Calculated Metabolic Equivalents: A Tool for Exercise Prescription Among African American Women

Abstract

Background:

Average daily steps (ADS) are a low-technology measurement of activity that is useful for exercise prescription. However, research demonstrates poor validity for ADS as a measure of exercise capability. We present a superior low-technology measure of exercise capability, which is easily applied by practitioners in clinical or nonclinical settings.

Methods:

Based on analysis of baseline data from an intervention study to test a sustainable approach to long-term physical activity improvement for employed African American women, between 2005 and 2008, we examined exercise tolerance metabolic equivalents (METs) and ADS of 158 participants and generated an alternative measure of exercise capacity. We conducted regression analysis to determine the impact of key health indicators on exercise capacity and examined associations between our predictive model and true (MET) exercise performance.

Results:

Using our predictive equation, 79.33% of participants were correctly categorized (very high, high, medium) based on our tool, with 10 women (6.67%) mischaracterized by one level higher than actual MET achievement and 21(14.00%) mischaracterized as one category lower than actual MET achievement. In contrast, using ADS alone resulted in 22.15% correctly categorized participants.

Conclusions:

The proposed tool is superior to existing low-technology measures of exercise capacity while retaining strong utility in nonclinical and low-resource settings.

Introduction

African American women are a population at high risk for morbidity related to a sedentary lifestyle.^1,2 Recent findings suggest that community-based, culture-appropriate, and gender-appropriate behavioral interventions can significantly increase physical activity, reduce weight, and reduce other chronic disease risk factors in the short-term among women in this population.³

African American women may have equally high or higher exercise self-efficacy but may be less likely to complete exercise programs.⁴ It is important, then, to consider the specific barriers and incentives for exercise that are specific to African American women and how this translates into program development. In focus groups, African American women cite time issues, family responsibilities, lack of social support, and monetary concerns as barriers to exercise.^5

–8 The few studies that have targeted increasing physical activity for African American women indicate that barriers may be overcome through culturally appropriate messages that focus on increased general activity rather than leisure time activity.^9,10 Based on the aforementioned barriers to exercise, it may be easier for working African American women to include exercise in ongoing activities at worksites, religious/community institutions, and other community activities in which they are already participating. However, programs in nonclinical settings may face technical limitations in ability to collect data to track patient need and progress.

Metabolic equivalents (METs) are commonly used for exercise prescription. Although measuring oxygen consumption to calculate METs via a treadmill stress test is the gold standard for establishing an individual's exercise capacity, many low-resource and nonclinical environments do not have the capability of using this measure. This suggests that a simple measure, such as average daily steps (ADS), may be a more attractive tool for measuring and increasing cardiovascular endurance. Multiple studies have demonstrated, however, that ADS/pedometry data are a poor substitute for METs and face significant validity and reliability concerns.^11
–13 Beyond ADS, existing alternatives in the literature include using an age-based nomogram for exercise capacity, but this measure is also quite limited.¹⁴ One study examined the utility of combining multiple regression models and step counts (ADS) to predict maximum oxygen consumption (Vo _2max) among Japanese women, finding validity for this process among this population.¹⁵ The variables included in the study by Cao et al.¹⁵ guide our multivariate analysis in this article.

The purpose of this analysis was to evaluate a new measure of exercise capability that requires little or no technology and is easily used by practitioners in nonclinical settings. We also thought it important to try to design a low-technology alternative to a treadmill stress test, which may not be feasible outside of a clinical setting or in programs with limited resources that might target African American women. To this end, we generated a new clinical tool that can be used for goal setting and tracking of physical activity in African American women. We hypothesize that a simple regression equation can be equally or more effective than existing methods (i.e., ADS or nomograms) at estimating exercise capability for African American women in nonclinical settings.

Materials and Methods

Study setting and design

This article is a cross-sectional analysis of data that were collected between 2005 and 2008 in the course of the P.R.I.S.E.^® fitness study (described below). Designed to test a sustainable approach to long-term physical activity improvement for employed African American women, P.R.I.S.E.^® was conducted by Emory University in collaboration with the Grady Health System in Atlanta and Meharry Medical College in Nashville, Tennessee, and consisted of a series of related components: PREPS (preparedness through individual and group counseling), REPS (weight training), INCREASED STEPS (increasing physical activity through walking/steps), and ENCOURAGEMENT (social support mechanisms). The study consisted of two phases. In phase I, we recruited 179 employees at Grady Health System, a large public hospital and health system in metro-Atlanta, where 80% of employees are African American. In phase II, we partnered with Meharry Medical College, a predominantly African American medical campus in Nashville, and recruited an additional 56 participants for the study. At baseline, women who indicated their interest in participating in the P.R.I.S.E.^® study were administered a battery of tests and instruments. Women were eligible for the study if they were between the ages of 18 and 55 years, currently walking an average of <10,000 steps per day or who performed below expectations on the stress test, had no history of negative cardiac events, and were employed by the relevant health systems. For the purposes of this analysis, we included data on all African American prospective participants with baseline treadmill data who did not have missing baseline ADS (n=158). We included those who demonstrated ADS >10,000 to account for variation in starting exercise capacity.

