Abstract
Purpose:
To estimate risks of incident Type 2 Diabetes (T2D) and Stage 2 and greater hypertension associated with self-reported and accelerometer-determined moderate-vigorous physical activity (MVPA) separately and adjusted for each other.
Methods:
The sample included 2,291 black and white men and women, ages 38–50, in the CARDIA (Coronary Artery Risk Development in Young Adults) Fitness Study, conducted during the Year 20 core CARDIA exam. Accelerometer-determined (Actigraph, LLC. model 7164) MVPA (MVPA-Acc), assessed at Year 20, was defined as mins/day of counts ≥2020/min. Self-reported MVPA (MVPA-SR) was assessed at Year 20 using the CARDIA Physical Activity History. Incident T2D was ascertained at Years 25 and 30 from fasting glucose, 2 hr GTT, HbA1c or diabetes medication; incident hypertension was ascertained at those same times from measured blood pressure or use of antihypertensive medications. Modified Poisson regression models estimated relative risk (RR) of incident (Years 25 and 30) T2D or hypertension, associated with middle and high tertiles of Year 20 MVPA-Acc alone, Year 20 MVPA-SR alone, and both, adjusted for each other, relative to bottom tertile.
Results:
In men, MVPA-Acc, but not MVPA-SR, was associated with a 37–67% decreased risk of incident T2D in a dose response relation that persisted with adjustment for BMI, Similar associations were observed in women, although the risk reduction was similar in the second and third tertiles, relative to the bottom tertile. In both men and women, MVPA-Acc was marginally associated with reduced risk of incident Stage 2 and greater hypertension, but only after adjustment for BMI, while MVPA-SR was not associated in either sex.
Conclusions:
Accelerometer-determined MVPA may provide more consistent risk estimates for incident diabetes than self-reported MVPA.
Keywords: Accelerometry, self-report, prospective analysis, epidemiology
INTRODUCTION
Decades of research have firmly established the benefits of regular physical activity (PA) for disease prevention and health promotion(1) (2). This research has relied largely on self-reported PA, even though the measurement errors and biases inherent in this approach are widely acknowledged (3;4). Because of the limitations of self-report (e.g. difficulty with recall, social desirability, cognitive challenges and lack of comprehensive ascertainment) (5–8), researchers have embraced accelerometry as a PA assessment tool (3;9), even though this approach also has sources of error (10;11). This shift in PA assessment poses a challenge for epidemiological cohort studies that span many years, namely consistency of measurement approach over time or adoption of more robust measures.
Adding to the challenge is the growing understanding that self-report and accelerometry assess different dimensions of physical activity (10;12), with self-report providing contextual details for intentional, moderate to vigorous intensity physical activity (MVPA), most typically in the recreational domain, and accelerometry providing information about the full spectrum of movement from sedentary behavior and light PA to MVPA in all domains. The implication of this is that epidemiological cohort studies might do well to include both types of measures in their protocols in order to gain a deeper and fuller understanding of the impact of physical activity on health outcomes. A growing literature has begun to evaluate this implication by examining whether associations of PA with disease biomarkers differ depending on the PA assessment method (13–19), and/or whether using both types of measures (20;21) or combining them into a single, composite measure (20) account for more of the variability in the exposure-outcome relation. However, to date, little is known about whether risk estimates of clinical endpoints, such as hypertension or diabetes differ by PA measurement approach.
In the CARDIA (Coronary Artery Risk Development in Young Adults) Fitness Study, an ancillary study conducted in conjunction with the Year 20 CARDIA examination, participants wore an accelerometer for up to seven days to measure physical activity and completed a self-report physical activity questionnaire as part of the core exam. Using these data, Gabriel et al found that having both self-reported activity and accelerometer-determined activity as separate independent variables adjusted for each other accounted for more of the variability in most, although not all, cardiovascular risk factors than either a composite measure that combined components of the two approaches or either measure alone (20)
In the current analysis, we continue to address the question of how the method of physical activity assessment impacts measures of association with disease risk factors and outcomes by examining whether the risk estimates for 10-year incidence of Type 2 diabetes (T2D) and Stage 2 and greater hypertension in the CARDIA cohort vary by PA measure. Underscoring the significance of this issue is the conclusion reached by the 2018 Physical Activity Guidelines Advisory Committee Report that insufficient evidence exists at this time to answer the question (2), a question that only becomes more relevant with advances in technology and assessment methodology. The specific aims of this paper are: first, to determine the association of each type of PA measure (i.e. self-reported and accelerometer-determined) with each outcome; and secondly, to examine whether the PA measures are independent of each other when they are considered jointly. The primary hypothesis is that associations of MVPA-Acc with incident T2D and Stage 2 and greater hypertension will be more consistent and stable than those between MVPA-SR and the two outcomes.
METHODS
Study Sample:
The design of the CARDIA study and the recruitment of the cohort have been previously described (22). Briefly, CARDIA is a multi-site, longitudinal, observational study of the development of coronary disease risk factors and subclinical cardiovascular disease in young adults. The cohort, recruited by random digit dialing in Birmingham, AL, Chicago, IL, and Minneapolis, MN and from the membership list of a large health maintenance organization in Oakland, CA., consists of 5,115 black and white men and women, ages 18–30 at the baseline examination in 1985–86. In order to ensure large enough sample sizes for meaningful comparisons, recruitment of the cohort was designed to be roughly balanced on race, sex, age (18–24, 25–30), and education (≤12 years, >12 years). Follow-up examinations were conducted in Years 2, 5, 7, 10, 15, 20, 25 and 30 with retention rates of 91%, 86%, 81%, 79%, 74%, 72%, 72%, and 71% of the surviving cohort, respectively. All CARDIA participants provide written informed consent and the institutional review boards at each participating center review and approve the study annually.
