European Journal of Cardiovascular Medicine

Tumor-infiltrating lymphocytes (TILs) have emerged as critical mediators of the host immune response within the tumor microenvironment, and their presence has been associated with favorable clinical outcomes in a wide range of epithelial malignancies (1,2). TILs reflect ongoing antitumor immune activity and have been recognized as both prognostic and predictive biomarkers, particularly in cancers such as triple-negative breast cancer, non-small cell lung carcinoma, and colorectal carcinoma (3,4). The Immunoscore—based on TIL density and distribution—has been proposed as a standardized approach to immune classification of tumors, offering complementary insights alongside traditional TNM staging (5).

With advancements in digital pathology and artificial intelligence (AI), the quantification of TILs has shifted from subjective visual estimation to objective, reproducible metrics derived from whole-slide imaging (6). Automated image analysis algorithms trained on histopathological features can accurately detect lymphocyte populations within stromal and intratumoral compartments, enhancing consistency and scalability of immune profiling across tumor types (7). This approach also enables high-throughput comparative studies across multiple cancer types, which is essential for identifying common and distinct patterns in tumor immune microenvironments.

Despite the growing interest in TIL-based biomarkers, limited studies have conducted large-scale, cross-sectional evaluations of TIL densities across different epithelial malignancies using standardized digital methods. The current study aims to fill this gap by quantitatively assessing TILs in major epithelial cancers using validated digital pathology algorithms. The goal is to establish comparative immunological baselines and explore the potential of TIL quantification in clinical stratification and immunotherapy planning.

This cross-sectional study included a total of 300 formalin-fixed, paraffin-embedded (FFPE) tumor specimens from patients diagnosed with epithelial malignancies, including breast (n=75), lung (n=75), colorectal (n=75), and head and neck cancers (n=75).

Each tissue block was sectioned at 4 µm thickness and stained with hematoxylin and eosin (H&E) using standardized protocols. The stained slides were then digitized using a high-resolution whole-slide scanner (40× magnification). The resulting whole-slide images (WSIs) were analyzed using a commercially validated, AI-driven digital pathology platform designed to identify and quantify TILs within the tumor-associated stromal regions.

The algorithm segmented stromal compartments and identified mononuclear lymphocytes based on morphological criteria. TIL density was expressed as the number of lymphocytes per square millimeter (cells/mm²) of stromal area. Three representative tumor regions per slide were selected for analysis to ensure consistency, and the average TIL density per case was calculated.

Statistical analysis was performed using SPSS version 25.0. Mean TIL densities were compared among cancer types using one-way ANOVA, followed by Tukey’s post hoc test for pairwise group comparisons. Statistical significance was considered at p < 0.05. Interobserver reliability of the digital platform was assessed using intraclass correlation coefficient (ICC) on a randomly selected subset of 30 slides reviewed independently by two pathologists.

The study included 300 cases of epithelial malignancies distributed evenly across four cancer types: breast carcinoma, lung carcinoma, colorectal carcinoma, and head and neck squamous cell carcinoma (HNSCC). The mean age of the patients was 58.3 ± 10.4 years, with a female predominance noted in breast cancer cases (Table 1).

Quantitative analysis of TILs using the AI-based digital pathology algorithm revealed significant variability in lymphocyte densities among tumor types (p < 0.001). The highest mean TIL density was observed in triple-negative breast cancer (TNBC) cases, with an average of 432.5 ± 65.3 cells/mm², followed by HNSCC at 387.4 ± 52.8 cells/mm². Non-small cell lung carcinoma (NSCLC) showed a moderate mean density of 296.8 ± 48.9 cells/mm², while colorectal adenocarcinoma exhibited the lowest mean TIL count at 215.1 ± 41.2 cells/mm² (Table 2).

Post hoc analysis confirmed statistically significant pairwise differences between colorectal cancer and the other three cancer types (p < 0.01). However, differences between TNBC and HNSCC were not statistically significant (p = 0.064). Interobserver reliability for TIL quantification was excellent, with an intraclass correlation coefficient (ICC) of 0.91 (95% CI: 0.86–0.95), indicating high consistency of the digital algorithm.

Furthermore, stratification of cases based on tumor stage revealed that early-stage tumors (Stage I–II) in breast and lung cancers exhibited higher TIL densities compared to advanced-stage tumors (Stage III–IV), suggesting an inverse relationship between TIL infiltration and disease progression (Table 3).

