Skip Navigation LinksHome > February 2009 - Volume 41 - Issue 2 > Self-Reported Physical Activity Improves Prediction of Body...
Medicine & Science in Sports & Exercise:
doi: 10.1249/MSS.0b013e318185d359
Basic Sciences

Self-Reported Physical Activity Improves Prediction of Body Fatness in Young Adults

ZANOVEC, MICHAEL1; JOHNSON, LISA G.2; MARX, BRIAN D.3; KEENAN, MICHAEL J.4; TUURI, GEORGIANNA1,4

Free Access
Article Outline
Collapse Box

Author Information

1School of Human Ecology, Louisiana State University, Baton Rouge, LA; 2Department of Kinesiology, Louisiana State University, Baton Rouge, LA; 3Department of Experimental Statistics, Louisiana State University, Baton Rouge, LA; and 4School of Human Ecology, Louisiana State University Agricultural Center, Baton Rouge, LA

Address for correspondence: Michael Zanovec, M.S., School of Human Ecology, Louisiana State University, 256 Knapp Hall, Baton Rouge, LA 70803; E-mail: mzanovec@agcenter.lsu.edu.

Submitted for publication May 2008.

Accepted for publication July 2008.

Collapse Box

Abstract

Purpose: To examine self-reported physical activity levels from the International Physical Activity Questionnaire (IPAQ) as an independent predictor of dual-energy x-ray absorptiometry (DXA)-measured percent body fat (%BF) from body mass index (BMI), gender, and race.

Methods: Two hundred and seventy-eight students, aged 18-24 yr, volunteered to participate. There were 133 males (85 white and 48 black) and 145 females (77 white and 68 black). Total activity levels were quantified in MET hours per week (MET-h·wk−1) using the IPAQ short form. Height and weight were measured, and BMI values were calculated (kg·m−2). %BF was assessed using DXA. Linear regression analysis was used to develop and compare a body fat prediction equation with (full) and without (reduced) the variable MET-h·wk−1. Both models included BMI, gender, and race as predictor variables. The prediction sum of squares (PRESS) statistic was used to cross-validate both models, and the individual predictive accuracy was compared using modified Bland-Altman plots.

Results: Mean ± SD values were as follows: BMI = 24.4 ± 4.1 kg·m−2, %BF = 24.5 ± 9.3%, and MET-h·wk−1 = 37.4 ± 21.9. Gender, BMI, and race explained 81% of the variance in %BF, with a root mean square error (RMSE) of 4.07. The full model with MET-h·wk−1 improved the prediction of %BF by 2% (R2 = 0.83, RMSE = 3.87). When cross-validated, the corresponding PRESS statistics for the reduced and full model were 4.10 and 3.90, respectively. Bland-Altman limits of agreement were greater for the reduced model compared with the full model (−8.09, 8.10 vs −7.67, 7.68).

Conclusion: These results suggest that %BF can be predicted with greater precision and accuracy in a young adult population when MET-h·wk−1 are included in addition to BMI, gender, and race.

The rising prevalence of obesity and physical inactivity during adolescence and young adulthood has focused attention on the need for accurate methods to estimate body composition in this population. Despite compelling evidence that a sedentary lifestyle contributes to the development of obesity (36), there continues to be a persistent decline in physical activity (PA), particularly among young adults (29,40). Furthermore, college-aged individuals have been identified as an important population for initiatives addressing lifestyle changes to decrease health risks (2). Because health complications associated with overweight and obesity are related to excess body fatness (40) and low levels of PA have been linked to obesity in adolescence and adulthood (36), it is critical to have accurate field-based prediction equations to estimate %BF that account for weight status and PA when more precise measures such as dual-energy x-ray absorptiometry (DXA) or hydrodensitometry (HD) are not available.

In epidemiological studies, overweight and obesity are typically determined from anthropometric measures of height and weight and calculated body mass index (BMI) (35). The use of BMI as a surrogate measure of adiposity is appealing because minimal equipment is needed and errors in measurement are small (8). Despite its convenience, BMI fails to account for the composition of body weight, which is comprised mainly of fat, lean tissue, and bone mineral (40). Previous studies have confirmed that the relationship between BMI and relative body fatness varies considerably by age, gender, and race/ethnicity (9,12,13). Males generally have less body fat than females, and blacks typically have a greater proportion of body weight as fat-free mass than whites (38). Additional factors including body build (7) and PA (17) have also been shown to influence the relationship between BMI and percent body fat (%BF), but these differences may not be detected when using BMI alone. For instance, in young adults, highly active individuals may have a greater proportion of body weight as lean-tissue mass (LTM) compared with sedentary counterparts at the same weight. Consequently, athletes are often misclassified as obese based on their BMI although their %BF may be well within a healthy range (23). Therefore, it is important to account for variations in PA level when developing body fat prediction equations for young adults.