The sampling frame for this analysis was the 3,549 members of the CARDIA cohort who attended the Year 20 core examination. Of these, 3001 consented to the CARDIA Fitness Study (CFS), a separately funded ancillary study designed to examine 20-year changes in physical fitness, physical activity and body weight in relation to 20-year changes in CVD risk factors. Excluded from analysis were 11 for missing self-reported physical activity data, 665 for missing valid accelerometer data, 11 for missing covariate data, 1 who was transgender, and 22 with biologically implausible values of BMI or PA (e.g. BMI≥138, MVPA-Acc=1440 mins/day of counts≥2020). The remaining 2,291 (347 black men, 619 white men, 603 black women, and 722 white women) constitute the sample for the present analysis.
Assessment of Physical Activity (Accelerometry):
As part of the CFS protocol, participants were asked to wear an accelerometer (Actigraph, LLC., model 7164) placed on an elastic belt around their waists for seven consecutive days, except when sleeping or doing water activities (e.g. showering, swimming). All accelerometers were calibrated prior to distribution and initialized with the epoch (data collection period) set to 1 minute. Upon return of an accelerometer, the data were downloaded using the manufacturer’s software and screened for wear time according to the methods described by Troiano et al (23). Time spent per day (min/d) in different intensity levels were defined using the NHANES cut points (MVPA ≥ 2020 cts/min; moderate PA = 2020–5998 ct/min; vigorous PA ≥ 5999 cts/min) (23) in order to be consistent with similar analyses conducted in the NHANES dataset (16). Daily estimates were then computed by averaging daily time across total number of days accelerometer was worn for those participants with ≥4 of 7 days with ≥10 hours per day of wear-time. Seventy four percent of the sample provided the full 7 days of recordings, thus capturing weekdays and both weekend days, while another 15% provided 6 days of recordings, capturing at least one weekend day.
For the purposes of this analysis, PA was defined as daily estimates of time spent in moderate, vigorous, or MVPA. Derived variables such as total or average accelerometer counts and time spent being sedentary or in light intensity physical activity were not considered since the PA questionnaire asked only about activities with a MET value of 3 and above.
Assessment of Physical Activity (Self-Report):
At every exam, CARDIA participants have completed the same Physical Activity History (PAH) questionnaire (24), which asks about participation in 13 mostly recreational activities (8 of vigorous intensity and 5 of moderate intensity) over the previous 12 months; data from the Year 20 PAH were used in this analysis. Because frequency and duration are not assessed specifically, the responses are summarized in terms of Exercise Units (EU), computed by multiplying the intensity of the activity in METs by the number of months of minimal participation (at least one hour total time) plus a weighting factor applied to the number of months of participation at a specified weekly frequency and duration and then summing over all activities. A score of 300 EU can be interpreted as approximately equal to meeting the PA recommendations of 150 minutes a week of moderate intensity activity or 75 minutes a week of vigorous intensity activity or an equivalent combination (1;25). Scores are computed for moderate intensity PA (activities with MET values of 3 to < 6), vigorous intensity PA (activities with MET values of 6 and above), and total MVPA (sum of moderate and vigorous PA). The CARDIA PAH has reasonable test-retest reliability and has been validated against accelerometers, PA diaries, and more detailed PA questionnaires (26), including one identical to the PAH except for inclusion of explicit questions about frequency and duration (25). It has also been indirectly validated by showing the expected relations with physical fitness, measures of body fat, and cardiovascular and metabolic outcomes (24;27;28).
Ascertainment of Incident Diabetes and Hypertension:
Incident diabetes was determined at the Years 25 and 30 core examinations based on the presence of any of the following: fasting glucose ≥ 7 mmol/l; use of medications for diabetes treatment; 2 hour glucose tolerance test (GTT) ≥ 11.1 mmol/l or HbA1c ≥ 6.5% (48 mmol/mol IFCC) (29). Serum glucose concentrations were measured using the hexokinase method. Consistent with recent categorizations from the American College of Cardiology (30), presence of Stage 2 and greater hypertension was defined as systolic blood pressure ≥ 140 mmHg, or diastolic blood pressure ≥ 90 mmHg, or use of antihypertensive medications and was ascertained at Years 25 and 30 based on measured resting blood pressure (Omron automated blood pressure device, calibrated to values obtained from a random zero sphygmomanometer) and reported medications. A standardized protocol was used to measure blood pressure that included 5 minutes of seated rest before the first measurement, followed by two additional measurements with a one minute rest in between; the average of the second and third measures were used in the analysis.
Those meeting the case definitions at either Year 25 or Year 30, but not at Year 20, were counted as incident cases of diabetes or hypertension respectively.
Covariates:
The covariates considered in this analysis were those that, based on the literature, might confound the risk of diabetes or hypertension associated with physical activity, specifically age, sex, race, education, smoking status, and body mass index (BMI). Sex and race were self-reported at the baseline exam. Education was reported at Year 20 and dichotomized as high school or less vs. more than high school. Age was calculated as the difference between date of birth, reported at the baseline exam, and date of the CFS exam. Smoking status, self-reported at the Year 20 CARDIA exam using a standard questionnaire that has been repeated at all CARDIA exams, was categorized as smoker vs. non-smoker. BMI was defined as weight in kg/(height in m)2 from measured height and weight using a calibrated scale and stadiometer and following a standardized study procedure. Study site was also considered as a covariate to account for any center differences, as was accelerometer wear time to account for differences in wear time.
Data Analyses:
Given evidence of differential relations between self-reported and objectively-assessed physical activity in men and women (31), and to be consistent with the initial sampling methodology, all analyses were stratified by sex. Year 20 characteristics of the CFS participants who were included in the analytic sample were compared to those who were excluded, using Student t tests for differences in means for continuous variables and Chi square tests for differences in proportions for categorical variables. The Year 20 characteristics of those in the sample were similarly compared by race and stratified by sex. The physical activity variables, both self-reported and from accelerometry (except for wear time), had highly skewed distributions and were described by medians and the inter-quartile range, and racial differences between blacks and whites, within sex, were assessed by the Wilcoxon rank sum test. Spearman rank order correlation coefficients, for the sample as a whole and stratified by race/sex group, provided measures of association of the self-reported physical activity variables with the accelerometer-determined variables.
