Non-histologic factors discriminating proliferative lupus nephritis from membranous lupus nephritis

Data from two independent LN cohorts from two tertiary referral hospitals in Seoul, Korea, were retrospectively reviewed. Both cohorts consisted of patients diagnosed with LN via renal biopsy between July 2006 and December 2018. All patients met the 1997 ACR classification criteria for SLE [8]. Renal biopsies were performed in accordance with the indications recommended by the ACR [7]. The patients were categorized into the proliferative LN (class III, class IV, class III + V and class IV + V) and membranous LN (class V) groups based on their renal biopsy reports. Given that the therapeutic strategy is similar between pure proliferative LN (class III and class IV) and mixed proliferative LN (class III + V and class IV + V), and that the therapeutic strategy in both is different from that in membranous LN [7, 9], pure proliferative LN (class III and class IV) and mixed proliferative LN (class III + V and class IV + V) were both categorized as proliferative LN in the primary analysis. As the purpose of this study was to identify the factors that discriminate proliferative LN from membranous LN, patients with class I, class II, and class VI were excluded. This study was approved by the Institutional Review Board (IRB) of Gangnam Severance Hospital (IRB No: 3-2019-0072). Owing to the retrospective nature of this study, the requirement for informed consent was waived.

Data on the following covariates at the time of renal biopsy were collected: age, sex, presence of hypertension and diabetes mellitus, manifestations of SLE other than those of LN, positivity of antibodies (Abs) to extractable nuclear antigens, anti-double-stranded DNA (anti-dsDNA) Ab, lupus anticoagulant, anti-β₂ glycoprotein Ab and anti-cardiolipin Ab, C3 and C4 levels, serum albumin and creatinine levels, glomerular filtration rate (GFR), urine PCR, urinalysis results, and SLE Disease Activity Index 2000 (SLEDAI-2 K) [10]. Autoantibodies were measured using an automated fluoroimmunoassay analyzer (EliA; Phadia, Uppsala, Sweden). Lupus anticoagulants were assessed using the IL Test TM LAC Screen/Confirm Kit (Instrumentation Laboratory Co., Bedford, MA, USA).

The patients’ characteristics were summarized using descriptive statistics. To compare the characteristics between the proliferative LN group and membranous LN group, Student’s t test or Mann-Whitney test was used for continuous variables and Fisher’s exact test or chi-square test (when appropriate) for categorical variables. Multivariable logistic regression models were constructed to identify the covariates associated with proliferative LN. Covariates with a p value of < 0.05 in the univariable logistic regression analysis were incorporated to the multivariable models. In the multivariable analysis, the variable inflation factor was tested to exclude multicollinearity among covariates. The Hosmer-Lemeshow test was used to assess the goodness of fit for the logistic regression models. Antibodies and complements (C3 and C4) were analyzed as binary variables (positive/negative for antibodies, and low/not low for complements) in univariable analysis and multivariable analysis (model 1) and were analyzed as continuous variables in multivariable analysis (model 2). We used a receiver-operating characteristic (ROC) analysis to assess the ability of the covariates identified in the multivariable models in discriminating proliferative LN from membranous LN. ROC curves were generated, and the associated area under the curve (AUC) for each covariate was determined. The statistical significance level was set at a p value of < 0.05. All analyses were conducted using the SPSS software (version 25.0; IBM Corporation, Armonk, NY, USA).

To test the robustness of our results, we performed several sensitivity analyses. First, we used a more restrictive definition of proliferative LN. Instead of including both pure proliferative LN (class III and class IV) and mixed proliferative LN (class III + V and class IV + V) in the proliferative LN group, we included only the pure proliferative LN (class III and class IV) in the proliferative LN group and performed multivariable logistic regression analysis and ROC analysis. Second, we compared mixed proliferative LN (class III + V and class IV + V) with membranous LN (class V). Third, as patients with classes I, II, and VI would also have to undergo a renal biopsy for diagnostic purposes, we included the patients with classes I, II, and VI and compared pure proliferative LN (class III and class IV) with non-proliferative LN (class I, class II, class V, and class VI).