A critical aspect regarding the accuracy of field methods lies in the use of appropriate regression equations. Field-based prediction equations are typically based on a statistical relationship between easily measurable body parameters (i.e., height, weight, and skinfolds) and body composition component as measured by a reference method such as DXA or HD (6). %BF from DXA has been shown to correlate well with a multicomponent model in young adults (26), particularly because DXA measurements are relatively unaffected by fluctuations in total body water in normal healthy adults (22,39). On the basis of several reviews of studies comparing DXA estimates of %BF to estimates obtained using a multicomponent molecular model, Lohman et al. (19) determined the precision of %BFDXA to be within 1-3% body fat.

Several equations for estimating %BF based on BMI have been developed and validated against DXA or HD for specific youth populations (19,24,25) and for generalized populations of adults (8,9,12,13). In addition, previous studies have examined the relationship between objective measures of PA or physical fitness and body composition (20,21,33). However, there has been far less effort to develop body fat prediction equations specifically for use with college-aged adults of different racial/ethnic groups. Furthermore, to our knowledge, there have been no studies that have examined the use of self-reported PA as an independent predictor of body fatness. Because PA has been shown to affect weight and body composition favorably by promoting fat loss while maintaining or increasing LTM (32,36) and because body composition is known to vary between sedentary and trained young and older males and females (17), it stands to reason that the inclusion of PA in combination with BMI should result in more accurate body fat prediction equations.

Although objective measures of PA are generally considered to be more precise, self-report techniques are the instruments of choice for population studies because they are practical and easy to administer, with relatively low cost and low participant burden (37). Furthermore, the main advantage of self-reports over objective methods such as accelerometers or pedometers is that they do not influence usual PA patterns (37). To obtain reliable self-reports of total PA levels, it is important to select an instrument that has been rigorously tested and is appropriate for the application for which it is intended. The International Physical Activity Questionnaire (IPAQ; www.ipaq.ki.se) was developed by a global working group of PA researchers in an effort to provide a comparable and valid instrument for assessing PA levels across populations and countries (5). This questionnaire has undergone extensive validity and reliability testing across 12 countries and is recommended for use in national population-based prevalence studies among 15- to 69-yr-olds (5). Using this standardized instrument, the aim of the present study was to develop a population-specific body fat prediction equation for young adults and to test the hypothesis that self-reported PA levels quantified from the IPAQ could be used to improve the prediction of DXA-measured %BF from BMI, gender, and race in a biracial young adult college-aged population.

Back to Top | Article Outline

METHODS

Recruitment and Data Collection

Participants were recruited from undergraduate classes and from flyers posted on the university campus with the goal of achieving equal representation within each gender and racial/ethnic group. Subject inclusion criteria were as follows: 1) apparently healthy; 2) white and black males and females; and 3) aged 18-24 yr. Race/ethnicity was self-reported. Exclusion criteria were as follows: 1) subjects over 6 feet 4 inches tall and/or weighing over 250 lb due to height and weight limits of the DXA device; and 2) individuals that reported race/ethnicity as Hispanic, Asian, or other. A total of 278 (48% male, 58% white) students participated and were included in the analysis. The research protocol was approved by the Louisiana State University institutional review board, and all subjects provided written informed consent.

All measurements were collected by one trained investigator (MZ) during a single laboratory session that lasted approximately 1 h. Appointments were scheduled via e-mail, and subjects were advised to refrain from exercising on the day of testing and to not eat at least 3 h prior. Subjects were also asked to wear lightweight, loose-fitting clothing free of metal and to remove all jewelry, hair clips, and shoes during measurements. Standing height was measured without shoes to the nearest 0.5 cm using a portable stadiometer (Shorr Inc., Olney, MD) with the subject's head positioned in the Frankfurt horizontal plane. Body mass was assessed to the nearest 0.1 kg with a digital scale (Seca, Model 880, Hanover, MD) calibrated daily using a 5-kg weight. Body mass index (BMI) values were calculated as weight in kilograms divided by height in meters squared.

Back to Top | Article Outline
Measures
DXA.

