Rheumatoid arthritis (RA) is a chronic autoimmune inflammatory disease characterized by infiltration of the synovium with neutrophils, macrophages, T cells, B cells, and dendritic cells, and progressive destruction of cartilage and bone, resulting in significant morbidity and shortened life expectancy [1,2].

The most commonly used markers of inflammation in clinical practice are C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR) to estimate the presence and activity of inflammatory diseases. However, these markers have some limitations such as reflection of short-term inflammatory activity and low discrimination ability. ESR is affected by age, sex, anemia, fibrinogen levels, hypergammaglobulinemia, and plasma viscosity, and reflects disease activity in the past few weeks [3], while CRP is less confounded by these factors, and reflects more short-term changes in disease activity [4]. There was an importance correlation between time-integrated CRP and ESR and radiological progression [4,5].

Neutrophils, lymphocytes, and platelets play a role in the control of inflammation, and systemic inflammation is associated with alterations in circulating blood cell quantity and composition, such as neutrophilia, lymphopenia, and thrombocytosis [6]. Recent studies have reported the numbers and ratios of complete blood cell subgroups in rheumatic diseases [7,8]. Neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and mean platelet volume (MPV) have recently been investigated as new inflammatory markers for the assessment of inflammation in many inflammatory, cardiovascular, and malignant diseases [9,10]. NLR is calculated as the absolute count of neutrophils divided by the absolute count of lymphocytes, and PLR is calculated as the absolute platelet count divided by the absolute lymphocyte count. As a novel marker of inflammation, NLR may be useful to estimate the activity of autoimmune and inflammatory diseases [11]. PLR is also used as an index for inflammatory status in diverse diseases [12]. MPV is the volume of the average circulating platelets in femtoliters and is a marker of platelet activation known to be associated with inflammation [13]. NLR, PLR, and MPV are inexpensive and easily obtainable laboratory markers of systemic inflammation. However, their roles in RA remain unclear. We hypothesized that NLR, PLR, and MPV may play a role as potential markers in assessing RA activity, because ESR and CRP have some limitations.

Studies on NLR, PLR, and MPV in RA patients in comparison with those in healthy controls, and on the relationship between hematological changes and RA activity have reported controversial results [7,8,14-27]. This may be because of the small sample sizes, low statistical power, and/or the presence of clinical heterogeneity. We performed the present meta-analysis to overcome the limitations of individual studies and resolve inconsistencies [28-30]. The aim of this meta-analysis was to evaluate the relationship between NLR, MPV, and PLR and RA, and to establish a correlation between the hematological indexes and RA activity.

Selection of eligible studies and data extraction

A literature search was performed to find studies that investigated NLR, PLR, or MPV in patients with RA and healthy controls. We searched the Medline, Embase, and Cochrane databases to identify available articles (until April 2017). We used the keywords and subject terms in this search were “neutrophil-to-lymphocyte ratio,” “mean platelet volume,” “platelet-to-lymphocyte ratio,” and “rheumatoid arthritis.” We also reviewed references cited in the selected articles to identify studies not covered by the electronic databases. Studies were included if (1) they were case-control, cohort, or cross-sectional studies; (2) they provided NLR, PLR, or MPV value in RA and controls; (4) they provided data on the correlation coefficient between NLR, PLR, or MPV and RA activity based on Disease Activity Score 28 (DAS28). No language or ethnicity restrictions were applied. We excluded studies if (1) they contained insufficient or duplicate data, or (2) they were review articles or case reports. Two independent reviewers extracted data from studies. Discrepancies in findings between the reviewers were resolved by consensus. We conducted the meta-analysis in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [31]. We extracted following information from selected studies: author, year of publication, country, ethnicity, subject number, mean and standard deviation (SD) of NLR, PLR, or MPV values, and correlation coefficient between NLR, PLR, or MPV and disease activity. We obtained the mean and SD using previously described formulae, if the data given were medians, ranges, or interquartile ranges [32,33].

