Association between Objective Activity Intensity and Heart Rate Variability: Cardiovascular Disease Risk Factor Mediation (CARDIA) : Medicine & Science in Sports & Exercise

Journal Logo


Association between Objective Activity Intensity and Heart Rate Variability: Cardiovascular Disease Risk Factor Mediation (CARDIA)


Author Information
Medicine & Science in Sports & Exercise 52(6):p 1314-1321, June 2020. | DOI: 10.1249/MSS.0000000000002259



We evaluated the associations between accelerometer-estimated physical activity (PA) intensity and heart rate variability (HRV) and examined mediation of these associations by glycemic control indices and other cardiovascular disease risk factors.


Data were from 1668 participants (age = 45.9 ± 3.5 yr, 58.0% female, 39.9% black) who participated in year 20 (2005–2006) of the Coronary Artery Risk Development in Young Adults Fitness Study. The ActiGraph 7164 estimated participants’ mean minutes per day of vigorous-intensity PA (VPA), moderate-intensity PA (MPA), and light-intensity PA (LPA) over 7 d. Three sequential 10-s 12-lead ECG strips were used to derive standard deviation of all normal RR intervals (SDNN) and root mean square of all successive RR intervals (rMSSD) HRV. Mediators representing glycemic control indices included fasting glucose, fasting insulin, and 2-h oral glucose tolerance, with other mediators being traditional cardiovascular disease risk factors. Multiple linear regression assessed independent associations of PA intensity with HRV per 1-SD. Mediation analyses computed the proportion of the PA–HRV association attributable to physiological mediators.


Participants averaged 2.7 ± 6.2 min·d−1, 33.0 ± 22.0 min·d−1, and 360.2 ± 83.8 min·d−1 of VPA, MPA, and LPA, respectively, with mean values for SDNN (32.6 ± 22.4 ms) and rMSSD (34.0 ± 24.8 ms) similar. After adjustment for demographic and lifestyle behaviors, VPA was associated with both HRV metrics (SDNN: std beta = 0.06 [0.03, 0.10]; rMSSD: std beta = 0.08 [0.05, 0.12]) and LPA with rMSSD only (std beta = 0.05 [0.01, 0.08]). Fasting insulin and glucose mediated 11.6% to 20.7% of the association of VPA and LPA with HRV, with triglycerides also potentially mediating these associations (range, 9.6%–13.4%).


Accelerometer-estimated VPA was associated with higher (i.e., improved) HRV. Light-intensity PA also demonstrated a positive association. Mediation analyses suggested these associations may be most attributable to glucose-insulin dynamics.


A publication error is reported in Table 1 of “Association between Objective Activity Intensity and Heart Rate Variability: Cardiovascular Disease Risk Factor Mediation (CARDIA)” (). The correct units of measure for “Average accelerometer counts” are (counts·min·d−1)−1.

Medicine & Science in Sports & Exercise. 52(9):2060, September 2020.

Heart rate variability (HRV) measurements represent a “gold standard” assessment of cardiac autonomic functioning (CAF) (1). Lower time-domain and/or impaired frequency-domain HRV values generally indicate increased influence of the sympathetic nervous system (SNS) on the heart, whereas higher/improved HRV values are indicative of greater stimulus by the parasympathetic nervous system (PNS) (2). Higher/improved HRV is generally regarded as an indicator of improved CAF. Indeed, prospective cohort studies and meta-analyses have noted that reduced CAF (i.e., lower time-domain and/or impaired frequency-domain HRV) is associated with higher risks of adverse cardiovascular outcomes and all-cause mortality (3–5).

Improved CAF can be achieved through healthy lifestyle behaviors. Physical activity (PA) is well-established as a key modifiable risk factor for improved cardiovascular health (6). Previous investigations have shown that individuals who report participating in higher amounts of moderate- and/or vigorous-intensity PA have higher time-domain and/or improved frequency-domain HRV values relative to their less active counterparts (7–10). Improvements in glycemic control have been hypothesized to explain how higher levels of PA and cardiorespiratory fitness contribute to more favorable HRV indices. Insulin resistance and associated hyperglycemic episodes have been documented as causing autonomic nervous system damage; thus disturbing the heart’s sympathetic–parasympathetic balance (11–15). Briefly, repeated hyperglycemic episodes can heighten the accumulation of advanced glycated end products (AGE). The receptor for AGE (RAGE) controls AGE accumulation, with RAGE signaling not only upregulated by continued circulation of AGE but also higher levels of proinflammatory cytokines (e.g., tumor necrosis factor-alpha, interleukin [IL]-6, c-reactive protein) (16–18). As proinflammatory cytokines continue to circulate at higher-than-normal levels, the PNS-controlled cholinergic anti-inflammatory pathway is further damaged and becomes less able to control proinflammatory cytokine levels within the body (19), with higher insulin resistance exacerbating this dysfunction (20,21). Importantly, in individuals with (22–24) and without (25) diabetes, research has noted lower, or impaired, HRV to be associated with higher levels of tumor necrosis factor-alpha, IL-6, c-reactive protein, and free radicals.

