Pretreatment radiomic imaging features combined with immunological indicators to predict targeted combination immunotherapy response in advanced hepatocellular carcinoma

Abstract

BACKGROUND

Early symptoms of hepatocellular carcinoma (HCC) are not obvious, and more than 70% of which does not receive radical hepatectomy, when first diagnosed. In recent years, molecular-targeted drugs combined with immunotherapy and other therapeutic methods have provided new treatment options for middle and advanced HCC (aHCC). Predicting the effect of targeted combined immunotherapy has become a hot topic in current research.

AIM

To explore the relationship between nodule enhancement in hepatobiliary phase and the efficacy of combined targeted immunotherapy for aHCC.

METHODS

Data from 56 patients with aHCC for magnetic resonance imaging with gadolinium-ethoxybenzyl-diethylenetriamine pentaacetic acid were retrospectively collected. Signal intensity of intrahepatic nodules was measured, and the hepatobiliary relative enhancement ratio (RER) was calculated. Progression-free survival (PFS) of patients with high and low reinforcement of HCC nodules was compared. The model was validated using receiver operating characteristic curves. Univariate and multivariate logistic regression and Kaplan-Meier analysis were performed to explore factors influencing the efficacy of targeted immunization and PFS.

RESULTS

Univariate and multivariate analyses revealed that the RER, neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, and prognostic nutritional index were significantly associated with the efficacy of tyrosine kinase inhibitors combined with immunotherapy (P < 0.05). The area under the curve of the RER for predicting the efficacy of tyrosine kinase inhibitors combined with anti-programmed death 1 antibody in patients with aHCC was 0.876 (95% confidence interval: 0.781-0.971, P < 0.05), the optimal cutoff value was 0.904, diagnostic sensitivity was 87.5%, and specificity was 79.2%. Kaplan-Meier analysis showed that neutrophil-to-lymphocyte ratio < 5, platelet-to-lymphocyte ratio < 300, prognostic nutritional index < 45, and RER < 0.9 significantly improved PFS.

CONCLUSION

AHCC nodules enhancement in the hepatobiliary stage was significantly correlated with PFS. Imaging information and immunological indicators had high predictive efficacy for targeted combined immunotherapy and were associated with PFS.

Key Words: Gadolinium-ethoxybenzyl-diethylenetriamine pentaacetic acid; Hepatocellular carcinoma; Targeted combination immunotherapy; Relative hepatobiliary enhancement ratio; Effect prediction

Core Tip: In recent years, molecular-targeted drugs combined with immunotherapy and other therapeutic methods have provided new treatment options for mid-to-advanced hepatocellular carcinoma, and predicting the effect of targeted combined immunotherapy has become a hot topic in current research. The degree of enhancement of advanced hepatocellular carcinoma nodules in the hepatobiliary stage on gadolinium-ethoxybenzyl-diethylenetriamine pentaacetic acid-enhanced magnetic resonance imaging was significantly correlated with progression-free survival. The imaging information of nodules in the hepatobiliary stage and clinical immunological indicators have high predictive efficacy for targeted combined immunotherapy and are associated with progression-free survival in patients.

INTRODUCTION

According to the latest cancer epidemiology research, the incidence of hepatocellular carcinoma (HCC) is in the fourth place, second only to colorectal cancer, and its mortality has reached the second place, seriously endangering human health[1,2], and accounting for 75%-85% of cases of primary liver cancer in China[3]. The onset of HCC is occult and early symptoms are not obvious; at the time of first diagnosis, more than 70% of patients are in the middle and late stages and cannot receive radical treatment, such as surgical resection[4-7]. In recent years, systematic anti-tumor therapies, such as molecular targeted drug therapy and immunotherapy, have provided new treatment options for mid-to-advanced liver cancer.

