The original trial found no difference in recurrence after randomizing 30 patients to standard thyroid removal (control group) and 31 to removal of both thyroid and lymph nodes (treatment group). However, the study has not noticeably changed clinical practice because of concerns about power and external validity. In this study, we use novel frequentist and Bayesian approaches to address these concerns, and we conclude that the results are robust despite small sample size and recruitment being limited to a single academic center. Using the Robustness of Inference to Replacement (RIR), we find that a large percentage of both the treatment and control groups would have to be replaced with patients who experienced similar probability of recurrence as the other group (RIR=28, requiring double switches). Additionally, in >99% of 1,000 bootstrap repetitions of the trial, we find no difference in recurrence. Finally, using Bayesian statistics, we find the study results would only change if the true effect of lymph node removal were a 100% reduction in recurrence relative to the standard surgery. We also estimate that repeating the trial would yield the same result ~70% of the time.