Although literature has suggested the importance of glycemic control to more favorable CAF (as assessed via HRV), notable literature gaps remain. First, few, if any, studies have assessed the association of accelerometer-estimated PA with HRV values (26)—noteworthy considering the low correlation between individuals’ self-reported PA and their accelerometer-estimated PA (27). Accelerometers also have the advantage of being able to more validly characterize PA across a range of intensity levels. Second, given the plausible physiological mechanism by which PA may impact CAF through improved glycemic control, there is a need to investigate mediation by glycemic control indices on the association between PA and HRV. It is also important to conduct these mediation analyses with cardiovascular disease (CVD) risk factors situated as mediators given literature suggesting how adiposity and blood lipids may contribute to reduced CAF via increased systemic inflammation (28–30). Such research would yield further insight into how PA influences not only CAF, but also the systemic inflammation which occurs partially as a result of imbalance between the SNS and PNS (19,31).

Using cross-sectional data from the Coronary Artery Risk Development in Young Adults (CARDIA) Study, we examined the independent associations of accelerometer-estimated vigorous-intensity PA (VPA), moderate-intensity PA (MPA), and light-intensity PA (LPA) with HRV, and investigated how measures of glycemic control and other CVD risk factors mediated associations of VPA, MPA, and LPA with HRV. We hypothesized that VPA and MPA would demonstrate significant independent associations with HRV and that fasting insulin and glucose would partially mediate these associations.


In 1985 to 1986, CARDIA investigators recruited 5115 black and white men and women, age 18–30 yr from four U.S. locations (Birmingham, AL; Chicago, IL; Minneapolis, MN; and Oakland, CA). The CARDIA Study has published detailed explanations of the study design, rationale, and protocol (32). The data for these analyses were collected during CARDIA year 20 (2005–2006), with 72% participant retention of the surviving cohort. All participants provided informed consent, and the institutional review boards of each participating institution approved the study.

During year 20 assessments, CARDIA investigators performed an ancillary study, the CARDIA Fitness Study (33). This study was part of the core clinic examination during which PA was assessed via accelerometry—one of only two examination years to have accelerometry data. A total of 2187 CARDIA participants enrolled in the CARDIA Fitness Study and wore accelerometers. From this sample, we excluded those without complete HRV data (n = 519). Finally, we verified that the remaining sample had validated wear time of ≥10 h·d−1 on at least 4 of 7 d (34). These exclusions resulted in a final analytic sample of 1668 participants.

PA assessment

A waist-worn ActiGraph 7164 (ActiGraph Corp.; Pensacola, FL) was used to assess participants’ PA for seven consecutive days during all waking hours except during participation in any water-based activities. Following participants’ return of the ActiGraph, CARDIA investigators downloaded these PA data, with wear time screening performed using the Troiano algorithm (35). For the current analyses, Freedson cut points (36) were used to categorize daily minutes of PA at the following intensities: LPA, 100 to 1951 counts per minute; MPA, 1952 to 5724 counts per minute and VPA, ≥5725 counts per minute. Despite physiological relevance, we did not present models adding sedentary behavior (SB) along with the preceding three PA intensities due to concerns regarding multicollinearity as well as biased or uninterpretable betas given that the four intensity categories comprise the entire accelerometry record (37,38). Indeed, with all variables in the model, principal components analyses indicated SB to have an eigenvalue of 0.39—falling below 1.0; thus, we excluded SB from the model as suggested (38). Further, literature has noted the ActiGraph to provide less valid SB measurements than other research-grade accelerometers such as the activPAL (39).

Heart rate variability

The GE MC1200 (General Electric Inc., Boston, MA) recorded resting ECG data at 500 Hz, with three 10-s ECG strips used to determine HRV. Research has reported on the validity and reliability of these short-term HRV measurements (2,40,41). HRV data were reported as the standard deviation of all normal RR intervals (SDNN; a measure of overall HRV) and the root mean square of all successive RR intervals (rMSSD; most representative of PNS activity) (2). The HRV measurements were performed in the supine position after quiet rest.


CARDIA personnel administered standardized questionnaires to obtain demographic information on sex, age, race (black and white), educational attainment, and CARDIA Field Center. These questionnaires also assessed smoking status, alcohol consumption (g·d−1) and use of beta-blockers and angiotensin-converting enzyme (ACE) inhibitors. For the current analyses, we used the a priori diet quality score (42) to assess dietary patterns. We used fasting glucose measurements and the American Diabetes Association guidelines to discern diabetes status (≥126 mg·dL−1, yes; <126 mg·dL−1, no), with the most recent American Heart Association blood pressure guidelines used to ascertain hypertension status (systolic ≥130 mm Hg or diastolic ≥80 mm Hg: yes; systolic <130 and diastolic <80: no). All other covariates were from year 20 (2005–2006); sex, race, and CARDIA Field Center were from year 0 (1985–1986). ActiGraph wear time (min·d−1) was also included as a covariate in all analyses.