Small-molecule tyrosine kinase inhibitors (TKIs) combined with programmed death-1 (PD-1) antibodies have demonstrated significant antitumor activity in the treatment of HCC[5,8-11]. Specifically, this combination effectively reduces tumor staging and enables surgical resection in patients who were previously considered inoperable. Sorafenib, a TKI, has been used for over a decade in patients with unresectable advanced HCC (aHCC). Recently, the application of other TKIs, such as lenvatinib, in the treatment of HCC has increased. Notably, the objective response rate (ORR) of patients with HCC treated with lenvatinib monotherapy is approximately 42.1% in Japan, 22.2% in China, and 18.9% in South Korea[12]. Moreover, among the various TKIs and PD-1 combination therapies available, the lenvatinib + PD-1 combination has demonstrated the highest ORR[13]. Nevertheless, the overall efficacy of targeted combination immunotherapy remains limited.

Resistance to target combination immunotherapies can be categorized into primary resistance and acquired resistance. Although Lenvatinib is effective in aHCC, its clinical application is considerably hindered by drug resistance, as more than 60% of patients with HCC develop resistance to lenvatinib within one year of treatment[14]. Therefore, targeting the resistance mechanism of these drugs and the tumor immune microenvironment through precise personalized therapy is a promising direction for future research[15-17]. However, some patients remain insensitive to TKI and PD-1 combination therapies and may experience various adverse events, including liver function damage, cardiotoxicity, and thrombocytopenia. Therefore, there is an urgent need for new biomarkers and methods to predict the efficacy of TKI and PD-1 combinations and to identify patients who are likely to benefit from these treatments.

Radiomics has been successfully applied in the diagnosis, staging, prognosis, and curative effect prediction of liver cancer owing to its non-invasive nature. Radiomics converts image information into structured data and establishes models to guide clinical decision-making[18]. Research has focused on radiometric feature models based on magnetic resonance imaging (MRI) to predict the therapeutic response to transarterial chemoembolization (TACE) for advanced liver cancer[19]. Notably, imaging omics can also predict the efficacy of PD-1 inhibitors or levatinizumab in the treatment of HCC[20]. Gadolinium-ethoxybenzyl-diethylenetriamine pentaacetic acid (Gd-EOB-DTPA)-enhanced MRI is the gold standard for the highly accurate diagnosis of HCC[21,22]. Gd-EOB-DTPA is a liver-specific contrast agent that is transported by the organic anion-transporting polypeptide 1B3 (OATP 1B3) on the surface of liver cells, taken up by hepatocytes, and subsequently excreted through the bile line and kidneys[23]. OATP 1B3 is one of the downstream targets of the Wnt/β-catenin signaling pathway[24], and its expression in HCC is closely related to the activation of this pathway[25]. In Gd-EOB-DTPA-enhanced MRI images of the hepatobiliary stage, OATP 1B3 expression is correlated with signal intensity (SI)[26]. Therefore, activation of the Wnt/β-catenin signaling pathway may enhance hepatobiliary phase images in Gd-EOB-DTPA-enhanced MRI, suggesting its potential as an ideal predictive imaging marker for the efficacy of targeted combined immunotherapy.

Many hematological indices can predict the prognosis of tumors. The prognostic nutritional index (PNI) predicts surgical risk, postoperative complications, and long-term prognosis by reflecting the nutritional and immune status of patients[27]. Neutrophil-to-lymphocyte ratio (NLR) and PNI can predict outcomes in patients with unresectable HCC after TACE[28]. Imaging omics models are widely used in the diagnosis and differential diagnosis of HCC, preoperative evaluation, prediction of microvascular invasion, and recurrence prediction[21-23]. This study aimed to explore the predictive effect of Gd-EOB-DTPA-enhanced MRI information, PNI, NLR, and platelet-to-lymphocyte ratio (PLR) on HCC-targeted combined immunotherapy and to provide new technical support for individualized treatment of patients with HCC.

MATERIALS AND METHODS

General information

According to the diagnostic criteria of the Guidelines for the Diagnosis and Treatment of Primary Liver Cancer (2022 Edition)[29], data from patients diagnosed with HCC who underwent Gd-EOB-DTPA-enhanced MRI examination in Henan Provincial People’s Hospital from January 2020 to January 2022 was collected, including sex, age, body mass index, alpha-fetoprotein, activated partial thromboplastin time, total bilirubin, albumin, aspartate transaminase, alanine transaminase, tumor diameter, cirrhosis degree, Eastern Cooperative Oncology Group score, Child-Pugh score, Barcelona Clinic Liver Cancer stage, and portal hypertension. This study was approved by the Ethics Committee of Zhengzhou University People’s Hospital and the subjects signed the informed consent was obtained from all participants (2024, Ethics No. 137).