Total percent body fat (%BF) estimates were obtained using a full-size Lunar Prodigy Pro (GE Lunar Corporation, Madison, WI) densitometer in conjunction with Encore 2004 software (version 8.10.027). The Prodigy Pro is a narrow-angle (4.5°) transversely scanning fan beam densitometer that uses a cadmium-zinc-telluride detector to directly convert x-rays into an electronic signal. The DXA machine was calibrated daily against a standard calibration block supplied by the manufacturer. The instrument automatically altered scan depth (standard or thick) based on the thickness of the subject as estimated from height and weight. All subjects were scanned in a supine position with the midline of the body centered on the table, arms by their side with palms face down, and feet held together by a Velcro strap. Approximate scan times were 6 and 10 min for standard and thick modes, respectively. All scans were performed and analyzed manually by one trained operator (MZ).

Back to Top | Article Outline
Physical activity.

The IPAQ short form was used to assess subjects' total health-enhancing PA and sedentary activity during the previous 7 d (www.ipaq.ki.se). Participants completed the short self-administered IPAQ version using an interview probe-type format directed by the researcher (MZ). A probe protocol similar to the one used by Rzewnicki et al. (28) was used to minimize the potential for overreporting. Participants were asked to recall the frequency (d·wk−1), duration (minutes), and level of intensity (vigorous, moderate, or walking) of PA undertaken during the previous 7 d within four domains: leisure-time PA, work-related PA, transport-related PA, and domestic and gardening (yard) activities. During the interview, respondents were asked to explain their responses and to give more detailed reports of all activities. Attention was given to the explicit criteria used in the IPAQ scoring protocol such as breathing cues and minimum duration of 10 min for individual bouts of PA. A sample question included, "You said that you were vigorously active 3 d last week for an average of thirty minutes each day. Can you please tell me about that activity?" Probe questions included, "Were these activities performed for at least 10 min consecutively?" and "How was your breathing affected?" Examples were provided for vigorous activity (i.e., heavy lifting or aerobics) and moderate activity (i.e., carrying light loads or bicycling at a regular pace). Subjects were told to think of vigorous intensity as activities that would increase their heart rate to the point where it would be difficult to sustain a conversation with someone. Moderate intensity was described as activities that required somewhat higher than normal breathing and did not include walking. The IPAQ included an additional question that asked about time spent sitting during a weekday; however, this question was not included in the analysis.

On the basis of the responses, each subject's total PA level was calculated and recorded in MET minutes per week (MET-min·wk−1) according to the IPAQ scoring protocol (www.ipaq.ki.se). The MET levels derived from the IPAQ validity and reliability study (5) were 8.0, 4.0, and 3.3 for vigorous-intensity, moderate-intensity, and walking activities, respectively. MET-min·wk−1 were computed as follows: MET level × minutes of activity per day × days per week. These values were recoded into MET hours per week (MET-h·wk−1) by dividing MET-min·wk−1 by 60 min. A detailed description of the IPAQ scoring protocol including the criteria for truncating extreme values is provided on-line (www.ipaq.ki.se).

Back to Top | Article Outline
Statistical Analysis
Description of sample.

There were 133 males (85 white and 48 black) and 145 females (77 white and 68 black). The physical characteristics of the subjects (mean, SD, and range) were calculated for the total sample and separately for males and females. Differences between racial/ethnic groups were examined using independent t-tests. Pearson's correlation coefficients were calculated to investigate the relations of %BF with body composition variables and self-reported PA in MET-h·wk−1. Data were examined using SAS (v. 9.1.3; SAS Institute, Inc., Cary, NC). The level of significance was set at P < 0.05.

Back to Top | Article Outline
Regression analysis.

Multiple linear regression analysis was used to describe the strength of association of DXA-measured %BF as the dependent variable with BMI, MET-h·wk−1, and their interaction terms as predictor variables. Gender and race were included as dichotomous variables to increase the total degrees of freedom and to account for gender and racial/ethnic differences in %BF. We applied stepwise regression techniques to assist in the selection of variables for the final multivariate models using the default criteria (α = 0.15). For this study, we chose not to evaluate alternative models for estimating %BF that included quadratic terms or linear transformations of BMI to simplify comparisons between models.

Back to Top | Article Outline
Model comparison.

A "reduced model" was developed to describe the strength of association of DXA-measured %BF as the dependent variable with BMI, gender, race, and their interaction terms as predictor variables. Also of interest was whether a "full model" that also included MET-h·wk−1 performed better than the "reduced model" without this variable. We examined this 1) by comparing the group predictive accuracy based on measures of goodness-of-fit statistics, including the R2 values adjusted for the degrees of freedom and the root mean square error (RMSE); 2) by comparing the precision of each model based on the total error (TE) of prediction obtained using the prediction sum of squares (PRESS) statistic (14); and 3) by comparing the within-group individual predictive accuracy of each equation by constructing modified Bland-Altman plots using the residual scores (measured − predicted) as a function of the criterion measure DXA %BF, with 95% limits of agreement based on the RMSE (4). Finally, the residuals from each model were correlated with DXA-measured %BF to determine whether the equations tended to under- or overestimate values at either end of the distribution for %BF.