We included as a priori hypothesized primary mediators the following measures of glycemic control: fasting glucose and fasting insulin. We also included values for 2-h oral glucose tolerance (2 h-OGT) as an a priori hypothesized primary mediator, assessed via measurement of serum glucose 2 h after oral ingestion of a 75-g glucose solution. We chose these CVD risk factors as primary mediators given the previously outlined physiological mechanisms by which PA may improve glycemic control and is subsequently hypothesized to improve CAF. Further, given this literature base, we did not include HOMA-IR given our desire to evaluate the independent mediation of glucose and insulin separately on the relationship between PA intensity and HRV. The following CVD risk factors were included as a priori hypothesized secondary mediators given PA’s established effects on these outcomes and emerging literature (28–30) on how these factors contribute to reduced CAF: total triglycerides, total cholesterol, low-density lipoprotein (LDL-C) and high-density lipoprotein (HDL-C) cholesterol, systolic blood pressure (BP), waist circumference, and body mass index (BMI). Fasting glucose and 2 h-OGT were measured via the hexokinase method while fasting insulin was evaluated via radioimmunoassay and lipids analyzed enzymatically using a trinder-type method on the Abbot Spectrum. Systolic BP was measured using the Omron HEM907XL (Omron Healthcare Inc.; Kyoto, Japan) calibrated to a random zero sphygmomanometer. The BP cuff was situated on the left arm and BP was measured three times in the seated position and the average of the last two BP measurements reported. Waist circumference was measured using a tape measure per established procedures. BMI was calculated as weight in kilograms divided by height in meters squared (kg·m−2), with participants classified per standard BMI categories (underweight: <18.5 kg·m−2; normal weight: 18.5–24.9 kg·m−2; overweight: 25.0–29.9 kg·m−2; Obese: ≥30.0 kg·m−2).

Statistical analyses

We used SAS 9.4 for all analyses. Data were first examined for normality through visual inspection of normality curves and associated Kolmogorov-Smirnov statistics. As a result of these analyses, we took the natural log of fasting glucose, fasting insulin, 2 h-OGT, triglycerides, HDL-C, SDNN, and rMSSD. Although nonnormality was observed for VPA and MPA, we did not log-transform these variables given that they represent the exposure variables. We examined covariates across categories of HRV, with continuous variables reported as means and SD and categorical variables as frequencies and percentages unless otherwise noted.

We then modeled our regression model covariate structure in a manner similar to an investigation in the Cardiovascular Health Study on the association between PA and HRV (43). Using multivariable-adjusted general linear regression (SAS PROC GLM), we determined the independent associations of VPA, MPA, and LPA on SDNN HRV and, in a separate model, rMSSD HRV. All models included VPA, MPA, and LPA simultaneously as predictors and SDNN or rMSSD HRV as the outcome. Our main regression model included demographic (age, sex, race, CARDIA Field Center, education), lifestyle behaviors (smoking, alcohol consumption, and the a priori diet quality score), and ActiGraph wear time as covariates. Within these models, VPA, MPA, and LPA were run as standardized values (e.g., [[each participant’s mean minutes of LPA − the sample’s mean minutes of LPA] / the sample’s standard deviation for mean minutes of LPA]) and the standardized beta coefficients then reported to represent per 1-SD increments for changes in the duration of time spent at a given PA intensity. Additive interactions of VPA, MPA, and LPA by race or sex suggested no significant interactions were present; thus we did not stratify the results by sex or race.

We then conducted separate mediation models for PA predictor variables that demonstrated statistically significant independent associations with HRV measures. These analyses were performed using PROC CAUSALMED—a recently introduced regression-based SAS procedure developed by Valerie and VanderWeele (44). Pursuant with our study aims, we first included, in separate mediation models, the a priori hypothesized primary mediators of fasting glucose, fasting insulin, and 2 h-OGT, with age, sex, race, education, study center, smoking status, alcohol consumption, the a priori diet quality score, and ActiGraph wear time included as covariates. The same approach was taken when modeling our a priori hypothesized secondary mediators: triglycerides, total cholesterol, LDL-C, HDL-C, systolic BP, waist circumference, and BMI. We only reported mediation percentages and associated 95% confidence interval (CI) if the total effect of the mediation model was statistically significant (P < 0.05).

Congruent with the aforementioned study by Soares-Miranda et al. (43), we then conducted sensitivity analyses for our regression and mediation models. Sensitivity analysis 1 included all covariates from our main regression and mediation models but added the following variables as covariates given their plausibility as biological intermediates that may have contributed to residual confounding in our main models: BMI status (underweight, normal weight, overweight, obese), diabetes status (yes/no), hypertension status (yes/no), and use of beta-blockers and ACE inhibitors (yes/no). We removed BMI status, diabetes status, and hypertension status from mediation models assessing mediation by BMI, fasting glucose/insulin or 2 h-OGT, and systolic BP, respectively. Finally, given the ambiguity in the literature regarding the adjustment for HR in analyses of HRV (45), we added HR as a covariate to all regression and mediation models as a covariate in a separate and final sensitivity analysis (i.e., sensitivity analysis 2), keeping in mind that resting HR is heavily influence by PA participation.


