Neutrophil-to-lymphocyte ratio in systemic lupus erythematosus is influenced by steroids and may not completely reflect the disease activity

Introduction Systemic lupus erythematosus (SLE) is an autoimmune disease characterized by multiple autoantibodies and varied clinical features including pancytopenia. Quantifying the disease for the purpose of disease assessment is still a challenge. Complement levels, ESR, presence of auto-antibodies and their titers, especially of the anti-ds DNA, are the laboratory parameters that are considered for disease assessment in routine practice.1 SLE disease activity index (SLEDAI), Safety of estrogens in lupus erythematosus national assessment (SELENA) and British Isles Lupus Assessment Group (BILAG) are the commonly used scores in clinical trials and in certain academic clinical settings. The SLEDAI score is commonly used in clinical practice.2


Introduction
Systemic lupus erythematosus (SLE) is an autoimmune disease characterized by multiple autoantibodies and varied clinical features including pancytopenia. Quantifying the disease for the purpose of disease assessment is still a challenge. Complement levels, ESR, presence of auto-antibodies and their titers, especially of the anti-ds DNA, are the laboratory parameters that are considered for disease assessment in routine practice. 1 SLE disease activity index (SLEDAI), Safety of estrogens in lupus erythematosus national assessment (SELENA) and British Isles Lupus Assessment Group (BILAG) are the commonly used scores in clinical trials and in certain academic clinical settings. The SLEDAI score is commonly used in clinical practice. 2 In the recent years, neutrophil-to-lymphocyte ratio (NLR) has emerged as one of the parameters useful for evaluating several inflammatory diseases. 3 Recent studies have highlighted the relationship of NLR with disease activity. [4][5][6][7] The NLR has been found to correlate with clinical disease activity in rheumatoid arthritis, Henoch-Schonlein purpura, malignancy and ischemic heart disease. [8][9][10][11][12] Accumulating evidence suggests that inflammatory markers such as NLR and PLR (platelet-tolymphocyte ratio) are significantly elevated in SLE patients and they have been often recommended as useful markers for assessing disease activity. [13][14][15] Literature also indicates positive correlation between NLR and pulse wave velocity, renal involvement and different classes of renal histological staging, and its use as an additive marker for diagnosing infection in SLE patients. [16][17][18] The current study evaluated NLR as a predictor of disease activity in SLE. The secondary objectives were to verify the relationship of NLR, ESR, CRP, and C3 with SLEDAI, and to examine the discriminatory ability of NLR and inflammatory parameters in classifying disease severity. The study included only non-renal SLE or renal lupus patients with no active renal disease. Since SLE per se alters the ratio of total count as well as lymphocyte count and the addition of steroid (used in majority of the SLE patients) is known to influence NLR, the study considered use of steroid as one of the parameters to analyse the relationship of NLR and disease activity. [19][20]

Subjects and methodology
The cross-sectional study recruited SLE patients fulfilling the systemic lupus international collaborating clinics/American College of Rheumatology (SLICC/ACR) criteria (2010) during the three-months period starting from December 2017. 21 The study was approved by the institutional ethics committee and the patients were enrolled consecutively on their routine visit to the center after obtaining informed consent. Subjects with active renal disease or suspicious infections were excluded from the study. Data values that are unusually far from the range of values of the variable in the study population were considered as extreme values. Inclusion of such data values would bias the measures of central tendency and test estimates and were excluded from analysis. The disease activity was assessed using SLEDAI-2K and the patients in remission were excluded from analysis. 22 Organ damage was ascertained using SLICC/ACR damage index (SDI). 23 The demographic, clinical and inflammatory parameters were recorded in a pre-structured proforma at the recruitment visit and all the analysis were performed as part of the study. Age, gender, total leucocyte count (TLC), differential leucocyte count, C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), complement factor 3 (C3), SLEDAI-2K and SDI were considered for the study based on clinical relevance. Data on immunosuppressants/disease modifying antirheumatic drugs (DMARDs), steroids and biologics received by the patients were documented. Hematological manifestations of leucopenia and lymphopenia were verified. Leucopenia was assessed by classifying patients into 3 TLC groups: leucopenia <3000 cells/mm 3 , normal ≥3000 ≤11000 cells/ mm 3 and leukocytosis >11000 cells/mm 3 . Patients with lymphocyte count <1500 cells/mm 3 were considered as those with lymphopenia and ≥1500 cells/mm 3 as normal. SLEDAI total score and NLR were calculated. NLR was calculated by dividing relative percentage values of neutrophil by lymphocyte. Patients were classified on the basis of SLEDAI total score (SELENA-SLEDAI groupings) into the following groups: minimum disease activity or remission (0-3), mild (4)(5)(6)(7)(8), moderate (9)(10)(11), and severe (≥12). 22 The patients were reclassified based on NLR values into 3 groups ≤2, >2-4 and >4. The NLR cut-off values were considered based on the previous studies by the same authors on NLR and rheumatoid arthritis disease activity. 24,25 Drug data of the study subjects was classified as single, double and 3 or more according to the types of immunosuppressants/DMARDs received. Steroid therapy was considered as currently on and not currently on and biologics were considered as given or not given.

