Hematological factors associated with immunity, inflammation, and metabolism in patients with systemic lupus erythematosus: Data from a Zhuang cohort in Southwest China

Abstract Introduction Hematological parameters play important role in multiple diseases. This study was to investigate the possible association of the routine hematological parameters involved in immunity, inflammation, and metabolism with systemic lupus erythematosus (SLE) in patients of Zhuang ethnicity in Guangxi, southwest China. Methods The medical records of 195 Zhuang SLE patients between January 2013 and November 2018 were retrospectively reviewed. Random forest algorithm and multivariate logistic regression were used to identify the feature hematological parameters in patients with SLE. Association rules were explored between each parameter and immunity‐ (IgG, IgA, IgM, C3, and C4), inflammation‐ (ESR, hs‐CRP, and CAR), and metabolism‐ (TG, TC, HDL‐C, LDL‐C, TP, PA, ALB, and UA) related indexes. Results Random forest algorithm and logistic regression analysis showed that neutrophil‐to‐lymphocyte ratio (NLR), red blood cell distribution width (RDW), and platelet‐to‐lymphocyte ratio (PLR) were the feature parameters for distinguishing SLE patients from healthy controls. According to the ROC curves, the optimal cutoff values to predict SLE were 1.98 for NLR, 13.35 for RDW, and 145.64 for PLR. Association rule analysis showed that NLR was strongly associated with C3, hs‐CRP, TG, ALB, and UA; RDW was strongly associated with C3, C4, hs‐CRP, TG, and ALB; PLR was strongly associated with IgG, hs‐CRP, HDL‐C, and UA. Conclusions Neutrophil‐to‐lymphocyte ratio, RDW, and PLR may serve as effective predictors of dysregulation in immunity, inflammation, and metabolism. These three indicators may be potential for cardiovascular risk assessment in Zhuang SLE patients in southwest China.


Systemic lupus erythematosus (SLE) is a chronic inflammatory auto-
immune disease affecting different organs and has various clinical manifestations. Hematological manifestation is confirmed as the most common initial presentation in SLE, including anemia, thrombocytopenia, leukopenia, and lymphopenia. 1 Nearly all SLE patients develop hematological abnormalities during their disease course, either isolated or in conjunction with other manifestations. 2 The cardiovascular involvement in SLE and the subsequent cardiovascular disease predispose to a significant morbidity and can raise the mortality risk, which occurs more often late in active SLE states. 3 The proportion of cardiovascular events is higher in SLE than in general populations of comparable age and sex. 4 Abnormal immune activation, chronic inflammatory state, endothelial dysfunction, and metabolic disorders have been proven to raise the risk of cardiovascular events in SLE. 5,6 Many different markers such as immunoglobulin, complement, c-reactive protein, erythrocyte sedimentation rate, interleukin, and interferon have been used to assess immune and inflammatory status in multiple diseases. Recently, hematological parameters have received more attention, which include red blood cell distribution width (RDW), neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), mean platelet volume (MPV), hematocrit (HCT), and eosinophil-to-lymphocyte ratio (ELR). These parameters have been novel biomarkers in diagnosis, prognosis, risk stratification, and predicting survival and mortality in a variety of diseases, such as cardiovascular diseases, 7 cancers, 8 autoimmune diseases, 9 and parasitic diseases. 10 In recent years, RDW, NLR, and PLR have played an active role in inflammations and had regulatory effects on the immune system in autoimmune diseases. Previous studies have showed that RDW, NLR, and PLR could be used for evaluating inflammatory response, disease activity, and infection in rheumatoid arthritis, dermatomyositis, and SLE patients. [11][12][13] Nevertheless, the correlation between hematological parameters and cardiovascular involvement in SLE patients has not yet been elucidated. Hence, the aim of the present study was to explore the feature hematological parameters associated with SLE and to investigate the possible association of hematological parameters involved in immune, inflammatory, and metabolic indexes related to cardiovascular risk in a Zhuang SLE cohort.  Additionally, SLE disease activity was evaluated by SLE Disease Activity Index (SLEDAI) score. 16

| Random forest algorithm
Random forest is an ensemble classifier used for data mining, it is composed of numerous decision trees, each one relying on the values of a random vector sampled independently. Based on the random forest algorithm, the out-of-bag (OOB) classification error rate was calculated. Subsequently, Mean Decrease Gini (MDG), the total decrease in node impurities measured by the Gini index from splitting on variables and averaged over all trees, was calculated. 17 It provides a way to quantify which parameters contribute most to the accuracy of classification, and greater MDG will suggest an important feature parameter. In this way, we used MDG to rank the important feature hematological parameters with SLE patients. Statistical analysis was carried out using randomForest package of R software (http://www.r-proje ct.org).