Statistical analysis

A meta-analysis was performed to examine NLR, PLR, or MPV value between patients with RA and healthy controls, and correlation coefficient between NLR, PLR, or MPV and DAS28. For continuity of data, we presented results as standardized mean differences (SMDs) and 95% confidence intervals (CIs). Within- and between-study variations and heterogeneity were assessed by using Cochran’s Q test [21]. We performed the heterogeneity test to investigate the null hypothesis that all studies evaluated the same effect. When the significant Q statistic (p<0.100) indicated between-study heterogeneity, We used the random-effects model in the meta-analysis [34]. When the significant Q statistic (p<0.100) did not show inter-study heterogeneity, we used the fixed-effects model. The effect of heterogeneity was quantified by I²=100%×(Q−df)/Q [35]. In the meta-analysis, we conducted statistical analysis by using the Comprehensive Meta-Analysis computer program (Biostat, Englewood, NJ, USA).

Sensitivity test, subgroup analysis, publication bias, and quality analysis

We performed a sensitivity test was to evaluate the influence of individual study on the meta-analytic SMD by deleting each study individually. Subgroup analysis to explore heterogeneity was performed using confounding variables including ethnicity, data type, and age/sex-adjustment because ethnic difference, calculated data, or non-adjustment for age/sex may affect the heterogeneity. Funnel plots are used to detect publication bias, but they needs diverse study types of varied sample sizes, and their interpretation may be subjective judgment. Thus, Egger’s linear regression test was performed to assess publication bias [36], which evaluated asymmetry of funnel plot. Quality of evidence was evaluated according to the Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology and divided into four categories: high, moderate, low and very low [37].

Studies included in the meta-analysis

We identified 171 studies by using electronic and manual search methods. Twenty of the studies were selected for full-text review on the basis of their titles and abstracts. Four of these were excluded because they had no data on NLR, PLR, or MPV [38-40], or a healthy control group [41]. Thus, 16 articles met the inclusion criteria (Table 1, Figure 1) [7,8,14-27]. Two reports contained data on two different groups [19,24]; hence, we analyzed these studies independently. We identified 9 comparison studies on NLR in RA and controls, 9 on PLR, 11 on MPV, and 4, 2, and 3 studies on correlation coefficients between NLR, MPV, or PLR and DAS28 (n=991), respectively (Table 1). The characteristic features of the studies included in the meta-analysis are summarized in Table 1.

Table 1 . Characteristics of the individual studies included in the meta-analysis.

Test	Author	Country	Number		Matched	DAS28	Result
			Case	Control			SMD*	Magnitude*	p-value
NLR	Maden, 2017 [23]	Turkey	82	61	NA	NA	0.888	Large	<0.001
	Zhang-1, 2016 [24]	China	125	126	Age, sex	NA	0.831	Large	<0.001
	Zhang-2, 2016 [24]	China	59	126	Age, sex	NA	0.557	Medium	0.001
	Gökmen, 2016 [25]	Turkey	84	60	Age, sex	NA	0.661	Medium	<0.001
	Yolbas, 2016 [7]	Turkey	91	55	NA	NA	1.219	Large	<0.001
	Tekeoğlu, 2016 [14]	Turkey	102	NA	NA	0.193	NA	NA	NA
	Zengin, 2016 [15]	Turkey	205	104	NA	NA	0.505	Medium	<0.001
	Mercan, 2016 [8]	Turkey	136	117	NA	0.310	0.300	Small	0.018
	Uslu, 2015 [16]	Turkey	104	51	Age, sex	0.345	0.715	Medium	<0.001
	Fu, 2015 [17]	China	128	78	Age, sex	0.250	1.601	Large	<0.001
PLR	Zhang-1, 2016 [24]	China	125	126	Age, sex	NA	0.806	Large	<0.001
	Zhang-2, 2016 [24]	China	59	126	Age, sex	NA	0.711	Medium	<0.001
	Yolbas, 2016 [7]	Turkey	91	55	NA	NA	1.248	Large	<0.001
	Zengin, 2016 [15]	Turkey	205	104	NA	NA	0.338	Small	0.005
	Mercan, 2016 [8]	Turkey	136	117	NA	NA	0.264	Small	0.037
	Uslu, 2015 [16]	Turkey	104	51	Age, sex	0.352	0.461	Small	0.008
	Fu, 2015 [17]	China	128	78	Age, sex	0.290	1.199	Large	<0.001
MPV	Maden, 2017 [23]	Turkey	82	61	NA	NA	1.454	Large	<0.001
	Talukdar-1, 2017 [19]	India	48	80	Age, sex	NA	0.614	Medium	0.001
	Talukdar-2, 2017 [19]	India	32	80	Age, sex	NA	1.861	Large	<0.001
	Gokmen, 2016 [25]	Turkey	84	60	Age, sex	NA	0.384	Small	0.024
	Yolbas, 2016 [7]	Turkey	91	55	NA	NA	0.064	No effect	0.708
	Tekeoğlu, 2016 [14]	Turkey	102	NA	NA	0.316	NA	NA	NA
	Tecer, 2016 [18]	Turkey	100	100	Age, sex	NA	0.951	Large	<0.001
	Cakir, 2016 [26]	Turkey	81	80	Age, sex	NA	0.749	Medium	<0.001
	Yildirim, 2015 [27]	Turkey	90	52	Age, sex	0.231	0.506	Medium	0.004
	Yazici, 2010 [20]	Turkey	97	33	Age, sex	0.270	0.615	Medium	0.003
	Gasparyan, 2010 [22]	UK	400	360	NA	NA	0.231	Small	0.002
	Kisacik, 2008 [21]	Turkey	32	29	Age	NA	1.660	Large	<0.001