Statistics
The independent variables were included based on their clinical relevance. The data are presented as mean±sd for normal distribution, median (min-max) for data without normal distribution and counts for categorical variables. Notched box-and-Whisker plots were used to verify the distribution of NLR, ESR, CRP, C3 and SLEDAI. The scales of the parameter were normalized by linear transformation to 0 -100 range scale using the formula: 100 * (observed value -minimum value)/ (maximum value -minimum value). The SLEDAI and NLR groups were compared for demographic, clinical and inflammatory parameters by ANOVA or Kruskal-Wallis test for continuous variable and chi-square test for categorical data. Pairwise comparison and adjusted residual method was performed to interpret variables with significant differences in ANOVA or Kruskal-Wallis test and chi square test respectively.
The relationship of the SLEDAI and NLR with demographic, clinical and inflammatory parameters was analyzed using bivariate Spearman's correlation. Agreement among the inflammatory parameters (NLR, ESR, CRP, C3 and SLEDAI) was verified by Mountain plot. SLEDAI and NLR subgroup level correlation and agreement were also verified for inflammatory parameters. SLEDAI was considered as standard reference for Mountain plots. The discriminating power of NLR, ESR and C3 was verified by constructing the receiver operating characteristic (ROC) curve using SLEDAI as binary standard. Patients with SLEDAI score ≥12 were classified as having severe disease and 4-11 as mild-moderate disease for ROC analysis. Univariate and multiple linear regressions were used to identify baseline predictors of NLR. To get a parsimonious model, the baseline predictors with P ≤0.15 in univariate regression were included in multivariate model. Sensitivity analysis was performed to verify the influence of steroid therapy on the variables in SLE patients by comparing patients currently on and not on steroid therapy. A two-tailed P <0.05 was taken as statistically significant for all the tests.   of illness, TLC, NLR, ESR, CRP, C3, steroids and biologics did not differ among the SLEDAI disease activity groups.
The sample size of patients reclassified on NLR into 3 groups were as follows: ≤2: n=49, >2-4: n=43 and >4: n=25 ( Table 2). The comparison of variables revealed TLC, lymphocyte count, CRP and steroids differed significantly across the groups. Age, gender, duration of illness, ESR, C3, SLEDAI, immunosuppressants and biologics did not differ among the groups.   The univariate regression estimates demonstrated CRP, immunosuppressants/DMARDs and steroids were statistically significant (P <0.05) factors associated with NLR (Table 3). TLC and biologics were within the P ≤0.15. Age, gender, duration of illness, ESR, C3 and SLEDAI were not significant. Multiple linear regression was performed with TLC, CRP, immunosuppressants/ DMARDs, steroids and biologics as predictors of NLR. The analysis demonstrated that CRP and steroids were predictors of NLR with weak association ( Table 3). The regression model was significant (F= 5.83, df= 5 and P <0.01). The model explained 20.8% of the variance (R2= 0.208). Multicollinearity and homoscedasticity diagnostics showed no violation of assumptions and the dataset did not include influential cases or outliers. The residuals were normally distributed. Sensitivity analysis showed patients on steroid therapy had significantly increased NLR than patients not on steroid therapy (Supplementary table 4). C3 levels were significantly lower in patients on steroid therapy.

Discussion
The present study demonstrated that NLR, ESR and C3 does not discriminate the patients with severe disease activity from those with mild-moderate disease activity. The study demonstrated that CRP and steroids were influencing the NLR and the association was weak. The NLR correlated only in the range of 2 to 4 with SLEDAI scores of moderate disease activity. The relationship of the NLR with SLEDAI was not consistent and this was reflected in Mountain plot analysis.
The study conducted by Wu et al. (2016) on SLE patients (n=116) has demonstrated a statistically significant positive correlation between SLEDAI and NLR. 26 The observations were in line with that noted by . 27  When subgrouped based on NLR, patients with >4 NLR had elevated TLC and CRP levels and reduced lymphocyte counts. Moreover, among subjects with >4 NLR, steroid usage was significantly higher compared to other NLR subgroups. This substantiates the fact that steroid has a stronger influence on the NLR. Influence of CRP and steroids on NLR was corroborated in regression analysis. The sensitivity analysis revealed patients on steroid therapy had significantly higher NLR. Chronic use of steroid influences NLR. Corticosteroids cause demargination of neutrophils contributing to increase in circulating neutrophil counts. It also causes depletion of lymphocytes. Increase in neutrophil counts and depletion of lymphocytes leads to skewing of the NLR. 30 In the current study SLEDAI was not significantly different among the NLR sub-groups.
The major limitation is the number of patients and even lesser number in the sub-groups. SLE being a heterogeneous disease, the organ involvement and treatment were heterogeneous. The single-centre study limits the generalizability of the results and recommends evaluation in a larger sample. The number of male patients was less to draw meaningful conclusions about influence of gender on SLEDAI and NLR. SLEDAI as a classified variable could have provided varied result. Penalized likelihood regression method provides more reliable estimates. The obtained findings could vary in treatment naïve SLE patients and needs to be validated separately. However, the parameters were analysed by different analytical approaches and the findings from these methods is the strength of the study. Each method excludes the bias of different nature. 31 In conclusion, NLR as a marker of inflammation or as a predictor of SLE disease activity was not consistent and needs further investigation. SLEDAI was not associated with NLR, whereas CRP was a predictor of NLR. It is important to consider the influence of steroids and immune-suppressive drugs, while interpreting the NLR. A better understanding of various measures of disease process could explain different aspects of disease and may be important in additional diagnosis and clinical management.