Predictive modelling of tumor response and progression in cervical and head and neck cancers: a retrospective and prospective validation study

INTRODUCTION

Cervical and head and neck cancers collectively represent a significant global burden, particularly in low- and middle-income countries where late-stage presentations are common.¹ Despite advances in chemoradiation and targeted therapies, predicting which patients will respond optimally to treatment remains a challenge.^2-5 Recent efforts in precision oncology have focused on leveraging imaging, histological, and molecular biomarkers to develop predictive models that can inform real-time clinical decision-making.^6-8

A promising approach involves quantifying tumor characteristics such as fat content, blood vessel density, necrosis, and immune cell infiltration. These parameters reflect the tumor microenvironment, which plays a pivotal role in determining therapeutic response. This study evaluates the utility of a composite scoring system incorporating these parameters in predicting tumor response, progression-free survival (PFS), and adverse outcomes in cervical and head and neck cancer patients.⁵

METHODS

Study design and participants

This study comprised two arms a retrospective analysis of 500 patients and a prospective validation cohort of 200 patients. All participants had histologically confirmed cervical or head and neck squamous cell carcinoma. For the retrospective arm, data were extracted from electronic health records from 2017 to 2023. The prospective arm included patients treated from January 2018 to April 2021. Ethical clearance from the KIMS medical college Amalapuram was sought to do the analysis as the same was done on the clinical data (both prospective and retrospective) without altering any of the patient treatment plans. Standard methods like CT scan, RECITS criteria and radiation as per global guidelines were followed.^6,9-12

Inclusion criteria

Age 20–72 years, ECOG performance status 0–2, Received curative intent treatment (chemoradiation or surgery and adjuvant therapy)

Exclusion criteria

Prior malignancy, incomplete imaging/histology records, lost to follow-up within 3 months.

Data collection

Demographic variables (age, gender), tumor site, stage (I–IV), and treatment details were recorded. Quantitative scores for fat content, blood vessel density, necrosis, and immune cell infiltration were assigned on a scale of 1–5 based on multiparametric imaging and digital pathology. An overall score was computed as the mean of these four values.

Clinical outcomes

Time to maximum response

Weeks from therapy initiation to radiologic nadir.

Tumor shrinkage

Percent reduction in largest tumor dimension.

Progression-free survival

Time from diagnosis to progression or death.

Adverse events

Graded per CTCAE v5.0.

Predictive model and concordance analysis

A supervised machine learning model trained on the retrospective cohort was used to predict TMR, tumor shrinkage, adverse event score and PFS based on input scores.

Predictions were compared with actual outcomes in the prospective cohort, and concordance was assessed using Lin’s concordance correlation coefficient and Bland–Altman analysis.

RESULTS

Patient characteristics

In the retrospective cohort (n=500), the median age was 46 years; 100% of cervical cancer patients were female, and 90% of head and neck cancer patients were male. Stage distribution was I (20%), II (33%), III (33%), IV (14%). In the prospective cohort (n=200), demographics and stage mirrored the retrospective cohort. All patients completed scheduled follow-up for at least 6 months. The demographics were described in Table 1.

Tumor microenvironment scores

Mean scores across both cohorts were represented in Table 2. And the predictive accuracy for the retrospective cohort is represented in Table 3 and prospective cohort metric were done in Table 4.

Predictive factors analysis

Multivariate regression identified, blood vessel density (β=0.42, p<0.001)\(1). Immune cell density (β=0.31, p=0.003)\(2) as strongest predictors of early response and prolonged PFS.

DISCUSSION

This study demonstrates the predictive utility of a composite score incorporating imaging and pathology-derived metrics in cervical and head and neck cancers. Notably, the predictive model achieved over 90% concordance with actual clinical outcomes, underscoring its translational potential. The inclusion of blood vessel density and immune infiltration echoes findings from prior studies emphasizing the importance of vascular and immune microenvironments in tumor control.^1,2 Our results suggest that tumors with high vascularity and robust immune presence are more responsive to therapy. Fat score and necrosis showed weaker correlations, possibly due to their variable presence across tumor types. Nonetheless, their inclusion improved overall model calibration.^3,13

Future work will explore integration with radio genomic and circulating tumor DNA (ctDNA) markers for enhanced prediction. Adaptive therapy protocols guided by score-based predictions may improve outcomes while minimizing toxicity.⁴

Limitations include single central analysis of the multi centric data- where the radiation planning included- both Varian, Electa and Cobalt machines data, limited ethnic diversity, and short-term follow-up in the prospective arm. External validation in multicenter cohorts with long-term outcomes is essential.

CONCLUSION

A multi-parameter scoring system based on fat, vascular, necrotic, and immune characteristics accurately predicts therapeutic response and progression in cervical and head and neck cancers. Prospective validation affirms its clinical relevance. This tool may support early treatment modification, ultimately improving patient outcomes. The Figure 1 represent a conclusion of correlation between tumor microenvironment scores and PFS.

Funding: No funding sources

Conflict of interest: None declared

Ethical approval: The study was approved by the Institutional Ethics Committee

REFERENCES

1. 1. Siegel RL, Giaquinto AN, Jemal A. Cancer statistics, 2024. CA Cancer J Clin. 2024;74(1):12-49.
2. 2. Jain RK. Normalization of tumor vasculature: an emerging concept in antiangiogenic therapy. Science. 2005;307(5706):58–62.
3. 3. Fridman WH, Pagès F, Sautès-Fridman C, Galon J. The immune contexture in human tumours: impact on clinical outcome. Nat Rev Cancer. 2012;12(4):298–306.
4. 4. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell. 2011;144(5):646674.
5. 5. Hajim WI, Zainudin S, Daud KM, Alheeti K. Optimized models and deep learning methods for drug response prediction in cancer treatments: a review. Peer J Computer Sci. 2024;10:1903.
6. 6. Eisenhauer EA, Therasse P, Bogaerts J. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer. 2009;45(2):228–47.
7. 7. Jia Q, Wu W, Wang Y. Local mutational diversity drives intratumoral immune heterogeneity in non-small cell lung cancer. Nat Commun. 2018;9(1):5361.
8. 8. O'Connor JP, Aboagye EO, Adams JE. Imaging biomarker roadmap for cancer studies. Nat Rev Clin Oncol. 2017;14(3):169–86.
9. 9. Binnewies M, Roberts EW, Kersten K. Understanding the tumor immune microenvironment (TIME) for effective therapy. Nat Med. 2018;24(5):541–50.
10. 10. Sun R, Limkin EJ, Vakalopoulou M. A radiomics approach to assess tumour-infiltrating CD8 cells and response to anti–PD-1 or anti–PD-L1 immunotherapy: an imaging biomarker, retrospective multicohort study. Lancet Oncol. 2018;19(9):1180–91.
11. 11. Kumar V, Abbas AK, Aster JC. Robbins and Cotran Pathologic Basis of Disease. 10th ed. Philadelphia, PA: Elsevier. 2020.
12. 12. Pérez-García VM, Calvo GF, Bosque JJ. Universal scaling laws rule explosive growth in human cancers. Nat Phys. 2020;16(10):1232–7.
13. 13. Huang C, Huang Y, Xu D. Immune cell infiltration and angiogenesis in cervical cancer: prognostic implications and therapeutic potential. Front Oncol. 2023;13:1172302.