Each week, we will be publishing layman’s abstracts of new articles from our prestigious portfolio of journals in statistics. The aim is to highlight the latest research to a broader audience in an accessible format.

The article featured today is from Pharmaceutical Statistics, with the full article now available to read in Early View here.

Law, M, Grayling, MJ, Mander, AP. A stochastically curtailed two‐arm randomised phase II trial design for binary outcomes. Pharmaceutical Statistics. 2020; 1– 17. https://doi.org/10.1002/pst.2067

Randomised controlled trials are considered the best way of assessing a new treatment, with one group of participants receiving the treatment of interest and the other receiving either placebo or standard care. However, in the early phases of assessing cancer treatments – where treatment outcome is binary (response or no response) – trials are often single-arm, meaning that all participants receive the treatment of interest. Consequently, the results from the new trial then have to be compared to results from another source. This can make it more difficult to estimate how well the new treatment works, as the two sets of results may come from groups that are different in various ways – for example, from trials conducted many years apart, in different countries, that included or excluded participants for different reasons and so on.

Though a number of reasons exist for choosing a single-arm trial, the primary reason is that single-arm designs require fewer participants than their randomised equivalents – that is, the required sample size is smaller. Therefore, the development of randomised trial designs that reduce sample size is of value to the trials community.

This paper introduces a randomised two-arm binary outcome trial design that greatly reduces expected sample size compared to typical randomised trial designs. Sample size is reduced in two ways: firstly, the design uses stochastic curtailment, which means allowing the possibility of stopping a trial before the final conclusions are known with certainty. Secondly, the proposed design involves the use of a randomised block design, which allows investigators to control the number of interim analyses, the points in the trial at which early stopping may occur. This proposed design is compared with existing designs that also use early stopping, in terms of maximum and expected sample size, through the use of a loss function. Comparisons are also made using an example from a real trial. The comparisons show that in many cases, the proposed design is superior to existing designs. Further, the proposed design may be more practical than existing designs, by allowing a flexible number of interim analyses. One existing design produces superior design realisations when the anticipated response rate is low. However, when using this existing design, the probability of incorrectly concluding that a treatment works can be higher than expected if the response rates are different to what was anticipated. Therefore, when considering randomised designs in cancer trials or for any trial with a binary outcome, we recommend the proposed approach be preferred over other designs that allow early stopping.