In addition to the established value of tumour tissue PD-L1 as a biomarker for ICI treatment (34), numerous researchers have recently become interested in the response-predictive potential of soluble PD-L1 (sPD-L1). Okuma et al (35) demonstrated that 59% of patients with NSCLC with low baseline sPD-L1 levels (<3.357 ng/ml) achieved complete or partial response to nivolumab compared with only 25% of patients with high sPD-L1 levels (>3.357 ng/ml) as summarized in Table I, which lists circulating immune checkpoint proteins and their associations with immunotherapy response. The latter group also exhibited shorter time to treatment failure (TTF) and overall survival (OS) compared with the former group. Costantini et al (36) observed that sPD-L1 concentration after 2 months of treatment was significantly higher in non-responders, with a median value of 67.64 pg/ml, compared with 32.94 pg/ml in responders; however, the authors did not observe a significant difference in baseline sPD-L1 levels between the two groups (Table I). Similarly, Zizzari et al (37) showed that the mean concentration of sPD-L1 after 3 months of treatment with nivolumab was lower among responders (1.70±0.06 pg/ml) than among non-responders (57.00±12.00 pg/ml). Incorvaia et al (38) showed contradictory results in metastatic renal cell cancer, where high baseline sPD-L1 levels (>0.66 ng/ml) were associated with longer progression-free survival (PFS) after nivolumab treatment. The same study showed a decrease in sPD-L1 (from 1.09 to 0.73 ng/ml) after two cycles of nivolumab among patients with a PFS of >18 months after two cycles of treatment. This conclusion is debatable, as the authors defined high sPD-L1 concentrations as >0.66 ng/ml based on receiver operating characteristic (ROC) curve-derived thresholds, whereas Okuma et al (35), applying the same method, reported an optimal threshold of 3.357 ng/ml (Table I). It may be hypothesized that such a difference in the cut-off value could be due to the different methods of sPD-L1 detection: Incorvaia et al (38) used a home-made test, whereas Okuma et al (35) used a commercially available kit. In addition to this potential cause of difference, the cancer type may also serve a role (namely, renal cancer was the focus of the former study, whereas lung cancer was the focus of the latter).

In patients with mismatch repair-proficient colorectal cancer treated with sintilimab and regorafenib, the baseline and end-of-treatment sPD-L1 levels were not significantly different between patients with durable clinical benefit (DCB) and those without. However, focusing on the change between baseline and fourth treatment cycle, the authors found a notable increase in sPD-L1 levels among patients with progression, whereas no marked change was observed in those who achieved DCB (39) (Table I). Zhou et al (40) observed that all patients with high levels of sPD-L1 (>1.4 ng/ml) before anti-CTLA-4 (ipilimumab) treatment experienced progressive disease. In addition, patients with a 1.5-fold increase in sPD-L1 levels after 5 months of treatment were more likely to experience only a partial response to pembrolizumab or to ipilimumab. In contrast to the aforementioned studies, Boschert et al (41) reported that higher baseline sPD-L1 levels predicted a better response to ICIs in head-and-neck carcinoma. Notably, the mean sPD-L1 concentration of the responder group was 740 pg/ml vs. 94.76 pg/ml in the non-responder group (Table I).

Taken together, these results indicate that baseline sPD-L1 levels and its dynamics over the first six cycles of anti-PD-1 treatment are worth further studying as biomarkers predictive of response or treatment failure.

