May autonomic indices from cardiovascular variability help identify hypertension?

Introduction: Altered profile of RR variability and reduced baroreflex gain, as autonomic proxies, are observed in hypertensive individuals. Aim: To assess whether using logistic models and cross-validation techniques autonomic proxies can be used to identify clinical hypertensive and normotensive groups. Methods: An observational study on 405 individuals (155 mild hypertensive and 250 controls). We used four steps for statistical analysis: preliminary descriptive statistics; logistic regression modelling; detection of best parsimonious set of variables; and concordance analysis between clinical and autonomic hypertension profile. Results: Accuracy index (rate of correct identifications of normotensive and hypertensive states), computed on each of the four gradually more complex logistic models (from A to D), reached its highest value (82.7%), in the most complete model D, including autonomic nervous system indices (RR variability and baroreflex gain), age and sex. Measures of predictive performance increased from the simplest model to the most complex one [model D, positive predictive value (PPV) = 0.767, negative predictive value (NPV) = 0.866], with higher specificity than sensitivity. A parsimonious set of autonomic proxies (Mean RR, Delta_RRLFnu – i.e. change from rest to standing up – baroreflex gain combined with age and sex) led to an accuracy index of 80.5%, thus proving to have discriminant and predictive powers in detecting hypertension very similar to the whole set of the explicative variables comprised in the complete model D. Conclusion: The clinical value of the observation that the information collectively carried by a small subset of indirect autonomic proxies may identify either hypertensive or normotensive groups needs to be further investigated.