Establishment and evaluation of a risk prediction model for venlafaxine plasma concentration exceeding alert levels

To investigate the factors influencing venlafaxine blood concentration exceeding the alert threshold in patients with depression and to develop a risk prediction model for elevated venlafaxine concentrations, providing a reference for individualized VEN therapy. A retrospective analysis was conducted on 590 hospitalized patients who received venlafaxine treatment and underwent TDM at the First Hospital of Hebei Medical University between January 2021 and August 2024. Patients were categorized into a target concentration group (100–400 ng/mL) and an above-alert group (> 800 ng/mL) based on their VEN plasma concentrations. Demographic and clinical variables, including sex, age, body mass index (BMI), average daily dose, plasma albumin level, concomitant medications, liver and kidney function, were collected and compared between groups. Logistic regression analysis was performed to identify independent risk factors associated with VEN concentrations exceeding the alert threshold. A nomogram prediction model was constructed based on the identified factors and was subsequently validated. Among the 590 patients, 516 were in the target concentration group and 74 were in the above-alert group. The proportion of females, patients with BMI < 24, average daily dose ≥ 225 mg, renal impairment, and concomitant use of CYP2D6 inhibitors was significantly higher in the above-alert group than in the target group (P < 0.05). Logistic regression analysis revealed that average daily dose ≥ 225 mg (OR = 26.628, 95 % CI: 12.912–54.916), renal impairment (OR = 2.429, 95 % CI: 1.215–4.854), and concomitant use of CYP2D6 inhibitors (OR = 5.232, 95 % CI: 2.781–9.844) were independent risk factors for VEN concentrations exceeding the alert threshold (P < 0.05). The nomogram model showed an AUC of 0.899 (95 % CI: 0.864–0.935), sensitivity of 48.65 %, specificity of 95.74 %, positive predictive value of 62.07 %, and negative predictive value of 92.86 %. Bootstrap validation demonstrated good consistency (Brier score = 0.072), and the Hosmer-Lemeshow test indicated good calibration (χ² = 3.16, P = 0.531). Decision curve analysis demonstrated clinical utility for threshold probabilities of 0.05–0.80. Average daily dose ≥ 225 mg, renal impairment, and concomitant use of CYP2D6 inhibitors are independent risk factors for VEN plasma concentrations exceeding the alert threshold. The constructed nomogram model effectively predicts the risk of venlafaxine concentration exceeding the alert range and has significant clinical application value.