Parkinson’s disease (PD) is a common degenerative disease, mainly caused by selectively and progressively loss of dopaminergic neurons in striatum and substantia nigra. Age, genetic predisposition, environmental risk factors, and inherent susceptibility of dopaminergic neurons were thought to be the inducements of PD. However, in recent years, studies have shown that inflammation in nerve-system is involved in the progression of PD [1]. In the animal model, someone observed the increment of lymphocyte subsets in the late stage of PD progression [2] and infiltration of CD4+ lymphocytes into the brain [3, 4], which illustrated the lymphocyte subsets are closely associated with Parkinson’s disease. Also several papers reported that natural killer cell (NK cell) was found in central nervous system [5] (CNS). However, these results are not consistent. Therefore, in this study, we carried a case-control study on the correlation of lymphocyte subsets/NK cell and PD by Meta-analysis of Observational Studies in Epidemiology rule [6] to provide new insights of PD pathogenesis.

Methods

Data source, search strategy, and inclusion criteria

SpringerLink, Web of Science, Ebsco-medline with full text, Pubmed, Elsevier-ScienceDirect, Ovid-lww-oup, and Wanfang Data were searched for original articles from inception to March 13, 2017 to identify all case-control studies. The following search items were indicated: (“lymphocyte” or “T cell” or “B cell” or “NK” or “CD4” or “CD8” or “CD3” or “CD19” or “CD56”) and (“Parkinson”). The language of the enrolled studies was restricted to English or Chinese. All the relevant articles were conducted on human subjects, and classified as case-control studies. We also conducted a hand search of the reference lists of the enrolled studies or reviews to identify additional eligible studies. Original trials were eligible for the present meta-analysis if they met the following criteria: (1) any study described the association between lymphocyte subsets/NK cell and Parkinson’s disease risk, (2) studies were case-control or cohort type, and (3) the number of the lymphocyte subsets and NK cell of the cases and controls were available. Studies were excluded if (1) there were no data regarding the associations between lymphocyte subsets/NK cell and PD risk, (2) they were duplicate of previous publications, and (3) they were reviews or abstracts.

Data extraction and study quality assessment

The following data from each article were extracted by two researchers independently: name of first author, year of publication, ethnicity of the participants, average age of cases, disease duration, number of controls and cases, Hoehn and Yahr staging (HY), Unified Parkinson’s disease rating scale III (UPDRS-III) scores, and so on. We extracted the data through the following ways: (1) from the articles directly and (2) from the supplementary materials. The quality of eligible case-control studies was assessed independently using Newcastle-Ottawa Scale [6, 7], which evaluates the studies based on following criteria: (1) was the case definition adequate? (1 point), (2) representativeness of the cases (1 point), (3) selection of controls (1 point), (4) definition of controls (1 point), (5) comparability of cases and controls on the basis of the design or analysis (2 point), (6) ascertainment of exposure (1 point), (7) same method of ascertainment for cases and controls (1 point), and (8) non-response rate (1 point). Studies with a score ≥5 points are considered to be high quality.

Data synthesis, statistical analysis, and heterogeneity

We used mean difference with corresponding 95% CI for continuous outcomes (percentages of each lymphocyte subsets of participants in the trials), then standardized mean difference with a 95% CI for continued outcomes when the unit was not unified (absolute numbers and percentages of cell populations of participants in the trials). All quantitative variables are listed in the form of mean ± SD. We calculated the average mean and SD of multiple groups by the methods described in previous studies [8,9,10]. All the statistical analyses were performed using Review Manager 5.2 (RevMan) from the Cochrane Collaboration. We used the χ2 tests and I 2 statistics to assess the magnitude of heterogeneity. We selected a fixed-effect model if there was no unexplained statistical heterogeneity. If heterogeneity existed, then the random-effect model was used [11]. When I 2 ≥ 50% was observed, we could take some measures to reduce the heterogeneity, such as perform subgroup analysis by age and sensitivity analysis. A two-sided P value <0.05 was considered to be statistically significant.

Results

Search results, study characteristics, and study quality

As presented in Fig. 1, the systematic literature retrieval identified a total of 2046 relevant publications from SpringerLink, Web of Science, Ebsco-medline with full text, Pubmed, Elsevier-ScienceDirect, Ovid-lww-oup, and Wanfang Data. After the titles and abstracts were checked, only 34 articles related to the association between lymphocyte subsets/NK cell and Parkinson’s disease remained. An additional 13 publications were excluded due to the lack of sufficient data. As a result, 21 publications were included in the present meta-analysis [12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32]. Overall, these studies had a relatively high quality judged by the Newcastle-Ottawa Scale (9 scores = 2, 8 scores = 6, 7 scores = 7, 6 scores = 5, 5 scores = 1). The characteristics and quality of the enrolled studies were present in Table 1.