As seen in Table 1, the majority of the sample was white (60.3%) and female (58.2%), with much of the sample normotensive (73.9%) and without diabetes (95.3%) at year 20. Participants averaged 2.7 ± 6.2 min·d−1, 33.0 ± 22.0 min·d−1, and 360.2 ± 83.8 min·d−1 of VPA, MPA, and LPA, respectively, with similar average SDNN and rMSSD HRV values (32.6 ms and 34.0 ms, respectively; correlation between HRV metrics: r = 0.85, P < 0.01). Table 2 presents mean SDNN and rMSSD HRV by each tertile of VPA, MPA, and LPA duration. Despite exclusion of SB in the regression analyses for reasons listed previously, Table 2 also presents HRV by tertile of SB.

Descriptive statistics for study sample (n = 1668): CARDIA, 2005–2006.
HRV by PA duration tertilea (N = 1668).

Main linear regression and mediation models

Table 3 includes the results of the main multiple linear regression models. Results from the fully adjusted models indicated that, per 1-SD, VPA was independently associated with both HRV metrics (SDNN: std beta, 0.06; 95% CI 0.03–0.10; rMSSD: std beta, 0.08; 95% CI, 0.05–0.12), whereas LPA was independently associated with rMSSD HRV only (std beta, 0.05; 95% CI, 0.01–0.08). MPA had positive, but nonsignificant, associations with both HRV metrics. We then included VPA and LPA as separate predictors in fully adjusted mediation models (Tables 4, 5, and 6). For our hypothesized primary mediators, fasting glucose and insulin were observed as mediators of the association between VPA and HRV. Specifically, fasting glucose mediated 16.6% (95% CI, 3.0%–30.3%) and 14.7% (95% CI, 4.2%–25.1%) of the associations between VPA and SDNN and rMSSD HRV, respectively, and fasting insulin mediated a slightly higher percentage of the association of VPA with SDNN (18.4%; 95% CI, 3.6%–33.1%) and rMSSD HRV (17.1%; 95% CI, 5.9%–28.4%). Two-hour OGT appeared as a weaker mediator and only of the association between VPA and rMSSD (6.9%; 95% CI, 0.8%–13.0%). Fasting insulin was the only observed mediator between LPA and rMSSD HRV (20.7%; 95% CI, 2.4–38.9). Of our hypothesized secondary mediators, we observed waist circumference as a mediator of the association between VPA and rMSSD (12.3%; 95% CI, 1.3%–23.2%), with total triglycerides also observed as a mediator of the association between VPA and both HRV metrics (SDNN: 9.6%; 95% CI, 1.6%–17.6%); rMSSD: 10.1%; 95% CI, 3.0%–17.3%). For LPA, we only observed a trend toward total triglycerides being a mediator for the association between LPA and rMSSD HRV (13.4%; 95% CI, −0.4% to 27.3%).

Multivariable linear regression analysis results for physical activity intensity as independent predictors of HRV.
Percent of total effect mediated by glycemic control indices and CVD risk factors between VPAa and SDNN HRV: Main model. b
Percent of total effect mediated by glycemic control indices and CVD risk factors between VPAa and rMSSD HRV: Main model. b
Percent of total effect mediated by glycemic control indices and CVD risk factors between LPAa and rMSSD HRV: Main model. b

Sensitivity analyses

Results from models including biologically plausible intermediates for the association between PA and HRV are shown in Supplemental Digital Content 1 (see Table, Supplemental Digital Content 1, Multivariable Linear Regression Analysis Results for Physical Activity Intensity as Independent Predictors of HRV, As expected, parameter estimates were slightly attenuated. Yet, the significant associations remained between VPA and both HRV metrics, as well as LPA and rMSSD HRV. Mediation results for these analyses are included in Supplemental Digital Content 2 and 3 (VPA) (see Tables, Supplemental Digital Content 2, Percent of Total Effect Mediated by Glycemic Control Indices and CVD Risk Factors between VPA and SDNN HRV, http://sv;Supplemental Digital Content 3, Percent of Total Effect Mediated by Glycemic Control Indices and CVD Risk Factors between VPA and rMSSD HRV, and Supplemental Digital Content 4 (LPA, see Table, Supplemental Digital Content 4, Percent of Total Effect Mediated by Glycemic Control Indices and CVD Risk Factors between LPA and rMSSD HRV, Fasting insulin remained a mediator of the association of VPA with SDNN and rMSSD HRV and, for LPA, rMSSD HRV, whereas fasting glucose remained a mediator of the association between VPA and both HRV metrics. Waist circumference remained a mediator for the association between VPA and rMSSD HRV only. Finally, in an analysis adding HR to the previously outlined covariates, we did not observe accumulated time spent in any PA intensity category to be significantly associated with HRV (see Table, Supplemental Digital Content 1, Multivariable Linear Regression Analysis Results for Physical Activity Intensity as Independent Predictors of HRV,; although a trend was observed for VPA. Analyses of all models excluding those with diabetes did not appreciably change the results.


