Abstract
Health care professionals require increased knowledge of health and disabilities to effectively implement health promotion initiatives at both the individual and group level for adults with intellectual disabilities (ID). The aim of this review is to examine the feasibility, reliability, and validity of various field-based measurements to assess body composition among adults with ID as compared to nondisabled controls. The literature was systematically searched from 1990 to 2017 for primary articles pertaining to the subject matter that were published in the English language and included only individuals ≥18 years of age. 1,989 studies were screened and 8 studies were included for review. Several field-based measurements for body composition are feasible and reliable yet none have been validated for use in adults with ID. Awareness of the various methods for assessing body composition in adults with ID in clinical practice, while simultaneously understanding their limitations, is necessary.
Introduction
The American Association on Intellectual and Developmental Disabilities (AAIDD) defines intellectual disability (ID) as encompassing significant limitations in intellectual functioning and adaptive behavior that originates before age 18 years (Schalock et al., 2012). In 2013, the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5; American Psychiatric Association, 2013) replaced the diagnosis of mental retardation with ID, using language that aligned with AAIDD’s classification. Numerous syndromes have been associated with ID with the more familiar being Down syndrome, Rett syndrome, 22q11.2 deletion syndrome, and Williams syndrome, among others (Toth, de Lacy, & King, 2016).
Individuals with disabilities, including ID, tend to have poorer health outcomes than those without (Emerson, Graham, Llewellyn, Hatton, & Robertson, 2011) and experience similar race and ethnic health disparities as the general population (Magana, Parish, Morales, Li, & Fujiura, 2016). Higher rates of chronic health conditions among those with cognitive impairments (Reichard, Stolzle, & Fox, 2011) places this population at greater health risk, some which could be prevented or reduced with improved care (Krahn & Fox, 2014). Healthy People 2020 organize social determinants of health into five domains (education, health & health care, community, environment, and economic stability) and individuals with disabilities, including physical, developmental, and intellectual are more likely to experience challenges in each of these domains as compared to individuals without disabilities (Office of Disease Prevention and Health Promotion [ODPHP], 2016). Strategies for managing health disparities in this population include an increased focus on addressing the social determinants of health in this population. For example, increased supports and services that help provide improved care and offer greater inclusion in the community can be just one avenue of approach. More broadly, however, improvement in health disparities for individuals with disabilities should also include bringing data on these disparities to the attention of policy decision makers (Krahn & Fox, 2014). This can include evidence from health surveillance and health indicators obtained via health services research (Krahn & Fox, 2014).
To help address these inequities, the World Health Organization (2008a) has put forth a call to action that includes expanding health care professionals’ knowledge of how to promote health to meet the needs of individuals with disabilities, as well as including those with disabilities, in health promotion activities. It is recognized that there is a strong need to implement evidence-based interventions in clinical and community settings for people with disabilities, including health and wellness programs (ODPHP, 2016).
The emphasis on including individuals with disabilities in health promotion programs, in addition to the increasing number of publications on nutrition and physical activity interventions targeted at this group (Brooker, van Dooren, McPherson, Lennox, & Ware, 2015; Spanos, Melville, & Hankey, 2013), underscores the need to ensure methods for assessing body composition are validated for use in this population. Body composition of individuals with ID may differ from those without in that they may have atypical fat distribution or greater fat mass; the origins of which may stem from underlying syndromes or comorbid conditions (Humphries, Traci, & Seekins, 2009). Understanding body composition of individuals with ID is important as obesity has been linked to an increased risk in developing cardiovascular disease, diabetes, hypertension, ischemic stroke, and some cancers (Mozaffarian et al., 2015; Wormser et al., 2011). Without valid ways of predicting or measuring body composition among those with ID, health intervention outcomes and health indicators cannot reliably be measured. Program implementation and success may be hindered without an understanding of how these factors influence health outcomes and unreliable health indicators will slow the progress of closing the gap on health disparities for this population.
