We retrospectively selected patients with unresectable HCC who received lenvatinib in combination with PD-1 inhibitors at Beijing Ditan Hospital, Capital Medical University, from July 2019 to January 2022. The following patients were included: (1) HCC diagnosed by histological or radiological criteria as defined by the American Association for the Study of Liver Diseases (AASLD) guidelines; (2) age 18 years or older; (3) patients with tolerable general status, Eastern Cooperative Oncology Group score (ECOG PS) 0-2, Child-Pugh class A or B; and (4) received lenvatinib combined with PD-1 inhibitors as the first-line of therapy. The Main exclusion criteria were as follows: (1) patients who received systemic drugs, including sorafenib, lenvatinib, PD-1 inhibitors, etc.; (2) patients who did not have enhanced abdominal computed tomography (CT) or magnetic resonance imaging (MRI) images before baseline, or the third lumbar vertebra (L3) was not within the imaging range; (3) baseline blood routine, c-reactive protein(CRP), and alpha-fetoprotein(AFP) were not performed within two weeks before treatment; and (4) having other malignancies or combined severe extrahepatic disease.

The study was conducted by the Declaration of Helsinki, and experienced clinicians determined patient eligibility for combined therapy based on guidelines. In addition, this study was approved by the Ethics Committee of Beijing Ditan Hospital, Capital Medical University (JDLKZ 2021-056-01).

CT or MRI before the first dose of PD-1 inhibitors was used to measure PMI and was independently assessed by two radiologists, and disagreements were resolved by a third experienced radiologist. The cross-sectional area of the psoas muscle was measured at the level of the L3. Axial images at the level of L3 were manually measured on a dedicated workstation (SliceOmatic software, version 5.0) for specific tissues (− 29 to + 150 Hounsfield units (HU) thresholds) ( Figure 1 ). PMI was determined by dividing L3 psoas cross-sectional area (mm²) by height squared (m²) (26). Given the absence of a standardized criterion for sarcopenia in China, we employed X-tile software (version 3.6.1) to determine the optimal cut-off value for psoas muscle index (PMI), which was separately selected based on gender to account for inherent gender disparities (20, 27). The patient cohort was divided into low and normal PMI cohorts, and patients in the low PMI cohort were considered to be in a state of sarcopenia. ΔPMI indicates the difference between baseline PMI and PMI at week four, showing dynamic changes in PMI after treatment.

Lenvatinib, in combination with PD-1 inhibitors, was administered to patients with unresectable HCC who met the criteria, and the treatment regimens are summarized below. Lenvatinib was orally administered daily, with the initial dose determined according to body weight (≥ 60 kg, 12 mg; < 60 kg, 8 mg); PD-1 inhibitor was intravenously injected at 200 mg every three weeks, and the drugs used included sintilimab, camrelizumab, and tislelizumab. This combination regimen has been used as the first line of therapy for included patients.

Patients were followed every 2-3 months until death or the end of the study. OS and PFS were calculated from the initiation of the PD-1 inhibitor. At the same time, ORR and disease control rate (DCR) were obtained based on the best radiographic response observed during treatment. ORR was defined as the proportion of patients achieving complete response (CR) and partial response (PR) according to modified Response Evaluation Criteria in Solid Tumors (mRECIST). At the same time, DCR increased the proportion of patients with stable disease (SD).

We retrospectively collected patients’ baseline data, including age, sex, etiology, previous treatment history, and BCLC stage. BCLC stage combines multiple information, including tumor size, number, ECOG PS, portal vein tumor thrombus (PVTT), extrahepatic metastases, and Child-Pugh class. In addition, the following parameters were recorded and calculated: neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), albumin-bilirubin (ALBI), prognostic nutritional index (PNI), AFP, and CRP. Individual measures such as NLR, PLR, PNI, CRP, and PMI were used as continuous variables to draw the receiver operating curve (ROC) for predicting the 1-year survival of patients and to compare AUC. The cut-off values of the above continuous variables were determined using X-tile software and subsequently categorized into binary classification indicators.