Separate multivariable modified Poisson regression models (32), stratified by sex, estimated the relative risk of incident diabetes or hypertension at Years 25 and 30 associated, first, with Year 20 self-reported MVPA (MVPA-SR) alone, then, with Year 20 accelerometer-determined MVPA (MVPA-Acc) alone, and finally with both, adjusted for each other. Because of the skewed distributions, both self-reported and accelerometer-determined MVPA were categorized into sex-specific tertiles with the bottom tertile used as the reference. All models were adjusted for age, race, and education. Models with diabetes as the outcome further adjusted for Year 20 fasting glucose and excluded those with prevalent T2D (N=152) while the models with hypertension as the outcome adjusted for Year 20 diastolic and systolic blood pressure and excluded those with prevalent hypertension (N=429). Models that adjusted for smoking status, study center, and wear time were also run. To provide evidence of the degree to which body fat confounded or mediated the relation between physical activity and risk of diabetes or hypertension, a second set of models adjusted further for BMI. Model fit was assessed by the Quasi-likelihood under Independence Model Criterion (QIC).
To examine the existence of a linear trend in the association of physical activity with risk of incident diabetes and hypertension, the categorical PA variables were entered into the respective models as continuous variables with the value of 1–3.
Non-missing exposure and covariate data were required for inclusion in the descriptive analysis; however, only participants in the analytic data set who had outcome data at Year 25 or 30 were included in the Poisson regressions. No adjustment was made for multiple comparisons because there is still controversy over this practice. As pointed out by Rothman, adjustment, while decreasing type I error, would increase the risk of type II error for associations that are not null (33). Nevertheless, as recommended by Rothman, all findings are interpreted with caution (34) All analyses were conducted in SAS v.9.3.
RESULTS
Compared to those included in the analytic sample, the CFS participants who were excluded were more likely to be black (56.4% vs. 35.9%, p<0.0001 for men, 58.6% vs. 45.5%, p<0.0001 for women), to be current smokers (27.1% vs. 17.5%, p=0.0009 for men, 22.2% vs. 15.2%, p=0.0007 for women), and to have a high school education or less (29.9% vs 23.1, p=0.01 for men, 28.0% vs. 20.5%, p=0.002 for women). Additionally, among the men, those excluded were younger (mean=44.8 years, sd=3.7 vs. mean=45.3 years, sd=3,4, p=0.03, and among the women, those excluded had a higher BMI (mean=30.8 kg/m2, sd=7.8 vs. mean=29.1 sd=7.4, p=0.0001), a higher systolic blood pressure (mean=116 mmHg, sd=17.4 vs. 113 mmHg, sd=15.2, p=0.001, a higher diastolic blood pressure (mean=74 mmHg, sd=12.8 vs. mean=71 mmHg, sd=11.4, p=0.0008), and were more likely to be hypertensive (28.5% vs. 19.3%, p=0.0001).
Among those in the analytic sample, blacks at Year 20 tended to be younger, have less education, higher BMI, higher SBP, and DBP, to be current smokers and to have prevalent hypertension and diabetes than whites in both men and women (Table 1). Black women also had higher fasting glucose than the white women. As shown in Table 1, the black women were also considerably less active on all measures of self-reported and accelerometer-determined physical activity than the white women. Among the men, there were no differences in self-reported MVPA, but based on accelerometry, black men spent less time in vigorous intensity PA than white men.
Table 1:
Year 20 characteristics of CARDIA participants included in analysis by race and sex
| Men | Women | |||||
|---|---|---|---|---|---|---|
| Black N=347 (35.9%) | White N=619 (64.1%) | p value1 | Black N=603 (45.5%) | White N=722 (54.5) | p value1 | |
| Age (yrs), mean (sd) | 44.7 (3.6) | 45.6 (3.3) | <0.0001 | 44.5 (3.8) | 45.9 (3.4) | <0.0001 |
| High School or less, N (%) | 119 (34.3) | 104 (16.8) | <0.0001 | 177 (29.4) | 95 (13.2) | <0.0001 |
| Body Mass Index, mean (sd) | 29.4 (6.1) | 28.3 (4.8) | 0.005 | 31.6 (7.4) | 26.9 (6.6) | <0.0001 |
| Current smokers, N (%) | 91 (26.3) | 77 (12.6) | <0.0001 | 116 (19.4) | 84 (11.7) | <0.0001 |
| SBP (mmHg), mean (sd) | 123 (13.8) | 117 (11.1) | <0.0001 | 118 (16.7) | 109 (12.3) | <0.0001 |
