Cancer-associated fibroblasts, clinicopathological characteristics and prognosis of liver cancer: A systematic review and meta-analysis based on real-world research

Abstract

BACKGROUND

Cancer-associated fibroblasts (CAFs), crucial components of the tumor microenvironment in primary and metastatic tumors, can impact the activity of cancer cells and contribute to their progression. Given their extensive interactions with cancer cells and other stromal cells, we aimed to evaluate the prognostic value of CAFs in patients with liver cancer (LC).

AIM

To investigate the association between CAF expression and clinicopathological characteristics as well as overall survival (OS) in patients with LC, including hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (iCCA).

METHODS

We performed a meta-analysis of cohort studies with available data on the effects of CAF expression on both clinicopathological characteristics and OS via hazard ratios (HRs) and risk ratios with 95% confidence intervals (CIs). Studies were subgrouped on the basis of CAF markers and cancer type, and the subgroup effects of CAF expression on both HCC and iCCA were analyzed through meta-regression. The Newcastle-Ottawa Scale was used to evaluate the included studies to guarantee their quality and minimize the possibility of bias.

RESULTS

Nine trials were selected and included a total of 1518 patients. According to our primary meta-analysis, the expression of CAFs in LC patients was significantly associated with a decrease in OS (LC: HR: 1.62; 95%CI: 1.34-1.97; P < 0.001; HCC: HR: 1.67; 95%CI: 1.34-2.07; P < 0.001; iCCA: HR: 1.47; 95%CI: 0.97-2.23; P = 0.07); nevertheless, it was not significantly associated with almost all clinicopathologic characteristics, including tumor size, venous infiltration, alpha-fetoprotein level, and differentiation grade. According to the subgroup analysis of smooth muscle actin (SMA) markers in both HCC patients and iCCA patients, high CAF expression in HCC (HR: 2.29; 95%CI: 1.01-5.22; P = 0.048) and iCCA (HR: 2.04; 95%CI: 1.09-3.81; P = 0.025) patients was a significant indicator of poor OS. Moreover, the clinicopathological characteristics were also verified by the SMA marker, which had a nearly significant effect on the venous infiltration of iCCA (risk ratio: 2.70; 95%CI: 0.97-7.49; P = 0.057).

CONCLUSION

High CAF expression, evaluated by both mixed markers and SMAs, is significantly associated with poor OS in patients with LC, including both HCC patients and iCCA patients. However, further research is necessary since how CAF expression and clinicopathologic features are related is yet unknown.

Key Words: Cancer-associated fibroblasts; Liver neoplasms; Prognosis; Liver cancer; Meta-analysis

Core Tip: This meta-analysis reveals that high expression of cancer-associated fibroblasts (CAFs), particularly marked by α-smooth muscle actin, is significantly associated with poor overall survival in liver cancer, including hepatocellular carcinoma and intrahepatic cholangiocarcinoma. CAFs serve as a robust prognostic biomarker, with stronger predictive value in hepatocellular carcinoma. While their impact on clinicopathological features remains inconclusive, these findings highlight CAFs as potential therapeutic targets. Further research is needed to clarify their role in tumor progression and optimize clinical applications.

INTRODUCTION

Liver cancer (LC) ranks as the fourth leading cause of mortality worldwide, resulting in more than 800000 deaths per year[1,2]. Hepatocellular carcinoma (HCC) accounts for almost 90% of all primary LCs, with intrahepatic cholangiocarcinoma (iCCA) and other primary liver malignancies being less common. There has been significant progress in the study and comprehension of all facets of LC, which encompasses HCC and iCCA. New developments in laboratory technologies have successfully identified cellular abnormalities that can ultimately lead to precise categorization and appropriate therapy for patients diagnosed with LC. The ability of tumors to become cancerous is believed to be solely attributed to cancer cells, as stromal cells are not undergoing differentiation or activated proliferation[3,4]. Stromal cells were formerly believed to enclose only cancer cells and play a noncancerous role. Recent research has provided a clearer understanding of the origins, characteristics, and functions of stromal cells[5-7].

