All deaths follow recurrence, but not all recurrences are fatal: a retrospective cohort study of recurrence timing and survival after curative resection of hepatocellular carcinoma

Background Postoperative recurrence is the main barrier to long-term survival after curative resection of hepatocellular carcinoma (HCC). Whether recurrence timing, rather than recurrence itself, distinguishes lethal from non-lethal trajectories remains unclear. Methods We analyzed 154 patients after curative HCC resection at a single centre. Preoperative circulating tumour cells (CTCs) were classified as epithelial, mixed, or mesenchymal (M-CTC); microvascular invasion (MVI) was graded M0-M2. We defined the dissemination axis (DA) as a composite of the preoperative M-CTC phenotype and postoperative pathological MVI grade. Because MVI is pathological, the complete DA can only be calculated after pathologic review, not before surgery. Early recurrence was defined as recurrence within 24 months. Overall survival (OS) and recurrence-free survival (RFS) were assessed with penalised Cox models, landmark and time-window analyses, competing-risk methods, and exploratory log-hazard mediation. Results Of 154 patients, 142 (92.2%) developed recurrence: 118 with early recurrence, 24 with late recurrence, and 12 remained recurrence-free. During the observed follow-up period, no deaths occurred without prior recurrence (0/154); 80 of 83 deaths (96.4%) were preceded by early recurrence. Median OS was 18.5 months for early recurrence and was not reached for late or non-recurrence groups (observed median survival, 49.0 and 39.5 months respectively). Across DA strata, 24-month recurrence rates increased from 22.2% (low risk) to 95-98% (high risk), while median OS decreased from 47 to 12 months. High DA status independently predicted poorer RFS (HR 3.72; 95% CI 2.28-6.07) and OS (HR 3.25; 95% CI 1.69-6.25). The effect was concentrated in the first 24 months. Early recurrence accounted for 27.5% of the log-hazard attenuation (95% CI 15.5-44.9%), but this exploratory analysis cannot establish causal mediation. 24-month RFS status discriminated subsequent OS (time-dependent AUCs 0.732-0.862). Conclusion In this exploratory single-centre retrospective cohort study, recurrence timing distinguished lethal from non-lethal outcomes. The DA, combining a preoperative CTC phenotype with the postoperative MVI grade, identified patients at high risk of 24-month early recurrence, supporting evaluation of risk-stratified surveillance and 24-month RFS as an early endpoint in perioperative HCC trials. Because no external validation was performed, these findings are hypothesis-generating and require independent validation before clinical use.