Derived inflammatory indicators for identifying preeclampsia in pregnant patients with systemic lupus erythematosus: a case–control study

Abstract

Background

Pregnant women with systemic lupus erythematosus (SLE) are at high risk of preeclampsia (PE). However, overlapping clinical features between the two conditions make it challenging to detect superimposed PE. Failure to promptly recognize PE significantly increases the risk of adverse pregnancy outcomes. In recent years, complete blood count–derived inflammatory indicators (e.g., platelet-to-lymphocyte ratio) have emerged as readily available, low-cost biomarkers with clinical utility. This study aimed to assess the discriminative capacity of these derived inflammatory indicators, measured within one week before delivery, for identifying superimposed PE in pregnant patients with SLE.

Methods

This retrospective single-center study included 157 pregnant women with SLE treated at Nanfang Hospital of Southern Medical University from April 2010 to October 2024. Clinical and laboratory data, including complete blood count, liver function tests, and coagulation parameters, were collected. Univariate and multivariate logistic regression were performed to identify risk factors for PE and construct combined models. Model performance was assessed using receiver operating characteristic (ROC) curves, Hosmer–Lemeshow test, calibration plots, and the integrated discrimination improvement (IDI) and net reclassification improvement (NRI) indices. Internal validation used the bootstrap method, and sensitivity analysis used E-values.

Results

Among the study patients, 35 developed PE. Univariate logistic regression analysis showed significant differences in monocyte-to-lymphocyte ratio (MLR) and platelet-to-lymphocyte ratio (PLR) between groups (P < 0.05). Multivariable logistic regression identified that lower MLR (OR = 0.223, 95% CI: 0.071–0.699) and PLR (OR = 0.344, 95% CI: 0.175–0.675) levels may be associated with an increased risk of superimposed PE in SLE pregnancies. The models based on MLR (AUC = 0.859, 95% CI: 0.782–0.936) and PLR (AUC = 0.860, 95% CI: 0.786–0.934) each showed discriminative capacity for identifying PE in pregnant patients with SLE.

Conclusion

MLR and PLR may serve as cost-effective and readily available markers for identifying superimposed preeclampsia in pregnant patients with SLE. These preliminary findings warrant validation in larger prospective cohorts, given the small sample size.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12884-026-09681-2.

Background

Systemic lupus erythematosus (SLE) is a chronic autoimmune disease marked by impaired immune tolerance, widespread inflammation, and multi-organ injury [1]. This disease predominantly affects women of childbearing age. In China, SLE is a serious clinical and public health burden, with approximately 106,610 new cases each year, which is the highest estimated global incidence of new cases in women [2, 3]. Pregnancy in women with SLE markedly elevates the risk of adverse pregnancy outcomes (12%–47%). These outcomes include preeclampsia (PE), preterm premature rupture of membranes, miscarriage, preterm birth, and fetal growth restriction [4, 5]. Among these complications, PE has a particularly high prevalence (11%–52%) in SLE pregnancies and is an important cause of maternal and neonatal mortality [6, 7].

Both SLE and PE have similar clinical features, such as hematological abnormalities, hypertension, and proteinuria. However, their treatments are significantly different. SLE is managed with immunosuppressive therapy, while the only definitive way to resolve PE is delivery. The overlapping clinical manifestations but different treatments make clinical diagnosis more difficult [8, 9]. In addition, the coexistence of SLE and PE can lead to worse pregnancy outcomes [10]. Therefore, the early and accurate identification of PE in pregnant women with SLE is critical for optimizing maternal and neonatal outcomes.