There are several field-based techniques that act as surrogate markers for body composition and assist in determining health risk. Body mass index (BMI) and waist circumference (WC) are some of the most common and easiest measurements performed in clinical practice (Verstraelen, Maaskant, van Knijff-Raeven, Curfs, & van Schrojenstein Lantman-de Valk, 2009). The Center for Disease Control and Prevention (2015) classifies adults with a BMI of 25.0 to 29.9 kg/m2 as overweight and ≥30.0 kg/m2 as obese. WC is the measurement of an individual’s waist and is used as an indicator of abdominal obesity (World Health Organization, 2008b). WC cannot measure body fatness but can indicate risk for comorbidities such as type 2 diabetes mellitus, coronary artery disease, and hypertension as a result of excess abdominal fat (Center for Disease Control and Prevention, 2015). Increased risk for metabolic complications can begin as early as a WC >94 cm in men and a WC >80 cm in women (World Health Organization, 2008b). It remains to be determined if using the standard BMI cutoff values for adults without ID is valid for use among adults with ID. A theorized barrier to using standard BMI cutoff values for those with ID is that body composition may differ from peers without ID (Temple, Walkley, & Greenway, 2010; Usera, Foley, & Yun, 2005), such as increased body fat or atypical fat distribution, thus questioning the sensitivity of BMI as an indicator of body fatness within this population.
Anthropometric girth measurements, skinfold (SF) measurements, air displacement plethysmography, and bioelectrical impedance analysis (BIA) are examples of clinical tools that have been examined among adults with ID within the literature (Havinga-Top, Waninge, van der Schans, & Jager-Wittenaar, 2015; Usera et al., 2005; Verstraelen et al., 2009; Waninge, van der Weide, Evenhuis, van Wijck, & van der Schans, 2009). SF measurements are based on a two-compartment model in which the body is divided into fat mass and fat-free mass (Usera et al., 2005). The measurements are used within a predictive equation that is appropriate for the individual being assessed (Usera et al., 2005). Girth measurements, which have been found to be simple, practical, and valid ways to assess body composition among obese, adults without ID (Daniell, Olds, & Tomkinson, 2010) have also been explored for their use in accurately estimating body composition among adults with ID (Kelly & Rimmer, 1987; Usera et al., 2005; Verstraelen et al., 2009; Waninge et al., 2009).
Similar to SF measurements, anthropometric girth measurement results are also used in predictive equations to estimate body composition. To date, only one study has validated a predictive equation for use in adults with ID using anthropometric girth measurements; however, it was only among a small group of residential males with ID between the ages of 18 and 40 years and only validated against another predictive equation versus a more rigorous criterion measurement (Kelly & Rimmer, 1987). Body adiposity index (BAI), an alternative to BMI, is a predictive equation that can estimate body fat percentage using the following equation: Body fat percentage = [hip circumference/height1.5]−18 (Bergman et al., 2011). Prior to the study by Nickerson et al. (2015), BAI had only been validated in adults without ID and not tested among individuals with ID, and more specifically among those with Down syndrome. BIA determines both fat mass and fat-free mass among individuals through the measurement of electrical impedance (Havinga-Top et al., 2015). Validated BIA predictive equations that take into account age, gender, and race can be used among healthy individuals with stable electrolyte and water balance (Kyle et al., 2004).
The importance of accuracy in these measurements for this population cannot be overstated. Using field-based body composition measurements that do not accurately reflect true body composition (or accurate change in body composition) may hinder the adoption or creation of meaningful health-promoting activities. If interventions cannot demonstrate desirable outcomes (e.g., decrease in body fat), support for their continuation may wane when in reality it may not be the intervention that is not yielding desirable outcomes but rather flawed measurement tools.
Aim and Methods of the Review
As the life expectancy of those with ID expands (Bittles & Glasson, 2004), there is a need for greater understanding of the validity and reliability of body composition assessment in clinical practice to support the goals of preventing chronic disease and maximizing health (ODPHP, 2016). Appreciating the unique aspects of body composition in adults with ID allows clinicians to develop the most effective nutritional care plans, including diet and exercise interventions, for successful health and wellness programs. Thus, the purpose of this narrative review is to determine the feasibility, reliability, and validity of assessing body composition among adults with ID and to explore the potential differences that may exist between this population and individuals without disabilities.