Table 1. Demographic Distribution of Study Population

Table 2. Mean Tumor-Infiltrating Lymphocyte (TIL) Densities across Cancer Types

Table 3. Mean TIL Densities by Tumor Stage in Breast and Lung Cancer

As shown in Tables 2 and 3, digital pathology effectively discriminated between cancer types and stages based on TIL density, underscoring the clinical relevance of automated immune profiling in oncology.

This study quantitatively evaluated tumor-infiltrating lymphocytes (TILs) across major epithelial malignancies using artificial intelligence (AI)-assisted digital pathology algorithms. Our findings revealed significant differences in TIL densities between tumor types, with the highest infiltration observed in triple-negative breast cancer (TNBC) and the lowest in colorectal adenocarcinoma. These results are consistent with existing literature suggesting that immunogenic tumors like TNBC and head and neck squamous cell carcinoma (HNSCC) often exhibit a more active immune microenvironment (1,2).

The observed elevated TIL densities in TNBC align with prior studies where high TIL levels have been linked to better prognosis and response to chemotherapy and immunotherapy (3,4). Similarly, high TIL levels in HNSCC have been associated with improved survival outcomes, possibly due to the viral etiology in a subset of these cancers, such as HPV-positive tumors, which are more immunologically active (5,6). In contrast, colorectal cancer showed the lowest TIL infiltration, particularly in microsatellite-stable tumors, which typically have a less immunogenic profile (7,8).

Digital pathology offers a reproducible and scalable method for TIL quantification. Traditional manual assessment is often subject to interobserver variability and lacks standardization (9). Our algorithm demonstrated a high intraclass correlation coefficient (ICC = 0.91), supporting the utility of AI in improving diagnostic accuracy and reproducibility. These findings echo previous reports validating machine learning tools for TIL evaluation in breast and lung cancer histopathology (10,11).

Interestingly, our stratified analysis revealed that early-stage tumors exhibited significantly higher TIL densities compared to advanced stages, particularly in breast and lung cancers. This inverse relationship may reflect immune evasion mechanisms adopted by tumors during progression, such as upregulation of checkpoint molecules and recruitment of immunosuppressive cells (12,13). These dynamics highlight the importance of assessing immune infiltration in the early disease course, potentially guiding therapeutic decisions.

The integration of TIL quantification into routine pathology could serve multiple roles—from risk stratification to predicting response to immune checkpoint inhibitors (14). Several clinical trials have demonstrated the predictive role of TILs in selecting candidates for PD-1/PD-L1 inhibitors, especially in lung and breast cancers (15,6). Moreover, the Immunoscore—based on TIL density and distribution—has been adopted in colorectal cancer as a complement to TNM staging and is associated with recurrence risk and survival (1).

Despite the strengths of our study, including a multicancer cohort and standardized AI-based analysis, certain limitations must be acknowledged. The retrospective design and lack of immunophenotypic classification of lymphocytes (e.g., CD8+, FOXP3+) limit mechanistic insights. Additionally, tumor heterogeneity and sampling bias in tissue sections could influence TIL measurements. Future studies incorporating multiplex immunohistochemistry and spatial transcriptomics may offer deeper insights into the functional immune landscape (8,9).

In conclusion, this study supports the feasibility and clinical relevance of AI-driven TIL quantification across epithelial malignancies. The differences in TIL infiltration patterns among cancer types and stages reinforce the need for tumor-specific immune profiling. As precision oncology evolves, integrating digital immunopathology into routine diagnostic workflows can enhance prognostic assessments and guide immunotherapeutic strategies.

