Can cancer researchers accurately judge whether preclinical reports will reproduce?

There is vigorous debate about the reproducibility of research findings in cancer biology. Whether scientists can accurately assess which experiments will reproduce original findings is important to determining the pace at which science self-corrects. We collected forecasts from basic and preclinical cancer researchers on the first 6 replication studies conducted by the Reproducibility Project: Cancer Biology (RP:CB) to assess the accuracy of expert judgments on specific replication outcomes. On average, researchers forecasted a 75% probability of replicating the statistical significance and a 50% probability of replicating the effect size, yet none of these studies successfully replicated on either criterion (for the 5 studies with results reported). Accuracy was related to expertise: experts with higher h-indices were more accurate, whereas experts with more topic-specific expertise were less accurate. Our findings suggest that experts, especially those with specialized knowledge, were overconfident about the RP:CB replicating individual experiments within published reports; researcher optimism likely reflects a combination of overestimating the validity of original studies and underestimating the difficulties of repeating their methodologies.Author summary: Science is supposed to be self-correcting. However, the efficiency with which science self-corrects depends in part on how well scientists can anticipate whether particular findings will hold up over time. We examined whether expert researchers could accurately forecast whether mouse experiments in 6 prominent preclinical cancer studies conducted by the Reproducibility Project: Cancer Biology would reproduce original effects. Experts generally overestimated the likelihood that replication studies would reproduce the effects observed in original studies. Experts with greater publication impact (as measured by h-index) provided more accurate forecasts, but experts did not consistently perform better than trainees, and topic-specific expertise did not improve forecast skill. Our findings suggest that experts tend to overestimate the reproducibility of original studies and/or they underappreciate the difficulty of independently repeating laboratory experiments from original protocols.