Formal Statistical Testing and Inference in Randomized Phase II Trials in Medical Oncology

Abstract

Formal statistical comparisons between arms of randomized phase II trials are frequently undertaken in medical oncology. The extent to which such a practice abrogates phase III testing is unknown.

Meta-Research on Oncology Trials: A Toolkit for Researchers with Limited Resources

“…As an interesting illustration of the appropriateness of limiting or not the number of journals analyzed, we have independently conducted research on the same topic using all the published literature or only selected journals . With regard to the period of interest, if the scope is to assess time trends in the literature, even shorter periods of time—separated by some years—may be analyzed, especially if no constraints are applied to the journals of interest . Limiting the search to English‐language primary sources appears to be a useful and acceptable filter, but one that arguably leads to no material difference, as most published clinical trials appear in English‐language journals.…”

“…Then, investigators describe their findings on the primary and secondary outcomes of interest in their study. For example, in a cross‐sectional survey on the proportion of randomized phase II trials that used inferential statistics to assess differences between trial arms, we observed that either p values or confidence intervals had been used by 89% as a form of statistical comparison; as a secondary outcome in that study, we tested predictive features of phase II trials that could be associated with use of any statistical comparison .…”