The literature was searched using Scopus, Ovid Medline, and CINAHL with the keywords and MeSH terms “Intellectual Disability,” “Developmental Disability,” “Down syndrome,” “Trisomy 21,” “Anthropometry,” “Body composition,” “Body Mass Index,” “Waist Circumference,” and “Adiposity.” Figure 1 provides details on the steps of the search strategy. In summary, search limitations included articles that were primary research, published in the English language, published between January 1, 1990 and May 1, 2017, and studies that included individuals with ID ≥18 years old. Duplicate articles from the initial search results were removed first, at which point the search limitations were applied by the primary author to the remaining titles and abstracts. Articles were excluded by the primary author if they were reviews, included participants ≤18 years old, were published outside the desirable date range or were not published in the English language. All remaining article titles and abstracts were examined by the primary author for relevance to the purpose of this narrative review. The most common reasons for exclusion at this step in the search process included using body composition measurements descriptively versus testing for feasibility, reliability, and/or validity and for including participants less than 18 years of age.
Process of Study Selection for Inclusion
Reference lists of the relevant articles as well as pertinent reviews were also screened to identify additional articles for inclusion. Eight studies were found to pertain to the present subject matter and a summary of the available evidence has been provided in Table 1. The strength of the evidence was analyzed using the quality criteria checklist, a risk of bias tool created by the Academy of Nutrition and Dietetics (2016) that facilitates critical appraisal of the research, as part of the evidence analysis process. Ratings for each study are provided in Table 1.
Evidence Table for Body Composition Assessment in Adults With Intellectual Disabilities (ID)
ADP = air displacement plethysmography; BAI = body adiposity index; BF = body fat; BF% = body fat % BIA = bioelectrical impedance analysis; BMI = body mass index; CI = confidence interval; DEBMI-BF% = BMI-based BF% equation from Deurenber et al (1991); DXA = dual-energy X-ray absorptiometry; FFM = fat-free mass; FM = fat mass; GABMI-BF% = BMI-based BF% equation from Gallagher et al. (2000); GMFCS = Gross Motor Function Classification System; HC = hip circumference; Ht = height; ICC = interclass correlation coefficient; ICD-10 = International Classification of Diseases, 10th revision; ID = intellectual disability; JABMI-BF% = BMI-based BF% equation from Jackson et al. (Jackson et al., 1980; Jackson & Pollock, 2004); LBM = lean body mass; LOA = level of agreement; SD = standard deviation; SIVD = severe intellectual and visual disabilities; StWC = standing waist circumference; SuWC = supine waist circumference; TL = tibia length; WC = waist circumference; WHO = World Health Organization; WOBMI-BF% = BMI-based BF% equation from Womersley and Dumin (1977); wt = weight.
Levels of evidence defined by the evidence analysis module. Retrieved from https://www.andeal.org/files/Docs/2012_Jan_EA_Manual.pdf. bQuality of evidence defined by the evidence analysis module. Retrieved from https://www.andeal.org/files/Docs/2012_Jan_EA_Manual.pdf, wherein “+” = positive; “−” = negative; “ø” = neutral.
Across eight trials, body composition measurements have been studied among 299 community-dwelling and residential adults with ID (Esco, Nickerson, Bicard, Russell, & Bishop, 2016; Havinga-Top et al., 2015; Nickerson et al., 2015; Temple et al., 2010; Usera et al., 2005; Verstraelen et al., 2009; Waninge et al., 2009; Waninge et al., 2010). Table 2 provides a comparison of the field-based body composition measurements that have been studied in adults with ID.
Body Composition Field Measurement Comparison Chart for Adults With Intellectual Disabilities (ID)
NOTE: BF% = body fat percentage; BIA = bioelectrical impedance analysis; BMI = body mass index; DXA = dual energy X-ray absorptiometry; ht = height; SF = skinfold; WHO = World Health Organization.
Rated feasible if found to be ≥95% feasible among participants across available studies. bRated reliable if found to have limits of agreement <10% of the mean of the first measurement in a test-retest design; cAmong adults without ID.
