A total of 164 patients with high-risk HCC were included in this study. Among these, 103 received HAIC-LEN-PD1 treatment and 61 received LEN-PD1 treatment. Generally, both groups had balanced baseline characteristics. Patient characteristics are listed in Table 1. The HAIC-LEN-PD1 group comprised 12 females and 91 males with a median age of 52 years. The LEN-PD1 group comprised 4 females and 57 males with a mean age of 56 years. The median tumor size was 12.2 cm in the HAIC-LEN-PD1 group and 13.5 cm in the LEN-PD1 group. Most patients had Child-Pugh A disease (HAIC-LEN-PD1, n = 84; LEN-PD1, n = 50) and underlying chronic liver disease caused by hepatitis B virus infection (HAIC-LEN-PD1, n = 93; LEN-PD1, n = 58) who all received antiviral therapy. Extrahepatic spread was observed in 32 and 20 patients from the HAIC-LEN-PD1 and LEN-PD1 groups, respectively. In the LEN-PD1 group, 18, 31, and 12 patients had TO ≥ 50%, Vp4, and both TO ≥ 50% and Vp4, respectively. In the HAIC-LEN-PD1 group, 33, 54, and 16 patients had TO ≥ 50%, Vp4, and both TO ≥ 50% and Vp4, respectively. In addition, no patients with bile duct invasion was observed in this study.

All 103 patients in the HAIC-LEN-PD-1 group received triple combination therapy with lenvatinib, PD1s, and HAIC. Among them, 62, 26, and 15 were treated with camrelizumab, tislelizumab, and sintilimab, respectively. In the LEN-PD1 group, 32, 20, and 9, were treated with camrelizumab, tislelizumab, and sintilimab, respectively.

The median follow-up time was 16.3 months in the HAIC-LEN-PD1 group and 24.1 months in the LEN-PD1 group. A total of 78 (75.7%) patients in the HAIC-LEN-PD1 group and 55 (90.2%) patients in the LEN-PD1 group experienced the disease progression or death. The median PFS time was significantly longer in the HAIC-LEN-PD1 group (9.6 months, 95%CI: 8.5–10.8) than the LEN-PD1 group (4.9 months, 95%CI: 3.6–6.1, HR = 0.48, p < 0.001) (Fig. 1a). A total of 83 deaths were observed (HAIC-LEN-PD1 n = 42; LEN-PD1 n = 41), and the median OS of the HAIC-LEN-PD1 and LEN-PD1 groups were 19.3 (95%CI: 11.0–27.5) and 9.8 (95%CI: 5.7–13.8) months, respectively (HR = 0.43, p < 0.001) (Fig. 1b).

The progression-free survival (PFS) and overall survival (OS) of the two groups. Kaplan–Meier curves of PFS (a) and OS (b) for patients in the LEN-PD1 and HAIC-LEN-PD1 groups

Tumor responses are listed in Table 2. Based on the RECIST1.1 criteria, the DCR and ORR were significantly higher in the HAIC-LEN-PD1 group than the LEN-PD1 group (92.2% vs. 72.1%, p = 0.001; 64.1% vs.14.8%, p < 0.001, respectively). Based on the mRECIST criteria, the DCR and ORR were also significantly higher in the HAIC-LEN-PD1 group than the LEN-PD1 group (92.2% vs.72.1%, p = 0.001 and 76.7% vs. 23.0%, p < 0.001, respectively). The intrahepatic tumor responses are presented in Table 2. The HAIC-LEN-PD1 group showed a significantly higher intrahepatic ORR than the LEN-PD1 group according to the mRECIST (80.6% vs. 24.6%, p < 0.001). The best response for intrahepatic target lesions according to the RECIST1.1 criteria is shown in the waterfall plot in Fig. 2. Additionally, 20 patients (HAIC-LEN-PD1, n = 18; LEN-PD1, n = 2) achieved a complete response to intrahepatic lesions based on the mRECIST.

Waterfall plot for tumor size changes of intra-hepatic target lesions. Abbreviations: PD, progressive disease; PR, partial response

Additionally, we investigated whether there were differences in the efficacy of different high-risk features. The HAIC-LEN-PD1 group had superior OS (not estimable vs. 9.7 months, p = 0.002) and PFS (10.8 months vs. 5.8 months, p < 0.001) in patients with Vp4 than the LEN-PD1 group. For patients with TO ≥ 50%, the PFS was longer in the HAIC-LEN-PD1 group compared to the LEN-PD1 group (7.7 months vs. 5.1 months, p = 0.035), but there was no significant difference in OS (15.9 months vs. 11.8 months, p = 0.117). No significant difference was observed in PFS (8.1 months vs. 3.4 months, p = 0.178) and OS (14.5 months vs. 3.9 months, p = 0.081) between patients with both TO ≥ 50% and Vp4 although the HAIC-LEN-PD1 group had a much better PFS and OS. This maybe due to the small sample size of patients with these features (Fig. 3).

