Clinical significance of preoperative neutrophil‐lymphocyte ratio and platelet‐lymphocyte ratio in the prognosis of resected early‐stage patients with non‐small cell lung cancer: A meta‐analysis

Abstract Background Poor prognosis is linked to peripheral blood levels of preoperative platelet‐lymphocyte ratio (PLR) and neutrophil‐lymphocyte ratio (NLR) in many advanced cancers. Nevertheless, whether the correlation exists in resected early‐stage cases with non‐small cell lung cancer (NSCLC) stays controversial. Consequently, we performed a meta‐analysis to explore the preoperative NLR and PLR's prognostic significance in early‐stage patients with NSCLC undergoing curative surgery. Methods Relevant studies that validated the link between preoperative NLR or PLR and survival results were found via the proceeding databases: PubMed, Embase, Cochrane Library, and Web of Science. The merged 95% confidence interval (CI) and hazard ratio (HR) was employed to validate the link between the NLR or PLR's index and overall survival (OS) and disease‐free survival (DFS) in resected NSCLC cases. We used sensitivity and subgroup analyses to assess the studies' heterogeneity. Results An overall of 21 studies were attributed to the meta‐analysis. The findings indicated that great preoperative NLR was considerably correlated with poor DFS (HR = 1.58, 95% CI: 1.37–1.82, p < 0.001) and poor OS (HR = 1.51, 95% CI: 1.33–1.72, p < 0.001), respectively. Subgroup analyses were in line with the pooled findings. In aspect of PLR, raised PLR was indicative of inferior DFS (HR = 1.28, 95% CI: 1.04–1.58, p = 0.021) and OS (HR = 1.37, 95% CI: 1.18–1.60, p < 0.001). In the subgroup analyses between PLR and DFS, only subgroups with a sample size <300 (HR = 1.67, 95% CI: 1.15–2.43, p = 0.008) and TNM staging of mixed (I‐II) (HR = 1.47, 95% CI: 1.04–2.07, p = 0.028) showed that the link between high PLR and poor DFS was significant. Conclusions Preoperative elevated NLR and PLR may act as prognostic biomarkers in resected early‐stage NSCLC cases and are therefore valuable for guiding postoperative adjuvant treatment.


| INTRODUCTION
Non-small cell lung cancer (NSCLC), which makes up about 85% of lung malignancies, is the most repeatedly diagnosed cancer globally and the leading mortality reason among all other cancers. 1 Notwithstanding the recent rapid advancements in molecular therapeutic strategies and immunotherapy, surgery is still the go-to treatment for people with early-stage NSCLC. However, when contrasted to thoracotomy, minimally invasive surgery (MIS), including video-assisted thoracoscopic surgery (VATS), can significantly reduce the time patients need to stay in the hospital and the number of postoperative complications they experience. 2,3 Furthermore, the tumor-node-metastasis (TNM) classification system can be utilized to some extent for determining treatment strategy and prognostic evaluation in patients with NSCLC. 4,5 This is valuable for clinicians, for whom the prognosis of cases with lung cancer has always been the focus. Nevertheless, patients with similar TNM staging continuously have different clinic survival outcomes.
Due to accumulating studies, it has become increasingly evident that inflammation can affect the progression and invasion of lung cancer. 6,7 Systematic inflammatory cells, along with their calculated ratios, including plateletlymphocyte ratio (PLR), lymphocyte-monocyte ratio (LMR), and neutrophil-lymphocyte ratio (NLR), act as independent factors which influence the cases' prognosis with different cancers, involving lung, 8 breast, 9 and nasopharyngeal malignancies. 10 A published meta-analysis, involving 244 studies of cases with resected pan-cancers, demonstrated that a close association exists between raised NLR/PLR and poor OS or cancer-specific survival (CSS) and between raised LMR and poor OS. 11 Another metaanalysis, including cases with NSCLC medicated with immune checkpoint inhibitors (ICIs), also indicated that raised PLR and NLR were related to inferior progressionfree survival (PFS) and OS. 12 Nevertheless, the link between NLR/PLR and survival outcomes in cases with resected NSCLC stays indefinite. Therefore, a comprehensive meta-analysis is warranted to validate the NLR/PLR's prognostic significance in patients with NSCLC.
Accordingly, this meta-analysis includes numerous relevant studies to explore the link between peripheral blood NLR/PLR and survival results in resected early-stage NSCLC cases.

