More Than a Quarter of Completed Lung Cancer Trials Are Unpublished

By Marilynn Larkin

November 13, 2019

NEW YORK (Reuters Health) - Approximately 27% of completed lung cancer trials were unpublished as of 2016, raising concern about publication bias, a cross-sectional study suggests.

"Most unpublished results are from negative studies where there is often no interest on the side of the sponsoring company to publicize the results," Dr. Giuseppe Giaccone of Georgetown University in Washington, DC, told Reuters Health by email. "Sometimes negative publications also have a difficult time getting published in good journals that prefer positive studies. A way to alleviate the first issue is to oblige companies and investigators that have a trial listed in to provide a written report within a time frame from completion."

Dr. Giaccone and colleagues searched to identify registered lung cancer clinical trials with a completion date between 2000-2016, and used that information to search PubMed, Google Scholar and EMBASE. They also sent emails to investigators of unpublished trials to confirm status.

Trial results were categorized as (1) publication in a peer-reviewed journal and (2) no peer-reviewed publication (including conference proceedings and reporting of results on

As reported online November 6 in JAMA Network Open, 1,294 clinical trials were included in the study. Among them, 1,038 (80.2%) were completed and 256 (19.8%) were terminated. The median follow-up time after study completion or termination was 116 months.

Among the 1,038 completed trials, database searches suggested that 702 (67.6%) were published and 336 (32.4%) were unpublished.

Contact information was identified and emails were sent to investigators of 183 (54.5%) of the unpublished trials. A total of 102 responses (55.7%) were received: 51 reported peer-reviewed publication and 51 confirmed nonpublication.

This meant that overall, among the completed trials, 753 (72.5%) were published and 285 (27.5%) were not.

Industry-sponsored trials were less likely to be published than those sponsored by the US National Institutes of Health (odds ratio, 2.27) or academic institutions (OR, 1.55).

Further, multicenter studies were more likely than single-center studies to be published (OR, 2.78). Greater patient enrollment was associated with publication: >500 participants, OR, 3.43 and 100-500 participants, OR, 2.87 compared to studies with fewer than 100 participants.

By contrast, randomization status and trial phase were not associated with publication.

Trials were terminated for various reasons: low enrollment, 46.5%; lack of efficacy, 11.7%; toxic effects, 8.6%; miscellaneous, such as the principal investigator leaving the institution (14.8%); and unknown, 18.4%.

Among 256 terminated clinical trials, 72 (28.1%) were published in peer-reviewed journals; the rest remained unpublished.

Coauthor Dr. Chul Kim, also of Georgetown University, noted in a separate email that a recently launched preprint repository, medRxiv (, "provides an online platform for researchers to share their unreviewed work. Increasing utilization of medRxiv may help reduce the rate of nonpublication," he said.

Dr. Nicholas Rohs, a hematologist at The Blavatnik Family-Chelsea Medical Center at Mount Sinai in New York City, commented by email, "The publication process can be exhausting, frustrating, and lengthy, particularly with all of the other demands on a researcher's time. If you feel you may not get your work published in a journal with a high impact factor, it may not seem worth going through all of the work."

"The current pace of research may also have some role in rate of publications," said Dr. Rohs, who was not involved in the study. "By the time some trials (are completed), there has already been a major shift in treatment, making the results less meaningful - for example, the advent of first-line immunotherapy in lung cancer derailing many of the ongoing second-line immunotherapy trials."

"Finally, we also have to acknowledge the elephant in the room: many clinical trial publications are controlled by pharmaceutical companies," he said. "These companies are incentivized to highlight positive trials."

"However, I strongly agree with the authors that it is our responsibility to ensure all trials undergo some form of publication," he said. "We certainly owe it to the patients and families who participated in these trials and accepted the uncertainty and extra efforts that come with clinical research."

"While this is a difficult issue to fully address, I think ensuring that trials that reach a certain point, such as a certain percentage of accrual, should be obligated to publish at least raw data," he added. "As our ability to harness AI and other technology to process big data sets continues to improve, this information may become even more valuable."


JAMA Netw Open 2019.