In melanoma, Machiraju et al (42) observed that high baseline soluble PD-1 (sPD-1) concentrations (>167 pg/ml) were associated with resistance to a combination of nivolumab and ipilimumab. Similar results were observed in melanoma by Incorvaia et al (43): Significantly lower baseline concentrations were observed among the responders to nivolumab or pembrolizumab compared with the non-responders (10.3 vs. 16.6 ng/ml, respectively). Higher baseline sPD-1 concentrations (>11.24 ng/ml) were also associated with a shorter TTF. A few years earlier, Incorvaia et al (38) published contradictory results in metastatic renal cell carcinoma: High baseline sPD-1 levels (>2.11 ng/ml) were associated with improved response and survival. Furthermore, sPD-1 significantly decreased, from 13.25 to 1.23 ng/ml, after two cycles of nivolumab among the responders (38). Notably, the optimal sPD-1 cut-off markedly differed between the two studies: In 2020, it was calculated to be 2.11 ng/ml in a cohort of 22 patients, whereas in 2023 it was estimated to be 11.24 ng/ml in a cohort of 41 patients (38,43) (Table I). Therefore, the results of these two studies should be interpreted with caution. In NSCLC, Zizzari et al (37) reported significantly lower sPD-1 levels at 3 months of treatment with nivolumab in the responder group (31 pg/ml) compared with in the non-responder group (53 pg/ml). Subsequently, by evaluating dynamic changes in sPD-1 during treatment, a lower sPD-1 level after 3 months of treatment was found in responders due to a significant decrease (from 67 pg/ml at baseline to 31 pg/ml at 3 months of treatment) of the protein during this period compared with the pre-treatment values (Table I). Thus, similar to sPD-L1, lower levels of baseline sPD-1 and/or their decrease on treatment may be associated with improved response to ICIs.

Leung et al (44) demonstrated that patients with stage IV melanoma who derived a clinical benefit from ipilimumab had significantly higher baseline serum CTLA-4 (sCTLA-4) levels than patients without clinical benefit (2,417 vs. 208 pg/ml, respectively). Furthermore, the median OS time was 43.2 months for patients with baseline sCTLA-4 >200 pg/ml, whereas it was only 5.9 months for patients with sCTLA-4 <200 pg/ml. Similar results were obtained in a larger study: The mean baseline sCTLA-4 levels for the responder, stable disease and progression group were 440.4, 268.5 and 108.4 pg/ml respectively, and patients with baseline sCTLA-4 levels >200 pg/ml had improved response to ipilimumab and survival (45). Notably, the cut-off of 200 pg/ml was identified by the ROC approach by Leung et al (44), whereas it the median sCTLA-4 concentration was used as the cut-off value by Pistillo et al (45). The median level of sCTLA-4 in Pistillo et al (45) clearly reflected significantly higher sCTLA-4 levels in patients with melanoma than in healthy subjects, in whom sCTLA was in the range of 10-40 pg/ml. sCTLA-4 in melanoma likely originates both from alternative splicing of the CTLA-4 gene in tumour cells and from activated peripheral T cells, particularly regulatory T cells (46). Therefore, sCTLA-4 reflects the global CTLA-4 burden (namely, the availability of the ipilimumab target). In other words, patients with melanoma with >200 pg/ml CTLA-4 have their immune system strongly suppressed by CTLA-4, and administration of ipilimumab provides a stronger relief from this suppression than that in patients with lower sCTLA-4 levels. Zizzari et al (37) showed that patients with NSCLC who responded to nivolumab had significantly lower sCTLA-4 levels after 3 months of treatment than non-responders (61 vs. 86 pg/ml, respectively).

Together, the aforementioned data indicate that higher pre-treatment and lower mid-treatment levels of sCTLA-4 could predict a higher probability of response to ipilimumab or nivolumab (Table I).

Tumour tissue upregulation of the alternative immune suppressive pathways is one of the mechanisms responsible for acquired resistance to ICIs (47). These modifications can result in the release into circulation of immune checkpoint regulators other than PD-1 or PD-L1. In a previous small study (n=22) on nivolumab-treated patients with NSCLC, non-responders presented higher soluble T-cell immunoglobulin and mucin-domain containing-3 (TIM-3), lymphocyte activation gene-3 (LAG-3) and B- and T-lymphocyte attenuator-4 concentrations than responders after 3 months of treatment (37). Another previous larger study (n=48) on pembrolizumab-treated patients with melanoma showed that the non-responders had significantly higher baseline sLAG-3 levels (186 pg/ml) than patients with stable disease (85 pg/ml). Furthermore, in the resistant patients, sLAG-3 was significantly increased during the first 6 weeks of treatment. These associations between sLAG-3 and response to treatment were not observed in patients treated with ipilimumab (n=23) or with a combination of ipilimumab and nivolumab (n=42) (42) (Table I). The value of sTIM-3 and sLAG-3 quantification for personalized immuno-oncology remains to be determined; however, with advances in the clinical development of TIM-3 and LAG-3 inhibitors such as sabatolimab (48) and relatlimab (49), it is highly probable that LAG-3 will form part of analyte panels to be tested in upcoming clinical trials.

