For too long, researchers who don’t like the results of a clinical trial have simply failed to publish them. The US government has now taken solid steps to crack down on this problem. On 16 September, the US Department of Health and Human Services (HHS) issued a revision to the 2007 law that requires clinical trials to be registered and published.

The law was poorly worded and opaque, so researchers were able to hide behind exceptions and loopholes to avoid publishing results. As a consequence, results were often reported only when the product was approved or pending approval. In one analysis, 30% of a sample of trials on hadn’t reported their results even 4 years after completion (H. Saito and C. J. Gill PLoS ONE 9, e101826; 2014). And the percentage is probably much higher for initial safety studies, for which there is no requirement to publish at all. Pharmaceutical firms are often criticized for failing to publish results from trials if they have no plans to continue with a therapy, but academic researchers seem equally guilty.

The practice is bad for science because it means that the literature is biased towards positive results, and researchers will unknowingly repeat failed experiments. It is bad for medicine because dangerous side effects stay hidden. Most of all, it is bad for the volunteers who take part in trials, risking their health only to see their contribution confined to a desk drawer. Yet the US Food and Drug Administration rarely cracks down on offenders: according to bioethicist Jennifer Miller of New York University’s Langone Medical Center, the agency has never enforced the US$10,000-per-day non-compliance fine.

Under the new rule, researchers must register their trials within 21 days of enrolling the first patient, and publish data within a year of completion. In a surprise and welcome move, researchers must now disclose their exact methodology and how they plan to analyse the results. They must also record any changes to the protocol. This attempts to address ‘p-hacking’: shopping around for the statistical test that will yield the best results, or leaving out a group — such as women, or people with a disease — that changes the overall results.

The reforms are not as strict as many would like. They still do not require the reporting of safety studies. The US National Institutes of Health has built on the rules to require this of agency-funded researchers, but industry remains exempt. The HHS also opted not to require researchers to publish data for individual patients, which would have allowed independent analysis of results and made it more difficult to hide adverse side effects. But for now, it is a good first step, one that fosters scientific rigour and affords greater respect to patient commitments.