High Density of Cancer-Associated Fibroblasts Correlates with Increased Lymphangiogenesis and Poor Prognosis in Hypopharyngeal Carcinoma

Abstract

Objective

This study aimed to investigate the relationship between Cancer-associated fibroblasts (CAFs), microlymphatic vessel density (MLVD), clinicopathological factors, and survival outcomes in HPSCC patients.

Methods

Clinical data were collected from patients diagnosed with HPSCC in the Department of Otolaryngology and Head and Neck Surgery at the 980th Hospital of PLA Joint Logistics Support Force (Bethune International Peace Hospital) from January 1, 2010, to October 1, 2022. CAFs, MLVD, and related factors were analyzed.

Results

CAF numbers and MLVD were relatively high in tumor tissues, positively correlated with each other, and linked to clinicopathological factors and poor survival. N stage, number of CAFs, and MLVD affected prognosis, with N stage and MLVD identified as independent risk factors.

Conclusions and Significance

The correlation between CAF density and MLVD suggests a potential association between CAFs and lymphangiogenesis in HPSCC. MLVD may serve as a prognostic biomarker, and CAF-associated lymphangiogenesis warrants further investigation.

Introduction

Histologically, hypopharyngeal carcinoma is almost exclusively squamous cell carcinoma and is consistently linked to the most unfavorable prognosis among patients with head and neck cancers.¹ Despite continuous advances in diagnosis and treatment methods, most patients with hypopharyngeal squamous cell carcinoma (HPSCC) are diagnosed, have a poorer prognosis,² when the disease has reached middle or advanced stages (stages III–IV); these patients account for 70–85% of all patients with HPSCC. This is attributable primarily to regional and distant metastasis. Dysfunctional lymphangiogenesis plays a pivotal role in driving tumor pathogenesis^3–5 and contributes substantially to the failure of multiple treatment modalities. Lymphatic metastasis,⁶ a defining feature of HPSCC, is initiated by the invasion of cancer cells into peritumoral lymphatic vessels, followed by their dissemination to regional lymph nodes—the most common and earliest site of metastasis. At the molecular level, lymphangiogenesis is predominantly orchestrated by pro-lymphangiogenic factors,⁴ which bind to and activate their cognate receptors on lymphatic endothelial cells. This activation subsequently triggers downstream signaling cascades, thereby promoting lymphatic endothelial cell proliferation, migration, and survival.

The tumor microenvironment (TME) serves as a critical niche that fosters tumor growth through a complex, interconnected stromal network. Dynamic interactions between the diverse cellular and molecular components of the TME and cancer cells are central to the mechanisms driving tumor development and metastasis.⁷ Cancer-associated fibroblasts (CAFs) play important roles in carcinogenesis, and their density is often associated with unfavorable clinical outcomes. The CAF-associated mechanisms that have been linked to tumor metastasis⁸ include extracellular matrix remodeling, paracrine oncogenic signaling, immune exclusion, angiogenesis, lymphangiogenesis, pre-metastatic niche construction, and treatment resistance. CAFs are characterized by the upregulation of the expression of markers such as α-smooth muscle actin (α-SMA).⁹

CAFs are pivotal mediators of tumor-stroma crosstalk within TME and drive tumor progression.¹⁰ This study seeks to elucidate the role of CAFs in the prognosis and lymphangiogenesis of HPSCC. Our findings are expected to shed light on the clinical course of hypopharyngeal cancer and offer a novel perspective for its therapeutic management.

