Bias in Extension Studies Despite Their Value to FDA Safety Reviews: JAMA
Long-term clinical trial extension studies can provide important evidence about drug safety and durability of treatment effects, but their findings may be compromised by selection bias, patient attrition and the loss of randomized comparison groups, according to a 10/7 JAMA editorial.
The editorial highlights a challenge for FDA and drug developers: obtaining reliable long-term evidence when pivotal trials are often too short or too small to fully characterize delayed adverse events or sustained treatment benefits.
The gap is well documented. Most pivotal trials are powered for a primary endpoint at a fixed timepoint, not for durability over years, and rare or delayed adverse events in areas such as immunology, oncology and neurology may not appear within the observation window at all, as Clinical Trial Vanguard (CTV) noted in its coverage. ICH E1, in place since 1994, has long expected long-term exposure data for drugs intended for chronic use. Yet a 2021 analysis in Clinical and Translational Science found that only about 1,115 of 216,944 trials registered on ClinicalTrials.gov between 2009 and 2019, under 0.6%, were uncontrolled extension studies.
Extension studies frequently allow patients originally assigned to placebo to receive active treatment, eliminating the randomized comparison needed to establish causal treatment effects. Patients who discontinue treatment because of adverse events or lack of efficacy may also be underrepresented, potentially making outcomes appear more favorable.
Attrition can compound the problem, according to Clinical Trial Vanguard. Research on open-label extensions suggests that about 74% of trial participants enroll in an extension and only about 57% of those complete it. Applied to a hypothetical 500-patient pivotal trial, that would mean roughly 370 patients entering the extension and about 211 finishing, a shrinking and potentially self-selected group on which long-term conclusions would rest.
The JAMA authors discuss the potential use of external control groups and real-world data to address these limitations, while cautioning that differences between populations and methods of measuring outcomes can introduce additional bias.
Clinical Trial Vanguard notes that the cost of late long-term data is visible in checkpoint inhibitors. A FAERS analysis covering 2011 to 2023 identified 3,738 delayed immune-related adverse event cases among 147,854 reviewed, signals that emerged years after approval through passive surveillance rather than a designed extension study.
The concerns are particularly relevant to cell and gene therapies, for which FDA has emphasized the importance of collecting long-term safety and efficacy information after approval. FDA’s long-term follow-up guidance for human gene therapy products already calls for observation that extends beyond the scheduled assessments of pivotal studies, in some cases for many years.
In its 9/2025 draft guidance, Postapproval Methods to Capture Safety and Efficacy Data for Cell and Gene Therapy Products, FDA discusses approaches to postapproval evidence collection when preapproval studies cannot fully characterize a product’s long-term effects.
Industry observers argue the practical lesson is to plan extensions early. Clinical Trial Vanguard contends that sponsors should pre-specify the estimand, budget for retention and select sites capable of multi-year follow-up before the pivotal trial enrolls, rather than treating the extension as an afterthought.
Although extension studies can help address these evidence gaps, the JAMA authors emphasize that their results must be interpreted in light of design limitations, incomplete follow-up and potential confounding.