While immunomodulatory agents have shown promising results in treating refractory solid tumors ^[1[42]^], not all patients exhibit satisfactory responses. Immunotoxicity is a common occurrence, underscoring the need for predictive biomarkers to identify patients who would benefit from immunotherapy ^[2[43]^]. Furthermore, the high cost of immunotherapies poses a barrier to access, limiting this treatment option to a privileged subset of patients in low–middle-income countries.

PD-L1 expression is the currently used biomarker for predicting the response and clinical efficacy of ICIs in NSCLC ^[3[44]^]. Positive PD-L1 expression is associated with better outcomes and clinical response rates than those for patients with negative PD-L1 negative expression ^[4[45]^], regardless of smoking habits ^[5[46]^]. PD-L1 expression levels can be measured using the Tumor Proportion Score (TPS) or Combined Positive Score (CPS), where TPS represents the percentage of tumor cells expressing PD-L1 and CPS accounts for PD-L1 presence in both tumor and inflammatory cells. Pro-inflammatory signals associated with JAK2 transduction ^[6[47]^] and IFN-γ expression ^[7[48]^] may increase PD-L1 expression ^[8[49]^].

In a phase I study, PD-L1 surface expression levels > 5% were associated with a response to nivolumab, while patients with PD-L1 expression < 5% did not respond well to treatment ^[9[50]^]. In the phase III Keynote 024 study, patients with aNSCLC and high levels of PD-L1 expression (≥50%) demonstrated improved outcomes when treated with pembrolizumab compared to the platinum-based CT group ^[10[39]^]. Although patients with advanced squamous cell NSCLC showed better responses to nivolumab than to docetaxel regardless of PD-L1 expression, even patients with negative PD-L1 expression experienced increased PFS after ICI treatment ^[11[35]^].

However, despite these findings, PD-L1 TPS alone is not sufficiently accurate or reliable in predicting the response to immunotherapy ^[12[51]^]. The technical limitations of immunohistochemistry (IHC) in the context of PD-L1 expression analysis include variations in sample quality, as the measurements are conducted on tumor biopsies that may not fully capture tumor heterogeneity. Additionally, diverse staining protocols and the establishment of appropriate cut-off points for interpreting PD-L1 expression pose further challenges ^[13[52]^]. In addition, the predictive capacity of IHC in determining the response to ICIs across various histologies, such as SCC, remains limited ^[11][14[34,35]^]. In this context, the combination of TPS and CPS is a more robust biomarker ^[15[53]^].

PD-L1 identification in circulating tumor cells (CTCs) tends to be higher than in tissues, potentially enhancing quantification efficiency. However, results regarding its correlation with treatment response remain controversial, with studies reporting absence of or worse-lasting response to ICIs with higher pretreatment PD-L1⁺ CTCs, while others indicate the positive prognostic value of PD-L1 expression in CTCs ^[16][17][18[54,55,56]^]. Some studies found no correlation between PD-L1 expression in tumor biopsies and CTCs ^[19[57]^].

Tumor cells that do not express PD-L1 may still be accompanied by immune cells expressing this ligand. Additionally, tumor-infiltrating immune cells in the tumor environment demonstrate increased levels of PD-L1 expression, and the administration of ICI leads to an enhanced influx of CD8⁺ T lymphocytes in the tumor region ^[20[58]^].

Plasma soluble PD-L1 (sPD-L1) levels have been investigated as an alternative to PD-L1 TPS. While certain studies have found no correlation between pre-treatment levels of soluble programmed death-ligand 1 (sPD-L1) and OS ^[21[59]^], other investigations have established a significant association between elevated pre-treatment sPD-L1 levels and unfavorable outcomes, including higher rates of treatment failure ^[22[60]^]. Furthermore, increased or stable levels of soluble PD-1 (sPD-1) after two cycles of nivolumab were linked to a better prognosis ^[23[61]^].

Genetic variations were also evaluated as possible biomarkers. PD-L1 polymorphisms were associated with the efficacy of ICIs, being correlated with OS improvement ^[24[62]^]. Patients carrying HER2 exon 20 ^[25[63]^], ERBB4 ^[26[64]^], KRAS G12C ^[27[65]^], FGFR4 ^[28[66]^], ARID 1A, and ARID 1B ^[29[67]^] mutations showed ICIs benefit. However, patients harboring KRAS G12V ^[27[65]^], EGFR ^[30[68]^], or ALK ^[31[69]^] alterations tend to have poorer outcomes.