| DBP (mmHg), mean (sd) | 76 (10.7) | 72 (9.4) | 0.001 | 76 (11.5) | 68 (10.1) | <0.0001 |
| Fasting glucose, mean (sd) | 103.3 (26.0) | 101.5 (21.1) | 0.26 | 99.9 (28.0) | 93.4 (13.1) | <0.0001 |
| Hypertensive2, N (%) | 85 (24.5) | 99 (16.0) | <0.0001 | 177 (30.8) | 68 (9.7) | <0.0001 |
| Diabetic3, N (%) | 36 (10.4) | 39 (6.3) | 0.02 | 70 (11.8) | 22 (3.1) | <0.0001 |
| Self-reported PA Total MVPA4 (EU), med (I-Q range) Heavy PA (EU), med (I-Q range) Mod PA (EU), med (I-Q range) |
351 (187–602) 216 (88–402) 139 (63–208) |
362 (200–588) 224 (83–384) 136 (80–216) |
0.69 0.68 0.09 |
170 (72–328) 72 (6–216) 76 (27–144) |
294 (152–486) 139 (43–302) 139 (67–216) |
<0.0001 <0.0001 <0.0001 |
| Accelerometer PA Wear time (mins/d), mean (sd) MVPA5 (mins/day),med (I-Q range) Moderate PA3 (mins/day) med (I-Q range) Vigorous PA2 (mins/day) med (I-Q range) |
896 (96.6) 33.7 (20–52) 32.6 (20–50) 0.2 (0.0–1.0) |
888 (74.5) 35.4 (22–52) 33.3 (21.47) 0.3 (0.0–2.7) |
0.26 0.70 0.86 0.05 |
866 (89.4) 17.6 (11–30) 17.4 (10–28) 0.0 (0.0–0.2) |
882 (73.7) 28.2 (16–45) 26.6 (16–40) 0.1 (0.0–1.8) |
<0.0001 <0.0001 <0.0001 <0.0001 |
p value from Chi square test for difference in proportions, t test for difference in means, Wilcoxon rank sum test for difference in medians
Hypertensive defined as systolic blood pressure ≥ 140 mmHg, or diastolic blood pressure ≥ 90 mmHg, or use of antihypertensive medications
Diabetic defined as fasting glucose ≥ 7 mmol/l, use of medications for diabetes treatment, 2 h GTT ≥ 11.1 mmol/l or HbA1c ≥ 6.5% (48 mmol/mol IFCC)
. p value from Wilcoxon rank sum test for difference in medians
MVPA=moderate-vigorous physical activity
NHANES cut-points (MVPA = min/day ≥ 2020 cts/min; moderate PA = min/day=2020–5998 cts/min; vigorous PA = min/day ≥ 5999 cts/min)
Table 2 demonstrates that the self-reported PA variables had only a modest correlation with the accelerometer-determined PA variables (rho ranged from 0.16–0.36). Even self-reported vigorous PA had a correlation of only rho=0.33 with accelerometer-determined vigorous PA. Self- reported vigorous PA accounted for much of MVPA-SR (rho=0.92), which combines both moderate and vigorous intensity activity, while accelerometer-determined moderate PA accounted for virtually all of the variability in MVPA-Acc (rho=0.99). Similar correlations were observed in each of the race/gender groups (data not shown), although the correlations between the self-report and accelerometer variables were even lower in the men (range from rho=0.06 to 0.29) than in the sample as a whole.
Table 2:
Spearman correlation coefficients between self-reported and accelerometer-measured physical activity
| Whole Sample (N=2,018) | ||||||
|---|---|---|---|---|---|---|
| MVPA-SR | Mod PA-SR | Vig PA-SR | MVPA-Acc | Mod PA-Acc | Vig PA-Acc | |
| MVPA-SR | 1.00 | 0.72 | 0.92 | 0.36 | 0.33 | 0.32 |
| Mod PA-SR | 1.00 | 0.42 | 0.25 | 0.26 | 0.16 | |
| Vig PA-SR | 1.00 | 0.33 | 0.30 | 0.33 | ||
| MVPA-Acc | 1.00 | 0.99 | 0.56 | |||
| Mod PA-Acc | 1.00 | 0.47 | ||||
| Vig PA-Acc | 1.00 | |||||
Abbreviations: MVPA-SR=moderate to vigorous physical activity, self-report; Mod PA-SR=moderate intensity physical activity, self-report, Vig PA=vigorous intensity physical activity, self-report; MVPA-Acc=moderate to vigorous physical activity, accelerometer; Mod PA-Acc=moderate intensity physical activity, accelerometer; Vig PA-Acc=vigorous intensity physical activity, accelerometer
Note: All correlations significantly different from 0, p<0.0001
Risk of Incident Diabetes Associated with Physical Activity:
At Years 25 and 30, 147 new cases of diabetes (59 in men and 88 in women) were detected. The risk of incident diabetes was significantly lower among men by about 56% in the middle tertile and 67% in the upper tertile of MVPA-Acc, after controlling for Year 20 fasting glucose and race, age, and education (Table 3, Model 1b) and, additionally, MVPA-SR (Table 3, Model 1c). The risk estimates associated with MVPA-Acc were still significant and only modestly attenuated with further adjustment for BMI (Table 3, Models 2b and c). There was no significant association between incident T2D and MVPA-SR (Table 3, Models 1a, 1c). Similarly, among the women, incident diabetes was only associated with MVPA-Acc and not MVPA-SR. Evidence of a dose response relation with MVPA-Acc was present only in men.