| Association rules analysis
In order to systematically verify the associations between hematological parameters and immunity, inflammation, and metabolism indexes, association rules analysis was used to estimate the strength of the associations. The Apriori algorithm is often used to discover association rules. 18 For association rules models, the immune, inflammatory, and metabolic indexes and the feature hematological parameters are set as former and consequent item, respectively. An association rule reflects the interdependence and correlation between two variables. The strength of an association rule in the Apriori algorithm is determined by its support and confidence. By setting the minimum support and confidence thresholds, we aim to mine more meaningful association rules. We performed this statistical analysis with the arules package of R software (http://www.r-proje ct.org).

| Statistical analysis
The student's t test or the Mann-Whitney U test was performed to compare differences between the two groups based on distribution status. Further, Spearman's correlation coefficient was used to evaluate the correlations between two variables. A multivariate logistic regression was performed to determine which hematologic parameters were best associated with SLE, and ROC curves were created to analyze optimal cutoff value, sensitivity, and specificity of the parameters in predicting SLE P < .05 was regarded as statistically significant, and all statistical analysis was conducted using SPSS (version 17.0, SPSS Inc).

| Characteristics of the subjects
The demographic and clinical characteristics and the laboratory data of the study population are summarized in Table S1. In the patient group, WBC, neutrophils, lymphocytes, RBC, HGB, HCT, MCV, and PCT levels were significantly decreased compared with those in the control group, while RDW, NLR, and PLR levels were significantly higher ( Figure 1). In addition, hs-CRP, ESR, CAR, IgG, TC, TG, and UA levels were significantly higher and TP, PA, ALB, C3, C4, and HDL-C levels were significantly lower in the SLE group as compared to the controls.

| Random forest algorithm
We applied the random forest algorithm by constructing 5000 decision trees from which a relatively stable OOB classification error rate of 7.33% could be obtained. The multi-dimensional scaling (MDS) plot of the proximity matrix for the hematological parameters was depicted by this random forest, showing similarities among samples and their respective categories by projecting a high-dimensional measure to a two-dimensional surface. This graph displayed good classification effects between SLE patients and healthy controls ( Figure 2).

F I G U R E 1 Comparison of NLR
(neutrophils-to-lymphocytes ratio), RDW (red blood cell distribution width), and PLR (platelet-to-lymphocyte ratio) levels in SLE patients and healthy controls Based on MDG analysis, we found that NLR, RBC, RDW, HGB, and PLR had larger MDG values than the other hematological parameters (Table 1). This suggested that these five parameters were the most important hematological characteristics associated with SLE patients (Figure 3).

| Multivariate logistic regression
The statistically significant hematological parameters shown in Table   S1 were selected for multivariate logistic regression analysis. The results were presented in Table 2 Hence, by means of random forest algorithm in conjunction with multivariate logistic regression analysis, the results demonstrated that increased NLR, RDW, and PLR were the important feature parameters associated with SLE patients. optimal cutoff values for these three parameters were determined by the maximum Youden index accumulated by the ROC curves. Our results showed that the optimal thresholds for NLR, RDW, and PLR were

| Correlation of NLR, RDW, and PLR with immune, inflammatory, and metabolic indexes
As shown in Table 3

| Association rules between NLR, RDW, and PLR and immune, inflammatory, and metabolic indexes
For association rules models, the immune, inflammatory, and metabolic indexes and the feature hematological parameters were set as former and consequent item, respectively. The strength of an association rule was determined by its support and confidence. In this experiment, the minimum support and confidence degree were set at 20% and 50%, respectively. After executing the association model, 14 association rules were searched out ( Table 4). The results showed that all support levels were >20% and confidence levels were >50%, which indicated that NLR was strongly associated with C3, hs-CRP, TG, ALB, and UA; RDW was strongly associated with C3, C4, hs-CRP, TG, and ALB; and PLR was strongly associated with IgG, hs-CRP, HDL-C, and UA.

| D ISCUSS I ON
Cardiovascular events are proportionally higher in SLE patients compared with general populations of comparable age and sex. 4 Abnormal immune activation, chronic inflammatory state, and metabolic disorders have been proven to raise the risk of cardiovascular events in SLE. 5,6 The aim of the current study was to investigate the association of hematological parameters involved in immune,