| Search strategy
We searched the literature from the four following databases: PubMed, Embase, Cochrane Library, and Web of Science. The primary deadline for literature retrieval was March 2022. The search strategy was performed using the following terms: ("Lung neoplasms" OR "pulmonary neoplasms" OR "lung cancer" OR "Carcinoma, Non-Small-Cell Lung" OR "non-small cell lung cancer" OR "NSCLC") AND ("Surgical Procedures, Operative" OR "surgery" OR "operative therapy" OR "operation" OR "operative procedures" OR "invasive procedures") AND (("neutrophil lymphocyte ratio" OR "neutrophil-to-lymphocyte ratio" OR "NLR") OR ("platelet lymphocyte ratio" OR "plateletto-lymphocyte ratio" OR "PLR")). The specific retrieval strategy can be found in Supplementary Text S1.

| Exclusion and inclusion criteria
Studies gathering the following requirements were involved: (a) populations involving early-stage NSCLC patients; (b) populations of NSCLC patients for whom curative surgical resection was performed; (c) the link between NLR/PLR and prognostic indicators of OS and/or disease-free survival (DFS) were investigated; (d) the 95% confidence interval (CI) and hazard ratio (HR) could be obtained from the original studies; (e) the literature was written in English.
The exclusion criteria were as follows: (a) reviews, case reports, abstracts, letters, and expert opinions; (b) populations of patients with other primary tumors; (c) surgical resection was not performed; (d) the studies had insufficient data to conclude on the HR and 95% CI; (e) literature with Newcastle-Ottawa Scale (NOS) scores <6; (f) the literature was not published in English.

Conclusions:
Preoperative elevated NLR and PLR may act as prognostic biomarkers in resected early-stage NSCLC cases and are therefore valuable for guiding postoperative adjuvant treatment.

K E Y W O R D S
meta-analysis, neutrophil-lymphocyte ratio, non-small cell lung cancer, operation, plateletlymphocyte ratio, prognosis 2.3 | Literature's data extraction and quality validation Two reviewers (Weibo Cao and Haochuan Yu) independently extracted relevant data and evaluated the studies' quality. Any disagreements were then resolved through discussion and consensus.
The first author's name, the year of publication, study design, country of origin, study design, sample size, pathological category of lung cancer, TNM, median follow-up (months), NLR/PLR cut-off values, and survival data (DFS and/or OS) were all extracted from each study. When both multivariate and univariate analyses were run, the multivariate analysis's HRs were extracted first. Our way of determining cut-off values for subgroup analysis was exploited according to the median value from all of NLR or PLR cut-off in the studies recruited in this meta-analysis.
We evaluated the literature's quality involved according to the scoring system of the NOS. The NOS contains three aspects: patient selection, comparability, and outcome assessment. Literature with scores of ≥6 was considered of high quality. Studies with lower scores were regarded as low quality and therefore excluded from the analysis.

| Statistical analysis
Higgins I 2 and Cochran's Q test statistics were utilized to evaluate the studies' heterogeneity. p ≤ 0.10 or I 2 ≥ 50% were significant heterogeneity indicative, and in such cases, a random-effect model was utilized. Conversely, a fixedeffect model was utilized to evaluate studies that did not present significant heterogeneity. Subgroup analyses dependent on the sample size, TNM, and cut-off values were also performed to investigate potential factors influencing the NLR/PLR's prognostic significance. Sensitivity analyses were then performed to determine any heterogeneity's source and assess the results' stability. Publication bias assessment was conducted via Funnel plots and Egger's test, in which p < 0.05 was deemed statistically significant. All statistical analyses were done utilizing STATA 15.0 (Stata Corporation, College Station, TX).

| Literature search and study characteristics
As per the search approach, 1006 studies were initially identified, of which 290 were ruled out owing to duplication. By screening literature titles and abstracts, a further 634 studies were excluded based upon the exclusion criteria, resulting in 82 full-text studies. Of these, 21 studies met the inclusion criteria and were subsequently involved in the meta-analysis.  The literature selection process is outlined in Figure 1. Table 1 lists the studies' primary traits that made up the meta-analysis. In conclusion, case sample sizes in the studies ranged from 134 to 2027, and the studies were all published between 2014 and 2021. Out of the included studies, six focused on NLR only, one concentrated on PLR only, and 14 evaluated both NLR and PLR. We had 20 NLR studies with 13,915 lung cancer cases and 15 PLR studies with 7484 lung cancer cases in the meta-analysis. NOS scores for the quality of the involved literature varied from 6 to 9, demonstrating a high quality of literature among the 21 studies. Quality assessment of involved studies can be found in Supplementary Table S1.