Materials and Methods

Patient Cohort and Sample Collection

We recruited a cohort of patients with a diagnosis of HPSCC from the Department of Otolaryngology and Head and Neck Surgery at the 980th Hospital of PLA Joint Logistics Support Force (Bethune International Peace Hospital). The recruitment covered a period from January 1, 2010, to October 1, 2022. The enrolled patients were comprehensively reviewed according to the predefined inclusion and exclusion criteria. Patients were included if they had a consistent histopathological diagnosis of HPSCC before and after surgery and had not received any preoperative (neoadjuvant) chemotherapy or targeted agents. The exclusion criteria included patients with synchronous malignant tumors in other sites or recurrent hypopharyngeal cancer requiring reoperation. Additionally, patients whose final pathology was not confirmed as squamous cell carcinoma; who had a history of anticancer treatment; who had chronic diseases such as diabetes, hypertension, and coronary heart disease; who had a documented chronic medication history; and who had fewer than 250 tumor cells in the paraffin block were excluded. Participants lacking complete clinical pathological or follow-up data were also excluded to ensure the robustness of the study. The detailed flow chart is presented in Figure 1. No significant differences were observed between the included and excluded patients with respect to age, sex, or Clinical stage (P>0.05), suggesting that the exclusion process did not introduce substantial selection bias. The surgical management of hypopharyngeal cancer was personalized based on the location of the primary tumor and clinical TNM stage, aiming to minimize the impact of surgery on survival. The survival time was calculated from the date of surgery to the date of death or the last follow-up date.

Figure 1.

Flow diagram of patient enrollment and exclusion. A total of 147 patients were assessed for eligibility. After exclusion of 56 patients for various reasons, 91 patients were included in the final analysis.

Immunohistochemistry (IHC)

Tissue samples (tumor and matched adjacent normal tissues) were fixed in formalin and embedded in paraffin. Consecutive sections were prepared, dried, deparaffinized, and rehydrated. After antigen retrieval, sections were incubated overnight with primary antibodies (α-SMA [Abways, #CY1132, 1:100] and LYVE-1 [Abcam, #ab219556, 1:20,000]). Subsequently, sections were treated with a reaction enhancer and incubated with an enzyme-labeled goat anti-rabbit IgG polymer for signal amplification. DAB was applied for color development (2–8 minutes), followed by hematoxylin counterstaining, differentiation, dehydration, clearing, and mounting. Digital images were captured using the Pannoramic SCAN II slide scanner at the Public Experimental Platform of Hebei Medical University. Slide Viewer software was used to view and analyze images of slides that had been stained with anti-α-SMA and anti-LYVE-1 antibodies. The slides were reviewed by two experienced pathologists who were blinded to the patients’ clinical information. If the discrepancy between the two reviewers exceeded 10%, a re-analysis was conducted. Inter-observer agreement for CAF identification between the two pathologists was excellent (Cohen’s κ = 0.936).

Characterization of CAFs and Microlymphatic Vessel Density (MLVD) in HPSCC

We evaluated CAFs based on the immunoreactivity score (IRS) of α-SMA, which reflects both their spatial distribution and expression intensity. Fibroblasts were classified based on the proportion of α-SMA-positive cells. α-SMA expression was absent (<1% of cells) and was scored as 0, a 1–10% positive rate was scored as 1, an 11–50% positive rate was scored as 2, a 51–80% positive rate was scored as 3, and a score greater than 80% was scored as 4. The staining intensity was also considered, with 0, 1, 2, and 3 points indicating no, weak, moderate, and strong staining, respectively. The percentage of positive cells and staining intensity produce the IRS.

Lymphatic vessel endothelial hyaluronan receptor 1 (LYVE-1)¹¹ is a well-established marker for lymphatic endothelial cells and is commonly used to differentiate lymphatic from blood vasculature. MLVD was assessed by quantifying LYVE-1-immunopositive lymphatic vessels. In accordance with the method established by Schoppmann, the region with the highest lymphatic vessel concentration in the sample, which was the most densely vascularized area or “hotspot”, was identified under wide-field microscopy. The number of LYVE-1-positive structures was counted in three high-power fields (200×). The average of these counts represented the MLVD, a measure of the microlymphatic vessel density.