DAS28: Disease Activity Score 28, SMD: standardized mean difference, NLR: neutrophil-to-lymphocyte ratio, PLR: platelet-to-lymphocyte ratio, MPV: mean platelet volume, NA: not available. *Magnitude of Cohen’s d effect size: 0.2∼0.5, small effect; 0.5∼0.8, medium effect; ≥0.8, large effect..

Figure 1. Flow diagram of the study selection process. SLE: systemic lupus erythematosus.

Meta-analysis for comparison of NLR, PLR, and MPV between the RA patients and controls

NLR was significantly higher in the RA group than in the control group (SMD, 0.800; 95% CI, 0.542∼1.058, p< 0.001; Table 2, Figure 2). In addition, stratification by ethnicity showed a significantly elevated NLR in the RA group in Asian and Turkish populations (SMD, 0.994; 95% CI, 0.418∼1.519; p=0.001 and SMD, 0.695; 95% CI, 0.443∼0.948; p<0.001, respectively; Table 2). Stratification by data type revealed a significantly high NLR in both the original data and imputed data groups (SMD, 0.588; 95% CI, 0.389∼0.787; p<0.001 and SMD, 1.047; 95% CI, 0.607∼1.488; p<0.001, respectively; Table 2). Stratification by the adjustment for age and/or sex revealed a significantly higher NLR in the RA group, independent of the adjustment (Table 2). PLR was also significantly higher in the RA group than in the control group (SMD, 0.708; 95% CI, 0.401∼0.995; p<0.001; Table 2, Figure 2). Stratification by ethnicity showed a significantly elevated PLR in the RA group in Asian and Turkish populations (SMD, 0.904; 95% CI, 0.622∼1.185; p<0.001 and SMD, 0.560; 95% CI, 0.173∼0.947; p=0.005, respectively; Table 2). Stratification by data type revealed a significantly high PLR in both the original data and imputed data groups (SMD, 0.335; 95% CI, 0.182∼0.488; p<0.001 and SMD, 0.980; 95% CI, 0.720∼1.239; p<0.001, respectively; Table 2). Stratification by the adjustment for age and/or sex revealed a significantly high PLR in the RA group regardless of the adjustment (Table 2). However, MPV was not significantly higher in the RA group than in the control group (SMD, 0.049; 95% CI, −0.425∼0.524; p=0.838), regardless of data type and adjustment for age and/or sex (Table 2, Figure 2).

Table 2 . Meta-analysis of the association of NLR, PLR, and MPV with RA.