1. Jimenez-Rodriguez RM, Patil S, Keshinro A, Shia J, Vakiani E, Stadler Z, et al. Quantitative assessment of tumor-infiltrating lymphocytes in mismatch repair proficient colon cancer. Oncoimmunology. 2020;9(1):1841948. doi:10.1080/2162402X.2020.1841948. PMID: 33235819.
2. Williams DS, Mouradov D, Jorissen RN, Newman MR, Amini E, Nickless DK, et al. Lymphocytic response to tumour and deficient DNA mismatch repair identify subtypes of stage II/III colorectal cancer associated with patient outcomes. Gut. 2019;68(3):465-74. doi:10.1136/gutjnl-2017-315664. PMID: 29382774.
3. Kim JK, Chen CT, Keshinro A, Khan A, Firat C, Vanderbilt C, et al. Intratumoral T-cell repertoires in DNA mismatch repair-proficient and -deficient colon tumors containing high or low numbers of tumor-infiltrating lymphocytes. Oncoimmunology. 2022;11(1):2054757. doi:10.1080/2162402X.2022.2054757. PMID: 35481287.
4. Rohr UP, Herrmann P, Ilm K, Zhang H, Lohmann S, Reiser A, et al. Prognostic value of MACC1 and proficient mismatch repair status for recurrence risk prediction in stage II colon cancer patients: the BIOGRID studies. Ann Oncol. 2017;28(8):1869-75. doi:10.1093/annonc/mdx207. PMID: 28460000.
5. Sari M, Atag E, Demir T, Simsek ET, Coban E, Cikrikcioglu M. Deficient Mismatch Repair and Lymphocytic Response to Tumor as Prognostic Markers in Stage II Colon Cancer Patients. J Coll Physicians Surg Pak. 2022;32(2):186-92. doi:10.29271/jcpsp.2022.02.186. PMID: 35108789.
6. Lee H, Sha D, Foster NR, Shi Q, Alberts SR, Smyrk TC, et al. Analysis of tumor microenvironmental features to refine prognosis by T, N risk group in patients with stage III colon cancer (NCCTG N0147) (Alliance). Ann Oncol. 2020;31(4):487-94. doi:10.1016/j.annonc.2020.01.011. PMID: 32165096.
7. Chou M, Illa-Bochaca I, Minxi B, Darvishian F, Johannet P, Moran U, et al. Optimization of an automated tumor-infiltrating lymphocyte algorithm for improved prognostication in primary melanoma. Mod Pathol. 2021;34(3):562-71. doi:10.1038/s41379-020-00686-6. PMID: 33005020.
8. Lu J, Xu Y, Wu Y, Huang XY, Xie JW, Wang JB, et al. Tumor-infiltrating CD8+ T cells combined with tumor-associated CD68+ macrophages predict postoperative prognosis and adjuvant chemotherapy benefit in resected gastric cancer. BMC Cancer. 2019;19(1):920. doi:10.1186/s12885-019-6089-z. PMID: 31521128.
9. Liu S, Kong P, Wang X, Yang L, Jiang C, He W, et al. Tumor microenvironment classification based on T-cell infiltration and PD-L1 in patients with mismatch repair-proficient and -deficient colorectal cancer. Oncol Lett. 2019;17(2):2335-43. doi:10.3892/ol.2018.9826. PMID: 30675299.
10. Zlobec I, Terracciano LM, Lugli A. Local recurrence in mismatch repair-proficient colon cancer predicted by an infiltrative tumor border and lack of CD8+ tumor-infiltrating lymphocytes. Clin Cancer Res. 2008;14(12):3792-7. doi:10.1158/1078-0432.CCR-08-0048. PMID: 18559598.
11. Qin Q, Ying J, Lyu N, Guo L, Zhi W, Zhou A, et al. [Correlations between DNA mismatch repair (MMR) and prognosis and prediction of treatment efficacy in stage II/II colon cancer]. Zhonghua Zhong Liu Za Zhi. 2014;36(11):844-8. PMID: 25620482. Chinese.
12. Yoon HH, Shi Q, Heying EN, Muranyi A, Bredno J, Ough F, et al. Intertumoral Heterogeneity of CD3+ and CD8+ T-Cell Densities in the Microenvironment of DNA Mismatch-Repair-Deficient Colon Cancers: Implications for Prognosis. Clin Cancer Res. 2019;25(1):125-33. doi:10.1158/1078-0432.CCR-18-1984. PMID: 30301825.
13. Giampieri R, Maccaroni E, Mandolesi A, Del Prete M, Andrikou K, Faloppi L, et al. Mismatch repair deficiency may affect clinical outcome through immune response activation in metastatic gastric cancer patients receiving first-line chemotherapy. Gastric Cancer. 2017;20(1):156-63. doi:10.1007/s10120-016-0594-4. PMID: 26796888.
14. Faraji F, Fung N, Zaidi M, Gourin CC, Eisele DW, Rooper LM, et al. Tumor-infiltrating lymphocyte quantification stratifies early-stage human papillomavirus oropharynx cancer prognosis. Laryngoscope. 2020;130(4):930-8. doi:10.1002/lary.28044. PMID: 31070246.
15. Rakaee M, Kilvaer TK, Dalen SM, Richardsen E, Paulsen EE, Hald SM, et al. Evaluation of tumor-infiltrating lymphocytes using routine H&E slides predicts patient survival in resected non-small cell lung cancer. Hum Pathol. 2018;79:188-98. doi:10.1016/j.humpath.2018.05.017. PMID: 29885403.