Publication Bias in Gastrointestinal Oncology Trials Performed over the Past Decade

Publication Bias in Upper Gastrointestinal Oncology Clinical Trials

PURPOSE

Evidence-based medicine requires evaluation of the medical literature to guide clinical reasoning and treatment recommendations. The presence of publication bias towards exclusion of non-statistically significant clinical trials may be leading to an incomplete evaluation of the literature and cause potentially incomplete guidance for patients. We aimed to compare publication rates and impact of publications of positive and negative outcome clinical trials.

METHODS

We queried the US National Library of Medicine Clinical Trials database identifying clinical trials with reported results on the topics of pancreatic, liver, and gastric cancer. A "positive" trial was defined as having a statistically significant difference between the treatment arms, while a "negative" did not. Data collected included termination cause, intervention, funding type, publication rates, and journal characteristics.

RESULTS

In total, 535 clinical trials were examined, across all pathologies clinical trials with significant findings for the primary outcome were published at a higher rate (99%) compared to those with non-significant findings (77%) (p < 0.01). Significantly, more studies with significant findings reached at least 80% of their estimated enrollment goal versus non-significant studies, 72% and 53% respectively (p < 0.01). Three of four metrics for impact of publication showed no difference between significant and non-significant studies once they reached publication.

CONCLUSION

These findings suggest that clinical trials of three of the most common upper gastrointestinal malignancies have a publication bias towards studies with significant primary outcome findings. This study has implications to the way evidence-based medicine is practiced as the medical literature appears to be failing to capture important data for consideration of clinical decision making.

Availability of Results of Trials Studying Pancreatic Adenocarcinoma over the Past 10 Years

BACKGROUND

Pancreatic adenocarcinoma (PDAC) is a lethal cancer with few therapeutic options. Availability of results is a crucial step in interventional research. Our aim was to evaluate results availability for trials in patients with PDAC and explore associated factors.

MATERIALS AND METHODS

We performed a retrospective cohort study and searched the ClinicalTrials.gov registry for trials evaluating PDAC management with a primary completion date between 1 January 2010 and 1 June 2020. Then, we searched for results submitted on ClinicalTrials.gov and/or published. Our primary outcome was the proportion of PDAC trials with available results: submitted on ClinicalTrials.gov (either publicly available or undergoing quality control check) and/or published in a full-text article. The association of predefined trial characteristics with results availability was assessed.

RESULTS

We identified 551 trials of which 386 (70%) had available results. The cumulative percentage of trials with available results was 21% (95% CI, 18-25%) at 12 months after the primary completion date, 44% (95% CI, 30-48%) at 24 months and 57% (95% CI, 53-61%) at 36 months. Applicable clinical trials, required to comply with the 2007 Food and Drug Administration Amendments Act 801 and its final rule on reporting of results on ClinicalTrials.gov, were more likely to have available results over time (HR 2.1 [95% CI 1.72-2.63], P < .001). Industry-funded, small sample size, and terminated trials were less likely to have available results. Other trial characteristics showed no association with results availability.

CONCLUSION

Our results highlight a waste in interventional research studying PDAC.

Mapping evolving treatment effects for overall and progression-free survival shows patterns across 2109 randomized oncology trials

OBJECTIVE

To assess the patterns and time trends in overall survival and progression-free survival treatment effects across randomized controlled trials (RCTs) in oncology.

STUDY DESIGN AND SETTING

A PUBMED search for oncology network meta-analyses (NMAs) was carried (to September 30^th, 2021). Relevant hazard ratios were extracted for systemic treatments from RCTs in the NMAs. After removing duplicate results, relationships between treatment effects, year of publication, trial design and other features were explored.

RESULTS

From 241 oncology NMAs, 2109 unique eligible RCTs provided analyzable data. On average, there was a 12-14% reduction in hazard for overall survival and 27-30% reduction for progression-free survival, with substantial heterogeneity across different malignancies. Correlation between overall survival and progression-free survival treatment effects was modest (r=0.60, 95% confidence interval, 0.56-0.64). Over time, there was a suggestive trend of increased progression-free survival treatment effect, while overall survival treatment effects remained steady. Only one in five trials met criteria for clinically meaningful improvements in overall survival. Among 300 randomly selected trials, mean absolute improvement was 1.6 months for median progression-free survival and 1.4 months for median overall survival.

CONCLUSIONS

Broad patterns across the past 50 years of oncology research suggest continuous progress has been made, but few results meet clinically meaningful thresholds for overall survival improvement.