PD-L1 expression has limited value for predicting the efficacy of immune-checkpoint blockade (ICB). In the KEYNOTE-042 trial, the value of PD-L1 expression for predicting clinical benefit was low (AUC = 0.6–0.7) (19). In our previous study, only 48.10% of NSCLC patients with ≥ 1% PD-L1 expression can benefit from pCR, with an AUC of 0.651. Another problem is that intratumoral heterogeneity cannot be avoided when the biopsy tissue is taken from a specific site. In addition, biopsy does not provide real-time monitoring due to its invasive nature. In a trial of sintilimab NIO, there was no obvious correlation between PD-L1 expression in biopsy tissue before treatment and in the surgical sample after treatment (6). In the NEOSTAR trial, PD-L1 expression after NIO was not associated with MPR (13). Tumor microenvironment (TME) test may be also unsuitable for prediction of NCIO efficacy, because a small biopsy specimen was insufficient for assessing TME. TMB is also not a routine test in NSCLC, especially in lung squamous cell carcinoma. For immune cell subsets, there are so many subsets that no criteria have been established (7, 15). In addition, all of the above-mentioned tests need extra infrastructure or financial resources, which limit their routine clinical applications. Peripheral blood routine tests may compensate these shortcomings and provide similar predictive values (16, 17). Peripheral blood and TME were closely correlated through circulating immune cells, cytokines, soluble protein, and ctDNA (16, 17, 20). Immune cells, such as neutrophils, lymphocytes and monocytes can directly infiltrate the TME by innate or adapted immune response, and they may also be modified by the tumor cells (16, 17). Some soluble proteins or enzymes, such as LDH and AST, are produced by cancer cells or damaged normal cells (16), and coagulation factors are correlated with systematic inflammatory response, and all of them can reflect the changes of TME (18).
To our knowledge, this is the first study to report that the absolute or changed values (or rates) in routine blood cells, and biochemical, electrolytic and coagulation tests at week 6 after two cycles of NCIO, as well as coagulation parameters at baseline, were significantly associated with pCR/MPR. Of note, we found △monocyte count' rate and absolute basophil count at week 6 were exclusively decreased in pCR and MPR groups, but increased in their reciprocal groups, which indicated these two indicators may be ease-used and promising biomarkers. What' more, we found that AST and blood magnesium levels at week 6 were new biomarkers for NCIO or immunotherapy of cancer. The AUC value (0.6–0.7) of each significant factor was similar to that of PD-L1 expression. Based on the five independent predictive factors, the total predictive ability of models were higher than sole factor with AUC of 0.785 and 0.881 for pCR and MPR, respectively. After the cross-validations, the results indicated that the models were satisfactory and steady. The predictive accuracy of the MPR model was higher than that of the pCR model, which was consistent with the pathological criterion that MPR had more viable tumor cells (0–10%) than pCR (0%).
In our study, the baseline WBC and neutrophil counts did not differ obviously. However, they decreased at week 6 after NCIO. The absolute WBC and neutrophil counts at week 6, their changes and changes' rates from baseline to week 6 were all associated with pCR/MPR. △WBC' rate and △WBC were independent factors for pCR and MPR, respectively. pCR, non-pCR and MPR groups all had a significantly lower WBC count at week 6 compared with baseline (Supplementary Table 3). Neutrophils are a major component of WBCs. There was no related report on NIO or NCIO, and evidences from palliative ICB showed that a lower WBC or neutrophil count was associated with better clinical outcomes. In a study of 54 advanced NSCLC patients treated with nivolumab, baseline WBC count < 8800/µl and neutrophil count < 6000/µl were correlated with better disease control rate (DCR) (56% vs. 18%, P = 0.01) and longer OS (17.7 months vs. 3 months, P = 0.004) (21). In a study of 44 advanced NSCLC patients treated with ICB, higher baseline WBC count [hazard ratio (HR) = 3.03, P = 0.001] and neutrophil count (HR = 3.20, P = 0.001) independently predicted poor OS (22). In the study of Facchinetti et al, after the first cycle of ICB in patients with metastatic NSCLC, a lower neutrophil count was found in responders (3400/mm3) than in nonresponders (6300/mm3) (P = 0.02), although the difference of baseline values was not significant (23). In the study of Wen et al, lower WBC and neutrophil counts at week 6 after two cycles of ICB were correlated with higher objective response rate (ORR) (40.8% vs 9.8% and 43.2% vs 10.9%, P = 0.001) in 90 patients of advanced NSCLC (24). Our results were consistent to those of Facchinetti and Wen (23, 24). Blood neutrophil count is directly associated with the intratumoral neutrophil count (17), and elevated neutrophil count in blood can promote tumor growth, invasiveness, metastases and drug resistance through inhibiting cytotoxic T cells and secreting proinflammatory cytokines (17). What' worse, tumor cells can in turn stimulate the proliferation of neutrophils by secreting growth factors (17).