Cytokines, which are essential proteins for communication between immune cells, serve key roles in both the activation and suppression of the immune response. Their blood levels are one of the best reflectors of immune system activity (50).

Interleukins (ILs) are the cytokines most extensively investigated as predictors of immunotherapy response. Several previous studies have shown that high baseline serum IL-6 (sIL-6) concentrations are associated with a poor response to PD-1/PD-L1 blockade. In NSCLC, the sIL-6 cut-off levels were similar across numerous studies, ranging from 11.1 to 13.8 pg/ml (51-54). Patients with sIL-6 above these levels have been reported to have a lower probability of responding to anti-PD-1 drugs as summarized in Table II, which lists cytokines and their associations with immunotherapy response. In squamous NSCLC, an increase in sIL-6 concentrations between baseline and the fifth treatment cycle of nivolumab alone or nivolumab with ipilimumab was observed among non-responders by Parra et al (29). In gastric cancer, Qi et al (55) showed that sIL-6 levels >13.3 pg/ml after two cycles of chemotherapy and PD-1 blockade were notably associated with a shorter PFS (Table II).

The response to treatments combining ICIs and targeted therapy has been shown to be associated with lower baseline or on-treatment levels of sIL-6 in patients. In patients with hepatocellular carcinoma treated with atezolizumab and bevacizumab, the responder group presented with significantly lower sIL-6 levels at both the baseline and after 6 weeks of treatment compared with in the non-responder group (Table II) (56). In patients with renal carcinoma treated with pembrolizumab combined with axitinib, the baseline sIL-6 levels were significantly lower in responders (8.6 pg/ml) than in non-responders (84.1 pg/ml) (57). Similarly, in a trial of nivolumab and ipilimumab, with or without enzalutamide, in metastatic castration-resistant prostate cancer, lower baseline concentrations of sIL-6 were associated with improved outcomes (58). Finally, two studies conducted in advanced melanoma treated with ipilimumab or nivolumab showed contrasting results regarding sIL-6: Cristiani et al (59) showed that low concentrations of sIL-6 (<6.80 pg/ml) were associated with improved PFS and OS, whereas Yamazaki et al (60) revealed that the concentrations of sIL-6 were markedly higher in patients with an objective response to nivolumab than in patients who experienced disease progression after this treatment (Table II). The main difference between these two studies was the IL-6 quantification time point: IL-6 was detected before treatment in Yamazaki et al vs. at 2 months after treatment in Christiani et al. With that in mind, the former study results would reflect, as the authors evoked, a stronger antitumour immunity already present before treatment and further stimulated by anti-PD-1 or anti-CTLA-4 treatment. Whereas the results of the latter study may reflect a reduction in the burden of an IL-6-producing tumour. These hypotheses should be verified in future studies by assessment of both the tumour tissue and sIL-6 levels together with the activation status of peripheral lymphocytes.

Similar to sIL-6, serum IL-8 (sIL-8) has mainly been identified as an unfavourable biomarker of response and outcome in immunotherapy studies. Schalper et al (61) observed in 107 patients with squamous NSCLC and 255 patients with non-squamous NSCLC that baseline sIL-8 >23 pg/ml was associated with a poor response to nivolumab (61). Kauffmann-Guerrero et al (52) reported similar findings in a smaller cohort treated with anti-PD-1, using 19.6 pg/ml as the cut-off value for baseline sIL-8 levels. The majority of studies that have analysed sIL-8 dynamics during treatment with ICIs have shown that an increase in sIL-8 levels compared with baseline levels is associated with a lack of response and/or with progressive disease. For example, Sanmamed et al (62) assessed patients with NSCLC treated with anti-PD-1 monotherapy and the results revealed that sIL-8 levels decreased from 20.8 pg/ml at baseline to 6.5 pg/ml at the time of best response (BR), whereas they increased from the baseline level of 15 pg/ml to 51 pg/ml at the moment of progressive disease diagnosis. Similar findings were reported in patients with squamous NSCLC or melanoma treated with nivolumab alone or in combination with ipilimumab (29,59,62) (Table II). Sanmamed et al also defined a cut-off (9.2%) of sIL-8 level change between baseline and 2-3 weeks of treatment, which best separated responders from non-responders (Table II).

