A Statistician’s Reflection: FDA Guidance on Limiting Crossover in Oncology Studies 

Authored by

In August 2025, the FDA released draft guidance on Approaches to the Assessment of Overall Survival in Oncology Clinical Trials. Within the guidance they describe issues regarding the use and interpretation of overall survival data, particularly when the primary endpoint is progression free survival and overall survival (OS) data is immature at the time of assessment. In many Oncology studies, an important aspect of the protocol is that participants can cross-over from the control group to the experimental treatment at the point of progression or treatment failure. This design feature can make a trial more desirable to participants, improving recruitment and retention. However, it is also recognised that crossover can impact the interpretation of results because some participants receive both treatments which makes it difficult to conduct a fair comparison of the effects of one treatment versus another. This issue becomes even more challenging when a high proportion of participants cross over.  

In its draft guidance, the FDA have attempted to avoid the ambiguous results that can be caused by crossover in studies by stating that that use of crossover should be limited but with the caveat that it would be appropriate to allow in indications that have few treatment options available. They stopped short of specifying what the limits should be or how they would be applied but it seems likely that crossover is simply not allowed in the study, save for some exceptional cases. The idea of introducing a pre-defined cap on the level of crossover within a study would seem to be difficult to implement in practise. Nevertheless, this lack of clarity will create uncertainty that will need to be managed by all stakeholders.  

Impact on Recruitment 

Following the guidance, there is also concern that these limits may impact recruitment into oncology trials.  In an interview with Fierce, Jake Van Naarden, President of Eli Lilly’s Oncology division, highlighted the tension between these policies, noting that ‘we need to pick a lane’, since the policy goals may pull in opposite directions.  It is a fair point. These goals do seem to be contradictory and are likely to lead to uncertainty in the near-term. However, in science there is often another perspective and alternative way of looking at the same problem. Could it be possible for crossover, participation and statistical clarity to co-exist? Fundamentally, this is a statistical problem so any solution should consider the development and validation of statistical methods that allow for a fair comparison between the treatment groups and an unbiased assessment of the effects of treatment whilst adjusting for the confounding effects of crossover. 

Statistical Methods to Overcome Common Problems in Oncology Trials 

Over the past 24 months, statisticians at MMS have been involved in two important and relevant methodological development projects to evaluate statistical methods that can be used to address common problems in Oncology trials.  The first project involves a case study linked to the submission of Phase 3 data to the FDA. In this case, the amount of cross-over from the control group to the experimental treatment group was much higher than anticipated. This led to substantial difficulty in interpreting the overall survival data. We applied a 2-stage survival model which was able to adjust for the confounding effect of cross-over. Whilst the model worked well in terms of enabling a fair comparison between treatment groups and bringing clarity to the interpretation of overall survival, its use was not pre-defined in the protocol or statistical analysis plan. In addition, more research is needed to validate the model by assessing its performance in a range of different scenarios.  

In a second project, MMS statisticians are collaborating with Friends of Cancer Research (FOCR) to address a longstanding challenge in cancer drug development: how to assess long-term treatment benefit when overall survival data is immature. In many oncology studies, endpoints such as progression-free survival (PFS) and objective response rate (ORR) are used to support accelerated decision-making or approval. However, reviewing OS data at the same interim point risks distorting the assessment of long-term benefit or harm. MMS and FOCR are collaborating as part of a consortium of stakeholders to develop and evaluate the risks in different scenarios, including crossover effects. This project will provide a clear view of the potential risks and benefits associated with acting on immature data in oncology clinical trials and supporting more informed decisions earlier in the trial process. In addition, the project will enable the validation of new methods that overcome some of the challenges in Oncology studies, including the cross-over issue. Furthermore, the project will provide guiding principles that will allow key criteria relating to the assessment of treatment effects and interpretation of risks to be pre-defined in study documentation.   

Developing Methodological Solutions 

The draft guidance from the FDA attempts to deal with some real concerns around the challenges of interpreting survival data but in the case of limitation on crossover, the guidance may appear to be a direct contradiction to other initiatives to improve study participation. As with many challenges in our industry, there is often a technological or methodological solution which could take some time to develop. During that time, there will be calls to soften or change aspects the guidance because of the unintended consequences. The strategic statistical services group at MMS believes the issue is more a statistical problem rather than an intrinsic issue with the guidance itself. We have implemented a methodological solution that places statistics and interpretation at the heart of the problem; one that allows crossover, participation and statistical clarity to co-exist. Ultimately, this would be the ideal outcome for all stakeholders as the issue of a conflicting dual policy disappears. 

These dynamics highlight the increasing complexity of oncology development today. You can explore this further in our on-demand webinar, Making the Right Decisions with Incomplete Evidence in Early-Phase Oncology Trials, which focuses on how sponsors can approach early-phase decision-making with greater clarity.