Inclusion and exclusion criteria

The inclusion criteria were receipt of TKI molecular targeted drugs combined with immune (anti-PD-1 antibody) treatment, receipt of Gd-EOB-DTPA-enhanced MRI within 6 months before the start of treatment; having measurable lesions as per the immune response evaluation criteria for solid tumors with clear boundaries and a diameter > 1 cm, and having complete clinical and imaging data. The exclusion criteria were liver surgery, radiofrequency ablation, transcatheter arterial chemoembolization, or receipt of other related treatments before Gd-EOB-DTPA-enhanced MRI, refusal to participate, or poor image quality. A total of 56 patients were included. The patients were divided into two groups: One group received targeted immunotherapy and the other group did not (Figure 1).

Figure 1 Flow diagram of patient selection. HCC: Hepatocellular carcinoma; MRI: Magnetic resonance imaging.

MRI examinations

All MRI examinations were performed using a 3.0-T MRI scanner (Siemens Shenzhen Magnetic Resonance Ltd.). The scanning range covered the entire liver tissue. Gd-EOB-DTPA 0.025 mmol/kg (Bayer Pharmaceuticals) was injected intravenously before contrast imaging, followed by an injection of the same volume of normal saline. Using the fluorescence tracking method, the contrast agent was detected in the proximal abdominal aorta by visual inspection, and the region of interest was placed in the abdominal aorta (at the level of the sub-diaphragm) for automatic detection. After the first detection of the contrast agent, the patients were instructed to hold their breath for 5-10 seconds, after which the arterial-phase sequence was acquired. Hepatobiliary phase imaging was performed 20 minutes after the injection of the contrast agent.

Image processing

Two radiologists manually delineating target regions of interest in the plain scan and hepatobiliary sequence, then defined the target regions in the plain scan and hepatobiliary phase sequences using medical image processing software (RadiAnt DICOM Viewer) and measured the SIs of the intrahepatic HCC nodules and background liver. The SIs measured by the two researchers were averaged. Relative intensity ratio (RIR) was calculated as nodule SI/liver parenchyma SI. Relative enhancement ratio (RER) was calculated as hepatobiliary phase RIR/plain scan phase RIR. When RER = 0.90, the sensitivity and specificity of predicting Wnt/B-catenin-activated HCC were high. Therefore, RER ≥ 0.9 was defined as a high-enhancing nodule, and RER < 0.9 was defined as a low-enhancing nodule[30]. Enhanced MRI should be performed immediately before treatment as the starting time for observation to measure the maximum diameter of the target nodule (Figure 2).

Figure 2 Patient magnetic resonance imaging of hepatobiliary phase and plain scan phase. A: Patient one’s magnetic resonance imaging (hepatobiliary phase); B: Patient one’s magnetic resonance imaging (plain scan phase); C: Patient two’s magnetic resonance imaging (hepatobiliary phase); D: Patient two’s magnetic resonance imaging (plain scan phase). The yellow marked area represents the regions of interest.

Statistical analysis

Statistical analysis was performed using SPSS 26.0 (IBM). P < 0.05 was considered statistically significant. When the numerical variable data conformed to a normal distribution, independence, and homogeneity of variance, an independent sample t-test was used. The χ²-test was used for categorical variables. The Kaplan-Meier method and logistic regression model were used to compare the progression-free survival (PFS) of patients.

RESULTS

Clinical and pathologic characteristics of patients with HCC

As shown in Table 1, no significant differences in terms of sex, age, body mass index, alpha-fetoprotein, activated partial thromboplastin time, total bilirubin, albumin, aspartate transaminase, alanine transaminase, tumor diameter, cirrhosis, Eastern Cooperative Oncology Group score, Child-Pugh score, portal hypertension, or Barcelona Clinic Liver Cancer score were observed between the groups (P > 0.05) (Table 1). The results showed that the two groups of patients were comparable at baseline.