As part of multi-analyte panels, several other cytokines have been assessed in the serum of patients treated with ICIs. Haddox et al (63) detected higher serum IL-4 (sIL-4) levels at baseline and at 1 week of treatment in patients with sarcoma who responded to a combination of eribulin and pembrolizumab. Furthermore, patients with gastric cancer treated with immunochemotherapy who had an increase in sIL-4 after four cycles of treatment (>1,279 pg/ml) also exhibited significantly improved PFS (55). By contrast, serum IL-5 (sIL-5) concentration was decreased between baseline levels (7.07 pg/ml) and time of BR (3.78 pg/ml) in patients with gastric cancer treated with PD-1 blockade; similar dynamics of sIL-5 were observed in a cohort of patients with NSCLC analysed by the same authors (64). Serum IL-7 and IL-17 levels were investigated in prostate cancer, where low and high baseline concentrations, respectively, were associated with an improved response to a combination of nivolumab with ipilimumab, with or without enzalutamide (58). Serum IL-10 (sIL-10) levels below a threshold of 4.3 ng/ml at baseline and after two cycles of ICI was associated with improved response in patients with renal cancer (65). However, in patients with nivolumab-treated melanoma, higher baseline concentrations of sIL-10 were associated with improved response (60) (Table II).

Zhao et al (64) showed a decrease in serum interferon γ (sIFN-γ) levels from baseline to time of BR in responders, whereas they increased from baseline to progressive disease in both NSCLC and gastric cancer. In another study on patients with gastric cancer treated with anti-PD-1, patients with low baseline levels of sIFN-γ (<10.79 pg/ml) had a median PFS of 119 vs. 267.5 days for patients with high levels of sIFN-γ (>10.79 pg/ml) (55). Higher sIFN-γ levels were also observed at baseline in melanoma responders treated with nivolumab and in sarcoma responders treated with eribulin associated with pembrolizumab (60,63). Kauffmann-Guerrero et al (52) showed contradictory results in NSCLC, where patients with sIFN-γ >16.7 pg/ml at baseline had no response to anti-PD-1 treatment (Table II).

Serum transforming growth factor β has been investigated in patients with hepatocellular carcinoma treated with pembrolizumab. The responders had markedly lower baseline concentrations than the non-responders (141.9 vs. 1,071 pg/ml, respectively) (66). Serum colony-stimulating factor 1 was reported to increase before radiological progression in a cohort of 160 patients with squamous NSCLC treated with nivolumab alone or in combination with ipilimumab (29) (Table II).

The cytokine receptors CD25 and CD27 have been studied in patients treated with PD-1 inhibitors. Patients with melanoma who responded to treatment presented higher baseline serum CD25 levels (1,348 pg/ml) compared with non-responders (451 pg/ml) (67). In the same patients, serum CD27 (sCD27) was found to be significantly lower among the responders after 1 month of treatment (1,372 pg/ml) compared with the non-responders (2,493 pg/ml) (67). Sam et al (68) showed, in two large cohorts of patients with melanoma (namely PREDIMEL, n=74 and MelBase, n=210), that sCD27 levels >100 U/ml were associated with resistance to pembrolizumab alone but not when it was combined with anti-CTLA-4 treatment. In metastatic renal cell carcinoma treated with anti-PD-1 therapy, higher circulating sCD27 levels were associated with poorer overall survival across independent cohorts (Colcheckpoint cohort and BIONIKK-like cohort), although the optimal prognostic cutoff varied between studies (91.8 vs. 112.71 U/ml). Despite differences in ROC-defined cut-off values across cohorts, elevated sCD27 was consistently associated with worse survival outcomes in patients with metastatic renal cell carcinoma receiving anti-PD-1 therapy (69) (Table II). Finally, a comprehensive article on the CD27-CD70 axis in cancer immunotherapy has indicated that high blood levels of CD27 are associated with resistance to PD-1/PD-L1 blockade in several types of cancer (68). Thus, quantification of circulating CD27 may serve to select patients for ICI treatment escalation.

