“…In three studies (7.9% [2.7-20.8%]), correction of time-dependent bias could have made the permissive hazard ratio or relative risk statistically significant [47,48,86]. In 20 studies where the relative risk was less than 1 (52.6% [37.3-67.5%]), correction of timedependent bias could have nullified previously significant hazard ratios or relative risks or could have transformed the preventive relative risk into a permissive hazard ratio or relative risk [45,50,52,53,55,57,[61][62][63]65,71,77,78,81,83,[87][88][89][90][91].…”