Several biomarkers have been explored for the detection of superimposed PE in SLE pregnancies, including elevated 24-h urinary protein, mean arterial pressure ≥ 96.5 mmHg, low albumin, high uric acid, elevated sFlt-1/PlGF ratio, and low serum C4. However, their discriminative performance still needs further validation, and some of these markers are limited by high cost or restricted accessibility [11–14]. Specifically, sFlt-1/PlGF ratio, anti-dsDNA antibodies, complement levels, and uric acid are primarily used in suspected PE rather than as routine screening markers in SLE pregnancies [15–17]. Emerging evidence indicates that both SLE and PE are closely linked to immune dysregulation and excessive inflammatory responses, suggesting they may share similar inflammatory pathways [18, 19]. In recent years, several readily available indicators derived from complete blood count have been explored in clinical research. These include the systemic immune-inflammation index (SII), monocyte-to-lymphocyte ratio (MLR), neutrophil-to-lymphocyte ratio (NLR), and platelet-to-lymphocyte ratio (PLR). The changes in these indicators, whether the levels are elevated or reduced, can effectively reflect the imbalance in innate immune-inflammatory homeostasis. Previous studies have demonstrated their value in assessing SLE activity and predicting pregnancy outcomes, including PE in the general obstetric population [20–26]. However, their clinical utility in identifying PE in SLE pregnancies has been incompletely defined.

This study aimed to explore the discriminative capacity of complete blood count–derived inflammatory indicators (SII, NLR, MLR, and PLR) for identifying superimposed PE in SLE pregnancies based on laboratory parameters obtained within one week before delivery, and to provide preliminary evidence for their potential use as convenient, low-cost screening aids when more specialized tests are not readily available.

Methods

Subjects

This retrospective, single-center study analyzed 157 pregnant women with SLE in the second or third trimester treated at Nanfang Hospital, Southern Medical University, between April 2010 and October 2024, where diagnostic criteria and laboratory protocols for SLE and PE remained consistent throughout the study period.

Inclusion criteria were as follows: (1) Diagnosis of SLE in accordance with the classification criteria of the 1997 American College of Rheumatology; (2) Diagnosis of PE based on the American College of Obstetricians and Gynecologists (ACOG) criteria.

Exclusion criteria included: (1) Comorbidity of other autoimmune diseases; (2) Presence of other conditions affecting hematological inflammatory indicators, including hematological disorders, active infections, diabetes mellitus, pulmonary arterial hypertension, and hemolytic anemia; (3) Incomplete or uninterpretable laboratory data. Therefore, the final analytical dataset contained no missing data.

The study protocol was approved by the Ethics Committee of Nanfang Hospital, Southern Medical University, China (NFEC-2020–155), and all procedures were conducted in compliance with the Declaration of Helsinki.

Data collection

Data were collected for all participants in two categories: (1) Baseline demographic and clinical characteristics, including age, pre-pregnancy body mass index (BMI), gestational age at delivery, parity, obstetric history, mode of delivery, and adverse pregnancy outcomes; (2) Laboratory parameters measured within one week prior to delivery: complete blood count (white blood cell, monocyte, lymphocyte, neutrophil, and platelet counts), liver function tests (total protein and albumin), and coagulation tests (prothrombin time and prothrombin activity percentage).

Definition of derived inflammatory indicators

Derived inflammatory indicators were calculated based on monocyte, lymphocyte, neutrophil, and platelet counts obtained from the complete blood count. These indicators included the systemic immune-inflammation index (SII), neutrophil-to-lymphocyte ratio (NLR), monocyte-to-lymphocyte ratio (MLR), and platelet-to-lymphocyte ratio (PLR). The SII was computed using the formula: SII = (neutrophil count × platelet count)/lymphocyte count.

Definition of SLE disease activity

For new-onset SLE cases identified during pregnancy, SLE disease activity was assessed using a composite definition requiring the simultaneous presence of: (1) anti-dsDNA positivity; (2) low complement levels (C3 or C4); and (3) specific clinical manifestations, including mucocutaneous lesions, serositis, lupus nephritis, or cytopenias. Although not a standardized instrument, this composite definition aligns with the pregnancy-related SLE monitoring indicators recommended by the 2023 SMFM and EULAR guidelines, and was adopted as a clinically feasible alternative for assessing disease activity in this retrospective cohort.