Table 1 Clinical and pathologic characteristics in patients with hepatocellular carcinoma, n (%).

Clinicopathologic variable	Positive	Negative	P value
Sex
Male	24 (42.9)	15 (26.8)	0.315
Female	8 (14.3)	9 (16.1)
Age	55.18 ± 10.51	54.91 ± 8.25	0.154
BMI	22.56 ± 3.63	23.06 ± 3.03	0.585
AFP	3941.72 ± 16289.44	8512.91 ± 25755.92	0.124
APTT	12.72 ± 1.70	13.35 ± 2.60	0.107
TBil	16.68 ± 9.08	16.77 ± 11.09	0.689
ALB	35.98 ± 5.95	36.07 ± 5.17	0.780
AST	81.80 ± 95.81	62.92 ± 49.07	0.204
ALT	67.83 ± 99.19	62.30 ± 59.86	0.685
Tumor diameter	4.88 ± 3.14	5.76 ± 2.74	0.651
Cirrhosis
Yes	27 (48.2)	18 (32.1)	0.384
No	5 (8.9)	6 (10.7)
ECOG score
0	17 (30.4)	12 (21.4)	0.893
1	13 (23.2)	11 (19.6)
2	2 (3.6)	1 (1.8)
Child-Pugh score
A	1 (1.8)	0 (0)	0.408
B	26 (46.4)	18 (32.1)
C	5 (8.9)	6 (10.7)
BCLC score
A	1 (1.8)	1 (1.8)	0.070
B	8 (14.3)	13 (23.2)
C	23 (40.2)	10 (17.9)
Portal hypertension
Yes	16 (28.6)	12 (21.4)	1.000
No	16 (28.6)	12 (21.4)

BMI: Body mass index; AFP: Alpha-fetoprotein; APTT: Activated partial thromboplastin time; TBil: Total bilirubin; ALB: Albumin; AST: Aspartate transaminase; ALT: Alanine transaminase; ECOG: Eastern Cooperative Oncology Group; BCLC: Barcelona Clinic Liver Cancer.

Univariate and multivariate analyses to determine factors for targeted combination immunotherapy efficacy

Univariate analysis revealed that RER, NLR, PLR, and PNI were significantly associated with the efficacy of TKIs combined with immunotherapy (P < 0.05). Incorporating the above variables into the multivariate analysis revealed that RER [odds ratio (OR): 0.005, 95% confidence interval (CI): 0.000-0.498, P = 0.024], NLR (OR: 0.046, 95%CI: 0.003-0.813, P = 0.036), PLR (OR, 0.030, 95%CI: 0.001-0.793, P = 0.036), and PNI (OR: 0.014, 95%CI: 0.000-0.734, P = 0.035) were associated with greater efficacy of targeted combination immunotherapy (Table 2). The results indicate that the combination of radiomics and laboratory indicators is beneficial for predicting the effectiveness of targeted combined immunity.

Table 2 Univariate and multivariate analyses of factors predicting therapeutic efficacy (n = 56).

Clinicopathologic variable	OR	95%CI	P value
Univariate analysis
Age	5.308	0.594-47.429	0.135
HBV infection	2.280	0.675-7.705	0.185
AFP	1.000	1.000-1.000	0.433
Vascular invasion	0.680	0.231-2.001	0.484
Ascites	0.918	0.315-2.679	0.876
RER	0.029	0.006-0.128	0.000a
ALBI	1.400	0.070-28.120	0.826
NLR	26.600	6.314-112.054	0.000a
PLR	7.222	2.148-24.288	0.001a
PNI	3.667	1.108-12.137	0.033a
Multivariate analysis
NLR group (< 5 vs ≥ 5)	21.765	1.231-384.877	0.036a
PLR group (< 300 vs ≥ 300)	33.857	1.261-909.284	0.036a
PNI group (< 45 vs ≥ 45)	73.367	1.362-3951.454	0.035a
RER group (< 9 vs ≥ 9)	187.23	2.007-17464.635	0.024a

^aP < 0.05. OR: Odds ratio; CI: Confidence interval; HBV: Hepatitis B virus; RER: Relative enhancement ratio; ALBI: Albumin-bilirubin; NLR: Neutrophil to lymphocyte ratio; PLR: Platelet to lymphocyte ratio; PNI: Prognostic nutrition index.