The morphology of fibroblasts found in cancerous tissues is comparable to that of myofibroblasts triggered during the healing process of wounds[8]. Recent research has demonstrated the importance of communication between cancer cells and a specific type of fibroblast known as cancer-associated fibroblasts (CAFs)[9]. CAFs are mesenchymal cells that are characterized by their enormous spindle shape and positive immunostaining for vimentin, alpha smooth muscle actin (SMA), and established fibronexus[10,11]. According to previous research, CAFs can enhance the advancement and spread of cancer by releasing a range of soluble substances, including inflammatory cytokines, growth factors, and chemokines[12,13]. CAFs trigger the malignant transformation of human breast cells through the secretion of transforming growth factor-β and hepatocyte growth factor[14]. In addition, CAFs have been found to enhance the growth and invasive characteristics of HCC cells[15]. Nevertheless, the specific effects of CAFs in HCC and iCCA are not fully understood.

Understanding the expression of CAFs and their predictive significance in LC could be particularly valuable in the context of treatment. Several studies have revealed that stimulated fibroblasts secrete significant quantities of extracellular matrix proteins, particularly type I collagen, resulting in the development of liver fibrosis[16]. The presence of long-lasting inflammation in chronic hepatitis significantly contributes to the development of HCC[17-19]. Therefore, the increased presence of CAFs is likely linked to the progression of colorectal cancer, indicating a correlation between CAFs and unfavorable prognosis in individuals with colorectal cancer. In this study, we conducted a meta-analysis using existing data to address this uncertainty and address the variation in the predictive significance of CAF expression in LC among different studies.

MATERIALS AND METHODS

Research question

This meta-analysis of real-world research was performed according to PRISMA guidelines for meta-analyses and systematic reviews. This analysis was based on data from previous studies registered on PROSPERO (CRD42024540443)[20].

Search strategy

A literature search of published articles related to the expression level of CAFs in patients with LC was performed in PubMed, EMBASE, MEDLINE and Cochrane on July 21, 2025. The search strategy was designed by integrating the different expression levels of CAFs and the required prognostic indicators. The detailed keywords used in the search are listed in Supplementary Table 1.

Selection criteria

All the articles were independently screened by three investigators in the search for appropriate works in the literature for further analysis. The inclusion criteria were as follows: (1) Original articles that investigated the association between the expression of the referred index and the prognosis of LC patients; (2) Articles written in the English language; and (3) Articles with full text. In addition, reviews, meta-analyses, case reports, conference abstracts, correspondence letters, and studies without endpoint parameters were excluded. The main exclusion criterion was the absence of hazard ratios (HRs) and/or risk ratios (RRs) of CAF expression for clinicopathological characteristics and survival outcomes.

Quality assessment

To ensure that all the studies included in our meta-analysis were of high quality, we used a set of predefined criteria based on the Newcastle-Ottawa Scale (NOS) criteria[21] to evaluate the studies independently by 2 investigators in parallel. The criteria of the NOS include three aspects: (1) Selection: 0-4; (2) Comparability: 0-2; and (3) Clinical outcome: 0-3. The total scores of the NOS range from 0 (lowest) to 9 (highest). According to the NOS, the included studies were classified into two levels: Low-quality research, with a score of 0-5, and high-quality research, with a score of 6-9. A score of 6 or more was considered the inclusion criterion through discussion and consultation among all the investigators.

Data extraction

The general information of each article, patient characteristics, and endpoint parameters were extracted. The general information included the name of the first author, year of publication, country or area of study, study design, and statistical methods (univariate vs multivariate). Patient characteristics included the total number of patients, sex ratio, median follow-up time and number of positive/high and negative/Low cases. Various labeled antibodies, such as SMA, fibroblast activation protein and other proteins, were used as markers of CAFs, which we also extracted from all the studies. The primary outcome was the prognosis index, and the secondary outcome included a series of clinicopathological characteristics. The survival data from multivariate analysis were preferred when both multivariate analysis and univariate analysis were provided. If the study included only Kaplan-Meier survival curves, survival data were extracted using Engauge Digitizer (v4.1). For each curve, two investigators independently performed digitization, and each time point was extracted twice to minimize random error. The average of the two readings was used for analysis. Any discrepancies between investigators greater than 5% were rechecked and resolved by consensus. Extracted values were further cross-validated with reported survival rates when available, ensuring consistency and reliability.