Back to Top | Article Outline

RESULTS

The physical characteristics of the sample are provided in Table 1. The large variation in %BF (5% to 49%) indicates the diversity of the sample. Males had higher mean BMI values, lower %BF, and higher self-reported PA levels than females. Black males had higher BMI values than white males, and black females had greater body weight, BMI, fat mass (FM), lean-tissue mass (LTM), and bone mineral content (BMC) than white females.

Table 1
Table 1
Image Tools

The correlation coefficients for %BF versus body composition variables and MET-h·wk−1 are shown in Table 2. %BF was significantly associated with height in black females only. %BF was highly correlated with weight, BMI, and FM in all four subgroups. In white females, %BF was significantly related to BMC. %BF was negatively associated with MET-h·wk−1 in all groups, with coefficients ranging from −0.25 in white females to −0.39 in black males. Finally, the association between BMI and MET-h·wk−1 was −0.02 for white males and females, −0.31 (< 0.05) for black males, and −0.20 for black females (data not shown). A scatterplot of MET-h·wk−1 versus %BF is presented in Figure 1. It shows considerable variation between self-reported PA and level of body fat, with a tendency for males to report higher levels of PA and to have less %BF at any given level of PA. The correlation between MET-h·wk−1 and %BF for the total study sample was −0.44.

Table 2
Table 2
Image Tools
FIGURE 1-Scatterplot...
FIGURE 1-Scatterplot...
Image Tools

On the basis of the associations of %BF with BMI and MET-h·wk−1, multiple linear regression analyses were used to determine the impact of PA on the prediction of %BF from DXA. A preliminary analysis that included BMI, MET-h·wk−1, gender, race, and their interaction terms revealed no significant interactions among the predictor variables; therefore, only main effects are reported. A comparison of parameter estimates and cross-validation statistics for a "reduced model" without MET-h·wk−1 versus a "full model" with this variable is presented in Table 3. For both models, gender emerged as the most influential variable, accounting for 41% of the explained variance. The addition of BMI increased the adjusted R2 value to 0.81 and reduced the RMSE from 7.16 to 4.09. For the "reduced model," race was not significant (P = 0.06), although this variable was retained by the stepwise procedure. The overall adjusted R2 value and the RMSE for the "reduced model" with three predictor variables were 0.81 and 4.07, respectively. After internal cross-validation using the PRESS statistic, the TE of prediction for this model increased from 4.07% to 4.10%.

Table 3
Table 3
Image Tools

In comparison to the "reduced model," the "full model" with the additional independent variable MET-h·wk−1 improved the adjusted R2 from 0.81 to 0.83 and reduced the RMSE from 4.07 to 3.87 (Table 3). The standardized regression coefficients for MET-h·wk−1 and race were −0.15 and −0.06, respectively, thereby indicating that self-reported PA was a stronger contributor to the overall explained variance in %BF. Furthermore, the additional variance explained by the "full model" resulted in significant regression coefficients for all four predictor variables, including race, which slightly reduced the RMSE from 3.89 to 3.87. When cross-validated, the TE of prediction increased marginally from 3.87% to 3.90%. Overall, the "full model" performed better than the "reduced model" in terms of both goodness of fit (R2adj = 0.83 vs 0.81, RMSE = 3.87 vs 4.07) and cross-validation (PRESS = 3.90 vs 4.10). Additionally, there were no apparent problems observed with multicollinearity, and neither model violated the general assumptions of homogeneity of variance and normality.

To illustrate and compare the accuracy of each model in predicting %BF for individuals within the group, we constructed modified Bland-Altman plots with the residuals (measured − predicted) on the y-axis as a function of the criterion measure %BF from DXA on the x-axis. A plot of the residuals versus DXA-measured %BF is shown in Figure 2 for the "reduced model" and in Figure 3 for the "full model." The constant error or the average difference between the predicted and the actual means was zero for both models, indicating a lack of systematic bias in either equation. The correlation coefficient between the residuals and the %BF from DXA was 0.43 for the reduced model compared with 0.41 for the full model (P < 0.01 for both). These values indicate that both equations tended to overestimate %BF at the lower end and to underestimate %BF at the upper end, with a greater trend observed for the reduced model. Furthermore, the 95% limits of agreement were greater for the reduced model (−8.09, 8.09) compared with the full model (−7.68, 7.68), indicating greater individual predictive accuracy. Upon visual inspection, males were more likely to be outside the limits of agreement in both models. There were 16 subjects (13 males, 3 females) outside the limits of agreement based on the residuals from the reduced model (Fig. 2) compared with 9 subjects (8 males, 1 female) for the full model (Fig. 3).

