The use of percent change in RR interval for data exclusion in analyzing 24-h time domain heart rate variability in rodents

Emma Karey, Shiyue Pan, Amber N. Morris, Donald A. Bruun, Pamela J Lein, Chao-Yin Chen

Research output: Contribution to journalArticle

Abstract

While epidemiological data support the link between reduced heart rate variability (HRV) and a multitude of pathologies, the mechanisms underlying changes in HRV and disease progression are poorly understood. Even though we have numerous rodent models of disease for mechanistic studies, not being able to reliably measure HRV in conscious, freely moving rodents has hindered our ability to extrapolate the role of HRV in the progression from normal physiology to pathology. The sheer number of heart beats per day (>800,000 in mice) makes data exclusion both time consuming and daunting. We sought to evaluate an RR interval exclusion method based on percent (%) change of adjacent RR intervals. Two approaches were evaluated: % change from "either" and "both" adjacent RR intervals. The data exclusion method based on standard deviation (SD) was also evaluated for comparison. Receiver operating characteristic (ROC) curves were generated to determine the performance of each method. Results showed that exclusion based on % change from "either" adjacent RR intervals was the most accurate method in identifying normal and abnormal RR intervals, with an overall accuracy of 0.92-0.99. As the exclusion value increased (% change or SD), the sensitivity (correctly including normal RR intervals) increased exponentially while the specificity (correctly rejecting abnormal RR intervals) decreased linearly. Compared to the SD method, the "either" approach had a steeper rise in sensitivity and a more gradual decrease in specificity. The intersection of sensitivity and specificity where the exclusion criterion had the same accuracy in identifying normal and abnormal RR intervals was 10-20% change for the "either" approach and ~ 1 SD for the SD-based exclusion method. Graphically (tachogram and Lorenz plot), 20% change from either adjacent RR interval resembled the data after manual exclusion. Finally, overall (SDNN) and short-term (rMSSD) indices of HRV generated using 20% change from "either" adjacent RR intervals as the exclusion criterion were closer to the manual exclusion method with lower subject-to-subject variability than those generated using the 2 SD exclusion criterion. Thus, 20% change from "either" adjacent RR intervals is a good criterion for data exclusion for reliable 24-h time domain HRV analysis in rodents.

Original languageEnglish (US)
Article number693
JournalFrontiers in Physiology
Volume10
Issue numberJUN
DOIs
StatePublished - Jan 1 2019

Fingerprint

Rodentia
Heart Rate
Rodent Diseases
Pathology
ROC Curve
Disease Progression
Heart Diseases
Sensitivity and Specificity

Keywords

  • Electrocardiogram
  • HRV
  • Lorenz plot
  • RMSSD
  • ROC
  • SDNN
  • Tachogram
  • Telemetry

ASJC Scopus subject areas

  • Physiology
  • Physiology (medical)

Cite this

The use of percent change in RR interval for data exclusion in analyzing 24-h time domain heart rate variability in rodents. / Karey, Emma; Pan, Shiyue; Morris, Amber N.; Bruun, Donald A.; Lein, Pamela J; Chen, Chao-Yin.

In: Frontiers in Physiology, Vol. 10, No. JUN, 693, 01.01.2019.

Research output: Contribution to journalArticle

@article{df9d26cf34df4f1badc83e3e0caeb4d2,
title = "The use of percent change in RR interval for data exclusion in analyzing 24-h time domain heart rate variability in rodents",
abstract = "While epidemiological data support the link between reduced heart rate variability (HRV) and a multitude of pathologies, the mechanisms underlying changes in HRV and disease progression are poorly understood. Even though we have numerous rodent models of disease for mechanistic studies, not being able to reliably measure HRV in conscious, freely moving rodents has hindered our ability to extrapolate the role of HRV in the progression from normal physiology to pathology. The sheer number of heart beats per day (>800,000 in mice) makes data exclusion both time consuming and daunting. We sought to evaluate an RR interval exclusion method based on percent ({\%}) change of adjacent RR intervals. Two approaches were evaluated: {\%} change from {"}either{"} and {"}both{"} adjacent RR intervals. The data exclusion method based on standard deviation (SD) was also evaluated for comparison. Receiver operating characteristic (ROC) curves were generated to determine the performance of each method. Results showed that exclusion based on {\%} change from {"}either{"} adjacent RR intervals was the most accurate method in identifying normal and abnormal RR intervals, with an overall accuracy of 0.92-0.99. As the exclusion value increased ({\%} change or SD), the sensitivity (correctly including normal RR intervals) increased exponentially while the specificity (correctly rejecting abnormal RR intervals) decreased linearly. Compared to the SD method, the {"}either{"} approach had a steeper rise in sensitivity and a more gradual decrease in specificity. The intersection of sensitivity and specificity where the exclusion criterion had the same accuracy in identifying normal and abnormal RR intervals was 10-20{\%} change for the {"}either{"} approach and ~ 1 SD for the SD-based exclusion method. Graphically (tachogram and Lorenz plot), 20{\%} change from either adjacent RR interval resembled the data after manual exclusion. Finally, overall (SDNN) and short-term (rMSSD) indices of HRV generated using 20{\%} change from {"}either{"} adjacent RR intervals as the exclusion criterion were closer to the manual exclusion method with lower subject-to-subject variability than those generated using the 2 SD exclusion criterion. Thus, 20{\%} change from {"}either{"} adjacent RR intervals is a good criterion for data exclusion for reliable 24-h time domain HRV analysis in rodents.",
keywords = "Electrocardiogram, HRV, Lorenz plot, RMSSD, ROC, SDNN, Tachogram, Telemetry",
author = "Emma Karey and Shiyue Pan and Morris, {Amber N.} and Bruun, {Donald A.} and Lein, {Pamela J} and Chao-Yin Chen",
year = "2019",
month = "1",
day = "1",
doi = "10.3389/fphys.2019.00693",
language = "English (US)",
volume = "10",
journal = "Frontiers in Physiology",
issn = "1664-042X",
publisher = "Frontiers Research Foundation",
number = "JUN",

}

TY - JOUR

T1 - The use of percent change in RR interval for data exclusion in analyzing 24-h time domain heart rate variability in rodents

AU - Karey, Emma

AU - Pan, Shiyue

AU - Morris, Amber N.

