Background

Obesity affects over 600 million adults worldwide and the number continues to grow [1]. Along with the rise in its worldwide prevalence [2, 3], the evidence for its adverse effects on individuals’ health has been accumulating. Obesity has been identified as a risk factor for cardiovascular disease [4], pulmonary embolism [5], large joint osteoarthritis (OA) [6], and certain types of cancer [7]. It has also been associated with a decrease in general physical function [8], as well as cognitive abilities [9]. The diversity in the adverse outcomes attributed to obesity, the complexity in the mechanisms leading to them, and the multi-factorial nature of this disease require the joint effort of different scientific disciplines to better understand the scope of the problem and to limit its detrimental effects.

With the prevalence of obesity among the workforce being equal to that of the general population [10], the occupational safety and health discipline has shown interest and effectively contributed to obesity research. The effects of obesity on work performance, physical capacity, and physical and cognitive function have been the research focus of many ergonomists, work analysts, and occupational health experts. As such, employees who are obese have been found to have higher rates of sick leave [11] and workplace injuries [12], along with increased employer-paid healthcare costs [13]. As these efforts expand to evaluate the relation between obesity and work [14], it is essential to explore how obesity status is measured in this field (e.g., body mass index (BMI) and body fat percentage (%BF)) as well as the basis for classifying individuals into distinct risk groups (e.g., types I and II obesity). In general, the issue of obesity measurement is two-fold: 1) selection of the appropriate measurement and 2) properly carrying out the measurement process to minimize measurement error.

Measurement selection

The World Health Organization (WHO) defines obesity as abnormal or excessive fat accumulation that may impair health [15], and this definition should serve as the basis for measurement selection. While underwater weighing and dual energy x-ray absorptiometry (DEXA) directly measure body fat, many indirect measures of adiposity have been used to measure obesity status. Anthropometric measures such as the weight-for-height index, BMI, waist circumference (WC), waist–hip ratio (WHR), and body fat percentage estimated by skinfold thickness (ST) are widely accepted indirect measures. Since the 1990s, BMI has been widely used to classify overweight and obesity, both in adults and children [16]. BMI has been suggested as an ideal measure of adiposity since it is easy to measure and is closely associated with obesity related health risks [17].

However, indirect measures such as BMI, fail to distinguish between fat, muscle or bone mass and are prone to misclassification, particularly among muscular subjects [18]. Mullie et al. [19] compared %BF, measured by bipolar bioelectrical impedance analysis (BIA), and BMI, for a cohort of 448 male military candidates and found a statistically significant difference between classifications of normal weight versus overweight. Almost 40% of the subjects classified as overweight with BMI > 25 kg/m2 had a %BF corresponding to normal weight. Similarly, Deurenberg et al. [20] observed a higher rate of misclassification with BMI compared to DEXA in 416 European individuals. This study showcased how individual results based on a single classification method should be interpreted with caution.

Reliance only on BMI can also lead to misclassification of those with excess body fat, but BMIs corresponding to normal weight. These “metabolically obese but normal weight” [21] individuals share many health risks with those categorized as obese both based on BMI and %BF [22]. The elevated visceral fat observed in this category is accompanied by increased levels of both liver and muscle fat [23]. In a workplace study, comparing new industry hires from 1990 to 1992 and from 2000 to 2002, there was no significant difference in BMI but a significant difference in %BF, measured by ST [24]. There were also significant differences in physical fitness as measured by timed sit-ups and squats, suggesting that employers would miss information regarding their employees’ fitness with reliance on BMI only. BMI is also not independently representative of body fatness. Significant dependencies on age and sex were observed in the relation between %BF and BMI in a study of 706 adult men and women [25]. BMI also overlooks the distribution of fat, which is an important factor in disease risk. For instance, android fat distribution (also referred to as abdominal, central, visceral, or upper body fat distribution) causes increased risk of diseases such as cardiovascular disease and type 2 diabetes [26], while gynoid fat distribution (i.e. larger hip and thigh circumferences) does not seem to have similar effects [27]. Indices such as WC and WHR are useful in characterizing the obesity morphology, particularly for studies where a difference in anthropometry among subjects is relevant to consider.

Misclassification and measurement error may be exacerbated in small sample sizes, which are common in exploratory laboratory-based occupational safety and health studies. Piers et al. [28] showed that despite the significant correlation between BMI and %BF (measured by underwater weighing method) of the 117 healthy samples, BMI only explained, on average, 50% of the variance in %BF. The reported poor sensitivity (47.7%) and positive predictive value (67.7%) of BMI makes it an unreliable measure of obesity in individuals. These findings not only suggest the inadequacy of BMI in classification of obesity status, particularly for individuals near the cutoff values, but also point out the importance of a rigorous obesity classification in studies with small sample sizes.