F I G U R E 2
Multi-dimensional scaling graph of the hematological parameters. The abscissa and longitudinal coordinates indicate two dimensionalities; the red dogs and blue dots indicate SLE and healthy controls, respectively inflammatory, and metabolic indexes related to cardiovascular risk in a Zhuang SLE cohort. As an important finding, this study shows that increased NLR, RDW, and PLR are the important feature hematological parameters for distinguishing SLE patients from healthy controls. Another novel finding is that increased NLR, RDW, and PLR are significantly associated with immune, inflammatory, and metabolic dysregulation. These results imply that increased NLR, RDW, and PLR might be potential indexes for assessing cardiovascular risk in Zhuang SLE patients.
The diagnostic value of NLR, RDW, and PLR has been explored in this study. Based on the ROC curve, we found that the AUC value for NLR was 0.79 with 75.8% sensitivity and 74.4% specificity. RDW produced an AUC value of 0.76 with 75.4% sensitivity and 69.3% specificity. The AUC value for PLR was 0.72 with 73.8% sensitivity and 68.4% specificity. Our study demonstrates good sensitivity and specificity as shown by others. 19 After random forest algorithm and multivariate logistic regression analysis, we found that increased NLR, RDW, and PLR are the important feature parameters associated with SLE patients. These suggest that NLR, RDW, and PLR might serve as valuable complementary biomarkers for SLE.

Hematological abnormalities are very frequently involved in SLE.
Anemia is particularly common, followed by leukopenia and thrombocytopenia. 20 The mechanisms of hematological change in SLE are not fully understood, though some believe that it may be due to immune-mediated bone marrow depression, excessive peripheral cell destruction, drug damage, stress, or secondary infection. 21,22 Additionally, polyclonal B-cell activation is involved in autoimmunity, resulting in the production of cytokines and immunoglobulins.
Neutrophils and platelets participate in the production of these cytokines, which, in turn, may be conducive to the further activation of the neutrophils and platelets. 23 WBC and its subtypes have been used to assess inflammatory and infectious status. NLR, as the ratio of neutrophils over lymphocytes, reflects the balance between innate immunity and adaptive immunity. Although the mechanism is not clear, chronic inflammation appears to influence the relationship between NLR and various outcomes. 24,25 Recently, a further specific mechanism has been elucidated that NLR level is closely related to the phenotypic activity of granulocytic myeloid-derived suppressor cells (MDSCs). 26,27 Common myeloid progenitor cells are more likely to differentiate into MDSCs under chronic pathologic states such as inflammation, 28 cancer, 29 and cardiovascular disease. 30 An up-regulation in peripheral leukocytes by MDSCs may result in higher neutrophil levels, while lower lymphocyte levels may be due to their suppression by MDSCs, which shows a similar trend to that we observed for NLR values following SLE patients in our study.
An elevated NLR is associated with an increased degree of immune dysfunction and has been identified as a useful predictor of prognosis, survival, and mortality in several conditions including malignancy, 31 cardiovascular, 23 and renal diseases. 32 NLR has also been studied to evaluate inflammatory response and disease activity in autoimmune diseases such as rheumatoid arthritis, 11 dermatomyositis, 12  Platelets and lymphocytes are thought to be important factors in disease immunology and inflammation. Activated platelets not only produce growth factors but also release chemokines which play significant roles in inflammatory progress. On the other hand, lymphocytes are involved in immune surveillance and immunoediting, and decreased lymphocyte count and function can indicate impairment of immune surveillance and defense. 34 An increased PLR correlates with poor overall and progression-free survival in some malignancies such as lung cancer 30 and cholangiocarcinoma. 35 PLR has also been validated as an independent risk factor for brain metastasis of lung adenocarcinoma. 36  Abbreviations: NLR, neutrophils-to-lymphocytes ratio; PLR, platelet-tolymphocyte ratio; RDW, red blood cell distribution width.
inflammatory, and metabolic indexes. NLR, RDW, and PLR might be effective predictors in immune, inflammatory, and metabolic dysregulation and potential indexes for assessing cardiovascular risk in Zhuang SLE patients in southwest China.

ACK N OWLED G EM ENTS
We thank the staff of the Rheumatology of Minzu Hospital of Guangxi Zhuang Autonomous Region for assisting with data collection. This work was supported by Chongzuo Science and Technology Project of Guangxi (No. 2019013).