Group	Population	Study (n)	Test of association			Test of heterogeneity
			SMD*	95% CI	p-value	Model	p-value	I2
NLR
	Overall	9	0.800	0.542∼1.058	<0.001	R	<0.001	84.8
Ethnicity	Asian	3	0.994	0.418∼1.519	0.001	R	<0.001	91.1
	Turkish	6	0.695	0.443∼0.948	<0.001	R	0.001	75.4
Data type	Original	5	0.588	0.389∼0.787	<0.001	R	0.063	55.2
	Calculated	4	1.047	0.607∼1.488	<0.001	R	<0.001	87.5
Age/sex-matched	Yes	5	0.874	0.517∼1.230	<0.001	R	<0.001	84.6
	No	4	0.709	0.331∼1.086	<0.001	R	<0.001	84.9
PLR
	Overall	7	0.708	0.401∼0.995	<0.001	R	<0.001	85.5
Ethnicity	Asian	3	0.904	0.622∼1.185	<0.001	R	0.062	64.0
	Turkish	4	0.560	0.173∼0.947	0.005	R	<0.001	85.9
Data type	Original	3	0.335	0.182∼0.488	<0.001	F	0.656	0
	Calculated	4	0.980	0.720∼1.239	<0.001	R	0.037	64.7
Age/sex-matched	Yes	4	0.800	0.514∼1.087	<0.001	R	0.013	72.0
	No	3	0.599	0.075∼1.122	0.025	R	<0.001	90.6
MPV
	Overall	11	0.049	0.425∼0.524	0.838	R	<0.001	95.9
Data type	Original	9	0.116	0.525∼0.757	0.723	R	<0.001	96.2
	Calculated	2	0.248	1.208∼0.712	0.613	R	<0.001	96.6
Age/sex-matched	Yes	8	0.227	0.414∼0.868	0.487	R	<0.001	95.8
	No	3	0.418	1.356∼0.520	0.382	R	<0.001	97.1

NLR: neutrophil-to-lymphocyte ratio, PLR: platelet-to-lymphocyte ratio, MPV: mean platelet volume, RA: rheumatoid arthritis, SMD: standard mean difference, CI: confidence interval, R: random-effects model, F: fixed effects model.*Magnitude of Cohen’s d effect size: 0.2∼0.5, small effect; 0.5∼0.8, medium effect; ≥0.8, large effect..

Figure 2. Meta-analysis of the relationship of (A) NLR, (B) PLR, and (C) MPV to RA and the correlation coefficients between (D) NLR, (E) PLR, and (F) MPV and DAS28. NLR: neutrophil-to-lymphocyte ratio, PLR: platelet-to-lymphocyte ratio, MPV: mean platelet volume, RA: rheumatoid arthritis, DAS28: Disease Activity Score 28, Std diff: standard difference, CI: confidence interval.

Meta-analysis of the relationship between NLR, PLR, and MPV and RA activity

The meta-analysis identified that NLR was positively associated with RA activity based on DAS28 (correlation coefficient, 0.277; 95% CI, 0.190∼0.359; p<0.001; Table 3, Figure 2). In addition, stratification by ethnicity showed a significantly positive correlation between NLR and DAS28 in Asian and Turkish populations (Table 3). PLR was also positively associated with RA activity based on DAS28 (correlation coefficient, 0.318; 95% CI, 0.197∼0.430; p<0.001; Table 2, Figure 2). Stratification by ethnicity showed a significantly positive correlation between PLR and DAS28 in Asian and Turkish populations (Table 3). However, the meta-analysis of the correlation coefficients revealed no correlation between MPV and RA activity (correlation coefficient, −0.095; 95% CI, −0.435 to 0.269; p=0.615; Table 3, Figure 2).

Table 3 . Meta-analysis of the correlation coefficients between NLR, PLR, and MPV and RA activity (DAS28).