Results
Body Mass Index
Three body composition studies examined the reliability and/or validity of BMI and/or BMI-based predictive equations as indicators of adiposity among adults with ID (Esco et al., 2016; Temple et al., 2010; Verstraelen et al., 2009). Feasibility of BMI calculations ranged from 95% to 100% among 167 adults with ID from across three studies (Temple et al., 2010; Verstraelen et al., 2009; Waninge et al., 2009) with a test–retest level of agreement of <10% when measured among 45 residential adults with ID (Waninge et al., 2009). While successful, feasibility and reliability of BMI depends on the cooperation of the individual being measured as well as their ability to remain still during the measurement of weight and height (Verstraelen et al., 2009; Waninge et al., 2009).
When assessing the level of agreement between BMI and several measurement methods (WC, SF measurement, multifrequency BIA) on determining weight status of 76 residential adults with ID, only BMI and WC had an acceptable level of agreement among the total study population (k = .61, 90% confidence interval [CI] [0.46, 0.76]) and also among a subgroup of short-statured subjects (n = 40, k = .60, 90% CI: [0.40, 0.81]; Verstraelen et al., 2009). However, while the kappa values suggest there was sufficient agreement between BMI and WC among these groups, the lower limits of the 90% CIs (0.46 and 0.40, respectively) highlights the poor agreement of the measurements at the lower limits. Additionally, this study failed to compare the level of agreement between BMI classified weight status and weight status as classified by a more stringent criterion measurement.
In 2010 Temple et al. (2010) also examined the validity of using BMI as an indicator of adiposity among 46 community-dwelling adults with ID. In this study, where over one third of the participants had Down syndrome, BMI measurements were compared to dual-energy X-ray absorptiometry (DXA) results, and BMI was found to have a strong and significant correlation with fat mass (r = 0.91, p < .001) but a weak and nonsignificant correlation with lean mass (r = −0.12, p = .43; Temple et al., 2010). The correlation between BMI and DXA was higher (adjusted R2 = 0.82, p < .001 vs. adjusted R2 = 0.68, p < .001) when expressed in terms of total body fat compared to percent body fat, respectively. The relationship remained when stratified by gender and presence of Down syndrome (males: R2 = 0.71, p < .001; females: R2 = 0.89, p < .001; Down syndrome: R2 = 0.88, p < .001; without Down syndrome: R2 = 0.85; p < .001; Temple et al., 2010). This finding was later echoed by Esco et al. (2016) who reported a significant relationship between BMI and fat mass (r = 0.84, p < .001) but not fat-free mass (r = 0.31, p = .17) among 20 community-dwelling adults with Down syndrome. Temple and colleagues also observed that a BMI ≥30.0 kg/m2 had 57.1% sensitivity and 100.0% specificity in relation to obesity identified by DXA, with sensitivity improving with lower cutoff points (e.g., BMI ≥ 23.0 kg/m2 had 100% sensitivity and 58.8% specificity). The BMI cutoff value of ≥30.0 kg/m2 accurately classified adults who are not obese but misclassified some individuals found to be obese by DXA as not having obesity (Temple et al., 2010).
BMI-based equations and their ability to accurately predict body fat have also been examined in the literature with one study by Esco et al. (2016) comparing four equations (Deurenberg, Weststrate, & Seidell, 1991; Gallagher et al., 2000; Jackson et al., 2002; Womersley, 1977) to DXA results among 20 community-dwelling adults with Down syndrome (Esco et al., 2016). While only one equation (Jackson et al., 2002) did not differ significantly from DXA results (p = .66), the correlation coefficients were large for all four equations (range of r = 0.68 to r = 0.89; Jackson et al., 2002). Overall, all four equations provided wide limits of agreement with the standard error of estimates ranging from 5.1% to 8.4% and total errors ranging from 6.8% to 8.93%. The equations that had tighter limits of agreements and stronger correlations to DXA included age, gender, and race as prediction variables (Esco et al., 2016).
Waist Circumference
Like BMI, WC has been found to be a feasible measurement method to perform in adults with ID with feasibility ranging from 95% to 100% among 164 adults across three studies (Verstraelen et al., 2009; Waninge et al., 2009; Waninge et al., 2010) with a level of agreement of <10% when measured in a test–retest fashion among 88 individuals with ID across two studies (Waninge et al., 2009; Waninge et al., 2010).
