Development and verification of a nomogram for recurrence risk of Benign Paroxysmal Positional Vertigo in middle-aged and older populations

Abstract

Background: Benign Paroxysmal Positional Vertigo (BPPV) is the most common cause of peripheral vertigo, with frequent recurrence, particularly pronounced among middle-aged and elderly populations, significantly affecting patients' quality of life. This study aimed to identify predictive factors for recurrence in middle-aged and older patients with BPPV and to develop a nomogram prediction model based on these predictors.

Methods: This retrospective study included 582 participants aged ≥45 years who were selected from the electronic medical records system of the First Hospital of Changsha between March 2021 and March 2024. Randomly chosen participants (n = 407, 70%) constituted the training group, whereas the remaining participants (n = 175, 30%) formed the validation group. This study used LASSO binomial regression to select the most predictive variables. A predictor-based nomogram was developed to calculate the risk of BPPV recurrence. The performance of the nomogram was evaluated using the area under the receiver operating characteristic curve (AUC) and calibration curves with 1,000 bootstrap resampling validations. Decision curve analysis (DCA) was conducted to assess the clinical usefulness of the nomogram.

Results: According to findings from least absolute shrinkage and selection operator (LASSO) binomial regression and logistic regression screening, older age, higher levels of uric acid (UA) and homocysteine (HCY), diabetes, migraine, anxiety, and insomnia were identified as independent factors associated with an increased recurrence risk of BPPV. A nomogram model for predicting recurrence risk was developed based on these predictors. The nomogram achieved an AUC (C-statistic) of 0.8974 (95% CI: 0.8603-0.9345) in the training group and 0.8829 (95% CI: 0.8253-0.9406) in the validation group. Calibration curves, after 1,000 bootstrap resamples, demonstrated good agreement between the predicted and actual probabilities in the development and validation cohorts. DCA indicated that the nomogram had clinical utility.

Conclusion: The nomogram model incorporating age, UA, HCY, diabetes, migraine, anxiety status, and insomnia demonstrated a strong predictive capability for estimating the probability of BPPV recurrence in middle-aged and elderly patients. This tool is valuable for identifying individuals at high risk of BPPV recurrence and can aid physicians in making informed treatment decisions aimed at reducing recurrence rates.

Keywords: BPPV; Benign Paroxysmal Positional Vertigo; nomogram; recurrence; risk factors.

Figure 1

The flow chart of patient recruitment concisely outlines the sequential application of inclusion and exclusion criteria, ultimately defining the final study cohort.

Figure 2

(A) LASSO coefficient profiles of the 23 risk factors. (B) Risk factors were selected using LASSO regression analysis. The two dotted lines indicate the optimal scores according to the minimum criteria (including age, smoking, Alcohol drinking, diabetes, hypertension, migraine, hyperlipidemia, anxiety, insomnia, osteoporosis, UA, and HCY at minimum criteria; age, diabetes, migraine, anxiety, insomnia, UA, and HCY at 1-se criteria).

Figure 3

Nomogram for predicting the probability of BPPV recurrence after CRM. A red dot on the nomogram represents the specific characteristics of a patient.In this example, a 70-year-old individual with a history of insomnia and anxiety, but no migraines or diabetes, has a UA level of 279 μmol/L and a HCY level of 14.57 μmol/L. The calculated sum of these specific points is 267, which corresponds to a position on the total point line. From this point, a solid line is drawn vertically down to the survival axis, indicating a recurrence probability of BPPV of 9.8% for this patient.

Figure 4

ROC curves were generated to assess the predictive performance of the nomogram for the recurrence probability of BPPV following CRM. Panel A displays the ROC curve for the training set, whereas Panel B shows the ROC curve for the validation set. AUC, area under the ROC curve; ROC, receiver operating characteristic.

Figure 5

Calibration curves of the nomogram were plotted separately for the training set (A) and the validation set (B).

Figure 6

Decision curve analysis was performed for the training set (A) and the validation set (B). In these plots, a horizontal line signifies that no patients are predicted positive, resulting in a net benefit of zero. Conversely, an oblique line indicates that all patients are predicted positive. A backslash with a negative slope represents the net benefit.