Receiver operating characteristic curves analyses of RER in predicting efficacy of TKIs combined with PD-1 immunotherapy

The area under the curve of the RER for predicting the efficacy of TKIs combined with PD-1 in patients with aHCC was 0.876 (95%CI: 0.781-0.971, P < 0.05), optimal cut-off value was 0.904, diagnostic sensitivity was 87.5%, and specificity was 79.2% (Figure 3). The results indicate that radiomics has good sensitivity and specificity in predicting the efficacy of target immunotherapy.

Figure 3 Receiver operating characteristic curve of relative enhancement ratio. Receiver operating characteristic curve analysis relative enhancement ratio was used to predict the efficacy of tyrosine kinase inhibitors combined with programmed death-1 in patients with advanced hepatocellular carcinoma (area under receiver operating characteristic curve = 0.874). ROC: Receiver operating characteristic.

RER, PNI, PLR, and NLR were associated with PFS in patients taking TKIs combined with immunotherapy

The median PFS of patients with highly-enhanced hepatobiliary HCC nodules was 6.04 ± 2.85 months, and that of patients with low-enhanced HCC nodules was 4.43 ± 2.40 months (P < 0.05). Kaplan-Meier analysis showed that NLR < 5, PLR < 300, PNI < 45, and RER < 0.9 significantly improved the PFS of patients (Figure 4).

Figure 4 Kaplan-Meier survival estimates of progression-free survival. A: Comparison of patients with relative enhancement ratio < 0.9 and relative enhancement ratio ≥ 0.9; B: Comparison of patients with neutrophil-to-lymphocyte ratio < 5 and neutrophil-to-lymphocyte ratio ≥ 5; C: Comparison of patients with platelet-to-lymphocyte ratio < 300 and platelet-to-lymphocyte ratio ≥ 300; D: Comparison of patients with prognostic nutritional index < 45 and prognostic nutritional index ≥ 45. RER: Relative enhancement ratio; NLR: Neutrophil-to-lymphocyte ratio; PLR: Platelet-to-lymphocyte ratio; PNI: Prognostic nutritional index.

DISCUSSION

Imaging omics is an emerging field that uses various predefined image representation algorithms to convert standard medical imaging into high-throughput, discoverable, and quantitative features. With the aid of computer technologies, massive image features can be extracted from medical images. By applying statistical and artificial intelligence learning methods, valuable imaging features can be identified, and potential information can be uncovered through the analysis of image data. These features are then summarized into mathematical rules for clinical information analysis. The main applications of this approach include tissue qualitative research, tumor gene analysis, clinical grading and staging, efficacy evaluation, and prognosis assessment[31]. Medical images contain information that directly reflects the overall tumor load and each tumor lesion[32]. Thus, imaging omics can be viewed as a “digital biopsy” of the entire tumor, providing a noninvasive, repeatable, and comprehensive view of tumor biology and heterogeneity without the need for additional blood or tissue samples. In HCC, imaging omics can improve the diagnosis, prognosis, and prediction accuracy of the disease. Currently, imaging omics models are widely used in the diagnosis and differential diagnosis of HCC, preoperative evaluation, prediction of microvascular invasion, and recurrence prediction. Imaging omics can also be used to predict the responses of HCC to TACE, rectal and breast cancers to neoadjuvant therapy, lung cancer to systemic therapy, and solid tumors to immune checkpoint inhibitors[33].

