This study was a retrospective, single-center observational study and 220 patients with SLE were recruited from the Second Affiliated Hospital of Guangzhou Medical University between September 2019 and April 2024. All SLE patients were diagnosed according to the 2019 EULAR/ACR classification criteria for SLE (Aringer et al., 2019). The inclusion criteria for patients in the LN group were: (1) persistent proteinuria >0.5 g/day or >3+ by dipstick, and/or the presence of cellular casts (red cell, hemoglobin, granular, tubular, or mixed); or (2) renal biopsy demonstrating immune complex-mediated glomerulonephritis consistent with LN (Hahn et al., 2012). SLE disease activity was evaluated utilizing the systemic lupus erythematosus disease activity index (SLEDAI)-2K score (Gladman et al., 2002). The ocular inclusion criteria were defined as follows: (1) best-corrected visual acuity better than 0.1 log of the Minimum Angle of Resolution (LogMAR); (2) intraocular pressure <21 mmHg; (3) spherical equivalent < +2.5 D or >−6.0 D. Patients with infections, malignancies, and other inflammatory diseases were excluded for this study. The ocular exclusion criteria were defined as follows: (1) spherical equivalent > +6.0 D or <-6.0 D; (2) axial length ≥26 mm; (3) any ocular pathological changes detected on slit lamp, fundus color photography, or optical coherence tomography imaging; (4) history of previous ocular diseases such as glaucoma, cataract or ocular surgery, including refractive surgeries. Informed consent was obtained from all participants, and the study received approval from the Ethics Committee of the Second Affiliated Hospital of Guangzhou Medical University.

Demographic data, clinical manifestations, and laboratory data were retrieved from patients’ medical records. Patient characteristics encompassed age, gender, disease duration, C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), complement 3(C3), C4, neutrophil-to-lymphocyte ratio (NLR), NC3R. Kidney function was evaluated through the following biochemical assays: serum creatinine, glomerular filtration rate (GFR), and 24-h proteinuria. Additionally, we collected information regarding organ involvement in patients, including gastrointestinal involvement, neurological damage, arthritis, skin involvement, hematological abnormalities, and serositis. The NLR and NC3R are calculated by dividing the absolute neutrophil count by the absolute lymphocyte count or serum C3 value.

According to the ETDRS (1991), a standard LogMAR chart was used to assess the best-corrected visual acuity (BCVA) in each eye, thereby evaluating central visual acuity. In the analysis of ocular vasculature, the region of interest (ROI) in the macular region was characterized by circular areas centered on the macula with diameters of 1 mm and 2.5 mm, respectively. The optic disc region was a circular zone centered on the optic disc with a diameter of 1.5 mm, 2.5 mm, 3.5 mm, and 5 mm, respectively. The parafoveal region referred to macular area 0.5–1.5 mm from the foveal center and the peripapillary region encompassed the optic nerve head. The ratio of flow pixels to total pixels was used to calculate macular vessel density (VD). The macular vessel length density (VLD) was calculated using the ratio of vessel length to total area. The foveal avascular zone (FAZ) area and perimeter were calculated by manually delineating the central avascular zone of the macula. The whole image referred to the entire 3 × 3 mm² area centered on the macula/the entire 6 × 6 mm² area centered on the optic disc. This circular region was subdivided further into superior, inferior, temporal, and nasal regions (Li et al., 2024) (Figures 1A–C).

SPSS software (version 25.0) and GraphPad Prism (version 10; GraphPad software) and Hiplot tools were used for analysis. The normality of distributions was tested by the Kolmogorov-Smirnov test. Data that followed a normal distribution were expressed as mean ± standard deviation and analyzed using Student’s t-test, while non-normally distributed data were presented as median and interquartile range (IQR) and evaluated with the Mann-Whitney U-test. The association between NC3R and relevant clinical indicators in SLE patients was assessed using Spearman’s correlation coefficient. The univariate logistic regression and multivariate logistic regression models examined if NLR and NC3R were associated with increased or decreased risk of LN in diagnosed patients. LASSO was applied to manage multiple predictors and reduce overfitting by selecting the most significant variables. A 10-fold cross-validation process was used to identify the optimal lambda value, based on the minimum binomial deviance. ROC and calibration curves evaluated the efficacy of the model. The maximum Youden index, calculated as sensitivity + specificity-1, from the ROC curve analysis, was used to determine the optimal cutoff values for NC3R and NLR. p < 0.05 constituted statistically significant results.