Statistical analysis

Statistical analyses were conducted using IBM SPSS Statistics 25.0 and R software (version 4.2.3), with a two-sided significance level of α = 0.05. Normally distributed continuous variables were presented as mean ± standard deviation, non-normally distributed continuous variables as median (interquartile range [IQR]), and categorical variables as number (percentage). Group comparisons were performed using the independent-samples t-test for normally distributed continuous variables, the Mann–Whitney U test for non-normally distributed variables, and Pearson's chi-square test for categorical variables. Due to right-skewed distributions, SII, NLR, MLR, and PLR were log-transformed, with MLR multiplied by 100 to improve OR interpretability.

Variables with P < 0.05 in univariate analyses were selected for multivariable logistic regression analysis after assessing collinearity using variance inflation factor (VIF) and evaluation of clinical relevance. Given the limited number of PE cases (n = 35), an events-per-variable (EPV) ratio > 5 was maintained to avoid overfitting, capping each model at five independent variables (EPV = 7). The original cell counts (white blood cell count, monocyte count, lymphocyte count, neutrophil count, and platelet count) share a deterministic mathematical relationship with their derived indicators (SII, MLR, NLR, PLR). Including both would compromise the biological interpretability of the regression coefficients. Accordingly, only the derived indicators were entered. For each derived indicator, three models were constructed: Model 1 (unadjusted), Model 2 (adjusted for demographic characteristics), and Model 3 (further adjusted for additional laboratory parameters). Notably, albumin and prothrombin activity percentage were specifically included in Model 3 as adjustment variables because they have been established as PE-related confounders [27, 28] with mechanistic relevance to endothelial dysfunction and hypercoagulability.

Receiver operating characteristic (ROC) curves were constructed for Model 3, with the area under the curve (AUC) calculated. Model goodness-of-fit was assessed with the Hosmer–Lemeshow test and calibration plots. Improvements in discriminative performance were quantified using the integrated discrimination improvement (IDI) and net reclassification improvement (NRI). Internal validation was performed using the bootstrap method with 1,000 resampling iterations. Sensitivity analyses were performed using E-values, calculated based on the ORs and confidence intervals [29].

Due to the retrospective design and because these biomarkers are not routinely measured in all SLE pregnancies, the sFlt-1/PlGF ratio and uric acid had a missing rate exceeding 50%. To avoid sample size reduction and selection bias, these variables were excluded from model construction. Complement levels and anti-dsDNA antibodies, used in our composite SLE disease activity definition, were not included as independent variables due to incomplete records.

Results

Baseline characteristic of SLE pregnancies

This study included 157 pregnant women with SLE, of whom 35 were diagnosed with preeclampsia (PE group) and 122 without PE comprised the non-PE group. Key baseline characteristics of SLE pregnancies are summarized in Table 1, with the complete baseline characteristics available in Additional file 1: Table S1. Significant differences between the PE and non-PE groups were observed in age, pre-pregnancy BMI, white blood cell count, lymphocyte count, neutrophil count, platelet count, albumin, prothrombin activity percentage, MLR, and PLR (all P < 0.05). SII showed no significant difference between groups (P > 0.05), though its levels tended to be lower in the PE group.

Table 1.