Data collection

All eligible enrollees completed an in-depth questionnaire that included health and fitness history, including questions on medical history, energy balance/weight loss history, previous physical activity, and dietary habits. Notably, participants were asked how many previous weight loss attempts they had made and completed an 11-question scale on weight management efficacy and a 20-question scale on exercise confidence. For both scales, if >30% of responses were missing, we imputed the mean of the scale. We also collected demographic information and personality and psychosocial measures shown to influence physical activity. At baseline, anthropometric measurements were taken by research nursing staff; weight was measured in pounds on a standard medical scale; height was measured in inches with a standard height rod.

Participants also participated in an exercise treadmill test (ETT) on a standard electronic treadmill at baseline, using the P.R.I.S.E.® protocol, a graded exercise protocol, which was a refinement of the Naughton protocol. Our P.R.I.S.E.^® protocol increased speed or elevation by ∼1 MET per stage, in accordance with workload increases that are appropriate for deconditioned individuals. The ETT was conducted by an American College of Sports Medicine (ACSM) Registered Clinical Exercise Physiologist® or by trained nurses in a hospital environment. A 12-lead electrocardiogram was measured before, during, and after the ETT. Women were asked to walk until they felt they needed to stop but not to exhaustion, which correlated with a perceived exertion of 9 on a scale of 1–10. The ETT was halted if vital signs or orthopedic limitations warranted. Baseline information on physical activity was obtained by providing a Yamax Digiwalker pedometer (model SW 200, considered top of the line for ADS validity at the time of the study) to the enrollees, who were instructed to record their total steps every day in a diary over a 2-week period. These step logs were reviewed for completeness (they were considered complete if the participant filled out at least 5 days per each 7 day week) and used to calculate ADS for each study subject by dividing the total reported daily steps by the number of days for which daily steps were reported.

All study protocols were approved by the respective IRBs at Emory University and Meharry Medical College, and written informed consent was obtained from all study participants using standard procedures.

Statistical analyses

We examined exercise tolerance (METs) and ADS. We generated categorical variables (low, medium, high, very high) for the METs and ADS variables. The categories for METs follow the ACSM guidelines in which low METs=0–3, medium MET achievement=3.1–6.0, high MET achievement=6.1–10.0, and very high MET achievement->10). Similarly, we segmented the ADS variable into low (0–5000 steps), medium (5001–8000), high (8001–10000), and very high (>10,000). We further generated a variable based solely on age (an existing nomogram for predicting METs) to generate expectations of physical activity. We ran initial correlations among the three key outcome variables to determine how closely related each proxy measure (predicted METs and ADS) was to actual MET achievement.

Upon determining poor correlations and sensitivity/specificity for these measures, we designed an alternative measure that could still be used easily in a nonclinical setting, with easy to understand counseling recommendations for our target population. We conducted ordinary least squares regression to determine the significant impact of several key health indicators, which can be readily assessed in nonclinical environments, on the prediction of METs achievement. These include body mass index (BMI), if the individual reported exercising regularly (response to a question on the baseline survey which asks: Do you exercise regularly?), and if a physician had recommended additional physical activity (self-report). We also included age (used in the predicted METs nomogram). Because many of these covariates appear similar, we ran correlation coefficients and found only low correlations, not causing us to remove one or more from the regression equation.

The aforementioned regression model was used to create an equation for a new variable: estimated METs. We tested correlation of our estimated METs with actual METs and measured predictive accuracy of the model.

Results

The sample population for this analysis consisted of 158 African American women between the ages of 18 and 55 who were employed at two health systems. Table 1 details the demographics of this population. Initial correlation matrices indicated that the two recognized proxies for METs, ADS, and the predicted METs nomogram derived from age actually demonstrated low correlations (0.24 and 0.26, respectively) with achieved METs.

Table 1.