Table 3:
Relative risk1 of incident diabetes2 over 10 years of follow-up associated with self-reported and/or accelerometer-measured physical activity, adjusted for covariates, stratified by sex3
| Men | Women | |||
|---|---|---|---|---|
| Model 14 | Model 25 | Model 14 | Model 25 | |
| RR (95% CI)3 | RR (95% CI) 3 | RR (95% CI) 3 | RR (95% CI) 3 | |
| Exposure Variable(s)6 | ||||
| MVPA-SR (EU) | Model 1a | Model 2a | Model 1a | Model 2a |
| T1 | Ref | Ref | Ref | Ref |
| T2 | 1.24 (0.69–2.23) | 1.07 (0.74–1.55) | 0.88 (0.57–1.36) | 0.94 (0.69–1.29) |
| T3 | 0.81 (0.41–1.60) | 0.82 (0.56–1.21) | 0.63 (0.38–1.07) | 0.93 (0.65–1.33) |
| p for trend | 0.55 | 0.33 | 0.09 | 0.67 |
| MVPA-Acc (mins/wk, cts>2020) | Model 1b | Model 2b | Model 1b | Model 2b |
| T1 | Ref | Ref | Ref | Ref |
| T2 | 0.44 (0.24–0.81) | 0.63 (0.44–0.90) | 0.52 (0.32–0.86) | 0.71 (0.52–0.97) |
| T3 | 0.33 (0.17–0.67) | 0.52 (0.35–0.78) | 0.61 (0.34–1.08) | 0.71 (0.48–1.05) |
| p for trend | 0.0007 | 0.0008 | 0.05 | 0.05 |
| Both PA Measures | Model 1c | Model 2c | Model 1c | Model 2c |
| MVPA-SR | ||||
| T1 | Ref | Ref | Ref | Ref |
| T2 | 1.31 (0.74–2.30) | 1.13 (0.78–1.64) | 0.94 (0.60–1.47) | 0.97 (0.70–1.33) |
| T3 | 0.96 (0.50–1.84) | 0.90 (0.62–1.33) | 0.71 (0.41–1.26) | 1.04 (0.72–1.50) |
| p for trend | 0.95 | 0.67 | 0.25 | 0.92 |
| MVPA-Acc | ||||
| T1 | Ref | Ref | Ref | Ref |
| T2 | 0.45 (0.24–0.83) | 0.63 (0.44–0.91) | 0.55 (0.33–0.90) | 0.71 (0.52–0.97) |
| T3 | 0.33 (0.17–0.65) | 0.55 (0.35–0.79) | 0.67 (0.37–1.22) | 0.70 (0.47–1.06) |
| p for trend | 0.0006 | 0.001 | 0.12 | 0.05 |
Relative risk estimated from modified Poisson regression
Diabetes defined as fasting glucose≥7 mmol/l, 2 hour glucose tolerance test ≥ 11.1 mmol/l or HbA1c≥6.5% (48 mmol/mol IFCC (N=59 men, 88 women)
Based on those whose diabetes status at Years 25 or 30 could be ascertained (N=2000)
Adjusted for race, age, education, and Year 20 fasting glucose
Additionally adjusted for BMI
Three different PA exposure variables examined for each model: a) MVPA-SR only; b) MVPA-Acc only; and c) both MVPA-SR and MVPA-Acc, adjusted for each other
Risk of Incident Stage 2 and Greater Hypertension Associated with Physical Activity:
As Table 4 illustrates, MVPA-Acc was marginally associated with a 10–20% reduction in risk of incident Stage 2 and greater hypertension (N=441, 195 in men and 246 in women) in both men and women, but only after adjustment for BMI (Model 2). In contrast, MVPA-SR was not associated with reduced risk either when considered singly or with adjustment for MVPA-Acc and/or BMI. When Stage 1 hypertension (SBP between 130–139 mmHg or DBP between 80–89 mmHg) was included in the case definition (30), there were no statistically significant relations with either MVPA-Acc or MVPA-SR (see Table, Supplemental Digital Content 1, Table Risk of Stage 1 and greater hypertension associated with self-reported and accelerometer-assessed physical activity, stratified by gender).
Table 4:
Relative risk1 of incident stage 2 and greater hypertension2 over 10 years of follow-up associated with self-reported and/or accelerometer-measured physical activity, adjusted for covariates, stratified by sex3
| Men | Women | |||
|---|---|---|---|---|
| Model 14 | Model 25 | Model 14 | Model 25 | |
| RR (95% CI) | RR (95% CI) | RR (95% CI) | RR (95% CI) | |
| Exposure Variable6 | ||||
| MVPA-SR (EU), model a | Model 1a | Model 2a | Model 1a | Model 2a |
| T1 | Ref | Ref | Ref | Ref |
| T2 | 1.00 (0.76–1.27) | 1.00 (0.84–1.19) | 0.78 (0.61–1.00) | 0.96 (0.82–1.123) |
| T3 | 1.02 (0.79–1.33) | 1.01 (0.84–1.21) | 0.88 (0.70–1.11) | 0.95 (0.80–1.12) |
| p for trend | 0.90 | 0.92 | 0.25 | 0.51 |
| MVPA-Acc(mins/wk, cts≥2020), model b | Model 1b | Model 2b | Model 1b | Model 2b |
| T1 | Ref | Ref | Ref | Ref |
| T2 | 0.92 (0.71–1.18) | 0.89 (0.75–1.05) | 0.88 (0.70–1.11) | 0.90 (0.78–1.04) |
| T3 | 0.82 (0.63–1.07) | 0.84 (0.70–1.01) | 0.80 (0.63–1.02) | 0.85 (0.72–1.01) |
| p for trend | 0.06 | 0.07 | 0.05 | |
| Both PA Measures, model c | Model 1c | Model 2c | Model 1c | Model 2c |
| MVPA-SR | ||||
| T1 | Ref | Ref | Ref | Ref |
| T2 | 1.01 (0.77–1.33) | 1.02 (0.85–1.21) | 0.80 (0.63–1.03) | 0.97 (0.83–1.14) |
| T3 | 1.05 (0.81–1.36) | 1.04 (0.87–1.25) | 0.93 (0.73–1.18) | 0.99 (0.83–1.17) |
| p for trend | 0.71 | 0.66 | 0.48 | 0.85 |
| MVPA-Acc | ||||
| T1 | Ref | Ref | Ref | Ref |
| T2 | 0.91 (0.70–1.18) | 0.88 (0.74–1.05) | 0.89 (0.70–1.13) | 0.90 (0.78–1.05) |
| T3 | 0.82 (0.63–1.07) | 0.84 (0.69–1.01) | 0.82 (0.64–1.05) | 0.85 (0.72–1.02) |
| p for trend | 0.14 | 0.05 | 0.11 | 0.06 |
Relative risk estimated from modified Poisson regression
Defined as systolic blood pressure≥140 mmHg, diastolic blood pressure ≥90 or use of antihypertensive medication (N=195 mean, 246 women)
Based on those whose hypertension status at Years 25 or 30 could be ascertained (N=2014)
Adjusted for race, age, education, and Year 20 systolic and diastolic blood pressure
Additionally adjusted for BMI
Three different PA exposure variables examined for each model: a) MVPA-SR only; b) MVPA-Acc only; and c) both MVPA-SR and MVPA-Acc adjusted for each other
Further adjustment for other covariates, including smoking, wear time and clinical site did not substantively change the estimates in any of these models (data not shown). Findings were also similar when models were run combining the men and women and adjusting for sex (see Tables, Supplemental Digital Content 2, 3, and 4, Risk of diabetes, stage 2 and greater hypertension, and stage 1 and greater hypertension, respectively, associated with self-reported and accelerometer-assessed physical activity, combining men and women), although in the combined models, MVPA-Acc was more consistently associated with reduced risk of stage 2 and greater hypertension, with or without adjustment for BMI.
