Early postoperative hyponatremia after pituitary adenoma surgery: risk factors, predictive model, and clinical implications

Early postoperative hyponatremia (EPH, days 1–3) following pituitary neuroendocrine tumor (PitNET) surgery is a clinically significant complication with poorly defined risk factors. This study was designed to: (1) identify independent risk factors for EPH across all PitNET subtypes; (2) construct and validate a reliable predictive model using clinicopathological variables; and (3) develop a simplified risk scoring system to guide risk-stratified perioperative management. This retrospective cohort study analyzed data from 300 adult patients who underwent PitNET surgery (January 2021 – December 2024) at Nanfang Hospital. Demographics, tumor characteristics (size, Knosp grade, functional type), preoperative conditions, and surgical details were collected. EPH was defined as serum sodium < 135 mmol/L within 1–3 days postoperatively. Multivariate binary logistic regression with stepwise selection (entry criterion P < 0.10, removal criterion P > 0.10) was used to identify independent risk factors and construct the predictive model. Model performance was evaluated by the area under the receiver operating characteristic curve (AUC), calibration plot, Hosmer-Lemeshow test, and bootstrap resampling (1000 iterations) for internal validation. The optimal risk score cutoff was determined by the Youden index. This study followed the STROBE guidelines. EPH occurred in 70 patients (23.3%). Five independent risk factors were identified: tumor diameter > 25 mm (OR = 2.85, 95% CI: 1.52–5.34, P = 0.001), Knosp grade ≥ 3 (OR = 3.10, 95% CI: 1.65–5.83, P < 0.001), adrenocorticotropic hormone (ACTH)-secreting tumor (OR = 3.55, 95% CI: 1.41–8.94, P = 0.007), preoperative hyponatremia (OR = 4.20, 95% CI: 1.81–9.74, P < 0.001), and operative duration > 180 min (OR = 2.15, 95% CI: 1.13–4.08, P = 0.019). The predictive model exhibited excellent discriminative ability (AUC = 0.85, 95% CI: 0.75–0.87) and adequate calibration (Hosmer-Lemeshow test: χ²=8.56, df = 8, P = 0.38), with robust performance confirmed by bootstrap validation (AUC = 0.83). A simplified risk scoring system (0–6 points) stratified patients into low- (0–1 points, EPH incidence 5.0%), medium- (2–3 points, 35.0%), and high-risk (≥ 4 points, 82.5%) groups. The optimal cutoff value (3 points) yielded a sensitivity of 78.6% and specificity of 83.9%. Tumor size (> 25 mm), invasiveness (Knosp grade ≥ 3), ACTH-secreting subtype, preoperative hyponatremia, and prolonged operative duration (> 180 min) are independent predictors of EPH. To our knowledge, this is the first comprehensive predictive model and simplified scoring system for EPH across all PitNET subtypes. The validated tool enables accurate preoperative risk stratification, facilitating personalized perioperative monitoring and targeted interventions to reduce EPH-related complications and improve patient outcomes.