Gene-expression-based biomarkers have been successfully employed in the field of oncology for the clinical management of cancer patients ^[32[72]^]. In patients treated with anti-PD-1, the expression levels of CSF1 R and HCST positively correlated with PD-L1 levels and high infiltration of CD8⁺ T cells ^[33[73]^]. Nevertheless, other potential ICI response signatures have been described, including the expression levels of MAP1A/1B/1S/4/6/7D1/7D3 in ADC and SCC, where high expression of these genes was associated with favorable response to immunotherapy ^[34[74]^]. In NSCLC, KAT2B expression displayed a positive correlation with the levels of infiltrating immune cells and mRNA expression of immune checkpoint genes. Conversely, tumor tissues exhibited downregulation of KAT2B expression, which was associated with ineffective response to ICI and unfavorable prognosis in patients with lung ADC ^[35[75]^].

A TCR co-expression signature has been identified as a valuable predictor of prognosis in NSCLC patients undergoing immunotherapy. Elevated expression levels of this specific gene-set are indicative of more favorable treatment responses ^[36[76]^]. Additionally, a cancer-specific immune score model has been developed to predict ICI response with satisfactory performance, achieving an area under the curve (AUC) of 0.68, indicating its potential for accurately assessing ICI response ^[37[77]^].

Moreover, the gene expression profile within the TME holds predictive value for pathologic complete response and disease progression in patients receiving combined neoadjuvant chemoimmunotherapy. Tumors showing favorable response to treatment exhibited elevated levels of IFNG, GZMB, NKG7, and M1 macrophages whereas tumors prone to relapse following surgery exhibited heightened expression of genes such as AKT1, BST2, OAS3, and CD8B ^[38[78]^].

To construct an immune gene score, pivotal immune cells, human leukocyte antigens (HLAs), and immune checkpoints were selected and the immune-related genes of those three aspects of the TME were combined to construct the score. The score derived from these three aspects demonstrated the ability to predict the response to ICIs, achieving an area under the curve (AUC) of 0.737 at the 20-month mark. Remarkably, within this signature, patients exhibiting a higher hypoxia score demonstrated a stronger association with immunotherapy efficacy. A predictive model integrating this immune score, tumor mutational burden (TMB), and long non-coding RNA expression exhibited promising predictive potential for effective immunotherapy response, achieving an AUC of 0.814 at 20 months ^[39[79]^].

Efforts are ongoing to identify gene-expression-based signatures that can accurately predict response to immunotherapy. However, current studies are in the preliminary stages and require further validation before their potential for clinical implementation can be fully assessed ^[40[80]^].

TMB refers to the presence of non-synonymous mutations in the coding regions of the tumor genome ^[41[81]^]. High TMB indicates a greater number of neoantigens, which can activate the T-lymphocyte response ^[42[82]^]. In NSCLC, the mutagenic effects of smoking can lead to higher TMB rates ^[43[83]^]. Several studies have correlated TMB rates with response to ICIs ^[42][44[82,84]^], suggesting that a TMB of ten or more mutations per megabase is associated with improved PFS ^[45[85]^]. Patients treated with pembrolizumab and high TMB rates have demonstrated durable clinical benefits ^[46[86]^]. Additionally, in ADC patients, a high burden of clonal neoantigens has been linked to better outcomes and ICI responses ^[47[87]^].

However, it is worth noting that less than 10% of nonsynonymous mutations result in immunogenicity ^[48[88]^]. Additionally, there is currently no established cut-off point for non-synonymous mutations that can reliably predict clinical benefit ^[49[89]^]. In colorectal cancer (CRC) patients, elevated TMB levels and microsatellite instability (MSI) are associated with a more favorable prognosis ^[50[90]^]. However, the prevalence of MSI-high (MSI-H) status in NSCLC patients is rare, limiting its usefulness as a biomarker in this patient population ^[51[91]^]. Despite these limitations, the combination of TMB with other biomarkers such as PD-L1 expression or the neutrophil-to-lymphocyte ratio (NLR) holds promise, as it enhances prediction capabilities ^[52[92]^].

Absolute complete blood count (CBC) values have been evaluated as potential biomarkers for predicting the response to ICIs by analyzing patients’ medical records ^[53[93]^]. In NSCLC, the presence of local inflammation results in an immune infiltrate rich in neutrophils ^[54[94]^]. The increase in NLR ^[55][56][57[95,96,97]^], myeloid-to-lymphoid ratio (M:L), absolute-neutrophil-count-to-absolute-lymphocyte-count ratio (ANC:ALC) and absolute neutrophil counts (post-ANCs) is associated with lower PFS ^[58[98]^] and OS ^[59[99]^]. Conversely, low platelet-to-lymphocyte ratio (PLR) ^[56][60[96,100]^], monocyte-to-lymphocyte ratio (MLR) ^[56[96]^], and high rates of ALC ^[61[101]^], absolute eosinophil count (AEC) ^[62[102]^], and relative eosinophil count (REC) ^[63[103]^] are associated with better PFS.

