This was a single-center, retrospective, cross-sectional study. This study was approved by the Ethics Committee of Shanghai Tenth People’s Hospital, Tongji University School of Medicine (Approval No. SHSY-IEC-5.0/23K42/P01) and conducted in accordance with the principles of the Declaration of Helsinki. The ethics committee waived the requirement for individual informed consent on the grounds that the study was retrospective and non-interventional, posed no unnecessary risk, and utilized fully anonymized patient data.

Adult patients hospitalized in the Division of Nephrology at Shanghai Tenth People’s Hospital between January 2015 and December 2023 who met the diagnostic criteria for CKD and T2DM were eligible for inclusion. CKD was defined according to the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines (28), characterized by the presence of kidney structural or functional abnormalities for over 3 months and meeting at least one of the following criteria (28) (1): urinary albumin-to-creatinine ratio (UACR) ≥ 30 mg/g; (2) urinary sediment abnormalities; (3) histological or imaging evidence of structural kidney abnormalities; (4) history of kidney transplantation; or (5) eGFR < 60 mL/min/1.73 m². T2DM was defined by meeting any of the following criteria (29): (1) fasting blood glucose (FBG) ≥7.0 mmol/L; (2) 2-hour plasma glucose during an oral glucose tolerance test (OGTT) ≥ 11.1 mmol/L; (3) glycated hemoglobin (HbA1c) ≥ 6.5%; or (4) documented history of T2DM or current use of glucose-lowering medication. The exclusion criteria were as follows: (1) type 1 diabetes or other specific types of diabetes (e.g., steroid-induced diabetes); (2) primary glomerular diseases (e.g., IgA nephropathy) or other non-diabetes-related secondary glomerular diseases (e.g., lupus nephritis); (3) current renal replacement therapy (hemodialysis, peritoneal dialysis, or kidney transplantation); (4) comorbid severe cardiopulmonary dysfunction or history of malignancy; (5) presence of acute infection, pregnancy, or lactation; or (6) missing renal function data. A total of 994 patients were included in the final analysis. The study flowchart is presented in Figure 1.

Data were extracted from electronic medical records by uniformly trained researchers. Demographic and Clinical Characteristics: Age, sex, height, weight, blood pressure, duration of CKD, coronary heart disease (CHD), stroke, and hypertension (HBP). Laboratory Parameters: Fasting blood samples were collected for complete blood count, renal function, electrolytes, liver function, lipid profile, HbA1c, serum ferritin, etc. eGFR was calculated using the CKD-EPI 2021 equation (30). SII was calculated as platelet count × neutrophil count/lymphocyte count (12). Body Mass Index (BMI) was calculated as weight (kg)/height (m)². Hyperlipidemia was defined as total cholesterol ≥6.2 mmol/L, or LDL-C ≥4.1 mmol/L, or triglycerides ≥2.3 mmol/L (31). Anemia was defined as hemoglobin (HGB) <130 g/L for males and <120 g/L for females (32).

Serum PRL levels were measured as part of routine clinical laboratory testing in our hospital using an electrochemiluminescence immunoassay (ECLIA) on a Roche Cobas e 601 analyzer (Roche Diagnostics, Basel, Switzerland). Blood samples were collected under fasting conditions and analyzed within 24 hours of collection. PRL measurement was conducted in accordance with standardized operational procedures and internal quality control protocols. All reagent batches were validated for inter-batch consistency according to the manufacturer’s specifications and calibrated against the WHO International Standard 84/500. The inter-assay coefficient of variation (CV) remained <5% throughout the study period. The analytical measuring range was 1.00–10000 mIU/L. Reference ranges were 102–496 mIU/L for adult females and 66–324 mIU/L for adult males. Hyperprolactinemia was defined as PRL levels exceeding these upper limits. Throughout the study period, the same analytical platform and assay methodology were consistently used without changes in instrumentation or calibration standards.

All analyses were performed using R software (version 4.3.2). Continuous variables are presented as mean ± standard deviation (SD) or median (interquartile range, IQR), and categorical variables as frequencies (percentages). Comparisons across PRL quartiles were conducted using one-way ANOVA or the Kruskal–Wallis test for continuous variables, and the chi-square test for categorical variables, as appropriate.