Key baseline characteristics of the study population

Variables	Total (n = 157)	Non-PE group (n = 122)	PE group (n = 35)	P
Age (mean ± SD)	28.54 ± 4.14	28.16 ± 3.95	29.89 ± 4.56	0.029
Pre-pregnancy BMI (mean ± SD)	24.77 ± 3.51	24.40 ± 3.32	26.04 ± 3.89	0.014
Pregnancy, n (%)				0.618
Second trimester	27 (17.2%)	20 (16.4%)	7 (20.0%)
Third trimester	130 (82.8%)	102 (83.6%)	28 (80.0%)
SLE disease activity, n (%)				1.000
Yes	10 (6.4%)	8 (6.6%)	2 (5.7%)
No	147 (93.6%)	114 (93.4%)	33 (94.3%)
White blood cell count	8.83 (6.95, 10.59)	8.61 (6.51, 10.51)	9.68 (8.16, 11.58)	0.006
Monocyte count	0.52 (0.36, 0.65)	0.53 (0.36, 0.66)	0.47 (0.31, 0.61)	0.338
Lymphocyte count	1.40 (1.02, 1.90)	1.32 (0.99, 1.78)	1.73 (1.33, 2.36)	0.001
Neutrophil count	6.76 (4.94, 8.59)	6.37 (4.89, 8.33)	7.15 (6.06, 9.70)	0.018
Platelet count (mean ± SD)	186.23 ± 79.06	194.40 ± 75.52	157.74 ± 85.48	0.015
Albumin	33.80 (29.20, 36.20)	34.25 (30.80, 37.12)	29.10 (22.50, 33.80)	< 0.001
Prothrombin activity percentage	106.90 (96.70, 120.50)	104.30 (95.00, 115.78)	122.20 (105.60, 137.80)	< 0.001
SII	850.04 (495.01, 1288.38)	896.49 (601.38, 1287.71)	709.16 (294.45, 1297.50)	0.086
NLR	4.83 (3.17, 7.19)	4.78 (3.22, 6.83)	4.84 (2.98, 7.40)	0.863
MLR	0.36 (0.25, 0.49)	0.38 (0.28, 0.49)	0.24 (0.19, 0.39)	0.001
PLR	129.17 (92.77, 180.35)	139.71 (98.23, 188.42)	95.43 (46.46, 136.76)	< 0.001

Data are expressed as median (Q1, Q3) unless noted

Logistic regression analysis of risk factors for superimposed PE in SLE pregnancies

Due to the right-skewed distributions of SII, NLR, MLR, and PLR, natural logarithmic transformations were applied before analysis, and MLR was multiplied by 100 to improve OR interpretability [30, 31]. Univariate logistic regression analysis identified that several factors may be associated with the risk of superimposed PE in SLE pregnancies (as shown in Fig. 1). Advanced age, higher pre-pregnancy BMI, elevated white blood cell count, lymphocyte count, neutrophil count, and increased prothrombin activity percentage were positively associated with PE risk. In contrast, inverse associations were observed for platelet count, albumin, MLR (OR = 0.314, 95% CI: 0.141–0.701), and PLR (OR = 0.367, 95% CI: 0.214–0.630).

Fig. 1.

Forest plot of univariate logistic regression analysis

Separate logistic regression models were constructed for SII, MLR, and PLR. For each indicator, Model 1 was unadjusted; Model 2 was adjusted for age and pre-pregnancy BMI; and Model 3 was further adjusted for albumin and prothrombin activity percentage. In Model 3, all VIFs were below 2, indicating no significant multicollinearity (Additional file 1: Table S2).

As shown in Table 2, in the fully adjusted Model 3, lower MLR (OR = 0.223, 95% CI: 0.071–0.699) and lower PLR (OR = 0.344, 95% CI: 0.175–0.675) were significantly associated with superimposed PE, with OR < 1, suggesting that a decrease in MLR or PLR corresponded to higher PE risk.

Table 2.

