The association of neutrophil to lymphocyte ratio, mean platelet volume, and platelet distribution width with diabetic retinopathy and nephropathy: a meta-analysis

The aim of the present study was to investigate the correlation of neutrophil to lymphocyte ratio (NLR), mean platelet volume (MPV), and platelet distribution width (PDW) with diabetic nephropathy (DN) and diabetic retinopathy (DR). We searched for eligible studies from PubMed, Embase, Web of Science, and CNKI up to 1 December 2017. Standardized mean difference (SMD) was calculated with a confidence interval (CI) of 95%. A total of 48 studies were included in our meta-analysis. Compared with patients with type Ⅱ diabetes mellitus (T2DM) and without DR, NLR, MPV, and PDW were higher in patients with DR (SMD = 0.77; 95% CI: 0.49–1.05; P<0.001; SMD = 0.68; 95% CI: 0.36–0.99; P<0.001; SMD = 0.52; 95% CI: 0.28–0.76; P<0.01). Compared with patients with T2DM and without DN, NLR, MPV, and PDW were higher in patients with DN (SMD = 0.63; 95% CI: 0.43–0.83; P<0.001; SMD = 0.81; 95% CI: 0.36–1.25; P<0.001; SMD = 0.70; 95% CI: 0.50–0.90; P<0.001). We also found that MPV was strongly associated with the severity of DR, and NLR was closely related to the degree of DN. Our findings indicated that NLR, MPV, and PDW could be recommended as inexpensive diagnostic biomarkers for DN and DR. However, considering several limitations in the present study, further high-quality clinical studies should be performed to investigate the relationship of NLR, MPV, and PDW to DN and DR.


Introduction
Diabetes is a global health burden and its morbidity and mortality are increasing in recent decades, especially in populations from the low-and middle-income countries [1,2]. Diabetes is a systemic metabolic disorder disease that can result in several long-term micro-and macrovascular complications, including diabetic retinopathy (DR) and diabetic nephropathy (DN). It was reported that ∼28.5-40.3% of patients with type II diabetes mellitus (T2DM) suffered from DR and 25-40% had DN [3][4][5]. To make things worse, DR and DN are serious microvascular complications and have been considered as the major cause of blindness and end-stage renal failure, respectively [3][4][5]. Thus, it is very significant for clinicians and researchers to elucidate pathogenesis of DR and DN and identify high-risk patients.
Diabetes mellitus (DM) is a chronic systemic disease. A flood of literatures have suggested that chronic inflammation was involved in insulin resistance and DM development [6][7][8][9]. Moreover, accumulated evidence have implicated that inflammatory response also plays a key part in the Articles identified through database searching (n = 834) (Pubmed=311, Embase=276, Web of sience=123, CNKI=123) Additional articles identified through other sources (n =0) Articles after duplicates and multiple publications excluded(n = 457) Articles screened (n = 251) Full-text articles assessed for eligibility (n = 123) Articles excluded for title and abstract bias (n = 204), with following reasons: 1) Reviews, letters, editorials, meeting abstracts and case reports (n = 42) 2) Articles not pertinent T2DM= 151) Included articles: n=22 Full-text articles excluded (n = 204), with following reasons: 1) Control group subjects were not patient with T2DM (n= 50) 2) Insufficient of data (n = 34)

Statistical analysis
The statistical analysis was performed by using STATA version 12.0 (Stata Corporation, College Station, TX, U.S.A.). Standard mean difference (SMD) and its 95% confidence interval (CI) was used to depict the pooled continuous variables. If the data of NLR, MPV, and PDW were reported in non-proliferative DR/proliferative DR (NPDR/PDR) or microalbuminuria/macroalbuminuria (MIA/MA) groups, the data in NPDR or MIA were used to compare with control group. Based on the 95% CI not crossing 0, the SMD < 0 suggested that the specific hematologic inflammatory marker increased in control group. In addition, statistical heterogeneity amongst the studies was evaluated by Cochrane Q test (χ 2 ) and I 2 test. Moreover, a random-effect model was conducted to calculate the SMD when significant heterogeneity existed amongst some studies (P<0.05 and I 2 ≥ 50%), otherwise, fixed-effect model was used. Sensitivity analysis was performed to verify the stabilization of our results, by excluding the articles one by one in each corresponding group, and then obtained meta-analysis results from the remaining studies. Furthermore, the Begg's test and Egger's test were performed to evaluate the publication bias [26,27]. If the significant publications bias exist (P<0.01 for Begg's test and Egger's test), Duval wnd Tweedie's [28] non-parametric trim-and-fill method was conducted to evaluate the effects of bias on the results.

Study selection
The flow chart diagram of the process of study selection was shown in Figure 1. A total of 190 articles were yielded through primary study searching for PubMed, Embase, Web of Science, and CNKI. Of these studies, 44 were from PubMed, 67 from Embase, 52 from Web of Science, and 27 from CNKI. Amongst these studies, 17 articles were removed because of duplication, and 105 publications were excluded due to reviews, meeting abstracts, and studies not pertinent to DR or DN after screening the title and abstract. Thus, the remaining 68 articles were left for the full-text review. Of the 68 studies, 9 articles were excluded since control group's subjects were not DM patients, but healthy people, and 11 articles were removed owing to the lack of available data. Finally, 48 articles with 9118 patients were enrolled in this meta-analysis [21][22][23].  Tables 1-6. Moreover, the method-  ological quality of the included articles were assessed according to NOS, and the scores ranged from 5 to 7, indicating that the quality of selected studies were moderate to high.

