Doubling down on a single study approach to FDA approval
The FDA has long advocated the ‘two study rule’ as the default for marketing approval. In this model, sponsors have typically been required to provide evidence from two randomised clinical trials to support marketing approval.
As of February 2026, the FDA has shifted its thinking and publicly adopted the position that one adequate and well controlled study plus confirmatory evidence is the new default evidentiary model for approvali. Though they reserve discretion to require more where scientifically necessary.
This change in approach has created much discussion amongst the industry, raising questions as to whether conducting just one confirmatory study is a risky position for sponsors to take. If done well, however, the opposite is true. A single confirmatory study could be a strategic advantage.
Building confidence in a single-study approach
Consider starting from a different position that assumes a single confirmatory study is the norm. Sponsors and regulators both require strong evidence that a treatment is safe and effective. Instead of robust evidence being generated via repetition, it’s built by setting a higher scientific bar. For example, a single confirmatory study is designed with a type I error rate of 1% (more stringent than the 5% significance level most commonly used) and with 90% power to detect a pre-specified treatment difference. (Power being the ability to detect a difference at the desired significance level [1% in this case] if the pre-specified treatment difference truly exists.) In other words, 9 times out of 10, the study will be successful if the treatment is as efficacious as expected.
The power trade-off
Now, compare this single-study approach to a ‘new’ suggestion of running two studies, in parallel, using the same protocol but with different investigator sites (not an uncommon outcome in recent years following regulatory interaction). The two-study approach is then justified on the basis that replication is important. Sponsors then adopt a less stringent type I error of 5% for each study and power is set to 80%, which is currently de facto in drug development.
How might this second scenario play out?
The probability of both studies meeting the threshold for significance at the 5% level (if the treatment difference is as expected) is 0.8 x 0.8 = 64%. In effect, the sponsor has now reduced the 90% chance of success with the one study approach to a 64% chance of success. The one study approach has a 1.4 times greater chance of success even though the type I error is more stringent.
Comparing single-study and two-study designs
Sample size: Sample size calculations include a function of both the type I error and (1 - power), that is then squared. Once accounted for, the sample size for this large single study will be approximately 5% less than the combined sample size for these two studies.
Delivery speed: Not only is chance of success 1.4 times higher with one study, but 5% fewer patients are required, meaning the study can be completed quicker. Study size isn’t the only timeline factor. When two studies are run in parallel they seldom finish at the same time, and importantly, the critical path to regulatory submission is dependent on the last study to finish.
Data and variation: With one study there is one database, one analysis, one clinical study report – all leading to less documentation. In the two-study model, one study is essentially being split into two (equivalent to randomly splitting the sites across two studies). As such, due to random variation there will be different treatment differences on primary and secondary estimands when comparing the two studies that require explanation. The risk here is not dissimilar to the differences found in the famous ISIS-2 trial that found treatment differences varied across the star (zodiac) sign subgroupsii, illustrating the role of chance significant in subgroup analyses. If the two studies use an identical protocol, what is the scientific explanation for any differences observed between the two, other than random variation? It is not uncommon to find Sponsor scientists spending hours poring over efficacy tables trying to determine a scientific explanation for such observed differences between studies with identical protocols.
Replication vs duplication
Running studies in parallel with identical protocols is not really replication, it is mirroring. Even the retired FDA Deputy Director Robert Temple acknowledged that the intention of a second study was not exact replication of the first. There is a separate argument for enriching the corpus of evidence through the variety of data generated, which indeed has some merit. Varied evidence leads to more robust conclusions and that can be achieved through conducting different types of study. But in practice, replication has turned into duplication. Within one larger study it is often possible to explore treatment differences that may differ between different regions and countries depending on the country mix – more efficiently so than within two separate studies.
Redefining risk and success in a single-study approach
The ‘two study rule’ was never intended to lead to mirroring that has become more the norm in recent years. This shift has perhaps resulted from an over-interpretation of the original FDA thinking (including by the regulators themselves) around the need for “substantial evidence” originating from “adequate and well-controlled investigations” per the Kefauver-Harris Drug Amendments in 1962 to the 1938 Food, Drug and Cosmetic Actiii.
Mirroring may also reflect increased globalisation. As processes, procedures and practices have become more standardised and technology has advanced, it is now much easier to run identical global studies across countries and regions. However just because sponsors can duplicate trials doesn’t mean that sponsors should. Drug development is a risky business. Running two studies in parallel that aren’t designed to increase the variety of evidence isn’t a good way of managing risk. Neither is it cost-effective or quick.
Doubling down on a single confirmatory study designed for robust evidence generation with stricter significance levels and higher power can optimise evidence generation whilst increasing cost effectiveness and delivery speed.
Connect with us to explore how to create a coherent total evidence package using a single-study approach.
Author
Andrew Garrett PhD CStat
Executive Vice President, Scientific Operations
Past President, Royal Statistical Society
References
i Prasad V, Makary MA. One pivotal trial, the new default option for FDA approval—ending the two trial dogma. N Engl J Med. 2026;394(8):815 817. doi:10.1056/NEJMsb2517623.
ii ISIS 2 (Second International Study of Infarct Survival) Collaborative Group. Randomised trial of intravenous streptokinase, oral aspirin, both, or neither among 17 187 cases of suspected acute myocardial infarction: ISIS 2. Lancet. 1988;2(8607):349 360. doi:10.1016/S0140 6736(88)92833 4.
iii Kefauver Harris Drug Amendments of 1962, Pub L No. 87 781, 76 Stat 780 (1962) (codified as amended at 21 USC § 355(d)).
The Temple quote is from:
Peck CC, Wesler J. Report of a Workshop on confirmatory evidence to support a single clinical trial as a basis for new drug approval. Drug Information Journal 2002:36:517-534.