A total of 176 patients with biopsy-proven LN were included. The patients were predominantly women (90.9%), with a mean age of 36.7 ± 15.0 years. Among the patients, 150 patients (85.2%) had proliferative LN, and 18 patients (10.2%) had membranous LN. Of the 150 patients with proliferative LN, 122 (81.3%) patients had pure proliferative LN (class III, 38 patients; class IV, 84 patients) and 28 (18.7%) patients had mixed proliferative LN (class III + V, 17 patients; class IV + V, 11 patients) (Table 1). Three (1.7%) patients with class I, four (2.3%) patients with class II, and one (0.6%) patient with class VI were excluded for primary analysis.

The comparison of the characteristics between the two groups is shown in Table 2. Age (p = 0.269), sex distribution (p = 0.649), and the proportion of patients with hypertension (p > 0.999) and diabetes mellitus (p = 0.599) did not differ between the two groups. The proportion of patients with mucocutaneous manifestations (p > 0.999), musculoskeletal manifestations (p = 0.408), neuropsychiatric manifestations (p = 0.149), and serositis (p = 0.475) was similar in the two groups; conversely, hematologic manifestations were more common in the patients with proliferative LN (44.7% vs 16.7%, p = 0.023). In the comparison of serologic covariates, no significant differences were observed in the positivity for anti-Sm Ab (p = 0.384), anti-Ro Ab (p = 0.885), anti-La Ab (p = 0.258), lupus anticoagulant (p = 0.768), anti-β₂ glycoprotein Ab (p = 0.077), and anti-cardiolipin Ab (p = 0.566) and albumin levels (p = 0.800). The patients with proliferative LN were less commonly positive for anti-U1RNP Ab (48.7% vs 77.8%, p = 0.020) and more commonly positive for anti-dsDNA Ab (88.0% vs 50.0%, p < 0.001), and more commonly had low C3 (94.0% vs 61.1%, p < 0.001) and low C4 (79.3% vs 50.0%, p = 0.015), higher creatinine level (1.13 ± 0.79 mg/dL vs 0.70 ± 0.26 mg/dL, p < 0.001), and lower GFR (83.2 ± 37.0 mL/min/1.73 m² vs 105.1 ± 23.8 mL/min/1.73 m², p = 0.002). Regarding the urine laboratory data, the urine PCR (p = 0.778) and proportion of patients with pyuria (p = 0.053) and urine casts (p = 0.202) did not differ between the two groups. The proportion of patients with urine RBC of ≥ 5/HPF was higher in the proliferative LN group (72.7% vs 38.9%, p = 0.003). The SLEDAI-2 K was also higher in the proliferative LN group (17.1 ± 5.8 vs 12.2 ± 5.8, p = 0.001).

In the univariable logistic regression analysis, the presence of hematologic manifestations (unadjusted odds ratio [OR] = 4.036, 95% confidence interval [CI] = 1.121–14.527, p = 0.033), positive anti-dsDNA Ab (unadjusted OR = 7.333, 95% CI = 2.574–20.893, p < 0.001), low C3 (unadjusted OR = 9.970, 95% CI 3.117–31.891, p < 0.001), low C4 (unadjusted OR 3.839, 95% CI 1.405–10.486, p = 0.009), creatinine levels (unadjusted OR = 10.645, 95% CI = 1.350–83.953, p = 0.025), presence of urine RBC of ≥ 5/HPF (unadjusted OR = 4.718, 95% CI = 1.516–11.509, p = 0.006), and higher SLEDAI-2 K (unadjusted OR = 1.173, 95% CI = 1.063–1.294, p = 0.001) were associated with proliferative LN. The positivity of anti-U1RNP Ab (unadjusted OR = 0.271, 95% CI = 0.085–0.861, p = 0.027) and GFR (unadjusted OR = 0.980, 95% CI = 0.964–0.997, p = 0.021) were inversely associated with proliferative LN. These covariates were included in the multivariable models, except for the creatinine level, because of the multicollinearity with the GFR.