DISCUSSION
This analysis of self-reported and accelerometer-determined MVPA and the risk of 10-year incident diabetes suggests that the magnitude and consistency of risk reduction associated with the accelerometer-determined measure are greater and more robust than with the self-reported measure. Correlations between the two PA measures were weak. The inverse relation of MVPA-SR with incident diabetes was not significant, and point estimates were attenuated when adjusted for MVPA-Acc and/or BMI. No relations at all were observed between self-reported PA and incident hypertension. The observed measures of association with MVPA-Acc with both diabetes and hypertension also tended to be more consistent with dose response relations (e.g. reduction in risk associated with highest tertile of activity greater than middle tertile).
The low correlation between self-reported and accelerometer-determined PA has been recognized for some time (4;35;36), and was starkly demonstrated by Troiano et al in 2008 who found that 25–33% of the population met PA recommendations by self-report but fewer than 5% did with accelerometry (23). This discrepancy has been attributed, in part, to the different PA constructs captured by the different measurements (12), but also, at least to some degree, to measurement error that is particularly inherent in self report. Although some evidence exists of differential reporting of physical activity by respondent characteristics, such as BMI or socioeconomic status, (37–39), the apparent substantial over-estimation of PA by self-report has been assumed to be largely, non-differential. Thus, associations with health outcomes may be assumed to be biased towards the null, and true associations with physical activity to be of greater magnitude than has been observed.
The current study offers support for this assumption in that accelerometer-determined physical activity was associated with increased risk of incident diabetes and more marginally with increased risk of incident hypertension, but self-reported PA was not. Other studies, such as an analysis in the NHANES adult population that found greater associations between physiological and anthropometric biomarkers of health with accelerometer-determined PA, support this assumption as well, although in that study self-reported PA was also independently associated with some biomarkers and remained associated when adjusted for accelerometer-determined PA (16). Similar findings were also observed in the NHANES youth population (15) as well as in other study samples (17;19).
On the other hand, in a study of a multi-ethnic British clinic population at high risk for T2D, self-reported walking was more strongly associated with glucose regulation even when adjusted for pedometer-measured steps (21). Similarly, accelerometry-assessed sedentary time was not consistently associated with cardiometabolic risk factors in a population-based British sample, while self-reported sedentary time was (40). Within the CARDIA data set itself, Gabriel et al found that both a composite measure of PA constructed from self-reported and accelerometer-determined PA and adding all PA measures separately in the model accounted for more variability in a range of cardiovascular risk factors than either measure alone (20).
Importantly, as stated in the 2018 Physical Activity Guidelines Advisory Committee Scientific Report, there is strong evidence, based primarily if, not solely, on self-reported PA, for inverse, dose-response relationships between physical activity and incidence of both T2D and hypertension (2). The fact that associations with self-reported PA were not consistently observed in the current study may be due, in part, to limitations inherent in the self-reported measure; a more detailed physical activity diary may have, perhaps, resulted in stronger associations. However, previous CARDIA investigations using the same self-reported PA questionnaire have reported inverse associations between self-reported PA and incident hypertension over the first 15 (41) and 20 years of follow-up (42). Another explanation may be the differences in estimates of the relative risk between logistic regression and the modified Poisson regression used in the present study; given that hypertension, in particular, is not a rare outcome, the odds ratios reported in other studies may exaggerate the true relative risk (32). These inconsistencies in existing evidence suggest that, until there is further, more definitive research, both self-reported and accelerometer-determined PA may still be needed to provide a deeper understanding of the full impact of PA on health outcomes.
This study has some limitations that deserve mention. Since the CARDIA cohort does not include Hispanics or Asians or those of other or mixed race/ethnicity, findings cannot be generalized to those population subgroups. The findings may also not be generalizable to those more at risk of developing either hypertension or diabetes at a younger age, given that physical activity was first assessed with accelerometry only at Year 20 when the cohort was between the ages of 38–50. The implication of this “left” censoring is that the analysis was conditioned on having survived to Year 20 without having developed the disease outcomes of interest. Furthermore, selection bias could have influenced the findings since those who were excluded from the analysis were more likely to have risk factors for diabetes and/or hypertension, such as lower education and higher BMI and prevalence of smoking. However, this bias would likely attenuate differences between associations with accelerometer-determined PA and those with self-reported PA by preferentially excluding those more likely to under-estimate their true level of activity (43). Also, the time frame for the two different PA assessments was not the same. Although both are considered measures of usual activity, the time frame for self-reported PA was the past year (relying on a combination of memory and synthesis by the participant) while the accelerometer measures were collected only over a 7-day period. Finally, because the self-reported PA questionnaire asked only about moderate or vigorous intensity activities, the only data used from the accelerometer were the number of minutes ≥2020 cts/min in order to ensure some comparability between the measures in terms of intensity of activity. This limited the ability to examine the impact of accelerometer-determined sedentary behavior and light intensity PA on diabetes and hypertension, which is a significant area for future research (2). Indeed, a recent finding from the CARDIA study shows that substitution of sedentary time with light intensity PA is associated with reduced risk factors cardiometabolic risk factors (44).
This study also has several notable strengths. First is the large, well-characterized, bi-racial and geographically diverse CARDIA cohort, which enhances the generalizability of the findings. Secondly, ascertainment of study outcomes (diabetes and hypertension) was based on well-defined and standardized measurements. Finally, the prospective study design, in which incident diabetes and hypertension over a 10-year period were examined in relation to both self-reported and accelerometer-determined PA, is essentially unique; most, if not all, other studies addressing a similar question were cross-sectional and relied on biomarkers, rather than disease end-points, for the study outcomes (15–19).