The present study described, for the first time to our knowledge, independent associations of accelerometer-estimated VPA, MPA, and LPA with SDNN and rMSSD HRV, and identified possible physiological mediators in glycemic control indices and other CVD risk factors. Our main analyses indicated VPA to be associated with SDNN and rMSSD HRV and, for LPA, rMSSD HRV. These associations were maintained in sensitivity analyses which included plausible biological intermediates. Fasting insulin and total plasma triglycerides were the most consistent physiological mediators overall for the associations of VPA and LPA with HRV.

Our study corroborated the VPA observations of previous self-reported PA intensity literature using accelerometer-estimated assessments of PA intensity. Indeed, although research is limited, VPA has been most consistently demonstrated as predictive of higher HRV when examining the association between self-reported PA intensity and HRV. In a sample of 43 middle-age adults (mean age, ~61 yr), Buchheit et al. (7) noted that participants with the highest self-reported moderate-to-vigorous PA had higher SDNN and rMSSD HRV values, as well as higher high-frequency power (a frequency-domain HRV metric reflecting parasympathetic innervation), when compared with sedentary participants. A cohort study among older adults (mean age, ~71 yr) suggested that faster self-reported walking pace was associated with higher SDNN HRV (43). Similarly, self-reported VPA has been predictive of both time and frequency domain measures of HRV in young adults (mean age, ~23 yr) (8). Despite the agreement of our study with past literature, additional studies are needed replicating our observations using accelerometry-estimated PA.

The current study adds to the literature as we noted accelerometer-estimated LPA to be associated with higher rMSSD HRV. This observation is encouraging given that LPA participation is attainable for much of the population regardless of disease status or age, with the most recent PA Guidelines (6) advocating for greater engagement in LPA to replace sedentary time given the noted beneficial health outcomes for LPA (46). This study provides support for these LPA-related guidelines, and suggests that even activities done at a light intensity (e.g., casual walking at ~2.0 mph [2.8 metabolic equivalents], household chores such as light sweeping [2.3 metabolic equivalents]) may be associated with improved CAF. Further prospective trials and epidemiologic studies are warranted. Although we did not observe MPA as a significant independent predictor of either HRV metric, three points should be kept in mind. First, our participants’ range of LPA duration was much larger than that of MPA, with LPA constituting the majority of participants’ daily PA duration; thus, we may have had more power to detect a LPA association with HRV. Second, VPA, although more limited in duration range in our participants than either LPA or MPA, has markedly stronger physiological effects than LPA and MPA (47). This explanation, combined with the noted positive association self-reported VPA has with HRV, supports our observations. Finally, it is possible that the lack of statistical significance of the MPA observations was an artifact of random variation or residual confounding. However, we note that although MPA may not have been observed to be significantly associated with HRV, MPA was still positively associated with higher SDNN and rMSSD values (Tables 2 and 3; see Table, Supplemental Digital Content 1, Multivariable Linear Regression Analysis Results for Physical Activity Intensity as Independent Predictors of HRV,

Mediation analyses for our hypothesized primary mediators revealed fasting insulin to be the strongest physiological mediator for the associations of VPA and LPA with SDNN and/or rMSSD HRV. Studies (20,21) have suggested that hyperinsulinemia caused by higher insulin resistance or lower insulin sensitivity can upregulate SNS activity; thus lowering time-domain measures of HRV. Findings from prospective cohort studies (14,15) and case-control studies (13) have noted results congruent with this physiological mechanism as higher fasting insulin levels and/or greater insulin resistance/lower insulin sensitivity have been associated with worsened time- and frequency-domain HRV values. Our observations align with these studies. Further, although fasting glucose was only observed a mediator of the association between VPA and HRV in our main mediation models, this might be due to the low percentage of individuals with diabetes in the current sample (4.7%). Regular PA participation has long demonstrated a robust ability to control hyperglycemic episode frequency—particularly after meal consumption—due primarily to increased insulin sensitivity and glucose disposal (48,49). Indeed, given regular PA’s known impact on reducing hyperglycemic episode frequency (48,49), and how repeated hyperglycemic episodes may impair CAF via upregulation of proinflammatory cytokines (16–19), our mediation results provide preliminary evidence that LPA may be associated with improved insulin dynamics, beneficial in proper CAF, and thus more favorable HRV indices. Again, additional prospective trials and epidemiologic studies are warranted.