Meta-analysis results
The association of NLR with DN and degree of albuminuria A total of 12 studies involving 2404 patients were included for the meta-analysis of the association of NLR with the presence of DN [29,30,33,[36][37][38][39][40]44,46,48,49], and 5 articles reported the data regarding the relationship between NLR and degree of albuminuria (MIA and MA) [29,30,36,38,44]. Random-effect model was applied since the significant heterogeneity was present for the presence of      level was substantially higher in patients with DR (SMD = 0.77; 95% CI: 0.49-1.05) ( Figure 3A) than those without DR. Additionally, four articles with 443 patients were also included for the meta-analysis of the association between NLR and the degree of DR (NPDR and PDR). However, no significant association between NLR and the degree of DR was observed in this meta-analysis (SMD = 0.57; 95% CI: −0.34-1.47) ( Figure 3B).

The association of PDW with DN
Three articles with 588 patients were included in the meta-analysis of the association between PDW and DN [32,41,43].    (Figure 6).

The association of PDW with DR
A total of 12 articles with 1681 patients explored the relationship between PDW and DR [21,23,32,50,[58][59][60]65,[67][68][69]71]. Considering the significant heterogeneity amongst the 12 studies (I 2 = 79.3; P<0.01), random-effect model was applied when we performed the meta-analysis. As the results showed, PDW level was significantly higher in DR group (SMD = 0.52; 95% CI: 0.28-0.76) than in DM group ( Figure 7A). Furthermore, there were five studies with 539 patients, which referred to the relationship between PDW and the severity of DR.  Hence, in the present study, we also performed the meta-analysis to investigate relationship between PDW and the severity of DR. However, the result indicated that there was no significant association of PDW with the severity of DR (SMD = 0.08; 95% CI: -0.09-0.25) ( Figure 7B).

Sensitive analysis
For the purpose of verifying the robustness of the meta-analysis results, we conducted sensitivity analysis by excluding the articles one by one in each step, and then obtained the meta-analysis results of the remaining studies. The results implied that the synthesized SMDs of NLR for DN ( Figure 8A), NLR for DR ( Figure 8B), as well as MPV for DN ( Figure 8C), MPV for DR ( Figure 8D) and PDW for DR ( Figure 8E) did not change substantially when omitting any individual study, indicating that the results of meta-analysis were robust although significant heterogeneity existed amongst the included studies. Owing to the limitation of study number, sensitivity analysis was not applicable for the synthesized SMDs of the other data.

Publication bias
The assessment of publication bias of the included studies was performed by using Begg's and Egger's test. From the results of the publication bias test, we found that there was no evidence of publication bias for the pooled SMDs of NLR for DN (Begg's test, P=0.244; Egger's test, P=0.990) ( Figure 9A) and for DR (Begg's test, P=0.929; Egger's test, P=0.588) ( Figure 9B). Similarly, the publication bias for the pooled SMD of MPV for DN was not observed either (Begg's test, P=0.371; Egger's test, P=0.690) ( Figure 9C).
Nevertheless, publication bias might exist amongst the eligible studies referring to MPV (Begg's test, P=0.039; Egger's test, P=0.148) ( Figure 9D) and PDW (Begg's tests, P=0.06; Egger's tests, P=0.02) ( Figure 9E) for DR. Consequently, we used meta-trim method to figure out the effect of publication bias on the dependability of the meta-analysis results. The results of meta-trim method showed that the adjusted SMDs of MPV and PDW were (0.68; 95% CI: 0.36-0.99) and (0.28; 95% CI: 0.018-0.540), respectively, indicating that the publication bias did not substantially affect the stability of the pooled results of the MPV and PDW for DR. In addition, the publication bias test was not applicable for the pooled SMD of PDW for DN owing to the limitation of the number of the eligible studies.