There was different feature of basophils to WBCs or neutrophils. In the pCR and MPR groups, basophil count was exclusively decreased at week 6 compared with baseline [both 0.3 vs. 0.4 (*109/l), P = 0.006 and 0.010], but it was similar in the non-pCR and non-MPR groups at week 6 compared with baseline [both 0.04 vs. 0.04 (*109/l), P = 0.317 and 0.428] (Supplementary Table 3). This indicated that basophil count may be an ease-used and promising biomarker for efficacy of NCIO in NSCLC. The basophil count at week 6 was an independent predictor for MPR (HR = 5.05, P = 0.023). Basophils may be an useful biomarker for efficacy of immunotherapy or chemotherapy of cancer. In a study of chemoimmunotherapy in 63 advanced gastric cancer patients, the higher basophil count group (> 20.0/µL) at baseline had a lower ORR (17.4% vs. 67.5%, P = 0.0001), poor progression-free survival (PFS) (4.0 months vs. 15.0 months, P = 0.0003), and poor OS (both not reached, P = 0.027) (25), and high basophil count was an independent predictor for ORR, PFS and OS (all P < 0.05) (25). Similar results were not observed in the chemotherapy-alone group (P > 0.05) (25). In a study of 448 postoperative gastric cancer patients, a high number of tumor-infiltrating basophils was independently correlated with poor OS (HR = 2.67, P < 0.001) and poor chemotherapy response in stage III patients (P = 0.007) (26). A similar prognostic role of basophils was also shown in prostate cancer (27). Basophils have been extensively studied in inflammatory disease but its mechanism was poorly studied in cancer. A relationship between peripheral and tumor-infiltrating basophils has been observed (R = 0.683, P = 0.005) (25). Basophils in the TME can promote M2 macrophage polarization through secretion of interleukin (IL)-4. M2 macrophages act as immunosuppressive cells by enhancing tumor angiogenesis and metastasis (28). There was a positive correlation between basophils and IL-4 expression/M2 polarized macrophages, but a negative correlation with interferon receptor expression in gastric cancer tissue (28). However, the situation may be complicated, because basophils have also been reported to have an inhibitory role in cancer by secreting chemokine CC ligand (CCL)3, CCL4 and CCL5, and it was suggested that basophils may be new targets for treatment of those inflammatory cold tumors (28). In general, the biological role of basophils in cancer is largely unknown and needs further study.
