The treatment of HNSCC encompasses a variety of disciplines, including surgery, radiation therapy, and cytotoxic chemotherapy47,48. However, the recent emergence of immunotherapy has significantly changed both treatment procedures and outcomes. Known for its enduring efficacy, minimal side effects, and broad applicability, immunotherapy has shown outstanding performance in the treatment of HNSCC. Research has demonstrated that immunotherapy significantly improves the prognosis of recurrent metastatic HNSCC49. New immunotherapy-based treatments were also being applied to locally advanced cases. Furthermore, numerous ongoing trials are investigating the use of new immunotherapeutic agents, including durvalumab, atezolizumab, avelumab, tremelimumab, monalizumab, and others50. These findings underscored the efficacy of immunotherapy in the treatment of HNSCC. Looking ahead, it is highly probable that this therapeutic approach will become the predominant method of treatment.
Immunotherapy has been associated with more durable outcomes and fewer, less severe, or at least more manageable adverse effects compared to traditional cancer treatments. This innovative approach leveraged the body's inherent immune response to identify and attack cancerous cells, thus enhancing the specificity and efficacy of treatment while decreasing collateral damage to healthy tissues50. Additionally, the combination of immunotherapy with other treatment approaches, including vaccines and chemotherapy, remained a subject of particular interest50. Cancer vaccines have the potential to boost the immune system's capacity for identifying and combating cancerous cells, while chemotherapy can reduce tumor burden and alter the tumor milieu in ways that make it more susceptible to immune attack. The integration of these modalities aims to create a synergistic effect, thereby enhancing the overall therapeutic outcome51.
In the realm of HNSCC treatment, numerous clinical trials are exploring the integration7 of immunotherapy with other therapeutic modalities and pharmacotherapy to enhance treatment efficacy and safety52. For example, studies have demonstrated that integrating radiation therapy with immunotherapy offers a promising method for patients with locally advanced, recurrent, or metastatic HNSCC49. These researches offer mechanistic insights into the synergies and therapeutic benefits of such combinations. Several phase II/III clinical trials suggest that augmenting radiation therapy with immunotherapy is a safe option53. Researchers have also examined various immunotherapy regimens and alternatives for patients who are not eligible for immunotherapy. Discussions have focused on using alternative systemic treatments (such as EGFR inhibitors or different forms of immunotherapy) to reduce treatment intensity, including the possibility of excluding simultaneous chemotherapy for patients with a low risk51. In conclusion, a biological framework has been established to investigate the potential interactions between radiation therapy and systemic treatments, particularly emphasizing immunotherapy. A retrospective analysis of G/GEJ adenocarcinoma patients receiving a combination of PD-1 inhibitors and chemotherapy between October 2017 and May 2022 revealed significant findings. Specifically, patients with a PD-L1 CPS of ≥ 5 significantly benefited from this treatment, showing enhanced response rates and prolonged progression-free survival 54. Therefore, we foresee that the coming decade will present numerous opportunities for the development of enhanced prognostic methods for personalized and concurrent treatments. This progress would facilitate the personalized choice of the best order and mix of cytotoxic chemotherapy, specific drugs, and immune therapy7, customized to enhance advantages and reduce adverse effects51.
In this study, we successfully identified two distinct molecular subtypes using ten integrated multi-omics clustering algorithms. Through the application of ten machine learning methods to analyze multi-omics data, significant enrichment of immune suppression and immune rejection features was found in the CS1 subtype, while in the CS2 subtype, significant enrichment of immune function, immune therapy features, and ICGs was shown.This suggested that the CS1 subtype was more likely to be a "cold tumor" or "immune-suppressive tumor", while the CS2 subtype was more likely to be a "hot tumor" or "immune-activated tumor". Subsequently, by analyzing the markers of these subtypes, we identified their IRGs, which we referred to as key features. Based on these IRGs, we employed a combination of over 300 machine learning algorithms and utilized the StepCox [forward] + Ridge algorithm to construct a prognostic model named CMPIS, consisting of 16 key genes (CD6, CD5, EPHX3, MAP4K1, CALML5, KLRB1, PDGFA, WNT7A, CHGB, CAMK2N1, DKK1, PTX3, STC2, HOXA1, EFNB2, and GSTO1). Next, the CMPIS model was validated and compared across various datasets, including clinical characteristics, external datasets, and previously published models. Significant biological and molecular pathway differences among different CMPIS groups were observed, which might explain their distinct prognostic and treatment response patterns.A systematic analysis of the impact of CMPIS on immunotherapy in HNSCC revealed that the low CMPIS patients demonstrated better prognosis, indicating a greater benefit from immunotherapy. In conclusion, we validated and compared our results using clinical characteristics, external datasets, and published models, ultimately concluding that patients with lower CMPIS scores were more suitable for immunotherapy and chemotherapy, while those with higher CMPIS scores were better suited for radiation therapy and EGFR-targeted treatment. These findings underscored the importance of patient stratification based on CMPIS in enhancing the accuracy of treatment selection and potentially improving clinical treatment outcomes.
Given the current challenges in the field of immunotherapy, including the lack of effective immune markers, poor clinical efficacy, and the inability to accurately determine patient sensitivity to treatment, there is an urgent need to develop models based on large-scale multi-omics data and advanced algorithms to identify and predict biomarkers that may effectively respond to immunotherapy. Numerous researchers were dedicated to this goal, focusing on predicting and identifying key biomarkers. For instance, S.Trebeschi et al. utilized artificial intelligence technology to characterize lesions based on preprocessed contrast-enhanced CT data, successfully developing and validating a non-invasive machine learning biomarker that can distinguish between responders and non-responders to immunotherapy. Their study concluded that the radiological characteristics of lesions could serve as non-invasive biomarkers for assessing response to immunotherapy and might improve patient stratification in neoadjuvant and palliative treatments55. Additionally, other researchers have emphasized the importance of predictive biomarkers in guiding patients' choices of monotherapy or combination therapy and in developing alternative treatment strategies56. Michaël Duruisseaux et al. discussed currently available and promising biomarkers, highlighting the need for predictive biomarkers of the efficacy of immune checkpoint inhibitors57. Furthermore, studies have recognized CCDC71L as an independent prognostic marker for HNSCC, with high expression related to ICGs, oncogene mutations, and markers of genomic heterogeneity, indicating its significant role in the progression of HNSCC58. These research findings are highly consistent with our study conclusions, further validating the reliability of our research and underscoring the critical role of immunotherapy biomarkers in the treatment of HNSCC. By comparing our research results with the aforementioned existing literature, we can verify and confirm the reliability of our findings, demonstrating the value and significance of our research.
While our findings hold promise, we acknowledge several limitations in our study. Firstly, the difficulty in obtaining fresh clinical samples impeded our ability to conduct functional validations with fresh tumor samples. Secondly, as our research was retrospective in nature, further validation through multicenter prospective cohort studies is essential to confirm our findings. Additionally, our dataset was relatively small and predominantly comprised non-metastatic patients, potentially introducing bias in our experimental outcomes. Our proposed model may be more suitable for predicting the prognosis and guiding treatment for non-metastatic patients, which could limit its generalizability. Furthermore, a notable limitation was the lack of basic experiments, including cell and animal studies, to explore the functional expression of genes.