In multivariable analysis model 1, anti-U1RNP (positive/negative), anti-dsDNA Ab (positive/negative), C3 (low/not low), and C4 (low/not low) were analyzed as the categorical variables. An alternative multivariable analysis model (model 2) was also performed, in which anti-U1RNP, anti-dsDNA Ab, C3 level, and C4 level were analyzed as the continuous variables. Both the positivity (model 1) for and level (model 2) of anti-dsDNA Ab (model 1: adjusted OR = 11.200, 95% CI = 2.202–56.957, p = 0.004; model 2: adjusted OR = 1.008, 95% CI = 1.002–1.014, p = 0.014) were associated with proliferative LN; conversely, both the positivity (model 1) for and level (model 2) of anti-U1RNP Ab (model 1: adjusted OR = 0.176, 95% CI = 0.040–0.769, p = 0.021; model 2: adjusted OR = 0.985, 95% CI = 0.0976–0.994, p = 0.002) were inversely associated with proliferative LN. Higher GFR (model 1: adjusted OR = 0.973, 95% CI = 0.951–0.994, p = 0.013; model 2: adjusted OR = 0.969, 95% CI = 0.946–0.994, p = 0.014) was also inversely associated with proliferative LN (Table 3).

The ROC curves for anti-U1RNP Ab, anti-dsDNA Ab, and the GFR are shown in Fig. 1. Anti-dsDNA Ab had the highest discrimination ability (AUC = 0.806, 95% CI = 0.695–0.916), followed by anti-U1RNP Ab (AUC = 0.677, 95% CI = 0.527–0.827) and GFR (AUC = 0.662, 95% CI = 0.554–0.770). When a combination of anti-dsDNA Ab, anti-U1RNP Ab, and GFR was used as a composite parameter, the discrimination ability (AUC = 0.864, 95% CI = 0.792–0.937) was higher than when each parameter was used as a single parameter (Fig. 1d).

When using the stricter proliferative LN definition (i.e., pure proliferative LN: class III and class IV), the ORs of anti-dsDNA Ab (model 1: adjusted OR = 19.591, 95% CI = 2.518–152.431, p = 0.004; model 2: adjusted OR = 1.008, 95% CI = 1.002–1.015, p = 0.012), anti-U1RNP Ab (model 1: adjusted OR = 0.178, 95% CI = 0.034–0.922, p = 0.040; model 2: adjusted OR = 0.987, 95% CI = 0.978–0.996, p = 0.007), and GFR (model 1: adjusted OR = 0.966, 95% CI = 0.941–0.992, p = 0.010; model 2: adjusted OR = 0.967, 95% CI = 0.942–0.992, p = 0.011) remained significant (Table 4). Further, the discrimination ability of anti-U1RNP Ab (AUC = 0.673, 95% CI = 0.522–0.823), anti-dsDNA Ab (AUC = 0.822, 95% CI = 0.713–0.931), GFR (AUC = 0.688, 95% CI = 0.579–0.798), and combination of anti-dsDNA Ab, anti-U1RNP Ab, and GFR as a composite parameter (AUC = 0.873, 95% CI = 0.801–0.945) was similar to the primary analysis (Fig. 2).

In another sensitivity analysis where mixed proliferative LN (class III + V and IV + V) was compared with membranous LN (class V), positivity of anti-dsDNA Ab was still significantly associated with mixed proliferative LN (model 1: adjusted OR = 4.545, 95% CI 1.107–18.661, p = 0.036) (Table 5), although the effect size was attenuated compared with the primary analysis (anti-dsDNA Ab in model 1: adjusted OR = 11.200) (Table 3) and in the sensitivity analysis where pure proliferative LN was compared with membranous LN (anti-dsDNA Ab in model 1: adjusted OR = 19.591) (Table 4). The effect size of anti-U1RNP Ab (unadjusted OR = 0.286, 95% CI 0.075–1.086, p = 0.066) and GFR (unadjusted OR = 0.988, 95% CI 0.968–1.009, p = 0.273) was also attenuated and failed to reach statistical significance (Table 5).

In comparison between pure proliferative LN (class III and class IV) and non-proliferative LN (class I, class II, class V, and class VI), anti-dsDNA Ab (model 1: adjusted OR = 13.741, 95% CI = 3.058–61.753, p = 0.001; model 2: adjusted OR = 1.008, 95% CI = 1.003–1.012, p = 0.002) was associated with pure proliferative LN, and anti-U1RNP Ab (model 1: adjusted OR = 0.273, 95% CI = 0.085–0.873, p = 0.029; model 2: adjusted OR = 0.991, 95% CI = 0.984–0.998, p = 0.012) and GFR (model 1: adjusted OR = 0.970, 95% CI = 0.952–0.989, p = 0.003; model 2: adjusted OR = 0.972, 95% CI = 0.954–0.990, p = 0.002) were inversely associated with pure proliferative LN, supporting their value in predicting proliferative LN (Table 6).