FIGURE 2-Bland-Altma...
FIGURE 2-Bland-Altma...
Image Tools
FIGURE 3-Bland-Altma...
FIGURE 3-Bland-Altma...
Image Tools
Back to Top | Article Outline

DISCUSSION

The results of this study demonstrate that %BF can be accurately predicted in young adult college students from an equation based upon BMI, gender, and race and that the inclusion of daily PA improves the precision and the accuracy of the prediction. To our knowledge, this is the first study to use self-reported PA to predict body fatness in this population. Increasing PA and prevention of obesity are listed as the top two priority health indicators of the Healthy Campus 2010 initiative (2). Identifying sedentary and obese individuals by predicting body fatness from simple measures (i.e., height, weight, and self-reported PA) may provide a unique perspective for evaluating progress toward meeting the Healthy Campus 2010 goals.

In this study, a standardized instrument was used to estimate PA, and the results obtained are consistent with previous literature of exercise patterns in college students (2,16). In a review of college students' PA levels, Irwin (16) reported that half of university students were not sufficiently active (e.g., meeting current American College of Sports Medicine guidelines). In 2006, the American College Health Association found that only 44.2% of students reported exercising vigorously for 20 min or moderately for 30 min on three of the previous 7 d (2). Although the threshold used to define sufficient activity may differ from previous studies, the IPAQ identified 50% of students as not meeting the criteria for "sufficient total activity" consistent with current recommendations for health-enhancing physical activity (HEPA) (31). Using the IPAQ scoring protocol, students were classified as sufficiently active if they reported 3000 MET-min·wk−1 of total activity or 1500 MET-min of vigorous intensity (8 METs) on three or more of the previous 7 d. The HEPA threshold is meant to approximate to the amount of PA consistent with ≥10,000 steps per day (34).

BMI is a simple, easy-to-use method of estimating obesity in population-based studies. Several other studies (12,13) have developed generalized body fat prediction equations for adults based on BMI, and a meta-analysis has been done (9). Gallagher et al. (13) reported an R2 of 0.67 and an SEE of 5.68% when combining BMI with age and gender. In that study which included 504 white and 202 black males and females aged 20-94, race/ethnicity did not significantly influence the %BF-BMI relationship. Furthermore, significant differences were noted in weight, BMI, FM, %BF, and fat-free mass between white and black females. However, only waist circumference was different between the males. These findings are similar to those obtained from this study, with several differences noted in females but not males. In a subsequent study conducted in 2000, Gallagher et al. (12) used a multicomponent model to develop %BF ranges that corresponded to BMI guidelines. On the basis of a total sample of 1626 subjects (26% white, 16% African American, and 59% Asian), the authors reported an R2 of 0.79 with an SEE of 3.97% based on BMI−1, age, gender, and race/ethnicity. A meta-analysis of 32 studies examined the relationship of %BF and BMI in different ethnic groups, using Caucasians as the reference (9). The formula developed combined gender, age, and BMI and resulted in an R2 value of 0.88 and an SEE of 2.5%. It was determined that for blacks, a 1.3-unit increase in BMI values would be necessary to reflect the same level of body fatness as Caucasians.

Although the use of BMI has shown to be a reasonable measure of adiposity in adults, some research suggests that BMI may be a poor indicator of body fatness in certain population subgroups, such as racial/ethnic minorities (6), college-aged athletes and nonathletes (23), and individuals with a large body build (7). To determine the effect of race on the prediction of body fatness from BMI and MET-h·wk−1 in this study sample, we performed a separate analysis for whites and blacks (Table 4). In both groups, the combination of gender, BMI, and MET-h·wk−1 contributed significantly the explained variance in %BF. However, the model adjusted R2 value for blacks was 0.87 (RMSE = 3.72) as compared with 0.78 (RMSE = 3.97) for whites. Furthermore, BMI contributed 30% of the variance in %BF in whites versus 50% in blacks. These results provide further evidence of the apparent differences in body composition among racial/ethnic groups; however, the findings do not confirm whether PA and/or body build help partially explain the variations observed.