Statistical analysis

To evaluate the associations between the expression of CAFs and the survival outcomes of LC patients, pooled HRs with 95% confidence intervals (CIs) were evaluated for overall survival (OS). The clinicopathological characteristics were evaluated by RRs with 95%CIs. In terms of partial studies focusing on specific markers and cancer types, subgroup analyses of SMA, HCC, and iCCA were conducted. The heterogeneity and publication bias among the different studies were estimated via the I-square test and funnel plot. If I² < 50% and P > 0.05 were tested, the studies were considered indicators of mild heterogeneity. For mild heterogeneity, a fixed-effects model with inverse variance was utilized; otherwise, a random-effects model was used when I² > 50% and P < 0.05. The publication bias was tested via a funnel plot along with Egger’s test. In our study, two sided P values less than 0.05 were considered statistically significant. Analysis of the effect size of all the data were performed with 11.1 Engauge Digitizer and 12.0 Stata software.

RESULTS

Selection

We retrieved a total of 1507 articles through a literature search, including 455 from PubMed and 1052 from other databases, such as the Cochrane Library and Embase. After the duplicate screening, 488 articles were excluded. From the remaining articles, 404 nonreal-world studies, 162 experimental design studies, 167 reviews and 25 articles that were not relevant to LC were excluded on the basis of the title and abstract reading. We then conducted a full-text reading and eligibility assessment of the remaining 442 articles, with two reviewers independently reviewing and a third reviewer conducting a rereview of uncertain articles. Finally, we included a total of 10 articles. The inclusion process is shown in Figure 1. Twelve cohort studies were ultimately identified that met the inclusion criteria (Table 1)[22-31]. Overall, a total of 1518 patients were analyzed, and all the pooled results are shown in the following text (Table 2).

Figure 1  Flowchart and selection process of eligible studies.

Table 1 Summary of included studies in meta-analysis.

Ref.	Country	Study type	Total (Male:Female:Unknown)	Follow-up (months)	NOS score	Sample	Cut-off value	Detecting methods	Index (ECM target)	Number (High:Low) (n)	Tumor size (cm) (> 5)	High AFP (> 200)	Venous infiltration	TNM stage (advanced)	Differentiation	Prognosis	Analysis type
Fang et al[25], 2018	China	Retrospective cohort	90:11	41	7	HCC tissue	The mean value of thickness of the stroma > 238.6 μm	IHC staining	α-SMA	52:48	41:33	29:18	23:24	25:28	NA	NA	NA
Eguchi et al[24], 2023	Japan	Retrospective cohort	49:22	NA	7	iCCA resected specimens	no	IHC staining	α-SMA	22:49	NA	NA	17:27	15:27	NA	OS	Multivariate
Cioca et al[23], 2017	Romania	Cross-sectional study	50:22	NA	6	HCC tissue	no	IHC staining	Podoplanin	23:49	9:17	NA	6:19	5:21	NA	NA	NA
Lan et al[26], 2022	Japan	Retrospective cohort	88:77	34.8	7	iCCA	Positive (> 0)	IHC staining	CAV1	78:80	NA	NA	NA	NA	28:23	OS	Multivariate
Zhang et al[30], 2017	China	Retrospective cohort	33:38	1.3-63.2	7	iCCA	IHC staining score ≥ 2	IHC staining	α-SMA	45:26	27:12	NA	4:1	23:12	20:5	OS	Multivariate
							IHC staining score ≥ 2	IHC staining	FSP-1	60:11	33:7	NA	5:0	32:3	24:1	OS	Univariate
							IHC staining score ≥ 2	IHC staining	PDGFRβ	56:15	29:10	NA	4:1	29:6	19:6	OS	Univariate
Amer et al[22], 2022	Egypt	Retrospective cohort	47:13	NA	6	HCC tissue	NA	IHC staining	Podoplanin	41:19	37:15	NA	3:8	NA	NA	NA	NA
Guan et al[31], 2018	China	Retrospective cohort	105	NA	5	HCC tissue	30% positive staining	IHC staining	Cav-1	NA	NA	NA	NA	NA	NA	OS	Univariate
Peng et al[27], 2022	China	Retrospective cohort	253:121	NA	6	HCC tissues and normal tissues	According to the median expression value of RAB42	IHC staining	RAB42	186:187	NA	66:51	NA	121:139	NA	OS	Univariate
Peng et al[28], 2022	China	Retrospective cohort	253:122	NA	6	HCC tissues and normal liver tissues	According to the median expression value of RAB6B	IHC staining	RAB6B	186:187	NA	59:58	NA	125:135	NA	OS	Univariate
Wang et al[29], 2016	China	Retrospective cohort	124	60	5	HCC tissues	NA	IHC staining	α-SMA	NA	NA	NA	NA	NA	NA	OS	Multivariate