Sample Demographics and Characteristics (n=158)

	n (%)
Age
20s	29 (18.35)
30s	52 (32.91)
40s	65 (41.14)
50s	12 (7.59)
Mean (SD)	38.27 (8.36)
Range	20–55
Marital status
Married/living with partner	69 (43.95)
Single	53 (33.76)
Separated/divorced/widowed	35 (33.76)
BMI
Underweight (<18.5)	–
Normal weight (18.5–<25)	15 (9.49)
Overweight (25–<30)	46 (29.11)
Obese (≥30)	97 (61.39)
Mean (SD)	34.15 (8.14)
Range	19.85–63.59
Mean no. of weight loss attempts	2.70 (1.72)
Weight efficacy	3.57 (0.63)
Exercise confidence	3.83 (0.72)
Exercise regularly	40 (25.97)
Baseline ADS
Low (0–5000)	50 (31.65)
Medium (5001–8000)	70 (44.30)
High (8001–10,000)	23 (14.56)
Very high (>10,000)	15 (9.49)
Mean (SD)	6342.36 (2597.38)
Range	440.14–14076.62
Actual METs
Low (0–3.0)	–
Medium (3.1–6.0)	23 (14.56)
High (6.1–10.0)	112 (70.89)
Very high (>10.0)	23 (14.56)
Mean (SD)	7.86 (1.93)
Range	3.45–12.90
Predicted METs
Low (0–3.0)	–
Medium (3.1–6.0)	–
High (6.1–10.0)	93 (58.86)
Very high (>10.0)	65 (41.14)
Mean (SD)	9.72 (1.09)
Range	7.55–12.10

ADS, average daily steps; BMI, body mass index; METs, metabolic equivalents; SD, standard deviation.

As shown in Table 2, a variety of easy to measure predictive variables are significant predictors of actual MET performance. Each of these measures—BMI, age, and whether the individual exercises regularly—is a significant predictor of METs achieved in a treadmill stress test. The final regression model for estimated METs is: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*}{\rm Y}_{\rm estimated \ METs} = \beta_0 (13.83) + (- 0.14 \ \beta_{\rm 1bmi}) + (- 0.06 \ \beta_{\rm 2age}) + (0.63 \ \beta_{\rm 3exercise\_reg}) + {\rm C}_{\rm i} \end{align*} \end{document}

Table 2.

Regression Results

Variable	Coefficient	SE	t value	95% CI
BMI	−0.137	0.015	−9.11^***	−0.17–−0.11
Age	−0.060	0.014	−3.87^***	−0.08–−0.03
Exercise regularly	0.633	.271	2.33^*	0.10–1.17

***

p<0.001; ^** p<0.01; ^* p<0.05; R ²=0.43.

CI, confidence interval; SE, standard error.

where BMI=body mass index, age=continuous variable of year of age, Exercise_reg=self-report of whether participant exercises regularly.

Notably, we did not use ADS in the model. Although ADS is widely used in community interventions, the existing literature and our own data demonstrate that ADS is not a significant predictor of exercise capability. In fact, including ADS in the model (data not shown) actually decreased the precision of the estimates. The R ² coefficient on this equation was 0.43, suggesting that nearly one half of the variance in MET achievement is explained by these measures. Using the equation to create the estimated METs variable, we found a correlation coefficient of 0.66 between estimated METs and actual MET achievement.

As shown in Table 3, using our predictive equation, 79.33% of participants were correctly categorized based on our estimated METs equation, with 10 women (6.67%) mischaracterized by one level higher than actual MET achievement and 21 (14.0%) mischaracterized as one category lower than actual MET achievement. Notably, women who were categorized higher than actual achievement level (n=10) were more likely (90%) to be obese than those properly categorized (37.8%). Among women who were categorized lower than actual achievement level (n=21), BMI was not significant. Using ADS or the predicted METs nomogram (age) alone resulted in only 22.15% and 51.27% of correctly categorized participants, respectively. Notably the predicted METs nomogram (based solely on age) suggests that everyone is in the high or very high category, making no allowances for the actual circumstances of the participants. Therefore, in a population with more diverse health issues, using predicted METs for exercise prescription could be problematic because of a lack of variance. In spite of their practical utility, the predictive values of ADS and predicted METs suggest that use of these low-technology measures alone is not an appropriate substitution for calculating METs via a treadmill test.

Table 3.

Estimated Metabolic Equivalents Achievement Compared to Actual Metabolic Equivalents Achievement

Estimated METs category
Actual METs achieved	Medium	High	Very high	Total
Medium	11 (52.38%)	10 (47.62%)	0 (0%)	21
High	3 (2.80%)	104 (97.20%)	0 (0%)	107
Very high	0 (0%)	18 (81.82%)	4 (18.18%)	22
Total	14	132	4	150

Discussion

Measuring METs via a treadmill stress test is the gold standard for establishing an individual's exercise capacity, but many low-resource and nonclinical environments do not have the capability of using this measure. For African American women, who demonstrate additional barriers to physical activity,^6,7 it is increasingly important to be able to create physical activity interventions in the workplace, religious communities, social centers, and other locations where these women already spend time.