Measurement process and method

After selection of the suitable and hypothesis-relevant obesity measure, it is the researchers’ responsibility to ensure that the measurement guidelines are thoroughly followed to reduce measurement error. For instance, WC is widely accepted as a simple and reliable measure of obesity in general, and abdominal obesity, in particular. There exist guidelines to ensure WC is appropriately measured [29, 30]. However, Agarwal et al. [31] found significant differences in the measured WC across varying anatomical sites, phases of respiration, and time since last meal, when following either the WHO or the National Institute of Health (NIH) guidelines. Overlooking these details can lead to an increase in the measurement error and the steps taken to control them should be acknowledged in publications.

Similarly, the cut-off values used to classify subjects into distinct risk groups are also worth scrutiny. For instance, WHO identifies 25 and 30 kg/m2 as BMI cut points for overweight and obesity respectively. However, it has been shown that among certain populations (e.g., individuals of Asian descent) cardio-metabolic risk is increased at lower body mass indexes [32]. While some researchers advocate using international cut-offs [33], some find nationally and ethnically selected cut points, when available, more advantageous [34]. Overall, inconsistency in the cut-offs used across studies with similar topics is detrimental to the strength of the body of evidence.

This study aims to examine the obesity-related research in occupational health and safety regarding obesity measurement methods. The findings will show how researchers in the aforementioned fields are conducting obesity research and will inform future obesity research in the occupational safety and health domain.

Methods

Research strategy and study inclusion criteria

A systematic review of the PubMed database was undertaken with the following MeSH terms: (‘Work ‘or ‘Ergonomics’ or ‘Biomechanics\Biomechanical’ or ‘Occupational’ or ‘Motion’ or ‘Movement’) and (‘Obesity’ or ‘Skinfold Thickness’). In addition, a keyword search using Google Scholar and manual search of citations from relevant papers and literature reviews was done. The search was limited to journal articles dated between January 1, 2000 and December 31, 2014, published in English and studied human adults. An initial search was performed on 3 March 2015, and repeated on 24 September 2015 to update the search and results.

The study inclusion criteria were: (a) publication contributed to occupational health and safety rather than health promotion and (b) weight status was the independent variable or the major covariate included in the analysis and not the dependent variable. Review papers, simulation-based studies [35], and studies including normal weight subjects loaded with excess weight [36] were also excluded.

The initial search resulted in 3283 studies. The first author assessed all search outcomes by title and/or abstract, out of which 950 were selected based on relevance of the topic. A review of the abstracts reduced the number of studies to 111. Manual searches of the references from these studies and Google Scholar added 15 studies that were not initially captured. Overall, 126 studies were selected (see Fig. 1).

Fig. 1
figure 1

Flow Diagram of Study Selection. WHP indicates studies of Worksite Health Promotion that have addressed obesity. Search was also limited to studies that aimed to assess the effect of obesity on occupational or general physical functioning, rather than work-related risk factors for obesity. Studies using simulation software and loaded normal weight subjects were also excluded

Full size image

Selected studies review process and data extraction

Selected articles were reviewed and the following information was extracted: publication year, country of origin (based on the first author), main research question, design, outcomes measures, subject population, primary method of obesity measurement and the corresponding cut-offs used, whether or not subjects’ weight status was self-reported, additional obesity classification methods applied, the statistical method used, sample size and number of subjects in overweight/obese (OW/OB) subgroup, whether or not the study included women in the sample, the main finding and if the results indicated significance of the obesity (and overweight if applicable) effect, and finally if the authors mentioned any potential limitations of the obesity measurement method they have used. For studies carried out in a controlled lab setting where normal weight and overweight/obese subjects were compared, the mean BMI (or any other primary obesity classification measurement) of the overweight/obese group was extracted.

Analysis

The main research question was categorized into two types: 1) general physical or mental work-related functioning (GF) and 2) task or body part specific functioning (TBS). The summary of all included studies, together with details about the study relevant to obesity classification are presented in Tables 1 and 2.

Table 1 The 37 studies that explored effects of obesity on general job-related outcomes, ordered chronologically
Full size table
Table 2 The 89 studies that explored effects of obesity on task or body part specific functioning, are categorized into 7 groups based on their main focus and ordered chronologically within groups
Full size table

Results

Within the selected time period (2000–2015), there has been an increasing trend in the number of studies published (see Fig. 2), with 2013 having the maximum number of publications (n = 23). This increase is mostly owed to the expanding interest in the specific effects of obesity rather than the general effects, which have been steadily studied by, on average, n = 2.4(SD = 1.9) studies per year. Overall, among all included studies, 63% relied solely on BMI to distinguish obese from non-obese and further classify them into distinct obesity status sub-groups (see Fig. 4). This is particularly concerning because some of these were lab-based studies with sample sizes as small as 12 or used young adults or older adults enrolled in aging research as subjects. In the following sections, studies focused on general vs. specific effects of obesity are discussed separately.