For those who cannot stand or remain still when standing, Wagnine et al. (2010) explored using supine measured WC as a valid measurement in lieu of a standing WC. Supine WC measurements were significantly lower than standing WC measurements obtained from 160 healthy individuals without a disability (p < .001), suggesting the measurements cannot be used interchangeably. A correction formula, which was proposed by Wagnine and colleagues, would be needed if supine WC measurements were to be compared with international standards of standing WC measurements, however, this was only validated among the control group (p < .001, R/R2: 0.982/0.964; Waninge et al., 2010).
Skinfold Measurements and Anthropometric Girth Measurements
When comparing a seven-site (Jackson & Pollock, 2004; Jackson, Pollock, & Ward, 1980) and three-site (Lohman, 1981) SF measurement predictive equation as well as a girth measurement predictive equation (Kelly & Rimmer, 1987) to air displacement plethysmography using the BOD POD®, Usera et al. (2005) found that correlations between the BOD POD and the predictive equations ranged from r = 0.11 to r = 0.54 among 13 adults with Down syndrome and from r = 0.75 to r = 0.94 among 14 adults without Down syndrome (Usera et al., 2005). The average errors for the SF predictive equations ranged from 13.2% to 14.9% in those with Down syndrome while those same equations had average errors of 4.9% and 5.8% among those without Down syndrome (Usera et al., 2005). There were significant differences in accuracy of the measurements between the groups, Wilks’s Λ = 0.38, F(3, 23) = 12.34 (p < .01), partial η2 = 0.56. For those with Down syndrome, the SF predictive equations underestimated body fat percentage (Usera et al., 2005).
Body Adiposity Index
Nickerson and colleagues compared the predicted body fat percentage using the BAI equation to body fat percentages assessed by DXA in 20 community-dwelling adults with Down syndrome (Nickerson et al., 2015). They found that the mean difference between both body fat percentage measurements was significant (DXA = 39.94% ± 10.80%, BAI = 42.60% ± 8.19%; p < .001) with a small effect size (Cohen’s d = 0.27). There was a large and significant correlation between the two body fat percentage variables (r = 0.73, p < .001), suggesting variance patterns for both variables were similar. The standard error of estimate for BAI, however, was 7.79% and the total error was 7.86%, which are unacceptable levels for a predictive equation (Nickerson et al., 2015). BAI overestimated body fat percentage by 2.65% when compared with DXA with wide limits of agreement (CE = 2.65%, 95% CI: [12.21%, 17.52%]; Nickerson et al., 2015).
Bioelectrical Impedance Analysis
While work has been done with regard to validating the use of BIA in body composition measurements among various adult populations, little evidence exists as it pertains to adults with ID. Two of the studies included within this review addressed the feasibility of BIA measurements in adults with ID, however, neither explored its validity against a criterion measurement (Havinga-Top et al., 2015; Verstraelen et al., 2009). Both Verstraelen et al. (2009) and Havinga-Top et al. (2015) found feasibility of BIA measurements to be 86% and 88%, respectively, among residential adults with ID. Of note, Havinga-Top et al. (2015) excluded participants from their study due to behavioral or anticipated anxiety problems related to BIA measurements. Barriers to successful BIA measurements included inability to stay still for measurements, inability to understand directions, and contractures.
Of 62 adults who had BIA measurements performed by Verstraelen et al. (2009) 53% were classified as underweight based on their fat-free mass index compared with 13% (n = 10) being underweight according to BMI classification. Havinga-Top et al. (2015) found that there were no differences in impedance results (resistance, reactance, or fat-free mass) between the varying degrees of impairment (mobility, intellectual, visual, or auditory) and there was good agreement and correlation between two separate BIA measurements (ICC resistance: 0.965 (95% CI: [0.922, 0.984]); ICC reactance: 0.858 (95% CI: [0.705, 0.934]); ICC fat-free mass: 0.992 (95% CI: [0.982-0.996]).