| NLR's impact on DFS
Of the 21 studies, 20 reported the correlation between survival results and NLR in resected early-stage NSCLC cases. Of these, 13 studies provided information on DFS (Figure 2A), which recommended that raised NLR was correlated with poor DFS (HR = 1.58, 95% CI: 1.37-1.82, p < 0.001). Owing to the great heterogeneity detected (I 2 = 60.90%, p = 0.002), a random-effect model was utilized to evaluate the studies. From the subgroup analyses, a strong association between raised NLR and poor DFS was noted for all subgroups, indicating the reliability of the meta-analysis. The findings of this can be found in Table 2.
From sensitivity analyses ( Figure 2B, Supplementary  Table S2), it was conducted that a single study's omission could not statistically affect the NLR's impact on DFS. Nevertheless, after excluding the study of Yan et al. (2020), the heterogeneity was found to decrease significantly (I 2 = 39.60%, p = 0.077), with a pooled HR of 1.65 (95% CI: 1.47-1.86, p < 0.001).

| Influence of NLR on OS
An overall of 17 studies investigated the NLR's impact on OS, among which high heterogeneity was detected (I 2 = 68.00%, p < 0.001). Thus, the random-effect model was again utilized, resulting in a pooled HR of 1.51 (95% CI: 1.33-1.72, p < 0.001). Overall, results from these studies revealed that great NLR was linked with worse OS in cases with NSCLC ( Figure 3A). This can be seen from the subgroup analyses between OS and NLR outlined in Table 2, indicating a significant association between raised NLR and poor OS.
Furthermore, sensitivity analyses exploring the potential heterogeneity of OS ( Figure 3B) suggested that the omission of a single study would not be able to statistically affect the NLR's impact on OS in the metaanalysis. In terms of OS (Supplementary Table S3), it was found that heterogeneity decreased to some extent (I 2 = 61.00%, p = 0.001) after excluding the study of Huang et al. (2018), with a merged HR of 1.46 (95% CI: 1.29-1.65, p < 0.001).
In terms of DFS ( Figure 4B, Supplementary Table S4)

| PLR's influence on OS
Data relating to the PLR's effect on OS was provided by 13 studies (Figure 5A). High heterogeneity (I 2 = 59.00%, p = 0.004) was found among these studies. Thus, a metaanalysis was conducted utilizing the random-effect model, from which a strong relation was found between high PLR and worse OS (HR = 1.37, 95% CI: 1.18-1.60, p < 0.001). For the subgroup analyses between PLR and OS (Table 3), all subgroups indicated a significant link between higher PLR and shorter OS.

| Publication bias
Funnel plots were found not visually asymmetric as per the influence of NLR on DFS ( Figure 2C) and OS (Figure 3C), indicating a significant risk of publication bias. This observation was further validated by the findings of Egger's test for DFS (p = 0.001) and OS (p < 0.001). Subsequently, trim  and fill methods were done to explore the publication bias's impact on effect estimates. No statistically significant alterations were noticed in the findings (Figures 2D and   3D). As per the PLR's influence on DFS ( Figure 4C) and OS ( Figure 5C  bias, accommodated by funnel plots which were not asymmetrical from visual inspection. Moreover, trim and fill methods again indicated no statistically significant changes in the results (Figures 4D and 5D).