Statistical Analysis

Statistical analysis was conducted using SPSS Statistics 20. The nonparametric rank sum test was used to compare target factor expression among different tissue types. The chi-square test or Fisher’s exact test (when the theoretical frequency was less than 5) was used to determine associations between target factors and clinicopathological features. Based on survival curve analysis, X-tile software was used to establish optimal cutoff values for the immune response scores of CAFs and MLVD. To evaluate the robustness and potential overfitting of the X-tile-derived optimal IRS cut-off value, we performed a bootstrap internal validation with 1,000 resamples and a fixed random seed (seed = 42) was applied to ensure reproducibility. For each resampled dataset (sampled with replacement from the original cohort), X-tile analysis was reapplied to identify the optimal cut-off. The stability of the cut-off was assessed by examining the distribution of the optimal cut-off values across all resamples. Univariate and multivariate survival analyses were used to identify prognostic factors, and the Log rank test was used to compare survival between groups. A Cox proportional hazards regression model was used for multivariate survival analysis. For pairwise comparisons, t tests were used, and the Welch correction was appropriately applied when variances were unequal. Spearman correlation analysis was used to evaluate the relationship between the number of CAFs and the MLVD. One-way ANOVA was used for comparisons among multiple groups. A p value < 0.05 was considered to indicate a statistically significant difference.

Results

Distribution of CAFs and MLVD

In this study, involving 91 patients diagnosed with HPSCC, males were predominant, constituting 87.9% (n=80) of the cohort, whereas females accounted for 12.1% (n=11). The age range of the participants was 42 to 81 years, with an average age of 60.4 years. The follow-up time ranged from 2 to 155 months, with an average of 55.6 months. The tumor was located mainly in the piriform fossa (79.1%, n=72), followed by the posterior pharyngeal wall (16.5%, n=15) and the posterior cricoid region (4.4%, n=4). The histopathological grading of the tumors included 5.5% (n=5) grade I, 72.5% (n=66) grade II, and 22% (n=20) grade III lesions. The staging of HPSCC followed the 8th edition of the American Joint Committee on Cancer (AJCC) TNM classification, and the N stage was determined through postoperative lymph node pathological reports. The cohort included 5.5% T1 (n=5), 18.7% T2 (n=17), 24.2% T3 (n=22), and 40.7% T4 (n=37) stage tumors. A total of 72.5% (n=66) of the patients had lymph node involvement, with 27.5% (n=25) having N0 status. According to the “Head and Neck Cancer Diagnosis and Treatment Guidelines” released by the Chinese Society of Clinical Oncology (CSCO) in 2024, early-stage patients (T1–2N0) accounted for 11% (n=10), whereas 89% (n=81) were classified as advanced (T1–2 N+ or any N stage T3–4).

Immunohistochemical staining revealed that α-SMA was located mainly in tumor stromal fibroblasts, as well as in the cytoplasm of microvascular walls and glandular structures. α-SMA-positive CAFs were preferentially distributed in the tumor stroma (Figure 2A). The immunoreactivity of LYVE-1, a specific lymphatic endothelial marker, was localized to the cytoplasm and appeared as brownish-yellow to yellowish-brown under staining (Figure 2A). LYVE-1 staining revealed that the lymphatic vessels within the tumor nests were compressed and presented as atrophic bands, whereas the small and medium-sized lymphatic vessels in the tumor stroma presented identifiable lumens. The shapes of the lymphatic endothelial cells were irregular and often distorted and curved, and cells were occasionally found in the lumen. In contrast, adjacent normal hypopharyngeal tissues exhibited unobstructed and dilated lumens.

Figure 2.

(A) α-SMA expression patterns and MLVD in HPSCC tissues compared with adjacent nonneoplastic tissues. Comparative immunohistochemical images showing high and low levels of α-SMA and MLVD in HPSCC tissues compared with normal tissues are shown. The scale bar denotes 200 μm, serving as a reference for the microscopic fields that are shown. (B) X-tile analysis was conducted to determine the optimal cutoff values for the CAF density and MLVD in HPSCC patients. All the HPSCC patients were stratified into groups according to high and low CAF density and MLVD. The graphical representation displays the χ² log-rank statistics when the patient cohort was segregated into two groups. The optimal cutoff values, highlighted by black dots on the horizontal axis, are shown on histograms that correspond to the entire patient population. Both the CAF density and the MLVD were significantly associated with overall survival in HPSCC patients. The Kaplan‒Meier survival curves for CAF density and MLVD across all patients are shown. Scale bar: 200 μm. (C) Distribution and correlation of CAFs and the MLVD. Distribution of CAFs and the MLVD in HPSCC tumors (T) and adjacent normal hypopharyngeal tissues (N). (D) Analysis of the correlation between the distribution of CAFs and MLVD in HPSCC tumor tissues. (E) Kaplan‒Meier survival curve analysis was performed based on the definition of the CAF density and MLVD in HPSCC. (*** P<0.001).