The degree of enhancement of hepatobiliary nodules on Gd-EOB-DTPA-enhanced MRI is significantly correlated with resistance to the targeted combination immunotherapy, making it an ideal predictive imaging biomarker for treatment efficacy. In two recent studies, tumor radiological features were associated with response to anti-PD-1 antibodies or lenvatinib monotherapy plus TACE in aHCC[29]. The first study proposed a radiological model based on contrast-enhanced computed tomography (CT) to predict response to anti-PD-1 antibody monotherapy from a single center with a relatively small sample size (n = 58)[30]. However, MRI, specifically contrast-enhanced, remains the preferred method for evaluating systemic therapeutic response in HCC, as it has several advantages over CT, including excellent imaging resolution, absence of ionizing radiation, and superior soft tissue contrast compared to CT[34]. In the second study, tumor radiological features extracted from pre-treatment MRI predicted disease progression after lenvatinib monotherapy combined with TACE. This was also a single-center study with a small sample size (n = 61), and the area under the curve was 0.71, indicating moderate discriminating power[35]. In addition, the tumor radiological features extracted from pre-treatment MRI can provide incremental predictive value for clinical clinicopathological features. These features are related to overall survival (OS) and PFS and can effectively predict the ORR of lenvatinib plus anti-PD-1 antibody[36]. In the present study, we developed a simple radiomics method based on pre-processed MRIs to predict the ORR and PFS of patients with unresectable HCC to TKIs combined with immunotherapy. Our study found that NLR, PLR, and PNI predicted the efficacy of TKIs combined with immunotherapy and were associated with PFS. To the best of our knowledge, this represents a novel method for predicting the efficacy of combination immunotherapy in patients with unresectable liver cancer.

Radiological features of aHCC tumors are associated with the efficacy of PD-1 antibody therapy, lenvatinib monotherapy, and TACE[30,35]. Specifically, radiomics can be used to identify details and extract features that cannot be captured or quantified by the naked eye on MRI, accurately reflecting the biology of HCC. Compared with clinical pathological features, radiomics provides more detailed tumor information and more sensitive treatment response predictions[36]. Moreover, the enhancement of arterial phase tumors before immunotherapy is related to tumor progression[37]. Notably, while models based solely on clinicopathological features cannot satisfactorily predict objective responses, radiological models can.

A previous study showed for the first time that NLR < 3 is a favorable factor for OS in patients with HCC treated with lenvatinib[38]. In this study, NLR < 3 was associated with OS in both the training and validation cohorts of patients treated with lenvatinib and anti-PD-1 antibody, but not with PFS. Another study found that high NLR (≥ 2.92) and PLR (≥ 128.1) are useful prognostic factors for predicting the prognosis of patients undergoing HCC resection. A clinical study has shown that, compared to using the NLR or PNI scores alone, the combination of the two can better reflect the systemic inflammatory response in liver cancer patients after TACE. Through univariate and multivariate analyses, NLR < 5, PLR < 300, PNI < 45, and RER < 0.9 significantly improved the PFS of patients (Figure 4). Compared to existing radiomics research, our study proposes a new method for predicting the efficacy of targeted drug combined with immunotherapy in HCC patients by combining imaging information with laboratory test results. The advantage of this evaluation method is that it is relatively simple to apply and combines patient biological indicators to improve prediction efficiency. Our research results indicate that when patients meet one of the following criteria, NLR < 5, PLR < 300, PNI < 45, and RER < 0.9, the effectiveness of target immunotherapy is significantly increased, and disease-free survival is also prolonged.

This study has some limitations. First, the sample size was limited, meaning that there was a potential selection bias, which affects the extrapolation of our conclusions. Further research is needed to expand the sample size based on existing data. Owing to the heterogeneity of patients with HCC[39], the proposed radiomics model may not be applicable to all patients with HCC. Further research is needed to compare the differences in radiological characteristics among patients with HCC with different etiologies. Additionally, as this was a retrospective study, tumor tissues were not prospectively collected to explore the biological significance of the radiological features. Meanwhile, this leads to significant systemic bias in the research. The manual segmentation of tumors is also relatively subjective, particularly for tumors with blurred edges[40], which leads to bias when analyzing image information. Additionally, different TKI-drugs were used in the study population, which may have increased the internal variability within the limited sample.

CONCLUSION

This study suggests that tumor radiological features extracted from preprocessed MRI images can be used to predict the objective response of patients with unresectable or aHCC to TKIs combined with PD-1 antibody therapy. These may provide greater predictive value, compared with clinical pathological features related to PFS after treatment initiation.

ACKNOWLEDGEMENTS

We would like to thank Dr. Peng-Bo Zhao (Zhengzhou University) and Dr. Yu-Zhu Wang (Zhengzhou University) for their support for this project.