Consecutive patients who received LTP conversion therapy from November 2019 to September 2022 for iuHCC were retrospectively analyzed. HCC was diagnosed by clinical assessment or histological examination (22). Tumors were considered unresectable either because of technical unresectability or oncologic unresectability, or both. Technical unresectability is defined in the presence of insufficient future remnant liver volume and lesions assessed by the surgeon as unsuitable for R0 resection. Oncological unresectability is defined as failure to obtain a better prognosis after surgical resection of intermediate-to-advanced HCC, such as major vascular invasion, intrahepatic metastases, and extrahepatic metastases. Other inclusion criteria were as follows: age between 18 and 80 years; Eastern Tumor Collaborative Group physical status (ECOG-PS) score of 0 to 1; Child–Pugh class A or B; adequate organ function; the absence of previous systemic and local treatment history for HCC, and at least one measurable target lesion in accordance with mRECIST. Those who received no radiologic evaluation at the first follow-up 4–6 weeks after the initial treatment for any reason were excluded. The ethics committee at our institution examined and approved the study protocol (LW2022147), and written informed consent was obtained from each patient.

Superselective TACE was performed by specialists with extensive surgical experience in the interventional unit (details in supplementary material). Repeat TACE was conducted when the active lesion area exceeded 50% of the baseline while the liver function was reasonable. Lenvatinib (LENVIMA^®, Merck Sharp & Dohme, H20180052) and PD-1 inhibitors were given within 1 week after TACE, depending on the general condition and recovery of liver function. Patients were given either 8 mg (weight <60 kg) or 12 mg (weight ≥60 kg) of lenvatinib daily via the oral route. PD-1 inhibitors included either 200 mg of camrelizumab (AiRuiKa^®, Jiangsu Hengrui Medicine Co. Ltd, S20190027) or 200 mg of sintilimab (Tyvyt^®, Innovent Biologics and Eli Lilly and Company, S20180016) via intravenous injection every 3 weeks. All patients were treated until disease progression, intolerable toxicity, death, or withdrawal for any reason.

Hepatectomy was performed if patients met the criteria for resection and informed consent was obtained. The criteria for tumor resectability had to be satisfied, as follows: (1) adequate residual liver volume; (2) adequate cutting edge to achieve R0 resection; (3) tumor response assessment of complete response (CR) or partial response (PR) in accordance with mRECIST criteria, or patients with efficacy assessment of stable disease (SD) and maintenance for more than 8 weeks were considered to have controlled tumor biological behavior; (4) Child–Pugh grade A or B, and no other contraindications to hepatectomy.

The first follow-up was scheduled 4–6 weeks after the initial treatment and then every 2–4 months until October 20, 2022. Each follow-up evaluation includes tumor response and laboratory tests, such as blood tests, liver and kidney function, urine routine, tumor markers, myocardial enzymology examination, and thyroid function. Tumor response was evaluated on contrast-enhanced computed tomography (CE-CT) using mRECIST and RECIST1.1 by two senior hepatobiliary surgeons at our hospital.

Early tumor response was defined as patients with CR or PR at the time of the first follow-up (4–6 weeks) after initial treatment using mRECIST, whereas late tumor response was documented after the first follow-up. The non-early response included late tumor response, SD, or progressive disease (PD). Early α-fetoprotein (AFP) response was defined as a reduction of more than 75% in AFP levels following the initial treatment at the first follow-up (23).

We also evaluated the prognostic factors by using inflammatory indexes such as NLR, PLR, and systemic inflammation response index (SIRI). SIRI was calculated as neutrophil count × monocyte count/lymphocyte count (24). These inflammatory indices were split into two groups, based on their median value. Overall survival (OS) was measured from the initiation of the conversion therapy to death from any cause. Progression-free survival (PFS) was calculated from the initiation of the conversion therapy to progression, relapse, or death.

The data collected in this study were statistically analyzed using the software SPSS ver. 24.0 (IBM, Armonk, NY, United States) and R ver. 4.1.1 (http://www.R-project.org/). The Mann–Whitney U test or t-test was used to compare continuous variables, represented as median and quartiles or mean ± standard deviation. Categorical variables were presented as the number of cases and percentages by using the χ² test or Fisher’s exact probability method. Survival analysis was conducted using Kaplan–Meier methods and log-rank tests. The inverse Kaplan–Meier method was conducted to determine the median time of follow-up.

Potential predictive factors of successful conversion surgery were determined using binary logistic regression methods. In multivariate analysis, all factors with p<0.05 and clinically important variables in the univariate analyses were included via the enter method. Given the clinical correlation between early tumor response and AFP response, two models were used to include early tumor response and AFP response in separate multivariate logistic regression analyses to avoid collinearity. Considering that no patients with distant metastases had successful conversion surgery, we presented a sensitivity binary logistic regression analysis for patients without distant metastases.

Potential prognostic factors for PFS and OS were determined using the Cox proportional-hazards models. All factors with p<0.05 and clinically important variables for prognosis were included in the multivariate analysis via the enter method. Given the correlation between successful conversion surgery and early tumor response, we included two models in separate multivariate cox regression analyses to avoid collinearity. For all analyses, p<0.05 was considered statistically significant.