Previous literature has already well established the need for and importance of practical tools that can be used in diverse settings. To this end, the measure of ADS is widely used^{16

–23} and recognized for its ease of measurement and comprehensibility for participants. ADS is also a valuable tool for exercise prescription, providing a metric that is easy to comprehend, easy to measure via a relatively inexpensive pedometer, and easy to adapt as participants increase their exercise tolerance. Yet ADS has been shown to be a poor proxy measure for exercise capability and MET achievement.

For similar purposes, previous researchers generated a predictive METs nomogram based solely on the participant's age.²⁴ Although we find some value in this predictive tool as well, it has limited accuracy in its predictions, as age is not the only factor that determines exercise capability. This is particularly true given that the nomogram tends to group everyone of working age into the high and very high categories, leaving little accommodation for those who are demonstrating low exercise capability. Myers et al.²⁵ added the Veterans Specific Activity Questionnaire (VSAQ), a self-assessment of exercise capability, to generate a second generation nomogram and found a superior correlation. However, this tool is also limited, as it is designed for patients referred for exercise counseling because of clinical needs and is not, therefore, appropriate for the general population. The VSAQ may also not be applicable for underfit African American patients, as there were none in this sample. Further, the VSAQ has been largely validated on men. Similarly, the multiple regression to predict Vo _2max among Japanese women, has not been validated for any other population but does provide some insight into variables that should be considered for this methodology. However, whereas Cao et al.¹⁵ also use step measures (ADS), we did not find this to be significantly predictive among our population.

Our analysis confirms the poor correlations and predictive accuracy of ADS and predicted METs based on age. Therefore, it was our goal with further analysis to create a similarly useful tool with better accuracy. Our predictive tool, estimated METs, is superior to either of the existing measures currently available and has been shown to be predictive in a sample of African American women. The components of the tool—BMI, age, and whether the individual exercises regularly (self-reported)—are all measures commonly collected in most intervention environments. Intervention staff members can generate the estimated METs value for the individual with these data with the aid of a simple calculator, following the equation: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} \hbox{Estimated METs if participant reports regular exercise} = 13.83 - (0.14 \ {\rm BMI}) - (0.06 \ {\rm age} \ ({\rm in \ years})) + 0.63 \end{align*} \end{document}

Or \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} \hbox{Estimated METs if participant reports not exercising} = 13.83 - (0.14 \ {\rm BMI})- (0.06 \ {\rm age} \ ({\rm in \ years})) + 0 \end{align*} \end{document}

With >77% of the sample accurately categorized, our tool outperforms both ADS and predicted METs by a substantial margin. Of the 20.67% of participants miscategorized, 6.7% were off by one level over (e.g., medium to high) and 14% were categorized into a lower grouping than they should have been (by one level). Predicting a lower categorization is less problematic than a higher categorization, as it would result in suggesting exercise intensities below the participant's training capacity rather than above it. Therefore, we recognize that for some portion of the population, this tool may be less effective, but it is better to err on the side of safety. Additionally, we note that this study sample comprised employed African American women in healthcare settings in the southeast. The results of this analysis may not be generalizable to African American women who are either unemployed or employed in alternate settings. Similarly, both our study locations were in the southeastern United States, and results may not be generalizable to other geographic regions. Finally, we recognize that our sample for this analysis is relatively small. In spite of these limitations, our analyses suggest that the proposed estimated METs variable may be an improvement on existing tools for measuring exercise capacity when a treadmill stress test is not an option.

Conclusions

Given the need for a low-technology measurement of exercise capacity among African American women in community-based interventions, we believe the tool proposed herein has significant potential merit. Grouping of the metric into high, medium, and low categories is further useful to prescribe alternate activities of similar metabolic intensities. We believe this tool is a positive step in measuring exercise capacity without treadmill resources and should be further examined in effectiveness research trials in the future.

Footnotes

Acknowledgment

This research was funded by Centers for Disease Control and Prevention R01 DP000106-02 and the United States Department of Agriculture 2006-55215-16692.

Disclosure Statement

No competing financial interests exist.

References

Manson

, Skerrett

, Greenland

, VanItallie

. The escalating pandemics of obesity and sedentary lifestyle. A call to action for clinicians. Arch Intern Med, 2004; 164:249–258.

