Limited Generalizability of Retrospective Single-Center Cohort Study in Comparison to Multicenter Cohort Study on Prognosis of Hepatocellular Carcinoma

Introduction: We aimed to evaluate the generalizability of retrospective single-center cohort studies on prognosis of hepatocellular carcinoma (HCC) by comparing overall survival (OS) after various treatments between a nationwide multicenter cohort and a single-center cohort of HCC patients.
Methods: Patients newly diagnosed with HCC between January 2008 and December 2018 were analyzed using data from the Korean Primary Liver Cancer Registry (multicenter cohort, n=16,443), and the Asan Medical Center HCC registry (single-center cohort, n=15,655). The primary outcome, OS after initial treatment, was compared between the two cohorts for both the entire population and for subcohorts with Child-Pugh A liver function (n=2797 and n=5151, respectively) treated according to the Barcelona-Clinic-Liver-Cancer (BCLC) strategy, using Log rank test and Cox proportional hazard models.
Results: Patients of BCLC stages 0 and A (59.3% vs 35.2%) and patients who received curative treatment (42.1% vs 32.1%) were more frequently observed in the single-center cohort (Ps< 0.001). Multivariable analysis revealed significant differences between the two cohorts in OS according to type of treatment: the multicenter cohort was associated with higher risk of mortality among patients who received curative (adjusted hazard ratio [95% confidence interval], 1.48 [1.39– 1.59]) and non-curative (1.22 [1.17– 1.27]) treatments, whereas the risk was lower in patients treated with systemic therapy (0.83 [0.74– 0.92]) and best supportive care (0.85 [0.79– 0.91]). Subcohort analysis also demonstrated significantly different OS between the two cohorts, with a higher risk of mortality in multicenter cohort patients who received chemoembolization (1.72 [1.48– 2.00]) and ablation (1.44 [1.08– 1.92]).
Conclusion: Comparisons of single-center and multicenter cohorts of HCC patients revealed significant differences in OS according to treatment modality after adjustment for prognostic variables. Therefore, the results of retrospective single-center cohort studies of HCC treatments may not be generalizable to real-world practice.

Keywords: BCLC, UICC, liver cancer, retrospective cohort, external validation