Logistic regression analysis of derived inflammatory indicators

Variables	Model 1		Model 2		Model 3
	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
InSII	0.647 (0.410–1.019)	0.061	0.553 (0.337–0.908)	0.019	0.590 (0.328–1.062)	0.079
InMLR	0.314 (0.141–0.701)	0.005	0.185 (0.074–0.459)	< 0.001	0.223 (0.071–0.699)	0.010
InPLR	0.367 (0.214–0.630)	< 0.001	0.304 (0.166–0.558)	< 0.001	0.344 (0.175–0.675)	0.002

Model 1 was unadjusted; Model 2 was adjusted for age and pre-pregnancy BMI; Model 3 was further adjusted for albumin and prothrombin activity percentage

Trend analysis of derived inflammatory indicators

SII, MLR, and PLR were divided into quartiles (Q1–Q4), with Q1 serving as the reference. Due to the limited number of PE cases (n = 35), more complex nonlinear analyses could not be performed; a linear trend test based on group medians was employed. In Model 3, the ORs for MLR and PLR did not follow a strictly monotonic decreasing pattern across quartiles; however, the overall trend from Q1 to Q4 decreased significantly (P for trend < 0.05), suggesting inverse associations with PE risk. The observed departures from strict monotonicity are likely attributable to random variation in the small sample. Complete quartile-based results are available in Additional file 1: Table S3.

Discriminative performance of derived inflammatory indicators for superimposed PE in SLE pregnancies

Discriminative performance was evaluated through ROC curves, the Hosmer–Lemeshow test, calibration plots, and the IDI and NRI. The fully adjusted Model 3 yielded AUCs of 0.837 (95% CI: 0.758–0.915) for SII, 0.859 (95% CI: 0.782–0.936) for MLR, and 0.860 (95% CI: 0.786–0.934) for PLR (as shown in Fig. 2). Calibration curves showed acceptable agreement between predicted and observed probabilities, supported by non-significant Hosmer–Lemeshow test results (all P > 0.1; Additional file 1: Figure S1). Model 3 consistently achieved the highest IDI and NRI values across all models evaluated, suggesting that the incorporation of age, pre-pregnancy BMI, albumin, and prothrombin activity percentage provided incremental discriminatory capacity for identifying superimposed PE in SLE pregnancies. Specifically, the MLR-based and PLR-based Model 3 showed comparable discriminative performance (as shown in Table 3). It should be noted that IDI and NRI are best interpreted as statistical measures of incremental discrimination rather than direct indicators of clinical utility, and their estimates may be sensitive to limited sample sizes.

Fig. 2.

ROC curves of Model 3 based on SII, MLR, and PLR

Table 3.

Discriminative performance of derived inflammatory indicators

Variables	AUC (95% CI)	Continuous NRI (95% CI)	IDI (95% CI)	Sensitivity	specificity
InSII
Model 1	0.595 (0.479–0.712)	Ref	Ref	0.457	0.779
Model 2	0.696 (0.596–0.796) ***	0.380 (0.012–0.749) *	0.081 (0.023–0.139) **	0.771	0.598
Model 3	0.837 (0.758–0.915) ***	1.166 (0.855–1.477) ***	0.306 (0.205–0.407) ***	0.800	0.770
InMLR
Model 1	0.688 (0.575–0.800) **	Ref	Ref	0.600	0.820
Model 2	0.747 (0.646–0.847) ***	0.626 (0.269–0.983) **	0.120 (0.050–0.190) **	0.571	0.852
Model 3	0.859 (0.782–0.936) ***	1.092 (0.770–1.414) ***	0.299 (0.200–0.397) ***	0.743	0.869
InPLR
Model 1	0.708 (0.608–0.809) ***	Ref	Ref	0.743	0.598
Model 2	0.768 (0.677–0.860) ***	0.462 (0.102–0.822) *	0.081 (0.019–0.143) **	0.800	0.705
Model 3	0.860 (0.786–0.934) ***	1.109 (0.788–1.430) ***	0.274 (0.171–0.376) ***	0.829	0.811

Model 1 was unadjusted; Model 2 was adjusted for age and pre-pregnancy BMI; Model 3 was further adjusted for albumin and prothrombin activity percentage. *** P < 0.001; ** P < 0.01; * P < 0.05

Internal validation and sensitivity analyses

The stability of Model 3 was assessed through internal validation using the Bootstrap method with 1,000 resampling iterations. The models for SII, MLR, and PLR yielded AUCs of 0.837 (95% CI: 0.758–0.909), 0.859 (95% CI: 0.774–0.930), and 0.860 (95% CI: 0.778–0.926), respectively (as shown in Fig. 3a–c). These results were consistent with the original model estimates, suggesting good stability. E-values for MLR (point estimate: 8.05; CI limit: 2.22) and PLR (point estimate: 5.27; CI limit: 2.32) were calculated, further suggesting robustness against unmeasured confounders (as shown in Additional file 1: Table S4).