In accordance with the methods of a previous study, clinical research data were imported into the appropriate software. The optimal cutoff value was determined using R statistical software (http://molpath.charite.de/cutoff) to generate graphs of the data. For this evaluation, by using X-tile plots based on patient prognosis and assessments of CAF and MLVD, optimal cutoff values were established at 4.4 for CAF and 18 for MLVD (Figure 2B). High numbers of CAFs were defined as those for which the IRS was >4.4, whereas low numbers of CAFs were defined as those for which the IRS was ≤4.4. Similarly, the MLVD was classified as high if the value was >18 and low if the value was ≤18. Bootstrap internal validation with 1,000 resamples confirmed the stability of the X-tile-derived cut-off value of 4.4. In 88.2% of the resampled datasets, the optimal cut-off ranged between 4.0 and 5.0. Furthermore, Kaplan-Meier survival analysis using the 4.4 cut-off remained statistically significant in 83.7% of the resampled datasets, indicating that the cut-off was robust and not driven by sampling variability.

When the optimal cutoff value was used, the downregulation of α-SMA (IRS≤4.4) was observed in 19.8% (n = 18) of the samples, whereas the overexpression of α-SMA (IRS>4.4) was observed in 80.2% (n=73) of the samples. The IRS of CAFs in the tumor stroma of HPSCC patients was 7.92 ± 2.85, which was significantly greater than the IRS of CAFs in adjacent normal tissues (2.78±1.71) (P< 0.001) (Figure 2C). The number of microlymphatic vessels per high-power field (200×) was 21.92 ± 15.43, which was significantly greater than that in adjacent normal hypopharyngeal tissues (5.40 ± 3.92) (P < 0.001) (Figure 2C).

This correlation indicates that as the density of CAFs increased, the MLVD also increased accordingly. A chi-square test was subsequently performed to investigate the association between the distribution density of CAFs and MLVD based on the previously determined cutoff values. The results confirmed a significant relationship between the distribution density of CAFs and the MLVD (χ² = 4.643, P < 0.05) (Table 1). Specifically, high CAF density was consistent with high MLVD, and the same consistency was observed between low CAF density and low MLVD. Among patients who were stratified by the optimal cutoff value, the postoperative survival rate of the high-density CAF group was significantly lower than that of the low-density CAF group (χ² = 6.397; P < 0.05) (Figure 2D). Similarly, the survival rate of patients with high MLVD was significantly lower than that of patients with low MLVD (χ² = 6.883, P < 0.01) (Figure 2E).

Table 1.

Correlation Analysis Distribution of CAF and MLVD in HPSCC Tumor Tissues

		MLVD		χ2	P
		Low	High
CAFs	Low	19	11	4.643	0.031
	High	24	37

Abbreviations: Low, low density; High, high density; HPSCC, hypopharyngeal squamous cell carcinoma; CAFs, cancer-associated fibroblasts; MLVD, microlymphatic vessel density.

Effects of the CAF Distribution and MLVD on the Clinicopathological Characteristics of Patients with HPSCC

To further understand the clinical significance of CAFs and MLVD, the immunohistochemical staining results were cross-referenced with six clinicopathological features, namely, tumor size, histological grade, T stage, N stage, invasion depth, and nerve invasion, via the chi-square test. Among the 91 patients with HPSCC, a high distribution of CAFs was significantly associated with advanced clinical stage and T stage (P<0.01). In addition, a high distribution of CAFs was closely related to larger tumor size and greater invasion depth (P<0.05). Moreover, MLVD was significantly associated with clinical stage, T stage, N stage, tumor invasion depth, and nerve invasion (P<0.05) (Table 2). In addition, Spearman correlation analysis demonstrated a significant association between CAF density and MLVD (r = 0.261, P = 0.012).

Table 2.