In this study, R software (version 4.1.3, http://www.rproject.org) was performed for data analysis, and P < 0.05 was judged to be statistically different. Demographic data and disease characteristics were compared between patients in the low and normal PMI groups, continuous data were expressed as mean plus or minus standard deviation or median (interquartile range), and T-test and Mann-Whitney U test were used to compare the two groups. Categorical data were described as numbers (percentages) and compared using the Chi-square test. PFS and OS were calculated using the Kaplan-Meier method, and differences between groups were tested using the Log-Rank test. Cox multivariate analysis was performed to investigate important predictive variables affecting prognosis, and nomogram-based prediction models were constructed. In addition, the ROC and calibration curve were plotted separately to demonstrate the model’s performance, and the predictive value between different variables was compared by calculating the area under the ROC curve (AUC).

A total of 180 patients were enrolled in this study, and Table 1 documents the patient baseline characteristics of the overall cohort. Most patients were male (n = 151, 83.9%), with a median age of 57.5 [51.0, 64.0] years. Hepatitis B virus (HBV) infection was the leading cause in this cohort (n = 159, 88.3%). Most patients had Child-Pugh A liver function before initiating PD-1 inhibitor (n = 127, 70.6%). Approximately half of the patients have an ECOG PS score of 1/2 (n = 94, 52.2%), along with PVTT (n = 100, 55.6%) and distant metastasis (n = 100, 55.6%). Moreover, 144 (80.0%) patients had BCLC stage C. Previous surgical resection, TACE, and ablation were performed in 29 (16.1%), 158 (87.8%), and 74 (41.1%) patients, respectively.

The patient cohort consisted of 151 males and 29 females, for whom separate cut-off values were calculated based on gender. A PMI cut-off of 2.9 mm²/m² was determined for females, assigning 6 patients to the low PMI cohort and 23 patients to the normal PMI cohort. In contrast, males had a PMI cut-off of 3.9 mm²/m², classifying 40 patients into the low PMI group and 111 patients into the normal PMI group. Ultimately, 46 (25.6%) patients were considered to be in a state of sarcopenia ( Figure 2 ). Table 1 also presents patient characteristics for the normal and low PMI groups, facilitating a comparison of baseline differences between these two cohorts. Patients with sarcopenia were older (median 60.0 vs. 56.5, p = 0.0138), exhibited elevated serum CRP levels (median 21.0 vs. 6.3, p = 0.0001), demonstrated a lower frequency of prior ablations (26.1% vs. 46.3%, p = 0.026), and displayed reduced PNI levels (median 41.5 vs 43.1, p = 0.0215) compared to non-sarcopenic patients.

Among the overall population, 62 (34.4%) patients developed decreased PMI levels after four weeks of PD-1 inhibitor therapy, but interestingly two patients exhibited an elevation in PMI. Evaluation based on mRECIST criteria revealed CR in 9 patients (5.0%), PR in 38 patients (21.1%), SD in 83 patients (46.1%), and progressive disease (PD) in 50 patients (27.8%). Both of the two patients with increased PMI got an objective response as indicated by PR. The ORR and DCR were 26.1% and 72.2%, respectively. The patients with low PMI had significantly lower rates of ORR and DCR compared to those with normal PM, 4.3% vs 33.6% (p = 0.0002) and 39.1% vs 83.6% (p < 0.0001), respectively.

As of January 2023, 111 (61.7%) patients had died, and 135 (75.0%) had a progression event. Median PFS and median OS were 6.2 (95% CI, 5.5 – 7.9) months and 16 (95% CI, 13.2 – 19.2 months, respectively. According to sex-specific cut-offs, patients with sarcopenia had a significantly worse PFS compared to those without sarcopenia, 3.0 months [95% CI, 2.57 – 3.6] vs 8.3 months [95% CI, 6.7 – 11.3], respectively (HR, 0.18; 95% CI, 0.12-0.28; p < 0.0001). A similar difference was found in the median OS, 7.3 months (95% CI, 4.9 – 8.7) for the sarcopenia group and 21.6 months (95% CI, 17.1 – 27.6) for the non-sarcopenia group, HR, 0.19; 95% CI, 0.13-0.3; p < 0.0001. ( Figures 3A, B ).