A total of 220 Chinese SLE patients were recruited, including 206 women, and 14 men. They were divided into two groups based on the diagnosis of LN, 142 patients without LN and 78 patients with LN. Demographic and clinical characteristics were summarized in Table 1. The mean disease duration was no significant difference between the two groups. Participants with LN showed significantly higher levels of SLEDAI score (12.5 vs. 10.0, p = 0.017). Notably, LN patients showed a higher NC3R level compared to those without LN (5.9 vs. 4.5, p = 0.007) (Figure 2A). However, there were no significant differences observed in NLR between the two groups (Figure 2B). Serologically, LN patients had a significantly lower GFR, creatinine, and albumin level, and a higher 24-h proteinuria level (all p < 0.01). Gastrointestinal involvement was significantly higher in LN patients compared to non-LN patients (11.5% vs. 2.1%, p = 0.005). However, there were no significant differences in other organ involvements, including neurological damage, APS, arthritis, skin involvement, hematological damage, and serositis (Table 1). A total of 47 patients with LN underwent renal biopsy, including 2 with class II, 4 with class III, 2 with class III + V, 13 with class IV, 21 with class IV + V, and 5 with class V (Table 1). We further investigated the immunotherapy treatment in both groups. The results showed that the proportions of patients using prednisolone, hydroxychloroquine, and cyclosporine were similar between the two groups. However, a higher proportion of patients in the LN group received tacrolimus (7.7% vs. 0.7%, p = 0.009) and mycophenolate mofetil treatment (65.4% vs. 12.7%, p < 0.001). In contrast, a greater percentage of patients without LN were treated with methotrexate (11.3% vs. 2.6%, p = 0.024).

To further investigate the differences in NC3R levels among different types of LN, we compared NC3R levels across the various LN subtypes. Our results indicate that there were no statistically significant differences in NC3R levels between the groups (Supplementary Table S1).

ROC curve analyses were conducted to compare the diagnostic performance of NC3R, and NLR. The AUC for NC3R was 0.613 (95% CI: 0.535-0.690), which was higher than the AUC for NLR 0.570 (95% CI: 0.491-0.649) (Figures 2C, D). The cutoff value of NC3R in identifying LN was 6.40, with a sensitivity of 48.1% and specificity of 72.0%. The cutoff value of NLR was 2.55, with a sensitivity of 72.7% and a specificity of 42.9% (Supplementary Table S2).

We further investigated the relationship between NC3R, NLR, and various clinical parameters including 24-h proteinuria, GFR, C3, C4, ESR, CRP, and SLEDAI scores. NC3R exhibited a positive correlation with 24-h proteinuria (R = 0.329, p < 0.001) and SLEDAI score (R = 0.268, p < 0.001), but a negative association with C3 (R = −0.422, p < 0.001) and C4 (R = −0.408, p < 0.001). Similarly, we found that NLR exhibited a positive correlation with 24-h proteinuria (R = 0.148, p = 0.048), SLEDAI score (R = 0.152, p = 0.028), and CRP (R = 0.271, p = 0.002), but exhibited a negative correlation with GFR (R = −0.177, p = 0.012) as shown in Table 2; Figure 2E.

Univariate logistic analysis revealed possible variables correlated with LN. This was performed on the LN patients and non-LN patients. Results are presented in Table 3. NC3R and NLR were stratified based on a predefined cutoff value. Of the clinical features, NC3R ≥ 6.40 (OR = 2.37, p = 0.004), NLR ≥ 2.55 (OR = 2.01, p = 0.024), GFR (OR = 0.99, p = 0.005) and 24-h proteinuria (OR = 1, p < 0.001) were statistically significantly associated with LN. To avoid multicollinearity and overfitting, LASSO regression was applied following univariate logistic regression screening. A 10-fold cross-validation process identified the optimal lambda value (Figure 3A), with the minimum binomial deviance ensuring the best model performance (Figure 3B). Variables with non-zero coefficients, including NC3R (≥6.40) and NLR (≥2.55), were included in the final logistic regression model (Supplementary Table S3). Our findings revealed that NC3R (≥6.40) independently contributed to the risk of developing LN, with an odds ratio of 2.03 (95% CI: 1.09–3.78, p = 0.025).