Impacts of Distribution of CAFs and MLVD on the Clinicopathologic Characteristics of HPSCC

	n	CAFs		p	χ2	MLVD		p	χ2
		H	L			H	L
Tumor size
≤ 6.0cm2	28	15	13	0.014	6.01	12	16	0.263	1.252
> 6.0cm2	63	17	46			35	28
Clinical stage
Early	9	2	7	0.005*	9.075	2	7	0.04*	4.638
Advanced	82	59	23			49	33
T status
T1-2	22	7	15	<0.0001	16.282	7	15	0.033	4.569
T3-4	69	54	15			40	29
N status
N0	25	15	10	0.705	0.143	8	17	0.04	4.199
N1-3	66	45	21			37	29
Depth of invasion
Mucosa/ submusoca	51	28	23	0.018	8.068	23	28	0.048	6.086
Muscularis	20	14	6			14	6
Adjacent structure	20	18	2			7	13
Nerve invasion
Yes	12	9	3	0.529	0.397	9	3	0.047	3.931
No	79	54	25			36	43

Note: L, low density; H, high density; *, Fisher’s exact test; Underlined values indicate statistically significant P values (P < 0.05).

Abbreviations: HPSCC, hypopharyngeal squamous cell carcinoma; CAFs, cancer-associated fibroblasts; MLVD, microlymphatic vessel density.

Univariate analysis was conducted to identify important clinicopathological factors that might affect the prognosis of HPSCC, including age, sex, tumor size, tumor location, T stage, N stage, clinical stage, tumor histological grade, the presence of intravascular cancer thrombus, treatment era, CAF density, and MLVD density. Univariate survival analysis revealed that N stage (P=0.016), CAF density (P<0.05), and MLVD (P=0.009) were prognostic indicators for HPSCC (Table 3). The specific survival characteristics were as follows: N status: Patients in the N0 group had a median overall survival (OS) of 129 months and a mean OS of 104.73 months, which were significantly better than those in the N1/N2/N3 group (median 76 months, mean 74.13 months). MLVD: Patients in the low MLVD group achieved a mean OS of 110.69 months, with the median OS not reached (NR), which was significantly superior to that of the high MLVD group (median 61 months, mean 77.09 months). CAFs density: Patients in the low CAFs density group had a mean OS of 107.80 months, with the median OS not reached, which was significantly higher than that of the high CAFs density group (median 61 months, mean 85.31 months). In addition, factors such as tumor size, age, sex, tumor location, T stage, clinical stage, pathological differentiation, vascular tumor thrombus, postoperative radiotherapy, treatment era, and depth of invasion showed no statistically significant impact on patient survival (all P > 0.05).

Table 3.

Univariate (Log-Rank and Cox Regression) and Multivariate (Cox Regression) Analyses for Overall Survival

Variables	Univariate Analysis						Multivariate Analysis
	Log Rank Test					Cox Regression	HR (95% CI)	P
	95% CI	χ2	P	Mean Survival (Months)	Median Survival (Months)	HR (95% CI)
Size
≤ 6.0cm2	74.265–101.63	1.36	0.244	87.947	84	0.690 (0.367–1.296)
> 6.0cm2	55.408–95.846			75.627	63
Age(years)
≤ 60	66.816–102.143	0.01	0.921	84.479	86	1.606 (0.882–2.924)
> 60	69.43–99.951			84	76
Gender
Female	63.663–99.537	0.001	0.973	81	46.73	0.984 (0.383–2.531)
Male	71.614–96.991			84	76
Location
Pyriform sinus	70.361–96.546	1.352	0.509	83.453	83	1.029 (0.813–1.302)
Posterior pharyngeal wall	61.564–124.641			93.103	98
Postcricoid region	20.266–115.984			68.125	71
T status
T1-T2	55.408–90.377	1.596	0.206	72.893	63	0.865 (0.625–1.198)
T3-T4	74.166–102.159			88.162	86
N status
N0	82.528–126.932	5.768	0.016	104.73	129	0.576 (0.295–1.124)	0.394 (0.184–0.847)	0.017
N1/N2/N3	61.876–86.381			74.129	76
Clinical stage
Early	49.601–122.542	0	0.997	86.071	61	0.728 (0.282–1.877)
Advanced	71.619–96.226			83.922	83
Pathological differentiation
I	26.244–98.156	2.665	0.264	62.2	61	1.532 (0.934–2.513)
II	67.043–94.293			80.668	82
III	73.817–122.794			98.306	98
Intravascular tumor thrombus
No	73.011–99.063	0.477	0.49	86.037	82	1.271 (0.640–2.524)
Yes	53.343–104.701			79.022	83
MLVD
Low	95.171–126.199	6.886	0.009	110.685	UD	1.036 (0.580–1.849)	0.54 (0.296–0.986)	0.045
High	59.283–94.891			77.087	61
Radiotherapy after surgery
Yes	64-91.23	2.231	0.135	77.635	75	1.629 (0.851–3.118)
No	76.541–118.804			97.672	91
CAFs
Low	91.332–124.274	6.403	0.011	107.803	UD	0.803 (0.433–1.490)	0.458 (0.209–1.003)	0.051
High	70.34–100.282			85.311	61
Depth of invasion
Mucosa/submusoca	66.585–98.81	2.564	0.277	82.697	70	1.073 (0.726–1.586)
Muscularis	75.512–121.948			98.73	91
Adjacent structure	58.395–88.836			73.615	82
Nerve invasion
Yes	22.959–69.415	3.443	0.064	46.187	26	2.689 (0.903–8.001)
No	75.616–99.383			87.499	83
Treatment era
2010-2016	60.476–92.360	3.409	0.065	76.418	43	0.507
2017-2022	50.744–68.683			59.713	UD	0.243–1.062

