Heart rate variability as a screening tool for obstructive sleep apnoea

Introduction: Obstructive sleep apnoea (OSA) has been associated with increased cardiac morbidity and mortality. Autonomic nervous system (ANS) function, characterised by heart rate variability (HRV), has been shown to be an independent predictor of cardiovascular mortality. Repetitive episodes of hypopnoeas/apnoeas trigger ANS responses affecting heart rate. We hypothesised that HRV analysis may provide a screening tool for OSA.

Methods: Overnight polysomnography was performed on 239 consecutive patients with suspected OSA. ECG was recorded at a sampling frequency of 512 Hz. Template matching was used to extract RR intervals and artefacts/ectopic beats were filtered with remaining gaps interpolated using a local variance estimation algorithm. HRV was quantified using standard time and frequency measures as well as non-linear algorithms based on entropy, fractal dimension, long-range correlations, and symbolic dynamics. We constructed two different binary logistic regression models: screening for (1) moderate-severe OSA (AHI > 30) and (2) mild OSA (AHI > 15). Both models were controlled for age, gender, BMI, smoking, CAD, DCM, HT, DM, and alcohol consumption. HRV measures were stepwise entered in the model using forward likelihood ratio procedure.

Results: In model (1) HRV measures predictive of OSA were VLFI (very low frequency power of RR increment) p < 0.001, mean NN (normal-to-normal interval) p = 0.024, LF (low frequency power) p = 0.025, and phvar (measure based on symbolic dynamics) p = 0.032. Correct overall classification was 85% (sensitivity = 58% and specificity = 94%). In model (2) predictive HRV measures were VLFI p = 0.002, αVLF (log-range correlation) p < 0.001, fdLF (fractal dimension) p = 0.004, and Renyi4 (entropy) p = 0.04). Correct overall classification was 77% (sensitivity = 72% and specificity = 81%).

Conclusion: HRV analysis of overnight holter monitoring may provide a screening tool for OSA.