Fig. 1
figure 1

Preferred reporting items for systematic review and meta-analysis flowchart of the process of study selection

Full size image
Table 1 The characteristics and quality of the enrolled studies.(mean ± SD)
Full size table

Meta-analysis

The results of meta-analysis were presented in Tables 2 and 3. The decrement of the CD3+ and CD4+ lymphocyte subsets numbers were associated with the Parkinson’s disease risk. In intermediate and the late stage of PD, CD8+ lymphocyte subsets had a significant decrement. On the contrary, the number of NK cell increased in Parkinson’s disease. But no significant relevance was identified between the change of B lymphocyte subsets number and the overall Parkinson’s disease risk.

Table 2 The results of the subgroup analysis based on age
Full size table
Table 3 The results of the subgroup analysis based on HY scores
Full size table

We observed significant heterogeneity in the analysis for CD3+, CD4+, CD8+, B lymphocyte subsets, and NK cells (P < 0.00001, I 2 = 91%; P < 0.00001, I 2 = 93%; P < 0.00001, I 2 = 93%; P < 0.00001, I 2 = 98%; P < 0.00001, I 2 = 86%, respectively). I 2 from the I 2 test was ≥50%, so a random-effect model was used. To investigate the potential discrepancy, we divided the studies by age, disease duration, HY scores, and UPDRS-III scores into subgroups. In the age <60 group of analysis for CD3+ lymphocyte subsets, age <60 and age 60–65 groups for CD4+, age <65 group for B, and age <65 group for NK cells, heterogeneity decreased (P = 0.15, I 2 = 44%; P = 0.0002, I 2 = 82%; P < 0.0001 I 2 = 82%; P = 0.40, I 2 = 0%; P = 0.04, I 2 = 76%). The trend of changes of WMD/SMD in each subgroup was same as the trend of changes of WMD/SMD in total, except for age >65 group of NK cells, in which NK cells had no significant relevance with PD. And we failed to find any association between ages and CD8+ lymphocyte subset (Table 2).

We also carried out a subgroup analysis based on HY scores. By reading the paper carefully, we found that Gao Ping 2005 [15], Li Fengwu 2012 [20], Tan Feng 2002 [25], Zhao Xu 2013 [29], and Feng Xueyi 2016 [31] divided the PD patients into different subgroups by different HY scores. Unfortunately, Kinya Hisanaga 2001 [18], Lei Gesheng 2001 [19], Li Li 2015 [21], and Qi Jinxing 2004 [23] did not provide the HY scores. So we divided the papers into different HY subgroups by their mean HY scores. Subgroups with different HY scores in above papers were included in different HY subgroups, and papers with no HY scores were included in the “No HY” subgroup. Because the controls in above papers were repeatedly calculated, the total number of controls exceeded the real number. All subjects of the papers provided data for B cell has the mean HY scores in the range of 1 to 3, therefore they cannot be divided into different subgroups. In the HY 1–3 group, HY 3–4 group, HY 4–5 group, and No HY group of analysis for CD3+ lymphocyte subsets, HY 4–5 group for CD4+, HY 3–4, and HY 4–5 group for CD8+, HY 3–4, and HY 4–5 group for NK cells, heterogeneity decreased (P < 0.00001, I 2 = 79%; P = 0.006, I 2 = 72%; P = 0.06, I 2 = 60%; P = 0.2, I 2 = 38%; P = 0.01, I 2 = 74%; P = 0.0004, I 2 = 80%; P = 0.21, I 2 = 34%; P = 0.11, I 2 = 60%; P = 0.26, I 2 = 20%). In the HY 3–4 and HY 4–5 groups for CD8+ subsets, the decreased numbers of lymphocytes had the significant relevance with PD. On the contrary, in the HY 1–3 and HY 3–4 groups for NK cells, we observed no association between numbers of cells and PD. The rest of results of each subgroup is consistent with the total results (Table 3).

Sensitivity analysis and publication bias

By excluding one study at a time to evaluate the influence of an individual study on synthetic statistics, we referred this method as a sensitivity analysis (Table 4). In the subgroup analysis based on age, heterogeneity was still significant in CD3+ forest plots when we excluded FENG Xueyi 2016 [31], but the total heterogeneity and heterogeneity in age >65 group decreased (P < 0.00001, I 2 = 83%; P = 0.0002, I 2 = 78%). The WMDs of age <60 group and age >65 group were higher than that of age 60–65 group. When Claire H.Stevens 2012 [13] in the analysis for CD4+ lymphocyte subset was excluded, the total heterogeneity and heterogeneity in age >65 group decreased too (P < 0.00001, I 2 = 82%; P = 0.0002, I 2 = 74%). Correspondingly, the standardized mean differences made some changes. Like the WMDs of CD3+ analysis, the SMD of age 60–65 group was the lowest. When we excluded the Claire H.Stevens 2012 [13] in the B cell analysis, the total heterogeneity and the heterogeneity in age >65 group reduced (P = 0.11, I 2 = 50%; P = 0.03, I 2 = 80%). In the analysis for NK cells, the exclusion of Natalia Pessoa Rocha 2017 [32] made the total heterogeneity and the heterogeneity in age >65 reduced and disappeared (P = 0.31, I 2 = 16%; P = 0.81, I 2 = 0%). The SMD of age <65 is lower than that of age >65. However, the result in CD8+ did not alter obviously.