Test	Population	Study (n)	Test of association			Test of heterogeneity
			Correlation coefficient	95% CI	p-value	Model	p-value	I2
NLR	Overall	4	0.277	0.190∼0.359	<0.001	F	0.651	0
	Asian	1	0.250	0.080∼0.406	0.004	NA	NA	NA
	Turkish	3	0.287	0.186∼0.382	<0.001	F	0.474	0
PLR	Overall	2	0.318	0.197∼0.430	<0.001	F	0.605	0
	Asian	1	0.290	0.123∼0.441	0.001	NA	NA	NA
	Turkish	1	0.352	0.171∼0.510	<0.001	NA	NA	NA
MPV	Overall	3	0.095	0.435∼0.269	0.615	R	<0.001	90.0

NLR: neutrophil-to-lymphocyte ratio, PLR: platelet-to-lymphocyte ratio, MPV: mean platelet volume, RA: rheumatoid arthritis, DAS28: Disease Activity Score 28, CI: confidence interval, F: fixed-effects model, R: random-effects model, NA: not available..

Heterogeneity, sensitivity test, publication bias, and quality analysis

Between-study heterogeneity was identified during the meta-analyses of NLR, MPV, and PLR in the patients with RA (Tables 2 and 3). However, all the studies showed the same direction of the effect size, except for MPV. The sensitivity analysis revealed that none of the individual studies significantly affected the meta-analysis results, indicating robust results for this meta-analysis (Supplementary Figure 1). Publication bias results in a disproportionate number of positive studies and poses a problem for meta-analyses. However, we found no evidence of publication bias in the meta-analysis performed in this study (Egger’s regression test, p>0.1) (Supplementary Figure 2). For NLR and PLR, the quality of evidence according to GRADE criteria was low, since they were case-control studies. For MPV, the evidence was deemed to be of very low quality due to inconsistency, imprecision and indirectness (Table 4).

Table 4 . GRADE assessment of the meta-analysis results.

Test	Study (n)	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias	Quality
NLR	9	Case-control	No	No	No	No	Undetected	Low
PLR	7	Case-control	No	No	No	No	Undetected	Low
MPV	11	Case-control	No	Serious	Serious	Serious	Undetected	Very low
NLR (correlation)	4	Case-control	No	No	No	No	Undetected	Low
PLR (correlation)	2	Case-control	No	No	No	No	Undetected	Low
MPV (correlation)	3	Case-control	No	Serious	Serious	Serious	Undetected	Very low

GRADE: Grading of Recommendations Assessment, Development and Evaluation, NLR: neutrophil-to-lymphocyte ratio, PLR: platelet-to-lymphocyte ratio, MPV: mean platelet volume..

The inflammatory process in RA involves inflammatory cells and molecules that cause changes in the number, shape, and size of peripheral blood cells. Inflammatory cytokines such as interleukin (IL)-1, IL-6, and tumor necrosis factor (TNF)-α in RA modulates erythropoiesis [42]. Neutrophils account for 40%∼75% of leukocytes, and lymphocytes, including T cells and B cells, account for 20%∼45% of leukocytes in peripheral blood circulation; platelets are one of the most important immune cells. Neutrophils and platelets are involved in the production of inflammatory cytokines, which result in the activation of neutrophils and platelets [43]. Increased neutrophils and platelets may be attributed to the inflammation condition in RA, characterized by increased inflammatory cytokines [42]. Throughout this process, dysregulation in the control of apoptosis of the lymphocytes may lead to decreased lymphocyte production [44]. Owing to changes caused by the inflammation in neutrophils, platelets, and lymphocytes, NLR and PLR can be considered inflammatory markers.

Meta-analysis is a statistical procedure for combining of results from several studies to produce a single estimate of the major effect [21]. There are reasons why a meta-analysis can be a useful tool in this research, First, meta-analysis is to increase sample size, which may reduce the probability that random error will produce false-positive or false-negative associations. Thus, meta-analysis increases power over individual studies, and enhances the precision and accuracy of estimates of the effect size. Second, meta-analysis results can be generalized to a larger population. Generalizing the results from a meta-analysis makes more sense than from single studies. Third, inconsistency of results across studies can be quantified and analyzed, and the presence of publication bias can be investigated. Therefore, meta-analysis is an ideal and powerful tool for summarizing the results from different studies.