Low levels of hemoglobin (HGB) ^[64[104]^], red blood cell (RBC) counts, and hematocrit (HCT), which reflect anemia, are associated with shorter OS. In this case, the ICI response can be identified by the combination of NLR and HGB ^[55[95]^]. The Lung Immune Prognostic Index (LIPI) score, which takes into account the derived neutrophil-to-lymphocyte ratio (dNLR) and lactate dehydrogenase level (LDH), is also associated with OS, with patients scoring zero demonstrating favorable outcomes ^[45[85]^]. However, studies have demonstrated that high levels of dNLR are present in patients who experience early failure with ICI ^[65[105]^].

Peripheral blood biomarkers, particularly those obtained from routine examinations, represent a significant advancement in clinical practice ^[66[106]^]. Nonetheless, further investigation is necessary to fully understand and validate these findings.

Peripheral blood mononuclear cells (PBMCs) encompass various immune cell types, including monocytes, T cells, B cells, granulocytes, and natural killer cells (NK), and they play a crucial role in the initial immune defense against malignancies ^[67[108]^]. Regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSCs) are implicated in tumor growth and exert immunosuppressive functions in cancer patients ^[68[109]^].

In NSCLC patients treated with anti-PD-1 therapy, responders exhibit increased proliferation of T PD-1⁺CD8⁺cells with an effector phenotype (HLA-DR⁺ CD38⁺ BCL-2^low) and elevated expression of CD28 ^[69[110]^]. High proliferation of PD-1⁺CD8⁺ T cells ^[70[111]^], PD-L1⁻ expressing CD14⁺ monocytes ^[71[112]^] and Forkhead Box P3 (FoxP3⁺) Treg cells ^[72[113]^] may also be associated with treatment response. However, an augmented frequency of TIM-3⁺ T lymphocytes, whether CD4⁺ or CD8⁺ T cells, negatively correlates to PFS ^[73[114]^]. Likewise, high levels of CCR9⁺ or CCR10⁺CD4⁺ T cells or CXCR4⁺CD8⁺ T cells negatively correlate with ICI treatment survival outcomes ^[74[115]^].

Nivolumab-treated NSCLC patients with a high central memory/effector T cell ratio demonstrate prolonged PFS and higher tumor PD-L1 expression ^[75[116]^]. Furthermore, an increase in exhausted cells (TIGIT⁺) and a decrease in memory effector CD8⁺ T cells (CCR7⁻ CD45RA⁻) are associated with disease progression ^[76[117]^].

The ratio of Tregs to Lox-1⁺ PMN-MDSCs (TMR) can be employed after the initial immunotherapy infusion and exhibits higher sensitivity in predicting negative treatment responses ^[68[109]^]. In contrast, high levels of granulocytic myeloid-derived suppressor cells (Gr-MDSC) are related to a positive response to treatment with ICIs ^[57[97]^]. Nevertheless, further studies are necessary to elucidate these conflicting findings.

The TME comprises non-malignant stromal cells, bone-marrow-derived cells, and tumor-infiltrating lymphocytes (TILs) ^[77[119]^]. In resume, the TME immune population of consists of macrophages, dendritic cells, natural killer cells, B cells, effector T helper cells, and Treg and cytotoxic T cells ^[78[120]^].

Evidence suggests that PD-L1-positive NSCLC patients treated with ICIs who have stromal CD8⁺ effector T cells as the most abundant TIL subpopulation experience better outcome (Table 6)s. However, TILs are distributed heterogeneously, and their predictive value may be diminished in patients expressing multiple markers of T cell exhaustion ^[79[121]^], limiting their association with treatment outcomes.

The measurement of TIL populations using computational methods also shows promise, as the spatial distribution of TILs may be linked to ICI response ^[80][81[122,123]^]. Studies have classified cohorts into three main phenotypes based on TME inflammation: immune-inflamed, immune-excluded, and immune-deserted. Among these, immune-inflamed phenotypes exhibit better responses to ICIs. Compared to other biomarkers, TIL levels can be more predictive than TMB load for PD-L1-negative patients ^[82[124]^].

Proteins have also been investigated as potential biomarkers for treatment response. Examples include CD24, which co-stimulates clonal expansion of CD4 T-cells ^[83][84[125,126]^]; CD73, which is involved in lymphocyte differentiation ^[85][86[127,128]^]; and CD137, which is associated with cancer immunity ^[87][88[129,130]^]. However, CD24 positivity has been correlated with worse PFS in PD-L1 < 50% patients treated with ICIs based on IHC analyses ^[83[125]^]. In addition, the increase in CD73 and CD137 correlates with better PFS ^[87[129]^], regardless of PD-L1 status ^[85[127]^], but these results are still controversial ^[89[131]^]. The major limitation of this methodology is the scarcity of NSCLC tumor tissue ^[13[52]^].