Table 4 One-study-removed sensitivity analysis
Full size table

In the subgroup analysis based on HY scores, excluding one study at a time did not make the results in CD3+ change obviously. And the higher HY scores a group got, the lower WMD it had. The result in CD8+ is the same as that in CD3+. When we excluded Claire H.Stevens 2012 [13] in the CD4+ analysis, the total heterogeneity and the heterogeneity in HY 1–3 group decreased (P < 0.00001, I 2 = 87%; P < 0.0001, I 2 = 84%). Except for No HY group, the higher HY scores a group got, the lower SMD it had. The exclusion of Natalia Pessoa Pocha 2017 [32] made the total heterogeneity, and the heterogeneity in HY 1–3 group decreased (P < 0.00001, I 2 = 79%; P < 0.00001, I 2 = 88%). The number of NK cells in the HY 3–4 group has no significant relevances with PD, and the SMD of this group is the lowest.

In addition, all the enrolled articles were examined by using funnel plot of RevMan. No obvious biases were identified (Fig. 2).

Fig. 2
figure 2

Funnel plot depicting the risk of publication bias

Full size image

Discussion

Recently, case-control studies indicated that the number of lymphocyte subsets and NK cell altered in PD patients. In the present study, we performed the meta-analysis on a total 21 studies enrolling 943 PD patients. Our data showed that the decreased numbers of CD3+, CD4+ lymphocyte subsets, and the increased number of NK cell may lead to Parkinson’s disease. In the intermediate and late stage of PD (HY 3–5), CD8+ lymphocyte subsets had a significant decrement. Along with the growth of the age, the numbers of CD3+ and CD4+ lymphocyte subsets decreased more. But when age grows further, the decrease slows down. We also found that the higher HY scores a patient got, the lower the numbers of CD3+, CD4+, CD8+ lymphocytes he/she had. And the number of B lymphocyte subsets had no significant association with PD.

It was confirmed that the peripheral blood lymphocyte subsets had complete dopaminergic system. We guessed that changes in CD3+, CD4+ and CD8+ T lymphocyte subsets may be affected by the neuroendocrine-immune regulation network in the central nervous system after inflammatory reaction. Along with the progress of PD, which means PD patients get more HY scores, the lymphocyte subsets get more suppression. On the other hand, with the increment of age, CD3+, CD4+ T lymphocytes also changed in the peripheral blood of PD patients, the number of lymphocyte subsets gradually decreased compared to the normal control group. But with the further increase in age, the number of lymphocyte subsets in PD patients and the control group tend to be similar. Considering the immune function gradually decreased with the growth of age; therefore, lymphocyte subsets of elderly patients with PD decreased more slightly than that of controls. Someone suggested that the numbers of NK cells may be mediated by the ADCC of K cells.

PD incidence has something to do with genetic mutations. About 10–15% patients have positive family history. However, the enrolled studies included PD patients with PD diagnosed according to the UK PD Brain Bank criteria. So, the patients with family history were excluded. All subjects were sporadic PD patients.

We observed the large heterogeneity in the analysis for CD3+, CD4+, and CD8+ lymphocyte subsets and medium heterogeneity for B cell when some studies were excluded. The possible reasons could ascribe to the heterogeneity of the varying age or different stage of PD patients. As PD patients get older or more HY scores, their symptoms get more serious. When we divided articles by age or HY scores, the decrement of heterogeneity demonstrated that. Other reasons may be the clinical diversity of PD. For example, some patients’ main symptom was tremor, others’ main symptom might be the spasm, but all of them were diagnosed as PD. Therefore, heterogeneity would be expected. In the sensitivity analysis for CD3+, the elimination of Feng Xueyi 2016 [31] can reduce the heterogeneity. When we read the paper again, we found that the difference between the cases and the controls were larger than that in any other studies. However, the study design and interventions have no significant differences with others. The reason may be ascribed to the experiment error. Same situation as Claire H. Stevens 2012 [13] and Natalia Pessoa Pocha 2017 [32]: The elimination reduced heterogeneity in the analysis for CD4+, B cell, and NK cells. However, the characteristics have no significant differences either, and the experimental data extracted for the analysis supported the result. The reason may be due to the experiment error too.

Although we have conducted a general retrieval for all eligible studies, there are several drawbacks that should be mentioned. First, the results may lack statistical power because of the limited sample size and limited number of studies enrolled. Second, level of evidence is lower than cohort studies and RCTs as case-control studies. Third, few studies provided UPDRS-III scores. None of them described details about the UPDRS-III scores. Because of the lack of data, we cannot assess such a correlation at the light of non-motor and pre-motor symptoms of PD. Finally, not all studies use the same criteria to evaluate the severity of PD for each patient. Thus, well-designed studies are needed for further study of the actual relevance between lymphocyte subsets/NK cell and PD risk.