In addition, we analyzed the relationship between PMI change and survival after four weeks of treatment. Notably, patients with no significant change in PMI exhibited a significantly longer median PFS compared to those with decreased PMI (9.2 months vs. 3.3 months, p < 0.0001) ( Figure 3C ). Furthermore, this difference was observed in OS (24.2 months vs 9.4 months, p < 0.0001) ( Figure 3D ).

Multiple inflammatory nutritional indicators currently exist to assess patient prognosis, and the ROC curves in Figure 4 describe the discriminatory power of PMI, NLR, PLR, PNI, and CRP for patient survival prediction, with AUC values of 0.750 (95% CI, 66.8 – 83.2), 0.569 (95% CI, 0.476 – 0.662), 0.601 (95% CI, 0.508 – 0.693), 0.628 (95% CI, 0.540 – 0.716), and 0.732 (95% CI, 0.650 – 0.813), respectively. The cut-off values for continuous variables were determined by X-tile software, and the ideal cut-off values for NLR, PLR, PNI, and CRP were 2.8, 180.3, 42.1, and 9.9 mg/L, respectively. The above cut-off divided the patient cohort into two groups, and survival differences were almost observed in both PFS and OS ( Figure 5 ).

Tables 2 , 3 indicate the results of Cox regression for OS and PFS, respectively. Univariate analysis of OS revealed significant associations between patient death and various factors, including Child-Pugh class, BCLC stage, previous surgical resection, history of TACE and ablation, tumor number, PVTT, distant metastasis, NLR, PLR, PNI, AFP, CRP, and PMI. Subsequent multivariate analysis identified independent prognostic factors for OS as the presence of distant metastasis (HR, 1.63; 95% CI, 1.01 – 2.62; p = 0.0445), high PLR (HR, 1.61; 95% CI, 1.01 – 2.56; p = 0.0459), high CRP (HR, 1.73; 95% CI, 1.07 – 2.79; p = 0.0251), and normal PMI (HR, 0.26; 95% CI, 0.16 – 0.41; p < 0.0001). Next, we conducted a PFS-related analysis for HCC treated with the combination of lenvatinib and PD-1 inhibitor therapy, and univariate results showed tumor number, NLR, PLR, PNI, CRP, and PMI as potential predictors of PFS. However, only sarcopenia was considered a significant independent risk factor for progression in multivariate analysis (HR, 0.22; 95% CI, 0.14 – 0.35; p < 0.0001).

Prognostic models of patient survival were constructed using the four variables selected by Cox regression above. Nomograms were plotted to predict the probability of patient survival at one year ( Figure 6A ). Calibration curves showed good agreement between the model-predicted 1-year survival of patients and actual observations ( Figure 6B ). The discrimination of this model in the ROC curve was fair, with an AUC value of 0.853 (95% CI, 0.791 – 0.915) ( Figure 6C ).

Treatment-emergent AEs are detailed in Table 4 . In this study, the majority of patients (83.33%, n=180) experienced AEs, with 50% (n=90) reporting treatment-emergent AEs ≥ grade 3. Grade 3 or higher AEs were observed in 34 (73.91%) patients in the sarcopenia group and 56(41.79%) patients in the non-sarcopenia groups, indicating a significantly greater severity of AEs in the sarcopenia group compared to the non-sarcopenia group. In the sarcopenia group, four patients experienced immune-related adverse events (irAEs), including immune-related adrenal insufficiency, grade 3-4 ALT/AST elevation, and immune-related rash. Among the common treatment-emergent adverse events, fatigue (51.17% vs. 29.85%), elevated blood bilirubin (43.48% vs. 26.87%), and hypothyroidism (36.6% vs. 13.43%) were significantly different between groups. In addition, noteworthy disparities were noted in reported dose reductions (23.9% vs. 10.4%) or discontinuations (21.7% vs 5.2%) for the combined administration of lenvatinib and PD-1 inhibitor.