Note: Underlined values indicate statistically significant results (P < 0.05).

Abbreviations: HR, hazard ratio; CI, confidence interval.

The clinicopathological factors that were identified in the univariate analysis were subsequently incorporated into the multivariate Cox proportional hazards regression model to identify independent prognostic factors for HPSCC (Table 3). N status (HR=0.394, 95% CI=0.184–0.847; P<0.05) and MLVD (HR=0.54, 95% CI=0.296–0.986; P<0.05) were independent predictors of poor outcomes in HPSCC patients. Specifically: N0 status was a protective factor for patient survival, with the risk of death being only 39.4% of that in patients with N1/N2/N3 status. Low MLVD status was also a protective factor for survival, with the risk of death being 54% of that in the high MLVD group. However, the degree of CAF activation did not reach statistical significance in the multivariate analysis (HR=0.458, 95% CI: 0.209–1.003, P=0.051), suggesting that its prognostic impact might be confounded by other clinical factors. A collinearity diagnostic test was performed to assess whether the borderline significance of CAF density (P = 0.051) could be attributed to collinearity with MLVD. The variance inflation factor (VIF) for CAF density was 1.350, and for MLVD was 1.749, indicating no significant collinearity between the two variables. Therefore, the loss of statistical significance for CAF density in the multivariable model is unlikely to be explained by collinearity. In the multivariable Cox regression model, 3 variables (CAF density, MLVD, and N stage) were included. Among the 91 patients, 47 events occurred during follow-up, yielding an events per variable (EPV) of 15.7, which is above the recommended threshold of 10. Thus, the statistical power of the model is adequate. MLVD and lymph node metastasis were identified as independent prognostic factors.

Discussion

Head and neck cancer ranks as the seventh most common malignancy worldwide, with approximately 90% of cases being squamous cell carcinoma.¹² Despite considerable progress in understanding its epidemiology, pathogenesis, and treatment, the survival rate for head and neck squamous cell carcinoma (HNSCC) has shown little improvement over the past 40 years, with a 5-year survival rate remaining around 50%.¹³ This figure is notably lower for HPSCC, a particularly aggressive malignancy associated with dismal prognosis.¹⁴ Patients with hypopharyngeal cancer are often diagnosed at advanced stages and exhibit a high propensity for lymph node metastasis. The extent and number of local tumor deposits, together with the involvement of regional lymph nodes, remain the most critical determinants of prognosis.¹