Total plasma triglycerides were observed to be the most consistent hypothesized secondary physiological mediator of the association between VPA and HRV, with total plasma triglycerides trending toward mediating the association between LPA and HRV. Waist circumference was also observed to be a mediator between VPA and HRV. Addressing these observations in parallel is important. Recent evidence has suggested that cardiolipotoxicity may increase circulation of proinflammatory cytokines (e.g., IL-6) and disrupt proper CAF (28). Additionally, higher visceral adiposity is associated with higher circulating triglycerides and heightened SNS activity which would have the effect of lowering time-domain HRV metrics (29,30). From a behavioral perspective, our observations are congruent with accelerometry-based data from 10 different countries indicating individuals of lower BMI are more likely to participate in higher intensities of PA—as assessed by accelerometer counts per minute—compared with individuals of higher BMI (50). Taking this physiological and behavioral literature into consideration, our observations may therefore indicate that although LPA had a positive association with HRV, achieving healthful waist circumference/BMI levels is clearly important to consider in unison with LPA engagement to ensure long-term PA participation and the prevention of cardiolipotoxicity and subsequent systemic inflammation. More LPA-oriented research on this matter is warranted.

The current study has the following strengths: [1] accelerometry was employed to quantify time spent in different PA intensities; [2] the sample was composed of racially diverse middle-age adults; and [3] a novel mediation modeling procedure was used to discern the magnitude of mediation glycemic control indices and other CVD risk factors have on the association between PA intensity and HRV as a marker of CAF. However, all results should be interpreted while considering the following limitations. First, the current analysis was cross-sectional, therefore we cannot infer temporality and, thus, causation. Prospective trials and epidemiologic studies are needed to confirm these observations. Second, few individuals within the current sample had diabetes which contributed to a homogenous sample as it pertains to one measure of the primary physiological mechanisms we desired to examine. Thus, we cannot comment on whether the associations and mediations observed would hold true in samples with greater numbers of individuals with diabetes. Nonetheless, the current results are aligned with prior research which has noted impaired CAF in individuals with diabetes (22–24) and individuals who are normoglycemic but insulin resistant (25). Finally, the HRV metrics obtained in this study were ascertained from three sequential 10-s 12-lead ECG strips. Although a potential limitation, it should be noted that these measurements were performed under standardized conditions, with short duration HRV measurements validated against longer duration HRV measurements such as the “gold standard” Holter monitoring (2,40,41). Holter monitors could be employed in future prospective studies.


The current study suggested that accelerometer-estimated VPA and LPA have independent positive associations with SDNN and/or rMSSD HRV after robust adjustment for relevant demographic, lifestyle, and clinical factors as well as biologically plausible intermediates. This study also noted fasting insulin to be a consistent physiological mediator for the association of VPA and LPA with HRV. As PA has a notable influence on glycemic control (48,49), continued advocacy for LPA, which has demonstrated health benefits and is attainable for most of the population (6,46), is needed and may assist in the promotion of more healthful CAF. Based on the mediation results observed for fasting triglycerides, more research on how PA may reduce cardiolipotoxicity and the ensuing systemic inflammation may also be needed.

Z. C. P. is supported by the National Heart, Lung, and Blood Institute (NHLBI) of the National Institutes of Health (NIH) under Award Number T32 HL007779. The Coronary Artery Risk Development in Young Adults Study (CARDIA) is conducted and supported by the National Heart, Lung, and Blood Institute (NHLBI) in collaboration with the University of Alabama at Birmingham (HHSN268201800005I & HHSN268201800007I), Northwestern University (HHSN268201800003I), University of Minnesota (HHSN268201800006I), and Kaiser Foundation Research Institute (HHSN268201800004I). CARDIA is also partially supported by the Intramural Research Program of the National Institute on Aging (NIA) and an intra-agency agreement between NIA and NHLBI (AG0005). Data for this study were also from the CARDIA Fitness Study which was supported by the National Institutes of Health (R01 HL078972). This manuscript has been reviewed by CARDIA for scientific content. Finally, the results of the study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation. The results of the present study do not constitute endorsement by ACSM.

The authors have no conflicts of interest to disclose.