Fig. 3.

Results of 1000 bootstrap resamples for Model 3 based on SII (a), MLR (b), and PLR (c)

Discussion

Derived inflammatory indicators (SII, NLR, MLR, PLR) have been associated with the risk of PE and SLE in prior studies [32–36]. However, their expression patterns and clinical implications in SLE pregnancies with superimposed PE remain insufficiently characteristics. Because deviations from immunoinflammatory homeostasis, whether elevated or reduced, may reflect a state of exacerbated inflammatory dysregulation, this study compared these indicators between SLE pregnancies with and without superimposed PE to identify candidate discriminative markers. In our study, compared to the non-PE group, the PE group showed significantly elevated white blood cell, lymphocyte, and neutrophil counts, accompanied by lower platelet counts, MLR, and PLR. In multivariable models, lower MLR and PLR emerged as independent risk factors associated with superimposed PE, and the fully adjusted models incorporating MLR and PLR each exhibited satisfactory discriminative capacity in this cohort. Taken together, our findings suggest that MLR and PLR may have potential utility as cost-effective and readily available screening aids for identifying superimposed PE in pregnant women with SLE.

Our findings suggest that decreased MLR and PLR may serve as potential risk indicators of superimposed PE in SLE pregnancies. Previous studies [20, 37] have established that SLE without PE exhibits inherent inflammatory dysregulation, evidenced by markedly elevated SII, NLR, MLR, and PLR compared with healthy pregnancies, a pattern confirmed in our cohort. However, in our study, the PE group showed significantly lower MLR and PLR than the non-PE group, an association that persisted after multivariable adjustment, suggesting an inverse relationship with superimposed PE. The alterations in MLR and PLR may reflect the unique immune-inflammatory microenvironment in SLE pregnancies with superimposed PE. Building on the pre-existing immune dysregulation in SLE, the superimposed PE state likely exacerbates the imbalance between pro-inflammatory lymphocytes (e.g., Th1) and regulatory T cells (e.g., Tregs); concurrently, endothelial dysfunction may drive platelet activation and consumption [38–40]. The combination of lymphocyte redistribution and platelet consumption may result in elevated lymphocytes and decreased platelets, ultimately leading to reduced MLR and PLR. Our observations suggested potential synergistic or additive interactions between the immunopathological mechanisms of SLE and PE. Importantly, decreased MLR and PLR in the PE group should not be misinterpreted as evidence of attenuated systemic inflammation lower ratios should be regarded as markers of immune-coagulation dysregulation rather than conventional anti-inflammatory markers.

Lymphocyte and neutrophil reduction is a common feature of SLE, and NLR is usually elevated [41, 42]. However, in our study, SLE pregnancies with PE exhibited markedly different blood profiles: both lymphocyte and neutrophil counts were significantly higher than those without PE, a pattern that led to relatively stable NLR levels. This phenomenon may be explained by the inherent pathological processes of PE. Inflammatory factors (e.g., IL-6) and anti-angiogenic factors (e.g., sFlt-1) released from the ischemic and hypoxic placenta strongly activate the innate immune system and drive substantial neutrophil mobilization; meanwhile, the systemic inflammatory environment associated with PE also promotes the mobilization and proliferation of specific lymphocyte subsets [18, 43, 44]. The stability of NLR suggests a synchronous activation of both innate and adaptive immune systems under this condition. This finding suggests that, superimposed on the chronic immune dysfunction inherent in SLE, PE may reflect a more intense and distinct systemic inflammatory response, leading to unique alterations in the blood cell profile.

