Recorded Early Imaging Responses After Transarterial Chemoembolization Plus Lenvatinib Versus Transarterial...

Background/Objectives: Vietnamese evidence on transarterial chemoembolization (TACE) plus lenvatinib for unresectable hepatocellular carcinoma (HCC) remains limited. We examined recorded early imaging responses and uncertainty from missing outcomes, exposure classification, and confounding. Methods: This single-center secondary cohort included all eligible available records during a shared post-approval period (21 February 2024-19 August 2025). The endpoint was a documented complete or partial response according to the Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST v1.1) at the first recorded post-index computed tomography (CT) or magnetic resonance imaging (MRI), an ascertainment-dependent surrogate rather than a standardized clinical-trial objective response rate. We reported risk differences (RDs), risk ratios (RRs), odds ratios (ORs), 95% confidence intervals (CIs), nonparametric bounds, tipping-point analysis, sequential models, and a comparator co-therapy audit. Results: Among 108 patients (23 combination; 85 comparator), recorded responses occurred in 17/23 (73.9%) and 30/85 (35.3%): RD, 38.6 percentage points (95% CI, 18.0-59.2); RR, 2.09 (95% CI, 1.44-3.05); and OR, 5.19 (95% CI, 1.85-14.57). With 6 and 34 undocumented outcomes, response-difference bounds ranged from -1.4 to 64.7 percentage points. Reclassifying 12 undocumented comparator outcomes as responses made Fisher's p-value exceed 0.05; reclassifying 33 eliminated the point-estimate advantage when all six undocumented combination outcomes were nonresponses. The expanded complete-case estimate was attenuated (adjusted OR, 1.97; 95% CI, 0.45-8.60). Excluding five comparator records with explicit systemic co-therapy yielded an adjusted OR of 5.49 (95% CI, 1.93-15.61). Safety incidence and survival were not estimable. Conclusions: Recorded responses differed between strategies, but missingness, imaging ascertainment, treatment-selection confounding, and incompletely characterized comparator exposure prevented identification of a comparative treatment effect. This study identifies registry and implementation gaps and generates a hypothesis for prospective modified RECIST (mRECIST)-based evaluation.