1. Tesfaye S, Malik R, Boulton A, et al. Diabetic neuropathies: update on definitions, diagnostic criteria, estimation of severity, and treatments. Diabetes Care. 2010;33:2285–93.
2. Shaffer F, Ginsberg J. An overview of heart rate variability metrics and norms. Front Public Health. 2017;5:258.
3. Shah SA, Kambur T, Chan C, Herrington DM, Liu K, Shah SJ. Relation of short-term heart rate variability to incident heart failure (from the multi-ethnic study of atherosclerosis). Am J Cardiol. 2013;112(4):533–40.
4. Kubota Y, Chen LY, Whitsel EA, Folsom AR. Heart rate variability and lifetime risk of cardiovascular disease: the atherosclerosis risk in communities study. Ann Epidemiol. 2017;27(10):619–25.e2.
5. Fang SC, Wu YL, Tsai PS. Heart rate variability and risk of all-cause death and cardiovascular events in patients with cardiovascular disease: a meta-analysis of cohort studies. Biol Res Nurs. 2019;22(1):45–56.
6. U.S. Department of Health and Human Services. Physical Activity Guidelines for Americans. 2nd ed. Washington, D.C.: U.S. Department of Health and Human Services; 2018.
7. Buchheit M, Simon C, Charloux A, Doutreleau S, Piquard F, Brandenberger G. Heart rate variability and intensity of habitual physical activity in middle-aged persons. Med Sci Sports Exerc. 2005;37(9):1530–4.
8. May R, McBerty V, Zaky A, Gianotti M. Vigorous physical activity predicts higher heart rate variability among younger adults. J Physiol Anthropol. 2017;36(1):24.
9. Ghardashi-Afousi A, Holisaz M, Shirvani H, Pishgoo B. The effects of low-volume high-intensity interval versus moderate intensity continuous training on heart rate variability, and hemodynamic and echocardiography indices in men after coronary artery bypass grafting: a randomized clinical trial study. ARYA Atheroscler. 2018;14(6):260–71.
10. Macagnan FE, Feoli AMP, Russomano T. Acute physical effort increases sympathovagal balance responses to autonomic stimulation in metabolic syndrome. Metab Syndr Relat Disord. 2019;17(1):67–74.
11. Callaghan BC, Little AA, Feldman EL, Hughes RA. Enhanced glucose control for preventing and treating diabetic neuropathy. Cochrane Database Syst Rev. 2012;6:CD007543.
12. Bassi D, Arakelian V, Mendes R, et al. Poor glycemic control impacts linear and non-linear dynamics of heart rate in DM type 2. Rev Bras Med Esporte. 2015;21(4):313–7.
13. Svensson M, Lindmark S, Wiklund U, et al. Alterations in heart rate variability during everyday life are linked to insulin resistance. A role of dominating sympathetic over parasympathetic nerve activity? Cardiovasc Diabetol. 2016;15:91.
14. Saito I, Hitsumoto S, Maruyama K, et al. Heart rate variability, insulin resistance, and insulin sensitivity in Japanese adults: the Toon health study. J Epidemiol. 2015;25(9):583–91.
15. Schroeder E, Chambless L, Liao D, et al. Diabetes, glucose, insulin, and heart rate variability. Diabetes Care. 2005;28:668–74.
16. Takuma K, Fang F, Zhang W, et al. RAGE-mediated signaling contributes to intraneuronal transport of amyloid-beta and neuronal dysfunction. Proc Natl Acad Sci U S A. 2009;106:20021–6.
17. Witzke K, Vinik A, Grant L, et al. Loss of receptor for advanced glycation end products (RAGE) defense. Diabetes Care. 2011;34:1617–21.
18. Schmidt A, Yan S, Yan S, Stern D. The multiligand receptor RAGE as a progression factor amplifying immune and inflammatory responses. J Clin Invest. 2001;108:949–55.
19. Vinik A, Erbas T, Casellini C. Diabetic cardiac autonomic neuropathy, inflammation and cardiovascular disease. J Diabetes Investig. 2013;4(1):4–18.
20. Huggett RJ, Hogarth AJ, Mackintosh AF, Mary DA. Sympathetic nerve hyperactivity in non-diabetic offspring of patients with type 2 diabetes mellitus. Diabetologia. 2006;49:2741–4.
21. Thorp AA, Schlaich MP. Relevance of sympathetic nervous system activation in obesity and metabolic syndrome. J Diabetes Res. 2015;2015:341583.
22. Lieb DC, Parson HK, Mamikunian G, Vinik AI. Cardiac autonomic imbalance in newly diagnosed and established diabetes is associated with markers of adipose tissue inflammation. Exp Diabetes Res. 2012;2012:878760.
23. Ciobanu D, Craciun A, Veresiu I, Bala C, Roman G. Ambulatory heart rate variability correlates with high-sensitivity c-reactive protein in type 2 diabetes and control subjects. IFMBE Proceedings. 2017;59. doi: 10.1007/978-3-319-52875-5_4.
24. Herder C, Schamarek I, Nowotny B, et al. Inflammatory markers are associated with cardiac autonomic dysfunction in recent-onset type 2 diabetes. Heart. 2017;103:63–70.
25. Stein P, Barzilay J, Chaves P, et al. Higher levels of inflammation factors and greater insulin resistance are independently associated with higher heart rate and lower heart rate variability in normoglycemic older individuals: the cardiovascular health study. J Am Geriatr Soc. 2008;56:315–21.