NOS: Newcastle-Ottawa Scale; ECM: Extracellular matrix; AFP: Alpha-fetoprotein; TNM: Tumor-node-metastasis; HCC: Hepatocellular carcinoma; IHC: Immunohistochemistry; α-SMA: Α-smooth muscle actin; NA: Not available; iCCA: Intrahepatic cholangiocarcinoma; OS: Overall survival; CAV1: Caveolin 1; FSP-1: Fibroblast specific protein 1; PDGFRβ: Platelet-derived growth factor receptor; RAB42: Ras-related protein rab-42; RAB6B: Ras-related protein Rab-6B.

Table 2 All pooled results in meta-analysis.

Marker	Outcome	Cancer type	Number of cohorts	Pooled effect size		heterogeneity		Publication bias with Egger’s test (P value)
				OR/HR (95%CI)	P value	P value	I2 (%)
All markers	Tumor size (cm) (> 5)	LC	6	1.27 (0.81-1.99)	0.3	0.503	< 0.01%	0.832
		HCC	3	1.60 (0.86-2.96)	0.138	0.734	< 0.01%	0.506
		iCCA	3	0.98 (0.51-1.89)	0.949	0.275	22.50%	0.284
	High AFP (> 200)	HCC	3	1.32 (0.99-1.76)	0.061	0.257	26.50%	0.383
	Venous infiltration	LC	6	1.06 (0.63-1.77)	0.84	0.376	6.30%	0.416
		HCC	2	0.70 (0.37-1.33)	0.28	0.608	< 0.01%	NA
		iCCA	4	2.31 (0.95-5.61)	0.065	0.911	< 0.01%	0.355
	TNM stage (advanced)	LC	8	0.81 (0.63-1.05)	0.109	0.162	33.40%	0.061
		HCC	4	0.68 (0.51-0.90)	0.007	0.662	< 0.01%	0.278
		iCCA	4	1.66 (0.95-2.90)	0.077	0.776	< 0.01%	0.052
	Low differentiation	iCCA	4	1.62 (0.98-2.67)	0.061	0.169	40.50%	0.244
	OS	LC	9	1.62 (1.34-1.97)	< 0.001	0.03	52.90%	0.256
		HCC	5	1.67 (1.34-2.07)	< 0.001	0.022	65.10%	0.649
		iCCA	4	1.47 (0.97-2.23)	0.07	0.153	43.10%	0.402
SMA	Venous infiltration	LC	3	1.25 (0.67-2.34)	0.479	0.176	42.50%	0.544
		HCC	1	0.79 (0.36-1.74)	0.564	NA	NA	NA
		iCCA	2	2.70 (0.97-7.49)	0.057	0.921	< 0.01%	NA
	TNM stage (advanced)	LC	3	1.01 (0.60-1.72)	0.96	0.321	12.00%	0.098
		HCC	1	0.66 (0.30-1.46)	0.305	NA	NA	NA
		iCCA	2	1.44 (0.70-2.93)	0.321	0.624	< 0.01%	NA
	OS	LC	3	2.13 (1.30-3.51)	0.003	0.969	< 0.01%	0.652
		HCC	1	2.29 (1.01-5.22)	0.048	NA	NA	NA
		iCCA	2	2.04 (1.09-3.81)	0.025	0.902	< 0.01%	NA

LC: Liver cancer; HCC: Hepatocellular carcinoma; iCCA: Intrahepatic cholangiocarcinoma; AFP: Alpha-fetoprotein; TNM: Tumor-node-metastasis; OS: Overall survival; OR: Odds ratio; HR: Hazard ratio; CI: Confidence interval; NA: Not available.