Discussion
To our knowledge, this is the first meta-analysis focussing on the association of NLR, MPV, and PDW with DN and DR. In our meta-analysis, we found that NLR, MPV, and PDW was higher in patients with DR or DN. In addition, we also observed that MPV was strongly associated with the severity of DR, and NLR was closely related to the degree of DN.
White blood cells and their subtypes have been widely considered as inflammatory markers in various diseases including DM. NLR calculated by blood routine examination is readily available and inexpensive. It has better stability since it reflects the balance between the peripheral blood neutrophils and lymphocytes, which is less affected by various physiological and pathological status, as compared with independent neutrophils, lymphocytes, and total white blood cell count. In recent years, studies have reported that NLR might be a novel inflammatory marker, and there was a significant relationship of NLR to conventional markers of inflammation. For instance, several studies showed that there was a positive association of NLR with serum IL-6 and C-reactive protein [74,75]. Furthermore, it has been recognized that chronic inflammation has a significant role in the development and deterioration of DM-associated complications [76]. Several studies have reported that various inflammatory molecules such as adipokines, chemokines, adhesion molecules, and cytokines, could contribute to the development of DN [77]. A study showed that NLR, as a novel inflammation marker, is inversely associated with glomerular filtration rate (GFR) patients with chronic kidney disease [14]. Particularly, a 3-year follow-up study showed that NLR could serve as a predictor of worsening renal function in diabetes patients [78]. Additionally, numerous studies have shown that elevated NLR was associated with the presence of DN and MA [75,[79][80][81]. In consistence with that, this meta-analysis confirmed that DM patients with DN, albuminuria, or MA have higher NLR, as compared with those without DN, albuminuria, or with MIA, which indicated that NLR plays an important role in the development and progression of DN.
It has also been demonstrated that chronic inflammation plays an important role in the development of DR with the help of various systemic and local (vitreous and aqueous fluids) inflammatory factors [82][83][84][85]. Neutrophils may contribute to the development and progression of microangiopathy and inflammation, when they adhere to the endothelial cell wall [86,87]. A study by Woo et al. [88] showed that neutrophil count in circulation was higher in patients with DR, and was closely associated with DR grades, indicating the considerable role of neutrophil-mediated inflammation in the development and progression of DR. Furthermore, several recent literatures have proposed that NLR, as a novel inflammation marker, was found to be higher in patients with DR and linked with the severity of DR [13,18,61,88,89]. However, in a study by Ciray et al. [79], NLR was not found to be associated with the pathogenesis of DR. In our meta-analysis, it was observed that NLR might be involved in the presence of DR, but not linked with the DR grade, indicating that NLR might only play a role in the early stage of DR. However, only four studies with the low number of cases were in place for the meta-analysis of the association between NLR and the DR grade. Therefore, the role of NLR in the later stage of DR remains controversial and further studies are in need to figure out this issue.
Albuminuria can lead to some changes in the levels of numerous biomarkers in blood circulation, which could be applied to diagnosis and the risk evaluation of progression of kidney diseases [90,91]. Several studies have shown that MPV and PDW values are higher in DM patients with DN when compared with those without DN. Similarly, our meta-analysis also demonstrated that DM patients with DN have increased level of MPV than those without DN. These results indicated that MPV and PDW values could be used for the diagnosis of DN. Furthermore, a few studies showed that the MPV values increased with DN stage proceeding, which suggested that MPV could be applied to determine the stage of DN and monitoring the progress of DN. Certainly, the exact causal association of MPV and PDW with DN need to be further explored, which will help physicians make full use of MPV and PDW in clinical practices.
It has been demonstrated that persistent presence of hyperglycemia, dyslipidemia, and insulin resistance could lead to endothelial and pericyte injury, which leaves DM patients in a 'prothrombotic condition' [92]. Furthermore, the morphologic and functional forms of platelets, which are related to enhanced platelet activity, were found to be increased in patients with DM [92]. Platelets with higher MPV value have more metabolic and enzymatic activities than the lower and produce more thromboxan-A 2 , b-thromboglobulin, and adhesion molecules releasing [19,20], which implicated that MPV might be connected with elevated thrombotic potential. In consistence with numerous studies focussing on the correlation of MPV with DR, this meta-analysis also showed that increased MPV was associated with the presence of DR and its severity. In addition, PDW is a parameter reflecting the variability of platelet size, and its high value indicated increased production of larger reticulated platelet, indicating that PDW value, similar to MPV, is closely linked with thrombotic potential. Consistently, from the results of meta-analysis, we found that higher PDW was related to the presence of DR, but there were no differences in PDW between patients with NPDR and PDR. Considering that only five studies were available for the meta-analysis of the relationship of PDW to the severity of DR, the small sample size might partly account for the negative association of PDW with the severity of DR. Thus, further studies are needed to further explore the correlation between PDW and the severity of DR.
When interpreting the results of the present study, several limitations should be taken into consideration. First, although random-effect model was applied, significant heterogeneity amongst studies was still observed. It might be attributed to differences in some characteristics of the included studies including age, ethnicity, diseases' duration, and body mass index, in detecting instrument, in experimental designs and in quality of the included studies. Second, the number of included studies was insufficient for a persuasive meta-analysis of the association of NLR, MPV, and PDW with the DR grade. Third, the standard of the stage of DR/DN with T2DM patients was not uniform and may affect the results. Finally, this meta-analysis did not investigate the relationship of the NLR, MPV, and PDW to other factors that may influence the development of DR/DN, due to the lack of available data in the included studies.
In conclusion, this meta-analysis indicated that NLR MPV and PDW has a positive association with DN and DR. Furthermore, MPV is closely related to the severity of DR and DN. NLR, MPV, and PDW could be recommended as inexpensive diagnostic markers for DN and DR, and even antiplatelet agents might be effective to control the development of DN and DR. However, further high-quality clinical studies should be performed to investigate the relationship of NLR, MPV, and PDW to DN and DR.