Registered Trials vs Published Results

Registered does not mean proven

When a headline says a treatment is "in clinical trials," that usually means a study was entered in a public registry such as ClinicalTrials.gov. A registry entry shows that a study was planned. It does not show that the study finished, that its results were good, or that anyone has seen them. This page explains three separate things: registration, results posting and journal publication. It is education, not medical advice. For a step-by-step look at reading a record, see finding and understanding registered clinical trials.

What registration is for

Registration puts the plan on record before results exist: who is being studied, what is being tested and what the main outcome will be. That lets readers later check whether the published paper matches the plan.

In the United States, federal regulations require "applicable clinical trials" to be registered, and set the timing: no later than 21 calendar days after the first participant is enrolled (with an earlier-era cutoff for older trials). [1] Which trials count is defined narrowly. For drugs, an applicable trial is an interventional trial of a US FDA-regulated drug whose phase is other than phase 1, with a US link such as a US site. For devices, the primary purpose must be something other than a feasibility study. [2] Many studies outside those rules are registered voluntarily, and some are not registered at all.

Results posting

For applicable trials, the same regulations generally require results information to be submitted no later than one year after the primary completion date, with some delays allowed in certain situations. [3]

Compliance has been uneven. A 2015 analysis of 13,327 trials found that only 13.4% reported summary results within the 12 months required, and 38.3% reported at any time up to September 2013. [4] Timely reporting was more common for industry-funded trials, and the authors also estimated that a share of trials was not required to report at all. [4] Those figures describe a past period, and practice may have changed, so treat them as evidence that gaps exist, not as a current rate.

Publication in a journal

Publishing in a journal is a separate step. A 2009 analysis of completed, non-phase-1 trials registered at ClinicalTrials.gov found that fewer than half of a sampled group (311 of 677, or 46%) had been published, and only 31% of the published ones gave a citation to the paper in the registry. [5] Trials sponsored by industry were less likely to be published than those from nonindustry, nongovernment sponsors (40% vs 56%). [5]

When results never appear, the public sees only part of the evidence. That is one reason a single positive trial can give a skewed picture.

Registration also lets you spot changed outcomes

If a trial lists one primary outcome in the registry and reports a different one in the paper, the result may look better than the plan allowed. A study of 389 randomized trials with clearly registered primary outcomes found that 33.4% had at least one primary outcome change, and trials with a change reported a 16% larger intervention effect size on average. [6]

The COMPare project checked 67 trials in five leading journals against their registered or protocol outcomes. Of 97 prespecified primary outcomes, 76.3% were reported correctly, and 58 trials had discrepancies that required a correction letter. [7] The authors noted that most journals rejected correction letters, so readers can be misled even by well-known journals. [7]

A short checklist

• Is the trial registered, and does the registry say it finished?
• Are results posted, or is a paper linked?
• Do the paper's main outcomes match the registry's?
• Is this the only study, or have others repeated it? See replication and reproducibility.
• Has it appeared as a preprint rather than a peer-reviewed paper?

For where to search, see how to find emerging research and trial records, and for the stages a product goes through, clinical-trial phases explained. The wider Research Library holds study records.