In conclusion, this study found that accelerometer-determined PA is superior to self-reported PA in terms of understanding the impact of PA on risk of diabetes and hypertension, Although several decades of research using self-reported PA have provided the basis for the widespread acceptance of physical activity as beneficial for disease prevention and health promotion, the current study suggests that the greater precision offered by objective measurement of physical activity may provide more specific and accurate estimates of those benefits. With the elimination of recall bias, and the capacity to measure sedentary behavior and light activity, as well as MVPA, accelerometry has the potential to contribute information and insights needed for establishing new targets for sedentary and physical activity behavior and shaping effective interventions. The addition of accelerometer measures to established cohort studies may finally help us realize the promise of physical activity for reducing the burden of cardiometabolic disease in the population.
Supplementary Material
1. Supplemental Digital Content 1. Table presenting risk of Stage 1 and greater hypertension associated with self-reported and accelerometer-assessed physical activity, stratified by gender
2. Supplemental Digital Content 2, 3, and 4. Tables presenting risk of diabetes, stage 2 and greater hypertension, and stage 1 and greater hypertension, respectively, associated with self-reported and accelerometer-assessed physical activity, combining men and women
ACKNOWLEDGEMENTS:
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). Accelerometer data collection was supported by grants R01 HL078972 from the NHLBI. This manuscript has been reviewed by CARDIA for scientific content.
The authors thank the investigators, the staff, and the participants of the CARDIA study for their valuable contributions.
The results of this 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.
Footnotes
CONFLICT OF INTEREST:
The authors have no conflicts to declare.
REFERENCES
- 1.U.S.Department of Health and Human Services. Physical Activity Guidelines for America, 2008. Washington DC, US Deparatment of Health and Human Services, 2009. [Google Scholar]
- 2.2018 Physical Activity Guidelines Advisory Committee. 2018 Physical Activity Guidelines Advisory Committee Scientific Report. Washington, DC, U.S. Department of Health and Human Services; 2018. [Google Scholar]
- 3.Haskell WL. Physical activity by self-report: a brief history and future issues. J Phys Act Health 2012;9 Suppl 1:S5–10. [DOI] [PubMed] [Google Scholar]
- 4.Troiano RP, Pettee Gabriel KK, Welk GJ, Owen N, Sternfeld B. Reported physical activity and sedentary behavior: why do you ask? J Phys Act Health 2012;9 Suppl 1:S68–S75. [DOI] [PubMed] [Google Scholar]
- 5.Adams SA, Matthews CE, Ebbeling CB et al. The effect of social desirability and social approval on self-reports of physical activity. Am J Epidemiol 2005;161:389–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Ainsworth BE. Issues in the assessment of physical activity in women. Res Q Exerc Sport 2000;71:S43–S46. [PubMed] [Google Scholar]
- 7.Ainsworth BE, Caspersen CJ, Matthews CE, Masse LC, Baranowski T, Zhu W. Recommendations to improve the accuracy of estimates of physical activity derived from self report. J Phys Act Health 2012;9 Suppl 1:S76–S84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Altschuler A, Picchi T, Nelson M, Rogers JD, Hart J, Sternfeld B. Physical activity questionnaire comprehension: lessons from cognitive interviews. Med Sci Sports Exerc 2009;41:336–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Troiano RP. A timely meeting: objective measurement of physical activity. Med Sci Sports Exerc 2005;37:S487–S489. [DOI] [PubMed] [Google Scholar]
- 10.Arem H, Keadle SK, Matthews CE. Invited commentary: meta-physical activity and the search for the truth. Am J Epidemiol 2015;181:656–8. [DOI] [PubMed] [Google Scholar]
- 11.Dowd KP, Szeklicki R, Minetto MA et al. A systematic literature review of reviews on techniques for physical activity measurement in adults: a DEDIPAC study. Int J Behav Nutr Phys Act 2018;15:15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Pettee Gabriel KK, Morrow JR Jr., Woolsey AL. Framework for physical activity as a complex and multidimensional behavior. J Phys Act Health 2012;9 Suppl 1:S11–S18. [DOI] [PubMed] [Google Scholar]
- 13.Celis-Morales CA, Perez-Bravo F, Ibanez L, Salas C, Bailey ME, Gill JM. Objective vs. self-reported physical activity and sedentary time: effects of measurement method on relationships with risk biomarkers. PLoS One 2012;7:e36345. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Alessa HB, Chomistek AK, Hankinson SE et al. Objective Measures of Physical Activity and Cardiometabolic and Endocrine Biomarkers. Med Sci Sports Exerc 2017;49:1817–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Belcher BR, Moser RP, Dodd KW, Atienza AA, Ballard-Barbash R, Berrigan D. Self-Reported Versus Accelerometer-Measured Physical Activity and Biomarkers Among NHANES Youth. J Phys Act Health 2015;12:708–16. [DOI] [PubMed] [Google Scholar]
- 16.Atienza AA, Moser RP, Perna F et al. Self-reported and objectively measured activity related to biomarkers using NHANES. Med Sci Sports Exerc 2011;43:815–21. [DOI] [PubMed] [Google Scholar]
- 17.Tucker JM, Welk GJ, Beyler NK, Kim Y. Associations Between Physical Activity and Metabolic Syndrome: Comparison Between Self-Report and Accelerometry. Am J Health Promot 2016;30:155–62. [DOI] [PubMed] [Google Scholar]
- 18.Stamatakis E, Davis M, Stathi A, Hamer M. Associations between multiple indicators of objectively-measured and self-reported sedentary behaviour and cardiometabolic risk in older adults. Prev Med 2012;54:82–7. [DOI] [PubMed] [Google Scholar]
