Prognostic role of pretreatment blood lymphocyte count in patients with solid tumors: a systematic review and meta-analysis

Background To evaluate the prognostic value of pretreatment lymphocyte counts with respect to clinical outcomes in patients with solid tumors. Methods Systematic literature search of electronic databases (Pubmed, Embase and Web of Science) up to May 1, 2018 was carried out by two independent reviewers. We included Eligible studies assessed the prognostic impact of pretreatment lymphocytes and had reported hazard ratios (HR) with 95% confidence intervals (CIs) for endpoints including overall survival (OS) and progression-free survival (PFS). Only English publications were included. Results A total of 42 studies comprising 13,272 patients were included in this systematic review and meta-analysis. Low pretreatment lymphocyte count was associated with poor OS (HR = 1.27, 95% CI 1.16–1.39, P < 0.001, I2 = 58.5%) and PFS (HR = 1.27, 95% CI 1.15–1.40, P < 0.001, I2 = 25.7%). Subgroup analysis disaggregated by cancer type indicated that low pretreatment lymphocytes were most closely associated with poor OS in colorectal cancer followed by breast cancer and renal cancer. Conclusions Low pretreatment lymphocyte count may represent an unfavorable prognostic factor for clinical outcomes in patients with solid tumors.

blood lymphopenia prior to initial treatment is associated with poor prognosis in various types of cancers, such as advanced carcinomas and sarcomas, cervical cancer, renal carcinoma, and bladder cancer [1,[7][8][9]. However, the inconsistent effect of pretreatment blood lymphocyte counts in patients with some publications cannot be ignored [10][11][12][13][14][15]. Moreover, the prognostic impact of lymphopenia in non-hematologic tumors has not been systematically analyzed. In order to reach a more reliable conclusion, a systematic review and meta-analysis to synthesize the evidence pertaining to pretreatment peripheral blood lymphocytes in patients with solid tumors is indispensable.

Data sources and search strategy
The Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) were applied in the present study [16]. We conducted a systematic literature search in the PubMed, Web of Science, and Embase electronic databases to identify relevant studies published as of May 1, 2018. Combinations of the following keywords were used to retrieve articles: "lymphopenia", "lymphocytosis", "lymphocytes", "tumor", "carcinoma", "cancer" and "prognosis" or "survival".

Study selection criteria
Studies that qualified the following criteria were included: (1) original articles published in English language; (2) studies that enrolled patients with pathologically confirmed solid tumors who had not received any treatment; (2) lymphocyte counts were measured prior to the first treatment (surgery and/or chemotherapy or radiotherapy or palliative therapy); (3) pretreatment lymphocytes were reported as a dichotomous variable; (4) assessed the prognostic impact of pretreatment lymphocytes and had reported hazard ratio (HR) with 95% confidence interval (CI); at least provided Kaplan-Meier survival curves from which HRs and 95% CIs could be calculated.
In case of duplicate publications based on the same dataset, only the article with the largest sample size was included. Letters, reviews, case-reports, expert opinions and conference abstracts were excluded from the present study.
Titles and abstracts of articles retrieved on initial search were independently screened by two investigators (W.H. and Y.L.) to eliminate irrelevant articles. Full texts of the remaining articles were reviewed against the above criteria to identify eligible studies. In case of any disagreement between the two reviewers, the final decision was made by a third reviewer (J.Z.).

Data extraction and quality evaluation
Data pertaining to the following variables were independently extracted by two authors (W.H. and Y.L.): first author; publication year; region; study design; cancer type; sample size; disease stage; cut-off value; survival analysis; treatment details; and HR with corresponding 95% CI for OS and/or PFS. Survival outcomes obtained on multivariate analysis were accorded precedence over those obtained on univariate analysis.
Two investigators (W.H. and Y.L.) independently assessed the quality of each study according to the Newcastle-Ottawa Scale (NOS); any disagreement was resolved by consensus [17]. Newcastle-Ottawa Scale mainly includes selection, comparability, and evaluation of outcomes. On a scale of 0 to 9, a study with score of ≥ 6 was considered as a high-quality study. However, quality assessment was not an exclusion criterion for eligible studies.

Statistical analysis
We extracted the HRs and 95% CIs of the ratio for low pretreatment lymphocytes over high pretreatment lymphocytes from each eligible study for OS and/or PFS. The endpoints of survival were OS and/or PFS mainly because the two endpoints were frequently used in the included studies. Meta-analysis was performed to evaluate the prognostic effect of pretreatment lymphocytes in patients with solid tumors for each of the endpoints (OS/PFS). Extracted data were pooled using the Stata 12.0 (STATA Corporation, College Station, TX, USA). Cochrane Q test and the I 2 statistic were used to test the heterogeneity among the studies included in the pooled analysis. In the absence of significant heterogeneity (P > 0.1 and I 2 < 50%), the fixed effects model was used for pooled analysis [18]; otherwise, the randomeffects model was used. Pooled HR > 1 was considered indicative of worse survival outcome of patients with low baseline lymphocytes. If the 95% CI did not overlap 1, the result was considered statistically significant. Subgroup analyses were performed to investigate the association of pretreatment lymphocyte counts with variables such as region, cancer type, disease stage, cutoff value, survival outcomes, and treatment scheme. Moreover, sensitivity analyses were performed by sequential elimination of one study at a time to explore its potential impact on the heterogeneity. We further used funnel plots and Egger's test to examine the influence of publication bias on the pooled OS and PFS, respectively. All statistical tests were two-sided and P < 0.05 indicated statistical significance.

