A modified Charlson Comorbidity Index to improve management of patients with hepatocellular carcinoma: a step towards precision medicine

Background & aims Patients with hepatocellular carcinoma (HCC) frequently have comorbidities that limit access to treatments and might increase mortality. The Charlson comorbidity index (CCI) assigns a numeric score to 17 diseases according to their effect on mortality. In patients with cirrhosis and HCC, these variables might be excluded from the CCI (modified CCI, mCCI). Aim of the study was to evaluate the performance of CCI and mCCI for HCC patients’ characterization. Methods We used data from our retrospective monocentric study of patients with first HCC diagnosis. We evaluated the performance of CCI and mCCI in dissecting our population and predicting overall survival (OS). As secondary end-point, the impact of mCCI on adherence to Barcelona Clinic Liver Cancer (BCLC) staging and treatment allocation was evaluated. Results The study included 385 patients (289 males, 68 years median age, 78% Child Pugh (CPT) A, 66% viral etiology) with HCC (271 BCLC 0/A, 64 B, 40 C and 10 D). According to CCI, 94% of our patients were “high-risk” (CCI≥5), 5.7% “intermediate-risk” (CCI 3-4) and 0.3% “low-risk” (CCI 0-2). These cutoffs were applied to mCCI: 21% “high-risk”, 48% “intermediate-risk”, and 31% “low-risk”. Patients' mCCI correlated with CCI (Kendall's τ=0.47; p-value< 0.0001). Overall, patients belonging to a “high” or “intermediate-risk” mCCI classes had a shorter OS as compared to “low-risk” patients (median OS 37 vs 49 vs 67 months, p-value=0.007). This was confirmed in BCLC 0/A (p-value=0.002). Moreover, “high-risk” mCCI class, worse CPT class, presence of varices, BCLC stage and AFP>200 ng/mL were independently associated to mortality [HR1.36 (95%CI 1.09-1.70, p-value=0.006; HR1.62 (95%CI 1.14-2.30), p-value=0.007; HR1.58 (95%CI 1.11-2.24), p-value=0.01; HR1.96 (95%CI 1.66-2.31) p-value<0.001; HR1.80 (1.17-2.79) p-value=0.008]. mCCI was also an independent predictor of mortality in BCLC 0/A, with higher tumor burden and worse CPT. No differences were observed in adherence to BCLC2022 algorithm and access to treatment among the mCCI risk classes. Conclusions mCCI better stratified patients with HCC as compared to CCI and independently predicted mortality. Implementation of a multiparametric therapeutic approach to treatment allocation including evaluation of comorbidities and frailty is warranted.