As with cytokines, serum chemokines can provide information on the mechanisms underlying ICI resistance, often linked to the appearance of immune-related AEs (50). In melanoma, Kasanen et al (70) showed a significant increase in three C-X-C motif chemokine ligands (CXCLs), namely CXCL9, CXCL10 and CXCL11, between baseline and 1 month of nivolumab treatment among the responders, while no significant change was observed among the non-responders. In a larger cohort of patients with ipilimumab-treated melanoma, serum CXCL11 levels <35 pg/ml were associated with response to treatment (71). Similar observations were reported in another study regarding serum CXCL9 (sCXCL9) baseline levels, which were significantly higher in responders to PD-1 inhibition (75 pg/ml) than in non-responders (0.4 pg/ml); sCXCL9 was much higher in responders than in non-responders also after 1 and 2 months of treatment (67). This protein was also studied in patients with hepatocellular carcinoma treated with anti-PD-1. sCXCL9 levels <478 pg/ml were primarily associated with progressive disease, whereas levels exceeding twice the threshold value of 478 pg/ml were associated with improved response to treatment (72). Serum chemokines including CXCL8, CXCL9, CXCL10, CXCL12 and CXCL13 have been evaluated in ICI-treated NSCLC cohorts, with heterogeneous and sometimes conflicting results. While several studies have reported that increased serum levels of CXCL8, CXCL10 or CXCL12 are associated with poorer outcomes (reduced OS, PFS or DCB), other studies have shown that higher CXCL9 and CXCL10 levels are associated with improved response and PFS. In addition, an early decrease in CXCL8 levels during treatment has been associated with better survival outcomes in patients with NSCLC (72). Parra et al (29) observed an increase in CXCL13 among non-responding patients with squamous cell lung cancer between baseline and cycle 5 (week 9 of treatment with nivolumab in monotherapy or in combination with ipilimumab). Furthermore, CCL2 was the focus of a study on a cohort of 24 patients with melanoma treated with nivolumab; patients responding to treatment presented with concentrations of >446.72 pg/ml (59). A summary of the reported chemokines and their associations with immunotherapy response is presented in Table III.

Among the proteins involved in the regulation of cellular processes, the most investigated ones in terms of their response to ICIs are those involved in mesenchymal cell proliferation and angiogenesis, including fibroblast growth factor (FGF), hepatocyte growth factor (HGF), vascular endothelial growth factor (VEGF) and angiopoietin-2 (ANG-2).

Higher baseline concentrations of serum HGF (sHGF) have been observed in patients with melanoma who responded to PD-1 or CTLA-4 blockade and to their combination (73). Similarly, in a cohort including various cancer types, patients who responded to anti-PD-1 or anti-PD-L1 therapy showed significantly higher baseline levels of sHGF compared with the non-responders (2,365 vs. 1,769 pg/ml, respectively) (54). In patients with hepatocellular carcinoma treated with a combination of atezolizumab and bevacizumab, Yang et al (74) observed a significant increase in serum FGF-19 concentrations between baseline and the BR time-point, whereas serum ANG-2, together with serum VEGF-D, significantly increased at the time of progression in patients who initially responded to treatment. Cristiani et al (59) showed that patients with melanoma with better response to two cycles of nivolumab had serum VEGF levels <413.06 pg/ml at that time-point. Table IV summarizes circulating cellular modulators and their reported associations with response to immunotherapy. These findings indicate that circulating angiogenesis markers are worth further validation in the ICI treatment setting, particularly to indicate development of resistance to ICIs, which could be suppressed by angiogenesis inhibitors.