Table 4
Table 4
Image Tools

Unique to this study was the use of self-reported PA to predict body fatness. The IPAQ was chosen because it assesses all activities performed during the previous week across multiple domains. The inclusion of activities involved in occupation, transport, and housework in addition to leisure-time PA has been shown to improve the accuracy of assessment, which has led to more meaningful relationships with health outcomes being described (1). The IPAQ has been shown to be a valid and reliable instrument for self-reporting PA levels across a wide range of age groups and country of origin. Several studies (5,10,15) have examined the criterion validity of the IPAQ compared with data from accelerometers, and one study (11) examined the validity of the IPAQ against fitness measures. Results from the 12-country reliability and validity study of the IPAQ reported the criterion validity of the short version against CSA accelerometers to be comparable to most other self-reports (N = 781, median P = 0.30, 95% confidence interval = 0.23-0.36). Similar results have been reported by Ekelund et al. (10) for total PA (MET-min·d−1) against MTI Actigraph accelerometers (counts·min−1) (N = 185, r = 0.34, P < 0.001) and by Hagstromer et al. (15) (N = 46, P = 0.55, P < 0.001). Fogelholm et al. (11) compared the IPAQ to measures of cardiorespiratory and muscular fitness in adult males and found that those that were classified as HEPA active had superior fitness levels as assessed by maximal oxygen uptake and total number of sit-ups, push-ups, and repeated squats.

Several studies (3,20,21,27,33) have examined the relationship between %BF and PA, and all have observed similar findings. In adolescent girls, Lohman et al. (20) reported significant inverse relationships between %BF and various levels of accelerometry-determined PA (r = −0.17 for MET-weighted MVPA). Similarly, Tudor-Locke et al. (33) reported a significant correlation of −0.27 between pedometer-determined steps per day and %BF predicted from bioelectrical impedance in healthy Caucasian and African American adults. Furthermore, a prospective study conducted with 140 young adult male conscripts observed that %BF from DXA, but not BMI, was significantly related to running performance (21).

Significant negative correlations between self-reported PA and %BF have been observed in adults ranging from −0.13 (3) to −0.51 (27). In this study, the correlation between MET-h·wk−1 and %BF for the entire sample was −0.44. This further supports the use of self-reported PA rather than an objective measure as a valid instrument of activity levels in this sample. Further, there is currently no gold standard for evaluating the validity of self-reports of PA. Pedometers have been known to miss or underestimate certain activities such as bicycling or swimming (33). Moreover, the additional variance explained by self-reported PA in this study resulted in a prediction equation with less error and greater individual predictive accuracy. From a theoretical perspective, use of easily accessible measures (i.e., IPAQ and BMI) may enhance the applicability of body fat prediction equations in other populations as well.

Field-based prediction equations for estimating %BF offer an opportunity for examining the relationship between PA and body composition, provided that the equations are applicable to the population of interest and have been cross-validated. According to Lohman (18), the accuracy of a new method (or equation) to predict %BF as compared with a reference method should be evaluated based on the SEE (equivalent to the RMSE), with values between 3.5% and 4.0% considered good and values >4.0% considered fair to poor. On the basis of these criteria, inclusion of PA as a predictor variable resulted in a "good" model as opposed to a "fair" model without this variable. This further supports the notion that PA should be accounted for when developing population-specific body fat prediction equations. These results are further strengthened by the use of DXA to estimate body composition, which has been shown to agree well with %BF estimates from multicomponent models (22,26,39).

The limitations of this study include its cross-sectional design with a convenience sample of young adult volunteers and the use of self-reported information to derive a model to predict %BF. Such a design limits conclusions about causal inference and generalizability. Furthermore, use of a narrow age range and unequal group representation among whites and blacks limits the applicability of the prediction equations, and therefore these results should be considered population specific. Self-reports of PA are prone to bias and low reliability. In the present study, however, a standardized instrument (IPAQ) with a probe-type protocol was used to maximize the potential for acquiring accurate responses. Therefore, the results may be more reflective of the study population than if the survey had been administered freely without assistance from the investigator. Additional studies with longitudinal designs and a combination of the IPAQ with an objective measure of PA to predict %BF should lead to a better understanding of the dose-response relationship from the perspective of more novel methodologies.

The aim of this study was to develop a body fat prediction equation appropriate for use with a biracial college-aged population and to examine self-reported PA levels as an independent predictor of body fatness. The findings indicate that self-reported PA levels quantified from the IPAQ significantly improved the prediction of DXA-measured %BF in white and black college students. Given the practical appeal of using a standardized instrument for assessing PA levels, future studies should consider including IPAQ scores when developing body fat equations in other populations.

