A comparison of matched interim analysis publications and final analysis publications in oncology clinical trials

Background: Progression-free survival is an increasingly popular surrogate end point for overall survival. The strength of correlation between the two end points varies, raising questions about the correlation between results of interim analyses that report mature progression-free survival data with the subsequent final publication that report overall survival. Methods: We searched PubMed from 2005 to 2015 for randomized controlled trials that measured both progression-free survival and overall survival. We matched interim publications that reported mature progression-free survival data with their final analyses that reported overall survival. We included 25 matched pairs and 8 unmatched interim analyses whose final analyses are not published. Our primary objectives are to compare interim publications with matched final publications in terms of journal prominence and Altmetric score and to compare progression-free survival and overall survival effect sizes. Results: All interim analyses (n=33) were prespecified and there was a statistically significant progression-free survival benefit in 31 (93.9%). Only eight matched final analyses had statistically significant overall survival data. Interim analyses were more often published in top-5 general medicine journals (P<0.01) but not in top-5 oncology journals (P=0.26). Altmetric scores were higher in interim analyses (P<0.01). Progression-free survival effect sizes from interim analyses were a median of 31% larger than overall survival effect sizes from final analyses. Conclusion: Interim analyses with progression-free survival data may generate hype in oncology, as evidenced by journal impact factors and Altmetric scores. The cause of this hype may be due, in part, to large progression-free survival effect sizes. Regardless, in trials that investigate progression-free and overall survival, publishing interim analyses with mature progression-free survival data apart from the final analyses with mature overall survival should be cautioned.