# A clinical prediction model for blood pressure changes after renal denervation in patients with resistant hypertension

## TL;DR

This study created and compared models to predict blood pressure changes after a procedure for high blood pressure that doesn't respond to medication.

## Contribution

The study introduces a Ridge regression model that outperforms traditional models in predicting blood pressure changes after renal denervation.

## Key findings

- Ridge regression had lower prediction errors and higher accuracy than OLS in predicting systolic blood pressure changes.
- IMR, preoperative SBP, age, and creatinine were significant predictors of blood pressure changes.
- The Ridge model also showed better performance in predicting diastolic blood pressure changes.

## Abstract

To develop clinical prediction models to estimate blood pressure changes in hypertensive patients undergoing renal denervation (RDN).

This single-center, prospective interventional study enrolled 70 hypertensive patients undergoing RDN between July 2022 and December 2023, with clinical data collected systematically before and after the procedure. Variable selection for modeling was performed through a rigorous process incorporating univariate analysis and clinical relevance assessment. Subsequently, two distinct clinical prediction models were developed and subjected to comparative evaluation. The primary outcomes were the absolute changes in systolic blood pressure (SBP) and diastolic blood pressure (DBP) at 6 months after RDN.

In both Ordinary Least Squares (OLS) and Ridge regression models, seven variables [including index of microvascular resistance (IMR), preoperative SBP, age and creatinine] were significantly associated with SBP change, while four variables were significantly associated with DBP change. In the prediction model on SBP change, compared to the OLS model, the Ridge regression exhibited lower prediction errors [mean absolute error [MAE]: 6.40 vs. 6.95; mean squared error [MSE]: 65.58 vs. 76.15] and a higher R² (0.79 vs. 0.72). In the DBP model, the Ridge regression also achieved a lower MAE (3.62 vs. 3.73) and a higher R² (0.77 vs. 0.71).

This study developed and compared predictive models for estimating blood pressure response at 6-month follow-up after RDN in patients with resistant hypertension. The Ridge regression model exhibited superior predictive accuracy and model stability. The model indicated that IMR was a factor associated with postoperative blood pressure reduction.

ClinicalTrials.gov, identifier, ChiCTR2200058696.