Jebb

, Moore

. Contribution of a sedentary lifestyle and inactivity to the etiology of overweight and obesity: Current evidence and research issues. Med Sci Sports Exerc, 1999; 31,Suppl 11:S534–541.

Banks-Wallace

, Conn

. Interventions to promote physical activity among African American women. Public Health Nurs, 2002; 19:321–335.

Wilbur

, Miller

, Chandler

, McDevitt

. Determinants of physical activity and adherence to a 24-week home-based walking program in African American and Caucasian women. Res Nurs Health, 2003; 26:213–224.

French

, Jeíery

, Story

, Neumark-Sztainer

. Perceived barriers to and incentives for participation in a weight-loss program among low-income women in WIC. J Am Diet Assoc, 1998; 98:79–81.

Carter-Nolan

, Adams-Campbell

, Williams

. Recruitment strategies for black women at risk for noninsulin-dependent diabetes mellitus into exercise protocols: A qualitative assessment. J Natl Med Assoc, 1996; 88:558–562.

Young

, Gittelsohn

, Charleston

, Felix-Aaron

, Appel

. Motivations for exercise and weight loss among African-American women: Focus group results and their contribution towards program development. Ethnicity Health, 2001; 6:227–245.

Kirchhoff

, Elliott

, Schlichting

et al. Strategies for physical activity maintenance in African American women. Am J Health Behav, 2008; 32:517–524.

Buchholz

, Artinian

. Dimensions of physical activity in African American women. Health Care Women Int, 2009; 30:308–323.

10.

Whitt

, DuBose

, Ainsworth

, Tudor-Locke

. Walking patterns in a sample of African American, Native American, and Caucasian women: The Cross-Cultural Activity Participation Study. Health Educ Behav, 2004; 31,Suppl 4:45S–456S.

11.

Saris

, Binkhorst

. The use of pedometer and actometer in studying daily physical activity in man. Part I: Reliability of pedometer and actometer. Eur J Appl Physiol Occup Physiol, 1977; 37:219–228.

12.

Tudor-Locke

, Sisson

, Lee

et al. Utility of pedometers for assessing physical activity: Convergent validity. Sports Med, 2002; 32:795–808.

13.

Tudor-Locke

, Sisson

, Lee

et al. Utility of pedometers for assessing physical activity: Construct validity. Sports Med, 2004; 34:281–291.

14.

Gulati

, Black

, Shaw

et al. The prognostic value of a nomogram for exercise capacity in women. N Engl J Med, 2005; 353:468–475.

15.

Cao

, Miyatake

, Higuchi

, Ishikawa-Takata

, Miyachi

, Tabata

. Prediction of Vo _2max with daily step counts for Japanese adult women. Eur J Appl Physiol, 2009; 105:289–296.

16.

Ayabe

, Brubaker

, Dobrosielski

et al. Target step count for the secondary prevention of cardiovascular disease. Circ J, 2008; 72:299–303.

17.

Cavanaugh

, Coleman

, Gaines

, Laing

, Morey

. Using step activity monitoring to characterize ambulatory activity in community-dwelling older adults. J Am Geriatr Soc, 2007; 55:120–124.

18.

Croteau

. A preliminary study on the impact of a pedometer-based intervention on daily steps. Am J Health Promot, 2004; 18:217–220.

19.

Ewald

, Duke

, Thakkinstian

, Attia

, Smith

. Physical activity of older Australians measured by pedometry. Aust J Ageing, 2009; 28:127–133.

20.

Felton

, Tudor-Locke

, Burkett

. Reliability of pedometer-determined free-living physical activity data in college women. Res Q Exerc Sport, 2006; 77:304–308.

21.

Flohr

, Todd

, Tudor-Locke

. Pedometer-assessed physical activity in young adolescents. Res Q Exerc Sport, 2006; 77:309–315.

22.

Speck

, Jaster

, Pronk

, Becker

, Schmitz

. Assessment and outcomes of HealthPartners 10,000 Steps(R) program in an academic work site. Health Promot Pract, 2010; 5:741–750.

23.

Villanova

, Pasqui

, Burzacchini

et al. A physical activity program to reinforce weight maintenance following a behavior program in overweight/obese subjects. Int J Obes (Lond), 2006; 30:697–703.

24.

Morris

, Myers

, Froelicher

, Kawaguchi

, Ueshima

, Hideg

. Nomogram based on metabolic equivalents and age for assessing aerobic exercise capacity in men. J Am Coll Cardiol, 1993; 22:175–182.

25.

Myers

, Bader

, Madhavan

et al. Validation of a specific activity questionnaire to estimate exercise tolerance in patients referred for exercise testing. Am Heart J, 2001; 142:1041–1046.