Primary outcome

A total of 7 studies evaluated whether the number of CAFs influences survival outcome with respect to HR in LC patients, including 5 studies associated with HCC and 2 studies associated with iCCA. Among these studies, we also found that more than one cohort existed in some studies, which were then analyzed separately. Our analysis revealed a significantly greater risk of OS in LC patients with high levels of CAF expression (HR: 1.62; 95%CI: 1.34-1.97; P < 0.001; Figure 2). However, substantial heterogeneity was also detected in the pooled HRs of OS (I² = 52.9%, P = 0.03). Regarding the potential discrepancy between HCC and iCCA, we analyzed the prognostic impact of CAF expression on different cancer types (HCC: HR: 1.67; 95%CI: 1.34-2.07; P < 0.001; iCCA: HR: 1.47; 95%CI: 0.97-2.23; P = 0.07; Figure 2), suggesting that only HCC patients with high CAF expression may have a greater risk of poor prognosis.

Figure 2 Pooled hazard ratios of cancer-associated fibroblast expression for overall survival in liver cancer patients. CAF: Cancer-associated fibroblast; HCC: Hepatocellular carcinoma; iCCA: Intrahepatic cholangiocarcinoma; HR: Hazard ratio; CI: Confidence interval.

Secondary outcome

To explore whether high CAF expression leads to poor clinicopathologic characteristics in LC patients, we further evaluated the associations between the expression of CAF-associated markers and a series of clinicopathologic indices, including tumor size, venous infiltration, tumor-node-metastasis (TNM) stage, alpha-fetoprotein level, and differentiation grade. Our analysis included a total of 9 studies and extracted each required cohort. However, the analysis revealed a significant association between only the number of CAFs and advanced TNM stage in HCC patients (Figure 3) (RR: 0.68; 95%CI: 0.51-0.90; P = 0.007; Figure 3C). The results also revealed considerable heterogeneity with respect to OS (I² < 0.01%, P = 0.662).

Figure 3 Pooled risk ratios of cancer-associated fibroblast expression for clinicopathological characteristics in liver cancer patients. A: Tumor size; B: Venous infiltration; C: Tumor-node-metastasis stage; D: Alpha-fetoprotein level; E: Tumor differentiation. CAF: Cancer-associated fibroblast; LC: Liver cancer; HCC: Hepatocellular carcinoma; iCCA: Intrahepatic cholangiocarcinoma; OR: Odds ratio; CI: Confidence interval.

Subgroup analysis

Our study focused on the characteristics of various markers and their ability to represent the expression of CAFs. Among the markers used in all the studies, only SMAs were evaluated more than two times. To further evaluate the impact of different markers, we performed subgroup analyses for SMA with a total of 3 cohorts via multivariable Cox regression analysis (Figure 4). The results revealed that SMA could predict poor OS in LC patients (HR: 2.13; 95%CI: 1.30-3.51; P = 0.003; Figure 4C). Similarly, the association was significant in both HCC (HR: 2.29; 95%CI: 1.01-5.22; P = 0.048; Figure 4C) and iCCA (HR: 2.04; 95%CI: 1.09-3.81; P = 0.025; Figure 4C) patients. In terms of HR, the SMA may be a better marker of CAFs for evaluating the risk of survival in LC patients. The test for heterogeneity revealed a negative result (I² < 0.01%, P = 0.969), indicating that there was no significant heterogeneity in the results. Moreover, there was no significant subgroup heterogeneity. With respect to clinicopathologic aspects, we analyzed only venous infiltration and TNM stage, owing to the limited number of studies assessing the expression of SMA. However, we found that SMA expression was not significantly associated with venous infiltration, and only iCCA patients had a nearly significant association with venous infiltration (RR: 2.70; 95%CI: 0.97-7.49; P = 0.057; Figure 4A). These findings indicate that CAF expression cannot predict the clinicopathologic characteristics of LC patients.

Figure 4 Forest plots of the impact of smooth muscle actin expression on clinicopathological characteristics and overall survival in liver cancer patients. A: Venous infiltration; B: Tumor-node-metastasis stage; C: Overall survival. SMA: Smooth muscle actin; LC: Liver cancer; HCC: Hepatocellular carcinoma; iCCA: Intrahepatic cholangiocarcinoma; OR: Odds ratio; CI: Confidence interval.

Publication bias and study quality assessment

All included studies were subjected to bias risk assessment following the NOS criteria (NOS score shown in Table 1) and received the prespecified quality threshold (a score of 5) via independent assessment and discussion, affirming the stability of the pooled results. We also performed a sensitivity analysis randomly to assess the potential impact of each study on the overall results, which indicated that certain studies exerted a notable influence on the associations between CAF expression and both clinicopathological characteristics and prognostic outcomes in LC patients (Supplementary Figures 1-9).