- 19.Sabia S, Cogranne P, van Hees VT et al. Physical activity and adiposity markers at older ages: accelerometer vs questionnaire data. J Am Med Dir Assoc 2015;16:438–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Pettee GK, Perez A, Jacobs DR Jr, Lee J, Kohl HW, III, Sternfeld B. The Utility and Cross-Validation of a Composite Physical Activity Score in Relation to Cardiovascular Health Indicators: Coronary Artery Risk Development in Young Adults. J Phys Act Health 2018;15:847–56. [DOI] [PubMed] [Google Scholar]
- 21.Yates T, Henson J, Khunti K et al. Effect of physical activity measurement type on the association between walking activity and glucose regulation in a high-risk population recruited from primary care. Int J Epidemiol 2013;42:533–40. [DOI] [PubMed] [Google Scholar]
- 22.Friedman GD, Cutter GR, Donahue RP et al. CARDIA: study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol 1988;41:1105–16. [DOI] [PubMed] [Google Scholar]
- 23.Troiano RP, Berrigan D, Dodd KW, Masse LC, Tilert T, McDowell M. Physical activity in the United States measured by accelerometer. Med Sci Sports Exerc 2008;40:181–8. [DOI] [PubMed] [Google Scholar]
- 24.Jacobs DR Jr., Hahn LP, Haskell WL, Pirie P, Sidney S. Valdity and reliability of short Physical Activity History: CARDIA and the Minnesota Heart Health Program. J Cardiopulm Rehabil 1989;9:448–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Gabriel KP, Sidney S, Jacobs DR Jr, et al. Convergent validity of a brief self-reported physical activity questionnaire. Med Sci Sports Exerc 2014;46:1570–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Jacobs DR Jr, Ainsworth BE, Hartman TJ, Leon AS. A simultaneous evaluation of 10 commonly used physical activity questionnaires. Med Sci Sports Exerc 1993;25:81–91. [DOI] [PubMed] [Google Scholar]
- 27.Sidney S, Jacobs DR Jr., Haskell WL et al. Comparison of two methods of assessing physical activity in the Coronary Artery Risk Development in Young Adults (CARDIA) Study. Am J Epidemiol 1991;133:1231–45. [DOI] [PubMed] [Google Scholar]
- 28.Whitaker KM, Pereira MA, Jacobs DR Jr., , Sidney S, Odegaard AO Sedentary Behavior, Physical Activity, and Abdominal Adipose Tissue Deposition. Med Sci Sports Exerc 2017;49:450–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Carnethon MR, Sternfeld B, Schreiner PJ 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. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.American College of Cardiology. Web Site [Internet]. New ACC/AHA High Blood Pressure Guidelines Lower Definition of Hypertension. [cited 2018 May 7] Available from: www.acc.org.
- 31.Prince SA, Adamo KB, Hamel ME, Hardt J, Gorber SC, Tremblay M. A comparison of direct versus self-report measures for assessing physical activity in adults: a systematic review. Int J Behav Nutr Phys Act 2008;5:56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Zou G. A modified poisson regression approach to prospective studies with binary data. Am J Epidemiol 2004;159:702–6. [DOI] [PubMed] [Google Scholar]
- 33.Rothman KJ. No adjustments are needed for multiple comparisons. Epidemiology 1990;1:43–6. [PubMed] [Google Scholar]
- 34.Rothman K, Greenland S. Modern Epidemiology. Lipincott-Raven, 1998. [Google Scholar]
- 35.Semanik P, Lee J, Manheim L, DiPietro L, Dunlop D, Chang RW. Relationship between accelerometer-based measures of physical activity and the Yale Physical Activity Survey in adults with arthritis. Arthritis Care Res (Hoboken ) 2011;63:1766–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Dyrstad SM, Hansen BH, Holme IM, Anderssen SA. Comparison of Self-reported versus Accelerometer-Measured Physical Activity. Med Sci Sports Exerc 2014;46:99–106. [DOI] [PubMed] [Google Scholar]
- 37.Sabia S, van Hees VT, Shipley MJ et al. Association between questionnaire- and accelerometer-assessed physical activity: the role of sociodemographic factors. Am J Epidemiol 2014;179:781–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Fjeldsoe BS, Winkler EA, Marshall AL, Eakin EG, Reeves MM. Active adults recall their physical activity differently to less active adults: test-retest reliability and validity of a physical activity survey. Health Promot J Austr 2013;24:26–31. [DOI] [PubMed] [Google Scholar]
- 39.Cust AE, Armstrong BK, Smith BJ, Chau J, van der Ploeg HP, Bauman A. Self-reported confidence in recall as a predictor of validity and repeatability of physical activity questionnaire data. Epidemiology 2009;20:433–41. [DOI] [PubMed] [Google Scholar]
- 40.Stamatakis E, Hamer M, Tilling K, Lawlor DA. Sedentary time in relation to cardiometabolic risk factors: differential associations for self-report vs accelerometry in working age adults. Int J Epidemiol 2012;41:1328–37. [DOI] [PubMed] [Google Scholar]
- 41.Parker ED, Schmitz KH, Jacobs DR Jr., Dengel DR, Schreiner PJ. Physical activity in young adults and incident hypertension over 15 years of follow-up: the CARDIA study. Am J Public Health 2007;97:703–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Carnethon MR, Evans NS, Church TS et al. Joint associations of physical activity and aerobic fitness on the development of incident hypertension: coronary artery risk development in young adults. Hypertension 2010;56:49–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Winckers AN, Mackenbach JD, Compernolle S et al. Educational differences in the validity of self-reported physical activity. BMC Public Health 2015;15:1299. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Whitaker KM, Pettee GK, Buman MP et al. Associations of Accelerometer-Measured Sedentary Time and Physical Activity With Prospectively Assessed Cardiometabolic Risk Factors: The CARDIA Study. J Am Heart Assoc 2019;8:e010212. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
1. Supplemental Digital Content 1. Table presenting risk of Stage 1 and greater hypertension associated with self-reported and accelerometer-assessed physical activity, stratified by gender
2. Supplemental Digital Content 2, 3, and 4. Tables presenting risk of diabetes, stage 2 and greater hypertension, and stage 1 and greater hypertension, respectively, associated with self-reported and accelerometer-assessed physical activity, combining men and women