Cancer-associated fibroblasts (CAFs) are genetically non-malignant stromal cells residing within the tumor microenvironment (TME) of solid malignancies, where they can constitute up to 80% of the total tumor mass.¹⁵ These cells acquire biomechanical properties reminiscent of cancer cells, enabling functions such as matrix deposition and remodeling, chemokine secretion, and the facilitation of tumor invasion.¹⁶ Through the release of cytokines and exosomes, as well as modulation of the extracellular matrix (ECM), CAFs profoundly influence tumor progression.¹⁷

In our study, observations revealed that CAFs were densely aggregated in tumor tissues, which were also characterized by significantly higher MLVD than adjacent normal tissues. Clinical pathological analysis revealed that increased CAF distribution density was closely related to tumor size, clinical stage, tumor invasion depth, and T stage. Similarly, higher MLVD was closely related to clinical stage, N stage, tumor invasion depth, and nerve invasion. The positive correlation between the MLVD and N stage suggests that high MLVD predicts lymph node metastasis. Moreover, survival analysis confirmed that N stage and MLVD are independent risk factors for poor prognosis in patients with HPSCC, highlighting the potential for treatments that target lymphangiogenesis. However, these clinicopathological correlation analyses were exploratory in nature, and the reported P-values were not adjusted for multiple comparisons. Therefore, the results should be interpreted with caution and considered hypothesis-generating rather than confirmatory.

In the present study, univariate survival analysis confirmed that CAF density was significantly associated with overall survival in HPSCC patients, establishing CAF density as a clinically relevant prognostic factor. To explore the potential mechanism underlying this prognostic effect, we examined the relationship between CAF density and MLVD. Spearman correlation analysis revealed a significant positive correlation between CAF density and MLVD (r = 0.261, P = 0.012), suggesting an association between CAFs and increased lymphatic vessel density. When both CAF density and MLVD were included in the multivariable Cox regression model, MLVD remained an independent prognostic factor, whereas the significance of CAF density was attenuated to a borderline level (P = 0.051). This pattern—CAF density is significant in univariate analysis, correlated with MLVD, and loses independent significance when MLVD is adjusted for—is consistent with a model in which the prognostic effect of CAFs on HPSCC survival is at least partially transmitted through their association with lymphangiogenesis, as reflected by increased MLVD. We acknowledge that formal statistical mediation analysis was not performed. This was primarily because both CAF density and MLVD were analyzed as categorical variables, whereas classical mediation frameworks assume a continuous mediator. Nevertheless, the combination of three observations—(i) CAF density significantly predicts survival, (ii) CAF density significantly correlates with MLVD, and (iii) adjusting for MLVD attenuates the prognostic effect of CAF density—provides a coherent body of evidence consistent with a mediating role of MLVD. This interpretation is further supported by collinearity diagnostics, which confirmed that CAF density and MLVD are not redundant variables (VIF = 1.7 and 1.4, respectively), indicating that MLVD represents a distinct biological mediator rather than a mere proxy for CAF density. Importantly, the statistical power of the multivariable model was adequate, with 47 events and only 3 variables included, yielding an events-per-variable ratio of 15.7—well above the recommended threshold of 10. Thus, the borderline significance of CAF density is unlikely to be attributable to insufficient power, but rather supports the interpretation that its prognostic effect is largely mediated through MLVD. Nevertheless, given the relatively small total sample size (n = 91) and the single-center retrospective design, our findings should be interpreted with appropriate caution and require validation in larger, multi-center prospective cohorts.

Despite the insights provided, several limitations should be acknowledged. The strict exclusion criteria—including the exclusion of common chronic diseases and patients who had received systemic therapy—may have introduced selection bias and limited the generalizability of our findings, particularly to patients receiving adjuvant chemotherapy or immunotherapy. While these criteria were necessary to minimize confounding, our results require validation in larger, more diverse cohorts with broader inclusion criteria. Future multicenter studies should specifically evaluate the prognostic value of CAFs and MLVD in patients receiving systemic adjuvant therapy. Notably, external validation of the X-tile-derived IRS cut-off value of 4.4 in an independent cohort was not performed, primarily due to the limited availability of a suitable external HPSCC cohort with complete clinicopathological and follow-up data. To partially address this limitation and evaluate the potential risk of overfitting, we performed a Bootstrap internal validation with 1,000 resamples. In each resampled dataset (sampled with replacement from the original cohort), X-tile analysis was reapplied to identify the optimal IRS cut-off. The results demonstrated that the 4.4 cut-off was highly stable, and not merely a product of overfitting to our specific dataset. Nevertheless, confirmation in larger, multi-center independent cohorts is warranted to further establish its generalizability. Additionally, while the precise molecular mechanisms by which CAFs correlates with lymphangiogenesis remain to be experimentally validated, the observed association may suggest that CAF -associated lymphangiogenesis contribute to the unfavorable prognosis of hypopharyngeal cancer. This possibility will be explored to elucidate the underlying mechanisms in our future studies, such as co-culture assays, cytokine analyses and animal models.