| DISCUSSION
Surgery has continuity as the radical medication strategy preferred for cases with early-stage NSCLC. It is also essential in the comprehensive modality of several cases with advanced-stage NSCLC. 34 In recent decades, several independent prognostic biomarkers have been discovered in patients with resected NSCLC. [35][36][37] Previous studies have demonstrated that NLR and PLR, as host inflammatory markers, are often used as valuable prognostic indicators for some cancers. [38][39][40][41] However, it remains unclear if NLR and PLR still have a similar effect in cases with resected NSCLC.
In the present study, we merged 20 NLR studies with 13,915 lung cancer cases and 15 PLR studies with 7484 lung cancer cases to further find out the NLR and PLR's prognostic impact in resected early-stage cases with NSCLC. The meta-analysis found that increased preoperative NLR in peripheral blood was correlated with poor DFS and OS. Despite high heterogeneity, the prognostic effect of NLR was not weakened by all subgroup analyses dependent on sample size, TNM, and cut-off value, which supports the reliability of the meta-analysis. Our combined findings regarding PLR still showed that high PLR was connected to poor DFS and OS. In terms of OS, each subgroup revealed a connection between longer OS and higher PLR. For the subgroup analyses between PLR and DFS, only subgroups with sample size <300 and TNM staging of mixed (I-II) indicated a significant correlation between raised PLR and poor DFS. These results suggest that the pooled results between PLR and DFS lack stability, which may be attributed to the limited studies in the meta-analysis.
The underlying mechanisms of the link between great preoperative NLR/PLR and poor survival results in NSCLC cases remain unknown. It has been noticed that  of immunosuppression and thus impair anti-tumor immune function. 43 Furthermore, inflammatory neutrophils can promote tumor angiogenesis and tumor progression by releasing vascular endothelial growth factor (VEGF) and Gelatinase A (MMP-2). 44,45 Thus, raised neutrophils are often correlated with poor prognosis in cancer cases. Several studies have shown that absolute lymphocytes can anticipate long-term survival outcomes in cancer patients. 46,47 Consequently, lymphocytes play an essential role in immune defense, and reduced lymphocytes are often associated with poorer survival in patients. Platelets were also often associated with biological processes that allowed tumors to escape immune defense, thereby promoting tumor progression and dissemination. 48 It may therefore be concluded that NLR and PLR act essential functions in the inflammatory and anti-inflammatory balance in the human immune response. As a literature-dependent meta-analysis, several restrictions of this meta-analysis should be noted. First, most of the literature involved were retrospective studies; hence, selection bias is present. More prospective studies are therefore required to affirm our findings. Second, only literature published in English was included in our study, with studies in other languages and unpublished studies not included, which reduced the data available for analysis. It is also possible that unpublished studies contain results considered 'unfavorable' or 'negative' by the researcher or the publisher, which may otherwise have contributed to this meta-analysis. Third, characteristics such as sample size, region, and TNM differed throughout studies, which may also be the source of high heterogeneity. Fourth, we cannot rule out the possibility that non-tumor-related factors influence the blood markers of patients. Moreover, 19 out of 21 studies did not specify the pathological types of NSCLC and we could not perform subgroup analyses based on pathological type. Finally, there was a lack of standardization of cut-off values for NLR or PLR in the included studies in the meta-analysis. The cut-off values of NLR (range: 1.5-5.0) and PLR (range: 81.07-300) differed throughout the involved studies. These cut-off values were defined in different ways. For instance, in the study of Yuan et al. (2017), the appropriate cut-off values for NLR and PLR were termed by X-tile software. 20 In contrast, in the study of Chen et al. (2018), the optimal values of PLR and LMR were determined by receiver operating characteristic (ROC) curve analysis. 21 Therefore, the lack of standardization of NLR and PLR cut-off values limits the translation of study results to clinical application. It was necessary to standardize the cut-off values of preoperative NLR and PLR.

| CONCLUSIONS
This meta-analysis concludes that poor DFS and OS in resected early-stage cases of NSCLC are strongly correlated with preoperative peripheral blood high NLR or PLR. This suggests that levels of NLR and PLR may serve as prognostic biomarkers in the clinic and are valuable measures for guiding patients' postoperative adjuvant treatment. In the future, more prospective, multi-center, and large-sample studies are needed to affirm the findings of this metaanalysis and promote its clinical application, especially in combination with other prognostic biomarkers.

AUTHOR CONTRIBUTIONS
The authors all participated in designing the meta analysis. Weibo Cao, Shuai Zhu, and Ning Zhou participated in the literature search. Weibo Cao and Haochuan Yu independently extrated relevant data and evaluated the quality of the studies. Tong Li, Fan Ren, and Xi Lei performed the statistical analysis. Weibo Cao wrote the first draft. Song Xu and Lingling Zu revised the article. All authors agreed on the final version of the manuscript.

FUNDING INFORMATION
The present study was funded by the National Natural Science Foundation of China (82172776), Tianjin Science and Technology Plan Project (19ZXDBSY00060) and (303078100412), Tianjin Key Medical Discipline (Specialty) Construction Project (TJYXZDXK-061B), and Diversified Input Project of Tianjin National Natural Science Foundation (21JCYBJC01770).