Extracellular vesicles (EVs), which includes exosomes and microvesicles, play a crucial role in the cellular communication mechanism by transporting bioactive molecules that can influence the extracellular environment and the immune system ^[90[132]^]. EVs derived from tumor tissue have the potential to serve as non-invasive biomarkers due to their molecular composition, which reflects the complexity and heterogeneity of the tumor microenvironment ^[91[133]^]. Some molecules carried by EVs, such as PD-L1, TGF-β1, FasL, TRAIL, COX2, CD39/CD73, CTLA4, and NKG2D, are involved in tumor evasion and immunosuppression, and therefore hold promise as predictive biomarkers for immunotherapy response ^[92[134]^].

Studies have shown that responders to ICIs exhibit higher levels of tetraspanin co-stimulatory molecules, specifically CD9, CD81, and CD63, in EVs. These findings suggest that these molecules may serve as promising biomarkers associated with a better objective response rate (ORR) ^[93[135]^]. Conversely, another study identified the overexpression of TGF-β in EVs as a predictor of poorer outcomes and non-response to ICI treatment ^[94[136]^].

EVs can carry small molecules, such as microRNAs (miRNAs), that are widely studied and can also present predictive value ^[95[137]^]. For instance, EV-miR-625-5p has been described as an independent biomarker of response to ICIs in NSCLC patients with PD-L1 expression ≥50% ^[96[138]^]. Furthermore, pre-treatment concentrations of EVs with an endothelial phenotype (CD41a⁻/CD31⁺/CD45⁻) in the blood have been correlated with longer OS, PFS, and clinical response to ICIs. Proteomic analysis has revealed that responders to ICIs have distinct protein loading in EVs at baseline and during treatment ^[97[139]^].

In terms of conventional biomarkers for immune response, EVs offer an alternative method for measuring PD-L1 levels ^[98][99[140,141]^]. Studies show that increased exosomal PD-L1 expression is associated with better ORR, OS, and treatment efficacy ^[98[140]^]. However, elevated levels of extracellular vesicles are more common in non-responders ^[97][99[139,141]^], although they may not always predict a sustained response or survival ^[99[141]^]. A novel biochip has been proposed to quantify PD-1/PD-L1 proteins on the surface of extracellular vesicles and EV PD-1/PD-L1 mRNA (Au SERP). This tool has shown 72.2% accuracy in detecting ICI responders and non-responders ^[100[142]^].

Medical imaging techniques, such as positron emission tomography-computed tomography (PET-CT), can provide insights into the cellular and molecular properties of tumors. These scans have been used to correlate the expression of PD-1/PD-L1 and the response to treatment ^{[101][102][103][104}[143,144,145,146]^]. Quantifications occur by detecting ⁸⁹Zr probes that are linked to monoclonal antibodies (mAbs) administered during immunotherapy. Studies utilizing ⁸⁹Zr-nivolumab ^[103[145]^] and ⁸⁹Zr-atezolizumab ^[101[143]^] particles have demonstrated their effectiveness in predicting the response to ICIs. On the other hand, ⁸⁹Zr-durvalumab ^[104[146]^], although considered safe and feasible, only exhibited a weak correlation with treatment response.

The application of positron emission tomography-computed tomography using ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG PET-CT) has shown promise in determining the metabolic tumor volume (tMTV), which serves as a reliable predictor of pembrolizumab efficacy. Furthermore, tMTV may be a valuable tool for guiding treatment decisions in patients who require more aggressive therapeutic interventions, such as chemoimmunotherapy ^[102[144]^].

The respiratory tract microbiome has demonstrated its potential to predict the response to ICI in NSCLC patients ^{[105][106][107}[147,148,149]^]. A dysbiotic signature in the respiratory tract microbiome has been associated with tumor progression and a poorer prognosis ^[108[150]^]. Conversely, a more diverse lung microbiome is linked to higher levels of CXCL9, a chemokine associated with better ICI response ^[107[149]^].

Moreover, the application of 16S RNA sequencing has identified specific microbial enrichments in NSCLC patients with different ICI responses. Poor responders to ICIs showed an enrichment of Fusobacterium nucleatum ^[105[147]^], Haemophilus influenzae, and Neisseria perflava ^[106[148]^]. In contrast, patients with an enrichment of Veillonella dispar ^[106[148]^] and Akkermansia muciniphila (AKK) ^[109[151]^] demonstrated a favorable response to ICIs.

The use of antibiotics can alter the gut microbiome and is correlated with unfavorable clinical outcomes, as patients treated with antibiotics exhibit lower alpha diversity ^[110[152]^]. Evidence suggests that antibiotic treatment reduces the presence of CXCL9 in the lung tumor microenvironment, which may be associated with lower sensitivity to ICIs ^[107[149]^]. Further studies are needed to fully understand the intricate interactions within the cancer–microbiome–immune axis