To evaluate the associations between serum PRL and clinical parameters, multivariable linear mixed-effects models were constructed and adjusted for age, sex, BMI, eGFR, HBP, CHD, and stroke. PRL was modeled as a continuous variable per 100 mIU/L increment. To investigate the relationship between PRL and the systemic SII, four models were constructed. The crude model included PRL only. Model 1 was adjusted for age, sex, eGFR, HBP, CHD, and stroke. Model 2 further included HbA1c and hyperlipidemia. Model 3 additionally adjusted for HGB and serum albumin. In these analyses, PRL was examined both as quartiles and as a continuous variable (per 100 mIU/L increment). All regression models were implemented as linear mixed-effects models with random intercepts for patient ID, thereby accounting for within-subject correlation arising from repeated measurements.

Restricted cubic spline (RCS) models were applied to explore potential nonlinear associations between serum PRL and the SII. PRL was modeled as a continuous variable using RCS with four knots placed at predefined percentiles. Models were adjusted for age, sex, eGFR, HBP, CHD, stroke, hyperlipidemia, HbA1c, HGB, and serum albumin. The predicted SII values displayed in the spline plots represent covariate-adjusted estimates derived from the fitted multivariable models. Nonlinearity was evaluated using likelihood ratio tests comparing models with and without spline terms. When a significant nonlinear association was identified, segmented regression analyses were performed to estimate slopes on either side of the inflection point.

Trend and nonlinearity tests were derived from regression models treating PRL as a continuous variable. Subgroup and sensitivity analyses were conducted to assess the robustness of the findings. Statistical significance was defined as a two-sided p < 0.05. Exact p values were reported wherever possible. Values less than 0.001 were presented in scientific notation. For results falling below the machine precision of double-precision floating-point arithmetic in R (.Machine$double.eps ≈ 2.22 × 10^-16), p values were reported as p < 2.22 × 10^-16.

This cross-sectional analysis included 994 patients with T2DM-related DKD. As detailed in Table 1, the mean serum PRL concentration was 459.85 ± 692.70 mIU/L in the overall cohort, with a median concentration of 344.45 mIU/L (IQR: 258.80-463.15 mIU/L), suggesting a right-skewed distribution. When stratified by sex, the median PRL concentration was 363.00 mIU/L (IQR: 273.00-484.00) in females (n = 400) and 333.00 mIU/L (IQR: 249.00-454.00) in males (n = 594).

Participants were stratified into quartiles (Quartile1-Quartile4) based on PRL levels. Significant differences in renal function were observed across quartiles, including a decline in eGFR (p = 1.5 × 10^-4) and an increase in the proportion of advanced CKD (G4-G5), rising from 29% in Quartile1 to 54% in Quartile4 (p = 2.5 × 10^-6). Markers related to CKD complications differed significantly across PRL quartiles. HGB and red blood cell (RBC) decreased across higher PRL quartiles (p = 7.8 × 10–⁸ and p = 4.7 × 10^-7, respectively), while serum phosphorus and intact parathyroid hormone (iPTH) increased (p = 1.4 × 10–¹⁰ and p = 4.3×10^-8, respectively). Serum albumin was lower in the highest PRL quartile (p = 0.025). Glucose and HbA1c levels were lower across increasing PRL quartiles (p = 0.003 and p = 0.015, respectively). Among inflammation markers, neutrophil count increased and lymphocyte count decreased across PRL quartiles (p = 5 × 10–⁴ and p= 0.01, respectively), whereas no significant differences were found in CRP or platelet levels. PRL levels were associated with sex distribution (p = 0.021) but not with age, BMI, or history of stroke or hyperlipidemia.

Of the 994 observations, 684 individuals contributed repeated measurements (mean 1.43 visits per subject; 137 with ≥2 visits). To account for within-subject clustering, associations between serum PRL and clinical parameters were evaluated using linear mixed-effects models with random intercepts for patient ID. PRL was modeled as a continuous variable per 100 mIU/L increment.

As shown in Table 2, in the multivariable mixed-effects model adjusted for sex, age, BMI, eGFR, hypertension, stroke, and CHD, higher PRL levels were associated with several inflammation-related markers of inflammation and immune activation, including increased white blood cell count (β = 0.035, 95% CI: 0.019 to 0.050; p = 1.29 × 10^-5), neutrophil count (β = 0.033, 95% CI: 0.019 to 0.047; p = 5.18 × 10^-6), serum ferritin (β = 3.728, 95% CI: 2.011 to 5.444; p = 3.32 × 10^-5), as well as decreased serum albumin levels (β = -0.043, 95% CI: -0.083 to -0.004; p = 0.032). Positive associations were also observed for CKD-mineral and bone disorder markers, including serum phosphorus (β = 0.003, 95% CI: 0.001 to 0.006; p = 0.015) and iPTH (β = 2.295, 95% CI: 0.586 to 4.003; p = 0.009). In contrast, the previously observed univariate associations with HGB and red blood cell count were no longer statistically significant after multivariable adjustment.

