Sequence Method Based Baroreceptor Sensitivity: Differences Between Traditional And Multi-level Modelling Approaches

Baroreceptors are stretch-sensitive neurons with terminals located within the walls of great arteries (i.e., aortic arch and carotids). Their firing rate to the autonomic cardiac centers is modulated by mechanical stretch on the blood vessels due to changes in arterial pressure. A non-invasive method to estimate baroreceptor function involves beat-by-beat detection of sequences of consecutive changing systolic blood pressures (SBP) with corresponding changes in inter-beat time intervals (IBI). Regression of IBI on SBP for each sequence and the resulting average slope coefficient from within each subject is considered a baroreceptor sensitivity index (BRSI). This traditional (TRAD) method violates certain statistical assumptions such as independence of scores. To address this issue, use of statistical models that accommodate nested scores, such as multi-level modelling (MLM), can more appropriately analyze for BRSI sequence data. PURPOSE: To contrast BRSI estimation via the TRAD method against a MLM approach. METHODS: Resting 3-lead EKG and a finger SBP from 22 healthy subjects (30 - 18 yrs. old) were collected for 10 min each during two different visits. Signal processing and sequence detection and extraction were conducted with a custom LabVIEW program. MLM analyses were performed with the lme4 package in R. RESULTS: A total of 43 recordings accounted for 222 valid sequences with a total of 968 IBI-SBP beats across the sample. TRAD showed a mean BRSI of 15.88 ± 10.66 ms/mmHg, whereas random intercepts and slopes MLM obtained a mean BRSI score of 18.08 ± 4.01 ms/mmHg with nested factor adjustments within sequences (variation of ±4.96 ms/mmHg), and individuals (variation of ±7.85 ms/mmHg). Mixed model comparison of both techniques showed that TRAD showed significantly lower the BRSI scores vs MLM (-2.19 ms/mmHg; 95%CI: -3.52 to -0.86, p < 0.001, ES: -0.54). CONCLUSIONS: TRAD and the more appropriate MLM resulted in statistically significant different estimates of BRSI, with TRAD underestimating the overall BRSI in this sample.