The authors have no financial relationships to report.

The results of this study do not constitute endorsement by American College of Sports Medicine.

Back to Top | Article Outline

REFERENCES

1. Ainsworth BE, Richardson MT, Jacobs DR Jr, Leon AS, Sternfeld B. Accuracy of recall of occupational physical activity by questionnaire. J Clin Epidemiol. 1999;52:219-27.

2. American College Health Association. American College Health Association National College Health Assessment Spring 2006 Reference Group Data Report (Abridged). J Am Coll Health. 2007;55(4):195-206.

3. Bouchard C, Tremblay A, Leblanc C, Lortie G, Savard R, Theriault G. A method to assess energy expenditure in children and adults. Am J Clin Nutr. 1983;37(3):461-7.

4. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307-10.

5. Craig CL, Marshall AL, Sjostrom M, et al. International Physical Activity Questionnaire: 12-country reliability and validity. Med Sci Sports Exerc. 2003;35(8):1381-95.

6. Deurenberg P, Deurenberg-Yap M. Validity of body composition methods across ethnic population groups. Forum Nutr. 2003;56:299-301.

7. Deurenberg P, Deurenberg-Yap M, Wang J, Lin FP, Schmidt G. The impact of body build on the relationship between body mass index and percent body fat. Int J Obes Relat Metab Disord. 1999;23(5):537-42.

8. Deurenberg P, Weststrate JA, Seidell JC. Body mass index as a measure of body fatness: age- and sex-specific prediction formulas. Br J Nutr. 1991;65(2):105-14.

9. Deurenberg P, Yap M, van Staveren WA. Body mass index and percent body fat: a meta analysis among different ethnic groups. Int J Obes Relat Metab Disord. 1998;22(12):1164-71.

10. Ekelund U, Sepp H, Brage S, et al. Criterion-related validity of the last 7-day, short form of the International Physical Activity Questionnaire in Swedish adults. Pub Health Nutr. 2006;9(2):258-65.

11. Fogelholm M, Malmberg J, Suni J, et al. International Physical Activity Questionnaire: validity against fitness. Med Sci Sports Exerc. 2006;38(4):753-60.

12. Gallagher D, Heymsfield SB, Heo M, Jebb SA, Murgatroyd PR, Sakamoto Y. Healthy percentage body fat ranges: an approach for developing guidelines based on body mass index. Am J Clin Nutr. 2000;72(3):694-701.

13. Gallagher D, Visser M, Sepulveda D, Pierson RN, Harris T, Heymsfield SB. How useful is body mass index for comparison of body fatness across age, sex, and ethnic groups? Am J Epidemiol. 1996;143(3):228-39.

14. Guo SS, Chumlea WC, Cockram DB. Use of statistical methods to estimate body composition. Am J Clin Nutr. 1996;64(3 suppl):428S-35S.

15. Hagstromer M, Oja P, Sjostrom M. The International Physical Activity Questionnaire (IPAQ): a study of concurrent and construct validity. Pub Health Nutr. 2006;9(6):755-62.

16. Irwin JD. Prevalence of university students' sufficient physical activity: a systematic review. Percept Mot Skills. 2004;98(3):927-43.

17. Kohrt WM, Malley MT, Dalsky GP, Holloszy JO. Body composition of healthy sedentary and trained, young and older men and women. Med Sci Sports Exerc. 1992;24(7):832-7.

18. Lohman TG. Advances in Body Composition Assessment. Champaign (IL): Human Kinetics; 1992. p. 3-4.

19. Lohman TG, Harris M, Teixeira PJ, Weiss L. Assessing body composition and changes in body composition. Another look at dual-energy x-ray absorptiometry. Ann N Y Acad Sci. 2000;904:45-54.

20. Lohman TG, Ring K, Schmitz KH, et al. Associations of body size and composition with physical activity in adolescent girls. Med Sci Sports Exerc. 2006;38(6):1175-81.

21. Mattila VM, Tallroth K, Marttinen M, Pihlajamaki H. Physical fitness and performance. Body composition by DEXA and its association with physical fitness in 140 conscripts. Med Sci Sports Exerc. 2007;39(12):2242-7.

22. Mazess RB, Barden HS, Hanson JA. Body composition by dual-photon absorptiometry and dual-energy x-ray absorptiometry. Basic Life Sci. 1990;55:427-32.

23. Ode JJ, Pivarnik JM, Reeves MJ, Knous JL. Body mass index as a predictor of percent fat in college athletes and nonathletes. Med Sci Sports Exerc. 2007;39(3):403-9.