26. Chen L, Zmora R, Duval S, Chow L, Lloyd-Jones D, Schreiner P. Cardiorespiratory fitness, adiposity, and heart rate variability: the coronary artery risk development in young adults study. Med Sci Sports Exerc. 2019;51(3):509–14.
27. Dyrstad SM, Hansen BH, Holme IM, Anderssen SA. Comparison of self-reported versus accelerometer-measured physical activity. Med Sci Sports Exerc. 2014;46(1):99–106.
28. Ali A, Boutjdir M, Aromolaran A. Cardiolipotoxicity, inflammation, and arrhythmias: role for interleukin-6 molecular mechanisms. Front Physiol. 2019;9:1866.
29. Schlaich M, Straznicky N, Lambert E, Lambert G. Metabolic syndrome: a sympathetic disease? Lancet Diabetes Endocrinol. 2015;3:148–57.
30. Straznicky NE, Eikelis N, Lambert EA, Esler MD. Mediators of sympathetic activation in metabolic syndrome obesity. Curr Hypertens Rep. 2008;10:440–7.
31. Thayer J, Yamamoto S, Brosschot J. The relationship of autonomic imbalance, heart rate variability, and cardiovascular disease risk factors. Int J Cardiol. 2010;141(2):122–31.
32. Friedman GD, Cutter GR, Donahue RP, et al. CARDIA: study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol. 1988;41(1):1105–16.
33. Carnethon M, Sternfeld B, Schreiner P, et al. Association of 20-year changes in cardiorespiratory fitness with incident type 2 diabetes: the coronary artery risk development in young adults (CARDIA) fitness study. Diabetes Care. 2009;32:1284–8.
34. Lee I-M, Shiroma E, Kamada M, Bassett D Jr, Matthews C, Buring J. Association of step volume and intensity with all-cause mortality in older women. JAMA Intern Med. 2019;179(8):1105–12.
35. Troiano R, Berrigan D, Dodd K, Masse L, Tilert T, McDowell M. Physical activity in the United States measured by accelerometer. Med Sci Sports Exerc. 2008;40(1):181–8.
36. Sasaki J, John D, Freedson P. Validition and comparison of actigraph activity monitors. J Sci Med Sport. 2011;14(5):411–6.
37. Maddison R, Jiang Y, Foley L, Scragg R, Direito A, Olds T. The association between the activity profile and cardiovascular risk. J Sci Med Sport. 2016;19:605–10.
38. Schreiber-Gregory D. Multicollinearity: What Is It, Why Should We Care, and How Can It be Controlled? Cary, NC: SAS; 2017. [cited 2019 June 22]. Available from:
39. Kozey-Keadle S, Libertine A, Lyden K, Staudenmayer J, Freedson P. Validation of wearable monitors for assessing sedentary behavior. Med Sci Sports Exerc. 2011;43(8):1561–7.
40. Munoz ML, van Roon A, Riese H, et al. Validity of (ultra-)short recordings for heart rate variability measurements. PLoS One. 2015;10(9):e0138921.
41. Nussinovitch U, Elishkevitz K, Katz K, et al. Reliability of ultra-short ecg indices for heart rate variability. Ann Noninvasive Electrocardiol. 2011;16(2):117–22.
42. Meyer K, Sijtsma F, Nettleton J, et al. Dietary patterns are associated with plasma F2-isoprostanes in an observational cohort study of adults. Free Radic Biol Med. 2013;57:201–9.
43. Soares-Miranda L, Sattelmair J, Chaves P, et al. Physical activity and heart rate variability in older adults: the cardiovascular health study. Circulation. 2014;129:2100–10.
44. Valeri L, VanderWeele T. Mediation analysis allowing for exposure-mediator interactions and causal interpretation: theoretical assumptions and implementation with SAS and SPSS macros. Psychol Methods. 2013;18(2):137–50.
45. de Geus E, Gianaros P, Brindle R, Jennings J, Berntson GG. Should heart rate variability be “corrected” for heart rate? Biological, quantitative, and interpretive considerations. Psychophysiology. 2019;56(2):e13287.
46. LaCroix A, Bellettiere J, Rillamas-Sun E, et al. Association of light physical activity measured by accelerometry and incidence of coronary heart disease and cardiovascular disease in older women. JAMA Netw Open. 2019;2(3):e190419.
47. Kenney W, Wilmore J, Costill D. Adaptations to aerobic and anaerobic training. In: Kenney W, Wilmore J, Costill D, editors. Physiology of Sport and Exercise. 6th ed. Champaign, IL: Human Kinetics; 2015. pp. 261–91.
48. Schwarz PEH, Timpel P, Harst L, et al. Blood sugar regulation for cardiovascular health promotion and disease prevention. J Am Coll Cardiol. 2018;72(15):1829–44.
49. American Diabetes Assoiation. Standards of medical care in diabetes-2018. Diabetes Care. 2018;41(1 Suppl):S1–172.
50. Van Dyck D, Cerin E, De Bourdeaudhuij I, et al. International study of objectively measured physical activity and sedentary time with body mass index and obesity: IPEN adult study. Int J Obes (Lond). 2015;39(2):199–207.


Supplemental Digital Content

Copyright © 2020 by the American College of Sports Medicine