Monocytes may be another ease-used and promising biomarker for efficacy of NCIO in NSCLC. The changed feature of monocytes after NCIO was similar to and better than basophils, which was different from WBCs or neutrophils. After treatment, we found that △monocyte count' rate were exclusively significantly decreased in pCR and MPR groups, but increased in their reciprocal groups (Table 2), even the baseline count was higher in the pCR and MPR groups (Supplementary Table 1 or 3). The △monocyte count' rate was a common independent predictor for pCR and MPR. Elevations of monocyte count in blood were found in lung cancer patients compared with healthy donors (29). There are evidences of the predictive value of monocytes in palliative ICB cancer patients. In advanced melanoma treated with ipilimumab, lower baseline monocyte count (< 650/µL) were correlated with higher DCR (P < 0.05) and better OS (P = 1.35⋅10−8) (30). In a study of 59 patients with stage IV melanoma treated with ipilimumab, baseline monocyte count was higher in nonresponders than in responders (P = 0.04) (31). After the first cycle of treatment, monocyte count of nonresponders was obviously higher than baseline and in responders (P < 0.05), with similar findings for monocytic myeloid-derived suppressor cells (moMDSCs), but not granulocytic MDSCs (grMDSCs) (31). A higher peripheral monocyte count has been reported as an independent predictor for poor 5-year disease-free survival in hepatocellular carcinoma (32). Intratumoral monocytes can convert to tumor-associated macrophages that contribute to the deactivation of cytotoxic T cells through secreting numerous immunosuppressive cytokines, such as IL-10, transforming growth factor-β, and nitric oxide (31, 33). moMDSCs are a class of key immunosuppressive cells in the TME and hinder the activities of cytotoxic T cells, NK cells and antigen-presenting cells (20). moMDSCs also play an immunosuppressive role by expressing PD-L1. In the study of Gebhardt et al, after the first ipilimumab infusion, expression of PD-L1 on moMDSCs was not downregulated, but it was significantly downregulated on grMDSCs (31).
Although some studies reported a higher baseline lymphocyte count associated with higher response to ICB of advanced cancer (34, 35), there was no obvious difference between the groups at baseline in our study. Interestingly, after two cycles of NCIO, lymphocyte count was significantly lower in the pCR group than that in non-pCR group, and a lower tendency in the MPR group than that in non-MPR group. In a study of neoadjuvant treatment of locally advanced rectal cancer, the baseline peripheral lymphocyte count had no correlation with tumor response (36). However, the lymphocyte count reduction rate was obviously higher in good responders than poor responders (P < 0.05) (36). In a study of 90 NSCLC patients received PD-1-inhibitor-based therapy, a lower lymphocyte count at week 6 was independent predictor for better OS (HR = 2.182, P = 0.046) (37). Currently, it is unknown why low lymphocyte count after treatment is associated with better therapeutic response. A study of ICB in melanoma showed that it may be due to homing of effective T cells to the TME after immunotherapy. Melanoma patients with early disappearance of tumor-associated antigens and reactive T cells (CD4+ and CD8+) at a median 42 days after ICB had better PFS (not reached vs. 4 months, HR 0.13, P < 0.001) and OS (1-year: 95.2% vs. 75.9%, 2-year: 83.3% vs. 56.5%, HR 0.21, P = 0.021). More of these reactive T cells was observed in tumor tissue of patients with early disappearance of peripheral T cells, which demonstrated that it indeed correlated with better outcome (38). Similarly, in a study of NSCLC treated with ICB, fewer peripheral blood CD8+ T cells pretreatment was associated with a durable clinical benefit (accuracy = 70%) (39). In our previous study, lower peripheral blood CD+4 cells pretreatment was associated with pCR. Besides above possible reasons, we infer that a greater decrease in lymphocyte count may be associated with more sensitive treatment, but the mechanism needs further investigation. We did not find any correlation of neutrophil-to-lymphocyte ratio, platelet-to- lymphocyte ratio, and lymphocyte-to-monocyte with the efficacy of NCIO, no matter at baseline or at week 6 (data not shown).
AST may be a new, effective and promising predictor of response to ICB in cancer. AST level at week 6 was an independent, and the strongest predictor that associated with MPR (OR = 15.78, P < 0.001). Besides, compared to baseline, AST level at week 6 was also exclusively obviously increased in non-pCR and non-MPR groups (P < 0.05), but only slightly increased in pCR and MPR groups (P > 0.05) (Supplementary Table 3). To our knowledge, this has not been reported in any other study of immunotherapy in cancer. Some related reports were found. A study of NACT in esophageal carcinoma showed that a lower baseline alanine aminotransferase (ALT) level was found in the ORR group compared to the non-ORR group (12.0 U/L vs. 18.0 U/L, P = 0.003) (40). Other studies reported the prognostic value of AST/ALT ratio. In a study of renal cell carcinoma treated with tyrosine kinase inhibitors, a higher AST/ALT ratio (≥ 1.2) was independently associated with poor cancer-specific survival (HR = 1.61, P = 0.008) and OS (HR = 1.69, P = 0.003) (41). In a study of 99 patients with bladder cancer, a higher AST/ALT (≥ 1.3) was independently associated with lower OS (18.28 months vs. 40.84 months, P < 0.001, HR = 3.09, P = 0.01) (42). In contrast, a higher ALT/AST ratio was an independent adverse prognostic indicator in gastric cancer (n = 231) (43). Metabolic reprogramming provides sufficient energy and materials for cancer cell growth (41). AST contributes to cancer cell growth and drug resistance through reducing reactive oxygen species and increasing NADPH synthesis (44). AST is also involved in glycolysis, with an increase in AST/ALT ratio (41). AST is synthesized in various tissues, whereas ALT is mainly enriched in the liver. Therefore, a higher AST or AST/ALT ratio is common in cancer (41).