Through the heterogeneity test, we choose random effect models if I² > 50%. The heterogeneity and publication bias in our study were assessed via funnel plots and Egger’s test, which were performed with Stata software. Funnel plots were used to assess the heterogeneity for all the results, which showed relatively significant asymmetry in our studies. According to Egger’s test, the publication bias of all our results was assessed with both data (Table 2) and plot presentations, which also revealed no publication bias in any of our analyses.

DISCUSSION

LC is a prevalent form of cancer that is among the most common malignant tumors. It ranks sixth in terms of occurrence and fourth in terms of death[32]. LC can be classified into HCC and cholangiocarcinoma (CCA) on the basis of the specific source of cancerous cells. Because the majority of cases are detected at a late stage, the outlook for LC is unfavorable. The growth of a tumor is influenced by complex and dynamic interactions between different biological components inside the tumor microenvironment (TME)[33]. CAFs, which are the most plentiful elements of the tumor stroma, have been implicated in the progression of LC. The interaction between CAFs and tumor cells, immune cells, or vascular endothelial cells in the TME, either through direct physical contact or indirect chemical signaling, influences the onset and progression of malignancies[34].

Our results revealed a significant association between CAF expression and poor survival outcomes in LC patients, especially when SMA biomarkers were used. In addition, the impact of CAFs on the prognosis of HCC patients is greater than that of iCCA patients, suggesting that CAFs play a more important role in HCC progression. Many studies have demonstrated that CAFs play crucial roles in controlling various aspects of tumor cell behavior, including proliferation, invasion, migration, and metastasis[35]. CAFs play a role in enhancing tumor chemoresistance and recurrence in the treatment of HCC and establish tight communication with HCC cells through the secretion of various types of cytokines, growth factors, or extracellular vesicles, either through direct or indirect means. Unlike HCC, CCA is characterized primarily by extensive infiltration of desmoplastic stroma within the tumor. This dense stroma in CCA is caused mainly by the presence of CAFs[36]. Curiously, in contrast to other types of solid tumors, such as breast and pancreatic cancer, where a large amount of stroma is linked to negative patient outcomes, in the case of ICC, patients with a high proportion of the stroma area experience better disease-free survival[37]. These findings suggest that the desmoplastic stroma may have a protective effect. These findings may provide insight into why CAFs play a significant role in promoting tumor growth in HCC.

However, we did not find an obvious effect of CAF expression on any clinicopathological characteristic. Although there was a significant association between the expression of CAFs and advanced TNM stage in HCC patients, this association was not verified by SMA markers. Moreover, we found some opposite trends in HCC patients, since high CAF expression decreased venous infiltration and TNM stage progression. Although these results were negative, we assumed that CAFs had an impact on clinicopathologic features, as CAFs are involved in several aspects of the progression of LC, such as tumor angiogenesis. HCC is a tumor characterized by excessive blood vessel formation (hypervascularization), which promotes the spread of tumor cells, invasion, disease recurrence, and metastasis. On the one hand, CAFs release vascular endothelial growth factor and placental growth factor in nearby tumor areas to stimulate cell division[38]. On the other hand, specific CAF subsets can also have contrasting effects on tumor blood vessels through the secretion of proline and SPARCL1[39,40]. Unlike HCC, CCA is a tumor with a reduced blood supply, and the dense tissue formed by numerous CAFs causes the blood vessels in the tumor to collapse, resulting in a low-oxygen environment. In both HCC and CCA, CAFs have comprehensive effects on tumor angiogenesis but have different tendencies, which also confirmed our results. Our analysis revealed that CAFs have significant effects on the clinicopathologic features of patients with LC if more studies can be performed.

α-SMA is commonly expressed by several subgroups of CAFs and is typically utilized as a marker to detect CAFs in LC[9]. In a prior study, retrieved CAFs were identified via immunofluorescence labeling of α-SMA and vimentin, allowing observation of the morphological features of CAFs. Hence, we performed separate analyses using SMA markers to evaluate the impact of CAFs. On the basis of our results, we confirmed that CAFs promote poor survival outcomes in LC patients, including HCC and iCCA patients. Nevertheless, clinicopathologic characteristics, such as venous infiltration and TNM stage progression, were still not significantly related to CAFs. Interestingly, from the pooled effect size, we found opposite impacts of CAFs on venous infiltration and TNM stage progression in patients with HCC and iCCA, which was consistent with the findings of previous analyses. We suppose that the association exists as long as enough qualified studies are analyzed. In the future, additional markers, such as fibroblast specific protein 1 and platelet-derived growth factor receptor, can be utilized to detect CAFs in CCA. Among these markers, fibroblast specific protein 1 has the highest expression rate in CAFs, with a prevalence of 84.5%[30].