Survival outcomes of patients with different high-risk factors. (a), (b) and (c) were the PFS of patients with Vp4, TO ≥ 50% and both respectively. (d), (e) and (f) were the OS of patients with Vp4, TO ≥ 50% and both respectively

We also compared the survival outcomes between patients in the HAIC-LEN-PD1 and LEN-PD1 group who received different types of PD-1s. There were no significant differences among the three types of PD1s, either in terms of median OS or PFS in HAIC-LEN-PD1 and LEN-PD1 group (Supplementary material 1b). Moreover, subgroup analysis showed that HAIC-LEN-PD1 was associated with better median OS and PFS than LEN-PD1 across most patient subgroups (Fig. 4).

Forest plots of (a) overall survival and (b) progression-free survival in different patient subgroups. HR, hazard ratio; CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; ALBI, grade Albumin-Bilirubin grade; AFP, alpha-fetoprotein; Vp, Japan’s portal vein invasion classification; EHS, extra-hepatic spread

The results of the univariate and multivariate analyses of PFS and OS are shown in Supplementary material 1c. Multivariate analysis showed that the independent risk factors for PFS were associated with HAIC or extrahepatic metastasis. Furthermore, the independent risk factors for OS were combined with HAIC, sex, and Child-Pugh Class.

In this study, 103 patients were treated with 396 HAIC cycles (median, 4 cycles). Dose adjustment of lenvatinib was observed in 28 patients in the HAIC-LEN-PD1 group and 12 patients in the LEN-PD1 group; however, none of the patients discontinued the target drugs. At the cutoff date, 78 and 55 patients developed disease progression in the HACI-LEN-PD1 and LEN-PD1 groups, respectively. Seventeen and 8 patients in the HACI-LEN-PD1 and LEN-PD1 groups, respectively, did not receive second-linetreatment because of refusal or poor liver function. The details of further treatments after disease progression are shown in Supplementary material 1d.

No treatment-related deaths occurred (Table 3). The grade 3–4 adverse events (AEs) more common in the HAIC-LEN-PD1 group included alanine aminotransferase (ALT) elevation (20.4% vs. 8.2%, p = 0.039), neutropenia (11.7% vs.1.6%, p = 0.014), thrombocytopenia (16.5% vs.3.3%, p = 0.011), vomiting (9.7% vs.1.6%, p = 0.046), and hyperbilirubinemia (17.5% vs. 4.9%, p = 0.020). Any grade, including neutropenia, thrombocytopenia, ALT elevation, aspartateelevation, vomiting, and abdominal pain, was more frequent in the HAIC-LEN-PD1 than the LEN-PD1 group. For patients in the HAIC-LEN-PD1 group, the increases in serum aminotransferases and total bilirubin were most significant on the second day after the completion of HAIC, and most patients can recover to normal levels within one week. To better illustrate the effect of triple therapy on liver function, we have drawn a Sankey diagram of the dynamic changes in Child-Pugh classification for the two groups. As shown in Supplementary material 1e, the liver function in the HAIC-LEN-PD1 group deteriorates significantly after treatment. However, with proper hepatoprotective treatment, the majority of patients can recover their liver function to the pre-treatment level.

The most common potentially immune-related TRAE was grade 1–2 hypothyroidism (34.9%). Moreover, 2 patients developed grade 3 immune-related dermatitis and 2 developed grade 3 immune-relatedhepatitis and pneumonitis. After treatment with corticosteroids and the suspending of PD1s, patients with immune-related dermatitis immediately returned to normal, and patients with immune-related hepatitis and pneumonitis recovered after 1 month. 6, 2 and 5 patients suspend of camrelizumab, sintilimab and tislelizumab due to 3–4 grade immune-related AEs. In addition, treatment related adverse events of different immune checkpoint inhibitors are shown in Supplementary material 1 f.

Additionally, the high incidence of abdominal pain may be associated with oxaliplatin infusion during the HAIC procedure; the specific abdomen could be acute and severe but was quickly relieved byimmediate infusion of lidocaine via a microcatheter.

Furthermore, we found that the spleen volume increased significantly more often in the HAIC-LEN-PD1 group during treatment than at baseline. Among them, 19 underwent partial splenic embolization.