Baseline coagulation parameters were the only biomarkers that significantly differed between groups. They also significantly differed between groups at week 6 after NCIO. We observed a significantly higher baseline PT, PT ratio and INR in the pCR group, and a higher PT ratio and FIB in the MPR group compared with their reciprocal groups. Higher PT, PT ratio or INR all indicate a lower coagulation state, but higher FIB indicates a higher coagulation state. We carried out further study to determine why they occurred in the same groups. We observed a tendency towards higher tumor stage (higher tumor burden) in the pCR and MPR groups (Supplementary Table 4), which was consistent with the CheckMate 816 trial (14). Except for baseline FIB being higher than the upper normal limit (4g/l), all the other 3 factors were normal. We inferred that increased FIB may act more effective coagulation function than normal PT, PT ratio and INR level. At week 6 after NCIO, FIB decreased more stronger in the pCR and MPR groups than in their reciprocal groups. Finally, FIB became higher, and PT, PT ratio and INR became lower in non-pCR/non-MPR groups, all of them consistently indicated a relative higher coagulation status in the non-pCR and non-MPR groups. This result was further confirmed by a higher tendency of D-dimer level was observed in the non-pCR and non-MPR groups compared with their reciprocal groups (Table 2). Cancer patients used to have coagulation dysfunction with a higher coagulation status and poor prognosis (18, 45). Thus, we suggest that coagulation status after therapy may be more valuable to judge clinical outcomes than at baseline.
In addition, we found low blood magnesium level at week 6, a new biomarker in ICB of cancer, was correlated with pCR. There was no related study in chemotherapy or ICB of cancer. A meta-analysis of 1723 patients with metastatic colorectal cancer treated with cetuximab- or panitumumab-based chemotherapy showed that hypomagnesemia after therapy was correlated with better ORR (RR = 1.81), PFS (HR = 0.64), and OS (HR = 0.72) (all P < 0.05) (46). LDH is also an important biomarker that can reflect the efficacy of ICB in cancer. High LDH was reported to correlate with poor ORR, PFS and OS after ICB of various cancers (16, 20). Due to the small sample size, it is regrettable that LDH was not entered into our multivariate analysis.
The present study had some limitations. Firstly, we did not detect the immune cells in surgical specimens after NCIO at present. Secondly, we did not detect the changes in the various parameters over a longer time, due to different cycles of therapy and limited data. Thirdly, LDH was not entered into multivariate analyses, because of a limited number of cases. Fourthly, the correlations of these parameters and prognosis was not analyzed, became the mean follow-up time was only 1 year which was immature.
In summary, our study showed that the absolute or changed values (or rates) of peripheral routine blood cells, and biochemical, electrolytic and coagulation parameters at week 6 after therapy better predicted the efficacy of NCIO in NSCLC than at baseline. Among them, we found five independent predictive factors, and found that change' rate of monocyte count, and absolute basophil count, AST and blood magnesium level at week 6 may be promising biomarkers for response to NCIO. Finally, we established clinical predictive models with satisfied AUC values, which were consistent with the pathological criteria of NCIO. All of above results could help to distinguish the population that is suitable for NCIO in NSCLC, and provide the clinical data for further study of the biological rationale of NCIO.