24. Ogle GD, Allen JR, Humphries IR, et al. Body-composition assessment by dual-energy x-ray absorptiometry in subjects aged 4-6 y. Am J Clin Nutr. 1995;61(4):746-53.

25. Pietrobelli A, Faith MS, Allison DB, Gallagher D, Chiumello G, Heymsfield SB. Body mass index as a measure of adiposity among children and adolescents: a validation study. J Pediatr. 1998;132(2):204-10.

26. Prior BM, Cureton KJ, Modlesky CM, et al. In vivo validation of whole body composition estimates from dual-energy x-ray absorptiometry. J Appl Physiol. 1997;83(2):623-30.

27. Richardson MT, Ainsworth BE, Wu HC, Jacobs DR Jr, Leon AS. Ability of the Atherosclerosis Risk in Communities (ARIC)/Baecke Questionnaire to assess leisure-time physical activity. Int J Epidemiol. 1995;24(4):685-93.

28. Rzewnicki R, Auweele YV, De Bourdeaudhuij I. Addressing overreporting on the International Physical Activity Questionnaire (IPAQ) telephone survey with a population sample. Public Health Nutr. 2002;6(3):299-305.

29. Sallis JF. Age-related decline in physical activity: a synthesis of human and animal studies. Med Sci Sports Exerc. 2000;32(9):1598-600.

30. Sarkin JA, Nichols JF, Sallis JF, Calfas KJ. Self-report measures and scoring protocols affect prevalence estimates of meeting physical activity guidelines. Med Sci Sports Exerc. 2000;32(1):149-56.

31. Sjostrom M, Oja P, Hagstromer M, Smith BJ, Bauman A. Health-enhancing physical activity across European Union countries: the Eurobarometer Study. J Pub Health. 2006;14:291-300.

32. Toth MJ, Beckett T, Poehlman ET. Physical activity and the progressive change in body composition with aging: current evidence and research issues. Med Sci Sports Exerc. 1999;31(11 suppl):S590-S6.

33. Tudor-Locke C, Ainsworth BE, Whitt MC, Thompson RW, Addy CL, Jones DA. The relationship between pedometer-determined ambulatory activity and body composition variables. Int J Obes. 2001;25:1571-78.

34. Tudor-Locke C, Bassett DR. How many steps/day are enough? Preliminary pedometer indices for public health. Sports Med. 2004;34:1-8.

35. US Department of Health and Human Services. Clinical Guidelines on the Identifcation, Evaluation, and Treatment of Overweight and Obesity in Adults: The Evidence Report. Rockville (MD): National Institutes of Health, National Heart, Lung and Blood Institute; 1998. Report No.: 98-083. p. 262. Available from: http://www.nhlbi.nih.gov/guidelines/obesity/ob_gdlns.pdf

36. US Department of Health and Human Services. Physical Activity and Health: A Report of the Surgeon General. Atlanta (GA): US Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, International Medical Publishing; 1996. p. 300. Available from: http://www.cdc.gov/NCCDPHP/SGR/SGR.HTM

37. Valanou EM, Bamia C, Trichopoulou A. Methodology of physical-activity and energy-expenditure assessment: a review. J Public Health. 2006;14:58-65.

38. Wagner DR, Heyward VH. Measures of body composition in blacks and whites: a comparative review. Am J Clin Nutr. 2000;71(6):1392-402.

39. Wang ZM, Deurenberg P, Guo SS, et al. Six-compartment body composition model: inter-method comparisons of total body fat measurement. Int J Obes Relat Metab Disord. 1998;22(4):329-37.

40. World Health Organization. Obesity: Preventing and Managing the Global Epidemic. Geneva: World Health Organization, Report of a WHO Consultation, WHO Technical Report Series; 2000. >Report No.: 894>. p. 265. Available from: http://www.who.int/nutrition/publications/obesity/en/index.html

Cited By:

This article has been cited 1 time(s).

Nutrition Research
Dietary factors are associated with coronary heart disease risk factors in college students
Fernandes, J; Arts, J; Dimond, E; Hirshberg, S; Lofgren, IE
Nutrition Research, 33(8): 647-652.
10.1016/j.nutres.2013.05.013
CrossRef
Back to Top | Article Outline
Keywords:

BODY COMPOSITION; PREDICTION EQUATION; IPAQ; DXA; BMI; YOUNG ADULTS

©2009The American College of Sports Medicine

Login

Article Tools

Images

Share

Search for Similar Articles
You may search for similar articles that contain these same keywords or you may modify the keyword list to augment your search.

Connect With Us