AU - Bruun, Donald A.

AU - Lein, Pamela J

AU - Chen, Chao-Yin

PY - 2019/1/1

Y1 - 2019/1/1

N2 - While epidemiological data support the link between reduced heart rate variability (HRV) and a multitude of pathologies, the mechanisms underlying changes in HRV and disease progression are poorly understood. Even though we have numerous rodent models of disease for mechanistic studies, not being able to reliably measure HRV in conscious, freely moving rodents has hindered our ability to extrapolate the role of HRV in the progression from normal physiology to pathology. The sheer number of heart beats per day (>800,000 in mice) makes data exclusion both time consuming and daunting. We sought to evaluate an RR interval exclusion method based on percent (%) change of adjacent RR intervals. Two approaches were evaluated: % change from "either" and "both" adjacent RR intervals. The data exclusion method based on standard deviation (SD) was also evaluated for comparison. Receiver operating characteristic (ROC) curves were generated to determine the performance of each method. Results showed that exclusion based on % change from "either" adjacent RR intervals was the most accurate method in identifying normal and abnormal RR intervals, with an overall accuracy of 0.92-0.99. As the exclusion value increased (% change or SD), the sensitivity (correctly including normal RR intervals) increased exponentially while the specificity (correctly rejecting abnormal RR intervals) decreased linearly. Compared to the SD method, the "either" approach had a steeper rise in sensitivity and a more gradual decrease in specificity. The intersection of sensitivity and specificity where the exclusion criterion had the same accuracy in identifying normal and abnormal RR intervals was 10-20% change for the "either" approach and ~ 1 SD for the SD-based exclusion method. Graphically (tachogram and Lorenz plot), 20% change from either adjacent RR interval resembled the data after manual exclusion. Finally, overall (SDNN) and short-term (rMSSD) indices of HRV generated using 20% change from "either" adjacent RR intervals as the exclusion criterion were closer to the manual exclusion method with lower subject-to-subject variability than those generated using the 2 SD exclusion criterion. Thus, 20% change from "either" adjacent RR intervals is a good criterion for data exclusion for reliable 24-h time domain HRV analysis in rodents.

AB - While epidemiological data support the link between reduced heart rate variability (HRV) and a multitude of pathologies, the mechanisms underlying changes in HRV and disease progression are poorly understood. Even though we have numerous rodent models of disease for mechanistic studies, not being able to reliably measure HRV in conscious, freely moving rodents has hindered our ability to extrapolate the role of HRV in the progression from normal physiology to pathology. The sheer number of heart beats per day (>800,000 in mice) makes data exclusion both time consuming and daunting. We sought to evaluate an RR interval exclusion method based on percent (%) change of adjacent RR intervals. Two approaches were evaluated: % change from "either" and "both" adjacent RR intervals. The data exclusion method based on standard deviation (SD) was also evaluated for comparison. Receiver operating characteristic (ROC) curves were generated to determine the performance of each method. Results showed that exclusion based on % change from "either" adjacent RR intervals was the most accurate method in identifying normal and abnormal RR intervals, with an overall accuracy of 0.92-0.99. As the exclusion value increased (% change or SD), the sensitivity (correctly including normal RR intervals) increased exponentially while the specificity (correctly rejecting abnormal RR intervals) decreased linearly. Compared to the SD method, the "either" approach had a steeper rise in sensitivity and a more gradual decrease in specificity. The intersection of sensitivity and specificity where the exclusion criterion had the same accuracy in identifying normal and abnormal RR intervals was 10-20% change for the "either" approach and ~ 1 SD for the SD-based exclusion method. Graphically (tachogram and Lorenz plot), 20% change from either adjacent RR interval resembled the data after manual exclusion. Finally, overall (SDNN) and short-term (rMSSD) indices of HRV generated using 20% change from "either" adjacent RR intervals as the exclusion criterion were closer to the manual exclusion method with lower subject-to-subject variability than those generated using the 2 SD exclusion criterion. Thus, 20% change from "either" adjacent RR intervals is a good criterion for data exclusion for reliable 24-h time domain HRV analysis in rodents.

KW - Electrocardiogram

KW - HRV

KW - Lorenz plot

KW - RMSSD

KW - ROC

KW - SDNN

KW - Tachogram

KW - Telemetry

UR - http://www.scopus.com/inward/record.url?scp=85069203600&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85069203600&partnerID=8YFLogxK

U2 - 10.3389/fphys.2019.00693

DO - 10.3389/fphys.2019.00693

M3 - Article

AN - SCOPUS:85069203600

VL - 10

JO - Frontiers in Physiology

JF - Frontiers in Physiology

SN - 1664-042X

IS - JUN

M1 - 693

ER -