Search and selection of studies
As illustrated in Fig. 1, a total of 2631 articles were retrieved on initial database search. Of these, 2507 articles were removed as irrelevant and duplicate articles. After full-text review, 75 were excluded due to lack of available information. Seven studies that reported lymphocytes count as a continuous variable were excluded. Finally, a total of 42 studies with a combined study population of 13,272 patients were considered eligible for inclusion [1,[7][8][9][10][11][12][13][14][15]. The articles were published in the period from 2005 to 2018. The most common types of cancers in the included studies were lung cancer (n = 5), followed by nasopharyngeal cancer (n = 4) and renal cancer (n = 4). All the included studies    Table 1.

Progression-free survival
A total of 14 studies comprising of 18 cohorts (5147 patients) were included in the analysis of HRs for PFS. The median cut-off value for pretreatment lymphocytes was 1.50 (range: 1-3). In 9 articles, the HRs and 95% CIs were obtained by multivariable analysis; while 5 articles had calculated HRs and 95% CIs by univariate analysis. Overall, low pretreatment lymphocyte counts were significantly associated with worse PFS (Fig. 3). Owing to the lack of significant heterogeneity (I 2 = 25.7%), the fixed-effects model was used for pooled analysis. On subgroup analysis stratified by cancer type, low pretreatment lymphocytes was most closely associated with poor PFS in patients with breast cancer (n = 2, HR = 1.76, 95% CI 1.42-2.20, P ˂ 0.001, I 2 = 0) ( Table 3). Likewise, the funnel plot was roughly symmetrical and Egger's test revealed no significant influence of publication bias (P = 0.267 for PFS).

