Effect of Lead Time Bias in the Evaluation of the Effectiveness of a Screening Program: Example of Screening by Ultrasound for Hepatocellular Carcinoma (HCC) in Compensated HCV-Related Cirrhosis
Abbas Mourad1,2, Michael Schwarzinger2, Nathalie Ganne-Carrie3, Isabelle Rosa4, Philippe Mathurin1, 5, Sylvie Deuffic-Burbon1,2
1Inserm U995, Université Lille Nord de France' LILLE, France; 2Inserm ATIP/Avenir, U738, Université paris Diderot,, PARIS' France; 3Service d, Hepato-Gastroentérologie, Hôpital Jean Verdier, BONDY, France; 4ANGH CHANGH Study Group, Centre Hospifolier Intercommunol, Créteil' France; 5Service Maladies de I'Apporeil digestif et de la Nufrifion, Hôpital Claude Huriez, LILLE, France
Background: Non-randomized controlled trials are confronted with lead time bias, i. e. an apparent improvement in survival due to anticipated diagnosis. We illustrated the effect of lead time bias by assessing the impact of ultrasonographic screening on survival of patients with compensated HCV-related HCC. Methods: We adapted a simple method of correction for lead time (LT) bias in survival analysis to HCC screening, estimating LT as Dtx3xlog(du/ds)/log(2), where Dt the median value of tumor volume doubling time, du and ds median tumor diameters of screened and unscreened patients, respectively. A Markov model was developed to simulate the progression of a cohort of patients with HCV-related HCC, aware of their HCV status, from diagnosis until death. Simulated patients were distributed and treated according to the tumor and health status at diagnosis. The model estimates life expectancy (LE) according to 3 scenarios: S1, no screening (du=4. 28cm); S2, current practice of screening corresponding to 42% of patients diagnosed at an early stage (ds=2. 8cm); S3, optimal practice of screening corresponding to 87% of patients diagnosed at an early stage (ds=2. 2cm). Estimates of du and ds were obtained from two French cohorts, CHANGH for current practice and CHし」000 for optimal practice of HCC screening. Estimate of Dt was found to be 117 days and was varied in sensitivity analysis assuming a less aggressive tumor (Dt=171 days). Estimates of Dt were obtained from studies evaluating growth rates of HCC. Results: A) Baseline analysis (Figure): Compared to no screening (S1), current practice of screening (S2) increases LE by 23 months without correcting for LT bias and by 19 months after correction, and optimal practice of screening (S3) increases LE by 53 months without correcting for LT bias and by 43 months after correction. B) Sensitivity analysis: When assuming a less aggressive tumor (Dt=171days vs. 117days), the LT bias would increase. Consequently, compared to no screening, LE with current practice of screening would decrease from 19 to 17 months, and LE with optimal practice of screening from 43 to 39 months. Conclusions: The benefit of HCC screening is overestimated when LT bias is not considered. Our approach allows quantifying the magnitude of LT bias and provides more accurate estimates of the benefit of HCC screening.
Nathalie Ganne-Carrie - Advisory Committees or Review Panels: Roche, MSD; Speaking and Teaching: BMS, Gilead
Philippe Mathurin - Board Membership: Schering-Plough, Janssen-Cilag, BMS, Gilead; Consulting: Roche, Bayer, Boehringer
Sylvie Deuffic-Burban - Consulting: MSD, GSK, Gilead, Abbott; Grant/Research Support: Roche, Janssen Pharmaceuticals, Schering-Plough; Speaking and Teaching: Cellestis
The following people have nothing to disclose: Abbas Mourad, Michael Schwarzinger, Isabelle Rosa