Rapidly growing fields in medicine, particularly immunology, are especially vulnerable to publication bias. Compared with therapeutic successes, negative and inconclusive clinical results are underreported because they are commercially inconvenient. In March, the FDA contacted over 2,000 sponsors and researchers to report trial results, and reported that nearly 30% of trials subject to mandatory reporting had no results posted [1]. A parallel image emerges from a JAMA Network Open study by Mariani et al., which examined breast cancer immunotherapy trials completed before the end of 2022, a quarter of which never reported outcomes [2]. This is a problem regulators have begun to address, but journals have a lever they have not yet used consistently.
Thousands of registered trials drive the growth of immunotherapy, yet many are single-centre or of small sample size. This contributes not only to low-certainty evidence but, at times, to evidence that never enters the public record. The JAMA adds a further signal: single-centre trials were significantly more likely to go underreported than multicentre trials (35.2% versus 15%) [2]. Among the 19 randomised trials that reported results, 17 did not meet their primary endpoint [2]. The unreported evidence is, therefore, plausibly skewed toward negative findings. The consequence is that the evidence available to clinicians and patients is filtered. Selective publication gives birth to the bias, guidelines inherit it, and patients are exposed to treatments whose risk–benefit profile is incompletely characterised.
After describing and quantifying the issue, solutions can be engineered to reduce publication bias. The FDA’s approach is to remind sponsors of their legal obligation, seek voluntary compliance, and send notices if no response is received. This is a commendable strategy, but it operates usually after the studies have been published and indexed. We recommend four main measures that are at the level of submission, before the evidence base is filtered.
First, mandatory prospective registration should be verified at submission, not only declared. The registration number should be checked against the manuscript itself; the primary endpoint, eligibility criteria, and sample size should match the registered record. Registration alone is insufficient if the record can be modified without dated amendments, so investigators should be required to provide a time-stamped registration history rather than a snapshot.
Second, results deposition should be a condition of publication. For any trial whose primary completion date precedes submission by more than twelve months, journals should require a posted results entry before the manuscript proceeds to review. This converts the reporting of results from a post-publication expectation into a pre-publication requirement, which closes the gap the FDA is currently trying to close through enforcement alone.
Third, adequately powered negative trials should be given equal chances of publication. The Mariani data show that negative immunotherapy trials are less likely to reach the public record, which means journals are part of the filtering problem. A stated commitment to peer review negative trials within a defined timeframe, and to publish null primary endpoints without requiring a positive secondary finding to justify the paper, would begin to correct this. It is a change of culture as much as a policy, which makes it the most high-yield.
Finally, authors should be required to state plainly, in the abstract, whether the primary endpoint was met. A simple structural requirement can be enforced to accomplish this: one sentence in a fixed position, reporting whether the primary endpoint was achieved. Evidence from immunology and allergy RCTs suggests that over half of abstracts with null primary endpoints contain some element of spin; a mandatory statement would not eliminate this, but it would make it harder to sustain [3].
Each of these four measures asks journals to enforce values they already profess. Publication bias is not primarily a failure of individual integrity; it is the predictable result of a system that permits underreporting and offers little consequence for it. In such a system, the point of leverage matters. Journals sit at a point in that system where a small change could have a large effect. Immunotherapy is growing faster than the systems that should keep its evidence base honest. The measures proposed here are modest, but they would ensure that the field’s failures are as visible as its successes.