The basket trial design does, however, have limitations. According to Dr Gonen, “The primary limitations have to do with the complexity of addressing multiple questions in 1 protocol. Aggregation and borrowing strength, the primary statistical strengths of this design, require complicated designs, which can be less accessible than traditional alternatives.”

A major limitation is the increased risk for false positives—when an ineffective drug is deemed efficacious. “We believe that increased risk of false positives is currently underappreciated as a threat to the validity of these designs,” Dr Gonen told Cancer Therapy Advisor.


Continue Reading

He noted that in a traditional trial, the concept of a false positive is more black-and-white. In a basket trial, however, there are often multiple questions being asked, and efficacy and/or safety may be different in the various baskets of the trial. As a result, the false positive rate is relatively cumulative, and with a rate of about 5% per basket, a 5-basket trial would have a false positive rate of about 23% and a 10-basket trial a rate of 40%. In contrast, the false positive rate of a traditional trial design is about 7%.

Adjusting the sample size and decision rules can help control the false positive rate, but transparency of the overall false positive rate is imperative. According to Dr Gonen, “it is essential that statisticians are included in basket trial teams so that they can be designed in the most efficient way possible.”

Despite these limitations, Dr Gonen stated that, “we think basket trials will grow into the primary method of clinical trial design for addressing precision medicine questions.”

References

  1. Targeted cancer therapies. National Cancer Institute website.  https://www.cancer.gov/about-cancer/treatment/types/targeted-therapies/targeted-therapies-fact-sheet. Updated April 25, 2014. Accessed December 13, 2016.
  2. Cunanan KM, Gonen M, Shen R, et al. Basket trials in oncology: a trade-off between complexity and efficiency. J Clin Oncol. 2016 Nov 28 doi: 10.1200/JCO.2016.69.9751 [Epub ahead of print]