The use of α-SMA to identify CAFs warrants consideration as a methodological point in the present study, as this marker is not entirely specific to CAFs. Prior to the formal experiment, we performed a pilot study comparing three commonly used CAF markers—α-SMA, fibroblast activation protein (FAP), and fibroblast-specific protein 1 (FSP1)—in hypopharyngeal cancer tissues. In our pilot staining, FAP and FSP1 exhibited weak or inconsistent expression in the tumor stroma, raising concerns about potential false-negative results that could lead to underdetection of CAFs. Given the critical role of CAF identification in this study, we opted to use α-SMA, which demonstrated robust and reliable stromal staining in our preliminary evaluation. We acknowledge that α-SMA may also label smooth muscle cells and vascular structures. However, based on extensive experience in histopathological assessment, we have established reliable morphological criteria to distinguish CAFs from these structures. Specifically, vascular smooth muscle typically appears as a concentric ring-like pattern encircling the lumen, often accompanied by identifiable red blood cells. Smooth muscle tissue, by contrast, is characterized by thick, well-organized fascicles with a regular architecture. CAFs, in comparison, are dispersed within the tumor stroma with an irregular, elongated, or stellate morphology, distinct from the organized patterns of vascular and muscular structures. These morphological distinctions were consistently applied during our analysis to ensure accurate identification of CAFs. Given the weak expression of FAP and FSP1 and the reliable performance of α-SMA, we chose not to pursue FAP/FSP1-based staining in the full cohort.

In summary, our findings establish that CAF distribution and MLVD are key determinants of disease progression and prognosis in HPSCC. Elevated CAF density related to increased MLVD, potentially facilitating lymphangiogenesis and metastatic dissemination. Elucidating the crosstalk between CAFs and lymphangiogenesis may identify novel therapeutic targets for HPSCC. Nevertheless, this retrospective cohort spanning 12 years is subject to limitations arising from evolving treatment paradigms for HPSCC; temporal heterogeneity in postoperative comprehensive management may induce survival bias. The relatively small sample size precluded subgroup analyses stratified by treatment era to adjust for this confounding factor. Accordingly, our findings warrant cautious interpretation, and prospective studies employing standardized treatment protocols are warranted to validate these observations.

Conclusion

In conclusion, CAF density and MLVD are markedly elevated in HPSCC tumor tissues and positively correlated with one another. Elevated CAF density and MLVD are associated with adverse clinicopathological characteristics and inferior overall survival. Multivariable regression analysis identified N stage and MLVD as independent prognostic determinants, supporting their utility in risk stratification for patients with HPSCC. Further exploration of molecular signaling between CAFs and lymphatic endothelial cells is critical to translate these research findings into clinical practice.

Funding Statement

We are sincerely thankful for financial support from the Natural Science Foundation of Hebei Province (Grant Nos. H2024505017), the Medical Science and Technology “Incubation” Plan of the 980th Hospital of the Joint Logistic Support Force of the People’s Liberation Army of China (Grant Nos. 2024FYQNXM04).

Data Sharing Statement

All the data generated and analyzed in this study have been included in this article or can be obtained from the corresponding authors upon reasonable request.

Ethics Approval and Consent to Participate

The work was carried out in accordance with the World Medical Association’s Declaration of Helsinki. This study was approved by the Medical Ethics Committee of the 980th hospital of PLA joint logistics support force (No. 2026-KY-115). Written informed consent was obtained from all patients for the use of their tissue samples in this study.

Disclosure

The authors report no conflicts of interest in this work.