A total of 97 patients who received conversion therapy to treat iuHCC were assessed; 3 patients were subsequently excluded ( Figure 1 ). Among the 94 patients included, 84 (89.4%) had hepatitis B-associated HCC. At baseline, 85 patients (90.4%) were classified as Child–Pugh grade A, and the majority were assigned with an ALBI grade >1 (88.3%). Further, 60 (63.8%) were of Barcelona clinical liver cancer (BCLC) stage C and 19 cases (20.2%) were of BCLC stage B; 56 (59.6%) patients had an initial AFP > 400 ng/mL. In 70.2% of patients, the tumors measured >10 cm in diameter. Multiple tumors were found in 51 patients (54.3%). The laboratory tests were also summarized, including the results for the hepatobiliary enzyme, total bilirubin, NLR, PLR, and SIRI ( Table 1 ). Although adverse events in varying degrees affected all patients, they were within controllable levels ( Table S1 ).

The median follow-up period was 14.4 (10.7–18.2) months. For all patients, the overall response rate was 87.2%, and the disease control rate was 93.6% based on mRECIST ( Table 2 ). Among the patients, 32 (34.0%) underwent conversion surgery. The median time to surgery was 3.8 (3.1–5.4) months. All patients who were successfully converted had no distant metastases. Of the entire cohort, 68 (72.3%) patients showed an early tumor response, and 26 (27.7%) had no early tumor response. The patients who showed early tumor response had a significantly higher conversion surgery rate than those who showed no such response (44.1% vs. 7.7%, p=0.001). The patients with early tumor response had similar baseline characteristics to those with no early tumor response, in addition to the ECOG-PS score (p=0.030) and AST (p=0.020) ( Table 1 ). Representative cases are presented in Supplementary Figure 1 .

The first multivariate model incorporated early tumor response in the 94 patients included in the study. The result indicates that the only independent predictive factor for conversion surgery was early tumor response (OR=10.296; 95% CI: 2.076–51.063; p=0.004). Early AFP response was included in the second multivariate model; however, early AFP response was not a predictor of conversion resection ( Table 3 ). The result was confirmed in 83 patients with non-distant metastases (OR=9.659; 95% CI: 1.899–49.125; p=0.006)( Table S2 ).

The median PFS for the patients with early tumor response was 15.4 months, whereas the median PFS with no early tumor response was 7.8 months (p=0.005; Figure 2A ). However, early tumor response was not an independent predictive factor of PFS (HR= 0.576, 95% CI: 0.302–1.097; p=0.093) ( Table 4 , Multivariate model 1). The median OS of the patients with early tumor response was 23.1 months (p=0.004; Figure 2B ), whereas that of patients with no early tumor response was 12.5 months. In multivariate analysis, early tumor response was independently correlated with OS (HR= 0.404, 95% CI: 0.171–0.954; p=0.039) and jointly with the following conditions: baseline AFP>400 ng/mL (p=0.020), portal vein tumor thrombosis (PVTT) (p=0.011), and extrahepatic metastasis (p=0.001)( Table 5 , Multivariate model 1). The effects of early tumor response on PFS (HR= 0.569, 95% CI: 0.278 – 1.164; p=0.123) and OS (HR= 0.329, 95% CI: 0.119 – 0.910; p=0.032) were confirmed in 83 patients with non-distant metastases ( Tables S3 , S4 , Multivariate model 1).

Given that non-early tumor response was a risk factor for OS, we assigned different scores to each risk factor on the basis of the beta coefficient score in multivariate analysis. We further divided the patients into 4 groups, based on the risk factor. Survival curves showed that the best median OS was in the score 0 group, which declined as the risk factor scores increased ( Figure 3 ).

Median PFS and OS (not reached, not reached) were significantly longer in patients who underwent successful conversion surgery than in those patients who received no such surgery (9.8 months, p<0.001; 14.9 months, p<0.001; Figures 4A, B ). We further assessed the role of conversion surgery in patients with early tumor response. Results showed that for patients with early tumor response, those who underwent conversion surgery also had significantly longer median PFS and OS (not reached, not reached) than those who did not undergo conversion resection (11.2 months, p=0.004; 19.4 months, p<0.001; Figures 5A, B ). Furthermore, early responders combined with late responders had a similar median OS to that of early responders (p=1.000). Both had significantly longer OS than non-responders (p<0.001, p<0.001; respectively; Supplementary Figure 2A ). We further assessed the role of conversion surgery in late tumor responders. The results showed that late responders who underwent conversion surgery had no significantly longer median OS than those who did not (p=0.3; Supplementary Figure 2B ).

Multivariate results confirmed that successful conversion surgery (HR=0.248, 95% CI: 0.099–0.622; p=0.003) was independently correlated with PFS ( Table 4 , Multivariate model 2). In addition, multivariate analysis confirmed that successful conversion surgery was also independently correlated with OS (HR = 0.147, 95% CI: 0.039 – 0.554; p=0.005), in addition to PVTT (p=0.002) and extrahepatic metastasis (p=0.023)( Table 5 , Multivariate model 2). The results on the relevance of successful conversion surgery for PFS (HR= 0.265, 95% CI: 0.105– 0.669; p=0.005) and OS (HR= 0.088, 95% CI: 0.021 – 0.363; p=0.001) were confirmed in 83 patients with non-distant metastases ( Table S3 , S4 , Multivariate model 2).