This study also analyzed other risk factors for superimposed PE in SLE pregnancies. The results suggested that increased age, pre-pregnancy BMI, prothrombin activity percentage, and decreased albumin were also associated with superimposed PE. Among these, hypoalbuminemia (decreased albumin) and increased prothrombin activity percentage have been previously linked to PE [27, 28]. Hypoalbuminemia may exacerbate endothelial dysfunction, while a hypercoagulable state can promote microthrombus formation. Their synergistic interaction likely contributes to greater clinical complexity. Therefore, albumin and prothrombin activity percentage were included as covariates for controlling confounding from these mechanisms, allowing a clearer estimation of derived inflammatory indicators. Furthermore, the observed lower SII levels did not reach statistical significance in the PE group, which may be attributable to the limited sample size. As a composite biomarker of systemic inflammation, SII warrants further investigation in larger cohorts to evaluate its sensitivity in multifactorial clinical contexts.

To further explore their clinical utility, we developed models to assess the individual and combined discriminative performance of selected derived inflammatory indicators for superimposed PE in SLE pregnancies. Among all models, the fully adjusted Model 3 (adjusted for age, pre-pregnancy BMI, albumin, and prothrombin activity percentage) showed satisfactory discriminative capacity. The MLR-based and PLR-based Model 3 showed comparable performance, with AUC values of 0.859 and 0.860, compared to 0.688 and 0.708 in Model 1. Furthermore, they also yielded IDIs of 0.299 and 0.274 and NRIs of 1.092 and 1.109, suggesting potential improvement in reclassifying SLE pregnancies with or without PE. These findings suggest that the combined models may be useful for identifying superimposed PE in this cohort, though external validation is required. Because these inflammatory indicators are derived from complete blood counts, the models may represent a potential cost-effective and easily accessible approach for identifying superimposed PE. However, the limited sample size, particularly for multivariable logistic regression and AUC values, increases the risk of overfitting. Although the EPV rule of 10 has been challenged as overly conservative, with acceptable performance achievable even at EPV values of 5–7 in certain settings [45, 46], and bootstrap internal validation can mitigate overfitting concerns in small samples [47, 48], the reliability of these internal validation procedures may still be limited by only 35 events. Calibration fit may partly reflect overfitting rather than genuine accuracy. Similarly, the ROC curve may yield an overly optimistic AUC due to limited sample size, and the IDI and NRI estimates are likewise subject to considerable sampling variability. Notably, several studies in SLE pregnancy or PE research with comparable or smaller sample sizes have yielded clinically meaningful findings [26, 49], supporting the feasibility of our exploratory analysis. Nonetheless, these findings should remain strictly exploratory, and external validation in larger, adequately powered cohorts is essential to confirm generalizability. Additionally, although SII demonstrated limited stand-alone capacity, incorporating it into the combined models still offered measurable discriminatory capacity, possibly reflecting its broader capture of systemic inflammation. SII may still hold adjunctive value as an exploratory marker.