Discussion
To the best of our knowledge, this is the first systematic review and meta-analysis that comprehensively summarizes the association between lymphocyte count and cancer survival. Current meta-analysis included a total of 42 studies with a combined study population of 13,272 patients and provides evidence that low lymphocyte counts are associated with shorter OS and PFS in patients with non-hematologic tumors. There was moderate heterogeneity among studies in the analysis of OS (I 2 = 58.5%) but not that of PFS (I 2 = 25.7%). Subsequently, on subgroup analysis by tumor location, the highest effect size with respect to OS was observed in patients with colorectal cancer followed by those with breast cancer and renal cancer. Intriguingly, we found a significant reduction in heterogeneity in subgroups of patients with colorectal cancer (I 2 = 0), breast cancer (I 2 = 0) and renal cancer (I 2 = 24.3%) although moderate heterogeneity was observed (I 2 = 58.5%) in the pooled analysis. Moreover, when stratified by disease stage in the analysis of OS and PFS, low lymphocyte count was an adverse prognostic factor in both non-metastatic and metastatic subgroups. This suggests that lymphocytes are involved in several stages of cancer development. Moreover, the negative prognostic effect on OS and PFS was consistent in subgroups stratified by cut-off value and type of survival analysis.
Patients with pretreatment lymphopenia have significantly worse survival than those of patients with normal lymphocyte counts in the context of several malignancies [1,[7][8][9]. Lymphocytes are known to play a role in cellular Zhao et al. Cancer Cell Int (2020) 20:15 [4]. Therefore, lymphopenia may reflect poor host immunity against cancer and a favorable microenvironment for tumor growth. The underlying mechanism of pretreatment lymphopenia in solid tumors has not been fully clarified and is probably multifactorial. It is widely believed that lymphopenia may result from increased lymphocyte apoptosis and/ or altered lymphocyte homeostasis. Kim et al. demonstrated that increased expression of Fas ligand (FasL) in tumor cells mediated apoptosis of TILs as well as circulating lymphocytes, which conferred immune privilege to tumors [51]. Increased numbers of apoptotic peripheral T lymphocytes have been detected in patients with gastric cancer [52]. Over-production of immunosuppressive cytokines such as transforming growth factor (TGF-β) and IL-10 by tumor cells specially during tumor growth may suppress different effector pathways of the immune  [53,54]. Exposure to TGF-β reduced the expressions of apoptotic activators (such as perforin and granzyme A and B) on cytotoxic T cells that infiltrated the tumor tissues. Additionally, tumor growth increases the recruitment of CD4+ regulatory T cells that secrete IL-10 and TGF-β and suppress effector CD8+ T cell responses [55]. IL-10 exerts an inhibitory effect on major histocompatibility complex (MHC) class I antigen presentation. Dummer et al. observed excessive expression of immunosuppressive factor IL-I0 in metastatic lesions and in cultured cells from metastases; they inferred that this cytokine plays a key role in tumor progression [56]. Although numerous studies previously focused on T-cell-mediated immunity, B cells play an equally prominent role in modulating anti-tumor immune responses and in carcinogenesis. B cells are classically known for their role as producers of antibodies. Tumor-infiltrating B cells have relation to improved survival in cervical cancer and non-small cell lung cancer [57,58]. Results from these clinical observations suggest that the potential mechanisms underlying B-cell anti-tumor immunity may involve tumor-infiltrating B cells could recruit and retain T cells at the tumor site, thus facilitating and sustaining T-cell responses that inhibit tumor development. Moreover, tumor-infiltrating B cells may function as antigen-presenting cells to aid in anti-tumor immunity [57,59]. Thus, it may be possible to generate more amplified and prolonged immune responses at the tumor site by promoting cooperative interactions of B cells and T cells.
Collectively, these findings suggest that lymphopenia may be a result of cancer-induced immune suppression that drives tumor progression. Neutrophil-lymphocyte ratio (NLR) has been identified as an independent prognostic factor in many solid tumors; a high NLR ratio was shown to be associated with inferior outcomes [60][61][62]. Nevertheless, it includes  [63]. The prognostic impact of NLR may be explained by the association of high NLR with inflammation. However, at the same time, the authors admitted that the confounding effect of concurrent inflammatory conditions cannot be completely excluded because high NLR has also been shown to be of prognostic relevance in non-cancerous conditions such as acute pancreatitis [64] and cardiac events [65]. Joseph suggested that the prognostic value of high neutrophil-lymphocyte ratio may actually be driven by lymphocytopenia rather than neutrophilia in patients with bladder cancer [9]. Similar results have been reported elsewhere; lymphocyte count was shown to exert a stronger impact on the neutrophil-to-lymphocyte ratio in clear cell renal carcinoma and pancreatic cancer [8,32]. Therefore, based on these observations, we evaluated the prognostic value of pretreatment peripheral blood lymphocyte counts with respect to clinical outcomes in patients with solid tumors.
Lymphocytopenia is not just a parameter related to cancer survival but may also reflect a biological mechanism that promotes tumor progression. Of note, adjunctive treatment for reversal of lymphopenia or to increase lymphocyte counts has also been proposed by some authors. Restoration of lymphocyte homeostasis may lead to activation of effector cytotoxic and helper T cells and result in a more potent antitumor immune response. IL-2 was used for treatment of patients with metastatic melanoma. Recombinant human IL-7 (rhIL-7) was shown to improve the immune function of patients with lymphopenia by promoting peripheral T cell expansion and suppressing the immunosuppressive network [66].
In view of the possible impact of different cut-off values of pretreatment lymphocytes on prognosis, we observed the largest effect size in the cut-off ≤ 1.0 subgroup; the next was the 1.0 < cut-off ≤ 2.0 subgroup. Nonetheless, the cut-off > 2.0 subgroup was not associated with poor OS. Similar results were obtained on subgroup analysis of PFS. Hence, a relatively lower pretreatment lymphocytes cut-off value may have a better discriminative prognostic value. However, optimal pretreatment lymphocytes cutoff value for various types of cancers needs further research.
Undoubtedly, our research has several limitations. First, our meta-analysis was based on HR and 95% CIs extracted from retrospective studies. Due to the inherent limitations of retrospective studies including heterogeneity with respect to data selection and analysis, our pooled data might be susceptible to biases and may be biased towards positive results. Second, moderate heterogeneity was observed in the analysis of OS and the sources of this heterogeneity remain unclear; however, no significant heterogeneity was observed in the analysis of PFS. This is likely attributable to inclusion of more than 40 cohorts comprising of 13,000 patients with different tumors and from various countries. As yet, we have not found any meta-analysis that determined the prognostic value of pretreatment lymphocytes in any malignancy. Our goal was to gain a comprehensive understanding of the prognostic value of lymphocytes in patients with solid tumors. Therefore, the moderate heterogeneity observed in the analysis of OS is reasonably expected. Third, in 16 out of the 42 studies, the HRs were calculated on univariate analysis. Compared with data from multivariate analysis, HR and 95% CI calculated on univariate analysis is more likely to lead to an overestimation of the prognostic value. Therefore, we conducted subgroup analysis of univariate analysis and multivariate analysis and the statistical significance was stable; moreover, the multivariate analysis subgroup even had a larger effect size.

Conclusion
Peripheral blood lymphocytes is a simple and routine index in clinical work. To the best our knowledge, we have not found any meta-analysis that determined the prognostic value of pretreatment lymphocytes in any malignancy. Our meta-analysis provides evidence that pretreatment lymphocyte might be a potential biomarker for survival in patients with solid tumors. However, the present meta-analysis was based on observational studies; we could not demonstrate a cause-effect relationship between pretreatment lymphocyte and survival in patients with solid tumors. Further prospective largescale investigations are required to explore whether reversing lymphopenia can be a new target for cancer treatment and to increase the understanding of its role in disease pathogenesis.