As shown in Table 3, both quartile-based and continuous analyses using mixed-effects models with random intercepts for patient ID revealed a significant and robust positive association between PRL and SII. In the fully adjusted model (Model 3), compared with the lowest Quartile 1, the β coefficients (95% CI) for SII were 87.00 (2.84 to 171.16) in Quartile 3 and 197.85 (105.52 to 290.18) in Quartile 4, with a significant trend across ascending PRL quartiles (p for trend = 1.55 × 10^-5). Consistent results were observed in the continuous analysis. In Model 3, each 100 mIU/L increase in PRL was associated with a 7.10-unit increase in SII (95% CI: 3.13 to 11.07; p = 5.12 × 10^-4).

Main analysis: Association between prolactin and SII in the complete study population. A total of 994 measurements from 684 individuals were included. Data are presented as β coefficients (95% confidence interval) for SII. PRL quartiles analysis: Values are median (interquartile range) for prolactin levels in mIU/L for each quartile (Quartile 1–4). The overall median (IQR) for prolactin levels is 344.45 (258.80, 463.15) mIU/L. SII: median (IQR) = 455.51 (325.76–671.19); mean ± SD = 575.41 ± 502.04. PRL continuous analysis: β coefficients represent the change in SII per 100 mIU/L increase in prolactin. Crude model: unadjusted linear mixed-effects model with random intercept for individual ID. Model 1: adjusted for age, sex, eGFR, and history of hypertension, coronary heart disease, and stroke, with random intercept for individual ID. Model 2: Model 1 + hyperlipidemia and HbA1c. Model 3: Model 2 + albumin and hemoglobin. All models are linear mixed-effects models with random intercepts for individual ID to account for repeated measures. P values correspond to two-sided tests for the PRL term in each model and were reported in scientific notation when < 0.001. Abbreviations: ref, reference; SII, systemic immune-inflammation index; PRL, prolactin; ID, patient identification.

To further explore the nonlinear relationship between serum PRL and the SII, we performed RCS analyses. The spline model revealed a statistically significant nonlinear association (p for nonlinearity = 0.002), with an inflection point identified at 282.85 mIU/L (Figure 2A). Consistent with the continuous spline analysis, categorical analyses based on PRL quartiles showed a progressive increase in SII across quartiles (Figure 2B). Median SII values increased stepwise from Quartile 1 to Quartile 4, and the overall trend across quartiles was statistically significant (p for trend = 3.79 × 10^-7). Segmented regression analysis based on the spline curve indicated that the association between PRL and SII differed markedly across PRL concentration ranges (Supplementary Table S1). Specifically, in the segments below the inflection point, the slopes were negative (β = -0.65, p < 2.22 × 10^-16). However, when PRL concentrations exceeded 282.85 mIU/L, the association reversed to a positive direction (β = 0.48, p < 2.22 × 10^-16) and remained positive in the highest concentration segment (≥ 769.70 mIU/L, β = 0.05, p < 2.22 × 10^-16).

We performed systematic subgroup and sensitivity analyses to validate the robustness of our findings. Stratified analysis (Supplementary Table S2) showed that the positive association between PRL and SII was significantly stronger in subgroups of patients aged ≥ 65 years (p for interaction = 7.02 × 10^-4), those with advanced CKD (G4–G5; p for interaction = 0.003), and those with HBP (p for interaction = 0.007). In contrast, no significant effect modification was observed in subgroups stratified by sex, CHD, Stroke, BMI, hyperlipidemia, or anemia status (all p for interaction > 0.05).

To further assess the robustness of the primary findings, a series of sensitivity analyses were conducted. After excluding patients with conditions directly affecting PRL (pituitary adenoma or dopaminergic drug use; n = 965) or those with potential acute inflammation (CRP > 50 mg/L; n = 970), the positive association remained significant, with effect estimates highly consistent with the main analysis (β = 6.930 and β = 6.855, respectively; Supplementary Tables S3, S4). Notably, the magnitude of the continuous association increased markedly after excluding patients with extreme PRL levels (4240 mIU/L) and without confirmatory MRI (β = 27.230, 95% CI: 14.394 to 40.066; p = 3.44 × 10^-5; Supplementary Table S5).