Several SLE-related factors may confound our model estimates, including SLE disease activity, corticosteroid exposure, hydroxychloroquine treatment, aspirin prophylaxis, and immunosuppressive therapies. Available evidence suggests that SLE disease activity is the predominant driver of PE risk in SLE pregnancies, and this has been associated with a more than two-fold increased PE risk in a meta-analysis [50]. Although most patients in our cohort had planned pregnancies and the two groups were comparable with respect to disease activity surrogates (anti-dsDNA, complement levels, and clinical manifestations), residual confounding from unmeasured factors (e.g., disease fluctuations or medication dosages) cannot be fully ruled out. Some meta-analyses on hydroxychloroquine have yielded inconsistent results [51, 52], and corticosteroids are primarily linked to fetal risks rather than being independent drivers of PE [53]. Low-dose aspirin is universally recommended for all SLE pregnancies [54], suggesting comparable exposure between groups. Moreover, in our study, E-values for MLR (point estimate: 8.05; CI limit: 2.22) and PLR (point estimate: 5.27; CI limit: 2.32) suggest that the observed associations are reasonably robust against potential unmeasured confounders. Therefore, the impact of therapeutic confounders on our findings is likely limited. Given the retrospective design, the direction and magnitude of these biases are difficult to quantify. Nonetheless, our findings suggest that derived inflammatory indicators may have discriminative utility in identifying superimposed PE before delivery. While direct comparisons with established biomarkers (sFlt-1/PlGF ratio, complement levels, and uric acid) are lacking, MLR and PLR may serve as inexpensive and readily accessible candidate markers, whose clinical utility warrants further prospective evaluation.

Several limitations of this study should be acknowledged. First, all laboratory parameters were measured only within one week before delivery; therefore, our findings reflect associations with established, superimposed PE rather than supporting early prediction or screening. Nevertheless, prior studies have validated the short-term clinical utility of peripartum biomarker measurements for preeclampsia risk assessment and for distinguishing SLE flare from PE [55–57], supporting our peripartum measurement window. Second, the single-center retrospective design and relatively small sample size may limit the generalizability of our findings, and both incomplete medical records and the lack of external validation further restrict the reliability of the conclusions. Third, although most patients had planned pregnancies (suggesting relatively controlled baseline disease activity) and the two groups were comparable in SLE disease activity surrogates (anti-dsDNA positivity, complement C3/C4 levels, and clinical manifestations), we were unable to adjust for all SLE-related therapeutic confounders, including corticosteroids, hydroxychloroquine, aspirin, and other immunosuppressive therapies; residual confounding cannot be fully excluded. Fourth, due to the retrospective design, established biomarkers such as the sFlt-1/PlGF ratio, uric acid, complement levels, and anti-dsDNA antibodies were not included, limiting evaluation of our model’s incremental value. In summary, our findings should be considered exploratory and hypothesis-generating. Future prospective multicenter studies with larger cohorts and detailed longitudinal records are needed to validate our findings and establish reliable clinical thresholds.

Conclusion

This study provides preliminary evidence that the derived inflammatory indicators (MLR, PLR) may have potential discriminative capacity and may serve as cost-effective and readily available screening aids for identifying superimposed PE in pregnant patients with SLE near delivery. However, given the exploratory nature of this analysis and the limited sample size, these findings are not yet sufficient to guide clinical decision-making. External validation in independent, multi-center prospective cohorts is required before these results can be considered clinically applicable.

Authors’ contributions

J.H. and M.Y.Z. conceived and designed the study, performed the formal analysis, and wrote the original draft. H.H. contributed to the study methodology and reviewed the manuscript. Y.X. and S.L. curated the data and conducted the investigation. J.C., J.Z. and S.S. assisted with the investigation. M.Z., G.H., and M.S. supervised the research, acquired funding, and reviewed and edited the manuscript. J.H. and M.Y.Z. contributed equally to this work and are joint first authors. M.Z., G.H., and M.S. are joint corresponding authors. All authors read and approved the final manuscript.

Funding

This work was supported by the grants from National Natural Science Foundation of China (grant number: 82471716, 82504470), Guangdong Basic and Applied Basic Research Foundation (grant number: 2024A1515140102), and Guangdong Medical Science and Technology Research Foundation (grant number: A2024097).

Data availability

The datasets generated and analysed during the current study are not publicly available, but are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study involved human participants and the research protocol was reviewed and approved by the Ethics Committee of Nanfang Hospital, Southern Medical University, China (NFEC-2020–155). All participants provided written informed consent before taking part.

Consent for publication

Not applicable. (The manuscript contains no individual person’s data in any form.)

Competing interests

The authors declare no competing interests.