Discussion
Noninvasive body measurements such as height, weight, BMI, and WC are feasible measurements to obtain in adults with ID in a clinical setting (Verstraelen et al., 2009; Waninge et al., 2009; Waninge et al., 2010); however, individual patient compliance and ability to follow instructions can influence feasibility (Verstraelen et al., 2009; Waninge et al., 2009). BMI is a reasonable indicator of body fatness among adults with ID (Casey, 2013; Esco et al., 2016; Temple et al., 2010) but does not correlate well with fat-free mass (Esco et al., 2016; Temple et al., 2010). Since fat-free mass includes muscle, bone, blood, and water, using BMI as a health indicator in this population does not fully capture true body composition. This limits the ability to determine improvements in body composition (e.g., improvement in muscle mass) as a result of health interventions. Given the health risks associated with obesity, choosing a BMI cutoff point that has higher sensitivity may be advantageous in this population as it may better classify those who are obese; however, this needs to be explored further (Temple et al., 2010).
For adults who may be unable to stand or remain still for a WC measurement, supine WC measurements are feasible (Waninge et al., 2010). Unmodified supine WC measurements should not be compared to standing international WC reference standards and it remains unknown if an established equation to modify supine WC measurements is valid for use in adults with ID (Waninge et al., 2010).
Predictive equations that have been used to estimate body composition (through BMI, SF, or girth measurements) have not been validated for use in adults with ID and do not adequately estimate body fat percentage among this population. Limitations to their use include wide limits of agreements, low correlation with criterion measurements, high rates of error, and, in the case of SF measurement, poor feasibility (Esco et al., 2016; Usera et al., 2005). These findings indicate further research is required to establish a validated predictive equation for adults with ID and other neurological impairments (Casey, 2013; Esco et al., 2016).
The wide limits of agreement and overestimation of body fat percentage among lean adults with Down syndrome and underestimation of body fat percentage among those with higher fat mass suggests BAI should not be used within this population as an estimation of body fatness (Nickerson et al., 2015). It remains unknown if using this equation among other adults with ID (e.g., those without Down syndrome) would yield similar results or if adaptations to this equation for use among adults with ID is feasible.
BIA has not been validated for use within this population against a criterion measurement and is a less feasible measurement to be obtained compared to other field-based techniques (Verstraelen et al., 2009). Using other options for assessing body composition among adults with ID in the clinical setting may be warranted.
In summary, BMI may be a reasonable health indicator of body fatness but not necessarily of fat-free mass in adults with ID. For surveillance purposes, it is unclear if using a more sensitive BMI cutoff value is warranted as it may help detect more adults with ID who are at increased health risk. WC, predictive equations, BAI, and BIA have not been validated for use in this population and interpretation of these measurements should be done cautiously.
Limitations
There are several limitations to the available literature surrounding valid body composition assessment among adults with ID. Given the small sample sizes of the studies examined, it is likely that these trials were underpowered and many did not provide power calculations (Esco et al., 2016; Fernhall et al., 2005; Iwaoka et al., 1998; Nickerson et al., 2015; Temple et al., 2010; Usera et al., 2005; Verstraelen et al., 2009; Waninge et al., 2009; Waninge et al., 2010). Additionally, in many of the trials, the sample populations were heterogeneous, which limits generalizability and presents difficulties in interpreting the results (Lante, Reece, & Walkley, 2010; Verstraelen et al., 2009; Waninge et al., 2009; Waninge et al., 2010). It is unclear how body composition and fat distribution may have differed among individuals depending on the etiology of their ID (e.g., Down syndrome, Rett syndrome, etc.). Furthermore, in several studies the investigators failed to use a suitable criterion measurement for which to compare body composition results against, which limits the interpretation of the data (Havinga-Top et al., 2015; Verstraelen et al., 2009; Waninge et al., 2009; Waninge et al., 2010). Overall, the strength of the evidence presented within this review assessing the validity of body composition measurements is limited given flaws in the research designs, small sample sizes, and lack of generalizability.
Conclusions and Implications for Practice
There is limited research exploring body composition in adults with ID. As the life span among those with ID continues to expand, there needs to be more evidence-based information available for clinicians who will be managing these individuals in clinical practice. Health care professionals need to be aware of the various methods of assessing body composition in adults with ID but understand the limitations that exist with their use. More research is needed to establish valid, noninvasive, body composition assessment tools for adults with ID. Barriers to performing body composition measurements among this population, such as fear of measurement tools or inability to remain still, should be taken into consideration before being performed on patients.