Recent advances have highlighted the clinical relevance and mechanistic complexity of CAFs in LC progression. Spatial multiomics analysis has identified functionally distinct CAF subpopulations, such as the F5-CAF cluster, which colocalize with cancer stem-like cells and are enriched in aggressive HCC, correlating with poor prognosis and stemness maintenance[41]. In iCCA, CAF-derived interleukin-6 (IL-6) has been shown to suppress autophagy and reduce chemotherapeutic efficacy, supporting its dual role in both prognosis and therapy resistance[42]. Further evidence has demonstrated that CAFs promote epithelial-mesenchymal transition in HCC via the IL-6/IL-6R/signal transducer and activator of transcription 3/tissue transglutaminase 2 axis, facilitating metastasis and immune escape[43]. Moreover, transcriptome-based risk stratification models built from CAF-related gene signatures have proven effective in predicting patient outcomes, drug sensitivity, and immune checkpoint expression patterns in HCC cohorts, indicating their potential for integration into precision medicine frameworks[44]. Our findings, by quantitatively validating the negative prognostic role of CAFs, particularly α-SMA + CAFs, reinforce their value not only as biomarkers but also as therapeutic targets. Future clinical trials incorporating CAF profiling could improve risk stratification, inform combinatorial treatment strategies, and accelerate the development of CAF-targeted therapies in LC.

Moreover, the early-stage enrichment of certain CAF subtypes may reflect their role in establishing a protumorigenic niche that favors tumor initiation rather than dissemination. As the disease progresses, other CAF subsets, such as inflammatory CAFs or myofibroblastic CAFs, may dominate and facilitate immune suppression, angiogenesis, or resistance to therapy. These dynamic shifts in CAF composition over time could help reconcile the observed inverse relationship between TNM stage and survival outcomes. Recent spatial multiomics studies have shown that CAFs are not a uniform population but rather comprise distinct subtypes with divergent roles in tumor progression. For example, Ge et al[45] identified a subgroup of CD36⁺ CAFs in HCC that promote tumor cell differentiation and suppress epithelial-mesenchymal transition, thereby potentially restraining local invasion and early stage progression. However, the same population also remodeled the extracellular matrix and secreted proinflammatory factors, contributing to long-term immune evasion and metastasis - mechanisms that may ultimately worsen patient survival.

Several limitations in this meta-analysis need discussion. Given that the number of included studies is moderate in scale, the conclusion that CAFs serve as a prognostic biomarker should be interpreted with caution. Furthermore, due to the retrospective nature of this study, the grade of evidence is inferior than that of randomized controlled trials. The high heterogeneity observed in some analyses may be attributed to the following factors. First, patient characteristics such as age, sex, tumor grade, comorbidities, and treatment regimens were not fully adjusted for. Second, immunohistochemical detection methods and cutoff values varied across different trials. Third, studies investigating the association between CAF expression in different tissues and the prognosis of LC patients remain insufficient. Similarly, research exploring the influence of CAFs on the clinicopathological characteristics of LC patients is relatively scarce. Importantly, not only do CAFs have prognostic value in LC, but other factors in the TME, including the risk factors PXDNL (Peroxidasin Like) and LINC02038 and the protective factors SLC27A2, KLRB1, IGHV1-12 and IGKV1OR2-108, are also potential prognostic markers[46]. CAFs are overexpressed in LC; thus, exploring related inhibitors may be promising for the treatment of LC.

CONCLUSION

In conclusion, the present meta-analysis revealed a significant association between high expression of CAFs and poor prognosis in patients with LC, especially in patients with HCC. Although the same predictable effects on clinicopathologic characteristics were expected, we can only presume this trend without statistical confirmation. Given the paucity of studies included, larger samples and well-designed prospective confirmations of these findings are warranted.

ACKNOWLEDGEMENTS

We would like to acknowledge the participants and investigators of all the included studies. We are grateful to all the researchers and their participants.