In this meta-analysis, we combined the evidence for NLR, PLR, and MPV in RA and the correlations of NLR, PLR, and MPV to RA activity. This meta-analysis of 16 studies revealed that NLR and PLR, but not MPV, were significantly higher in the RA group than in the control group. NLR and PLR, but not MPV, had a positive correlation with RA activity measured by using DAS28. The correlations found between NLR and PLR and RA disease activity as expressed by DAS28 is positive and significant but weak. Elevated NLR and PLR values reflected significantly increased disease activity. The meta-analysis data suggested that NLR and PLR estimate the inflammatory status and activity of RA. This meta-analysis showed that NLR and PLR were new potential inflammatory markers that can be used to evaluate inflammatory status and disease activity in patients with RA. This increase in NLR and PLR can be explained by the fact that inflammatory cytokines cause increased neutrophil and platelet production in active RA as part of the inflammatory process [45].

ESR and CRP level are the most widely used markers for measuring acute-phase response to indicate inflammation in RA. However, they have some limitations. ESR react slowly to inflammatory conditions, and CRP level lacks specificity [3,4]. NLR and PLR are relatively more stable than individual white blood cell parameters [46]. The positive correlation of NLR and PLR with DAS28 can help us to estimate the activity of RA. NLR and PLR are cheap and readily available objective markers for the assessment of inflammation and disease activity in RA. As easily measurable and available laboratory parameters, NLR and PLR may be useful in clinical practice. Time-integrated CRP is associated with greater radiologic progression in RA [5], but there have been no data on relationship between NLR and PLR and radiologic progression in RA. Therefore, further studies needed to investigate whether NLR and PLR are likely to help with discriminatory ability of bone erosion in RA. MPV is another marker used in the assessment of inflammation [13]. The association between MPV and RA remains unclear. We failed to observe high or low levels of MPV in RA or a correlation of MPV with disease activity.

This meta-analysis has some shortcomings that need to be considered. First, most of the studies included had small sample sizes, and only a small number of studies evaluated the correlation coefficient between the hematological indexes and RA activity. Thus, the meta-analysis may be underpowered. Second, the studies included patients with heterogeneous demographic characteristics and clinical features. NLR, PLR, and MPV values may be affected by multiple factors. Heterogeneity and confounding factors such as age, sex, drugs used (e.g., corticosteroids and disease-modifying antirheumatic drugs), or comorbidities (e.g. hepatopathy, obesity) may have affected the present results. For example, glucocorticoids may affect the count, size, and function of neutrophils, lymphocytes, and platelets, leading changes in NLR, PLR, and MPV [7], and anti-TNF therapy reduces NLR and PLR values in RA [15]. There was a negative correlation between the NLR and body mass index [47], HCV-related liver diseases had lower PLRs than the healthy controls [48], and MPV level was significantly higher in the inactive hepatitis B surface antigen (HBsAg) carrier group than in the control group [49]. Additional subgroup analyses were needed to justify the clinical heterogeneity between studies: by number of previous relapses, by previous failure to previous biological treatment, by current treatment. However, these limited data did not allow further analysis. Nevertheless, this meta-analysis also has its strengths. To the best of our knowledge, our meta-analysis is the first study to provide combined evidence for NLR, PLR, and MPV in RA patients. Compared with individual studies, our study provides more accurate data on the relationship of NLR, PLR, and MPV to RA by increasing the statistical power and resolution through pooling of the results of independent analyses.

In conclusion, this meta-analysis demonstrated that NLR and PLR are significantly higher in patients with RA and that NLR and PLR were significantly positively but weakly correlated to RA activity. Our meta-analysis suggested that NLR and PLR may be possible indexes for determining the extent of inflammation and evaluating the disease activity of RA. Further studies are necessary to elucidate that NLR and PLR can serve as biomarkers for monitoring RA activity in clinical practice.

Supplementary data can be found with this article online at https://doi.org/10.4078/jrd.2018.25.3.169.

No potential conflict of interest relevant to this article was reported.