Fig. 2
figure 2

Over the period of the review, there was an increasing interest in obesity research

Full size image

Category 1: General physical or mental work-related functioning

Among the 126 reviewed studies, 37 were related to the general effects of obesity as they pertain to occupationally-relevant outcomes such as performance, disability and discharge rate, healthcare cost, and overall well-being (see Table 1). The majority (64%) of the studies were from North America (see Fig. 3). Over the period of the review, the topic of general studies has gradually moved from work performance and workplace costs associated with obesity to the potential reasons behind elevated costs and poor performance, such as musculoskeletal symptoms and mental health issues. These studies applied a wide range of designs, with cross-sectional being the most frequent (15), followed by longitudinal prospective studies (13). Participants in 10 studies were army personnel, police officers, or career firefighters and the rest were either civilian labor force (20) or their occupational status was not reported or relevant to the topic (7).

Fig. 3
figure 3

The origin of the studies. As expected, North America had the highest number of publications in both GF and TBS categories

Full size image

With regard to the measurement of obesity and group classification, in over 71% of these studies BMI was the only obesity measure used to distinguish obese from non-obese (see Fig. 4), with about 57% of these studies using self-reported weight and height to calculate BMI. About 13% of general studies used additional anthropometric measures such as WC and WHR to enhance obesity measurement accuracy. Finally, of the 6 studies using a direct adiposity measure, 5 were studies of army personnel, fire fighters, or police officers. Four studies reported using cut-off values other than 25 and 30 kg/m2 to categorize subjects into distinct BMI subgroups, out of which two were army studies, one included Asian participants, and one provided no justification to use BMI ≤ 26 kg/m2 as the cutoff for grouping. The median sample size was 1284 (14–69,515). With the exclusion of two studies that did not report the number of obese/overweight subjects included in their sample, on average 55.7(0.2) % of the samples consisted of OW/OB. Only 5 studies in this category provided information regarding the mean body mass index (or any other primary obesity classification measurement) of the OW/OB group. Overall, in 11 studies, the authors discussed the possibility of subject misclassification due to a reliance on BMI as the sole indicator of obesity status, either as a justification to use additional measures (2 studies) or as a limitation.

Fig. 4
figure 4

The obesity measurement methods used in studies. Regardless of the study category, BMI was the most frequently used method of obesity measurement

Full size image

Category 2: Task or body part specific functioning

The majority of the reviewed studies (89) investigated a wide range of specific effects of obesity (see Table 2). North America and Europe contributed by 45 and 39% of such studies, respectively (see Fig. 3). Authors from Italy in particular contributed 16% of the publications, ranking higher than Asia and Australia, with five and six studies respectively. It should be noted that region of origin did not systematically affect the measurement approach used. Although the majority of the studies in this category were laboratory-based observational studies, BMI was still the most frequently used measure, with 59% of the studies relying solely on it to distinguish obese from non-obese. Study topics varied broadly, however, they were categorized in seven groups based on their main hypothesis and research focus (shown in Table 2). These groups, ordered based on number of studies, are discussed in more details as follows.

Twenty-three studies (~ 26%) discussed how obesity alters outcomes related to gait, such as metabolic cost, preferred speed, spatio-temporal parameters, and joint moments. From 2006 to 2014, at least one study related to the effects of obesity on gait was published each year. All of the studies in the gait category were lab-based observational studies. While other studies recruited subjects from a general population, obese subjects in three European studies were females, sampled from obesity clinics. The median sample size was 28 (10–244). More than half of the included subjects (56.7%) were categorized as OW/OB (only two had an overweight group). There were 11 studies which used BMI only (see Fig. 5). With the exclusion of the studies which reported sex or condition-stratified averages (4), the average BMI for nine studies were ≤ 35 kg/m2, four were ≤ 40 kg/m2 and four were > 40 kg/m2. All but four studies reported a significant main effect for obesity or overweight on their outcomes of interest. It is noted that three of the studies reporting non-significant results used BMI as the sole obesity measure.

Fig. 5
figure 5

The obesity measurement methods used across the 7 sub-categories of studies that explored effects of obesity on task or body part specific functioning. With the exception of studies in gait categories, anthropometrics, in particular BMI, were still the most frequently used measures of obesity

Full size image

The next largest group focused on the prevalence, incidence, burden, and changes in symptoms of diseases such as carpel tunnel syndrome, osteoarthritis, low back pain (LBP), asthma, and sleep disorders in association with obesity. This category included some large scale public health studies, hence there was more diversity in terms of study design. The median sample size was 384 (30–38,924). With the exclusion of two studies that did not report the proportion of OW/OB, on average 54 (24) % of the samples were obese or overweight. Six studies used patients and hospital admits as participants and six studies reported subjects belonging to a certain occupation. Eighteen studies relied solely on BMI, two added %BF and one added WC. Sixteen studies failed to report the obesity class of the obese group. In the four that did, all but one had mean BMI ≤ 35 kg/m2. Only two studies, which both had one additional obesity measure, mentioned the inadequacy of BMI.

Changes in functional capacity were the topic of 16 studies. Functional capacity encompasses all topics related to muscle strength, endurance, functional reach, range of motion (RoM), and motor control behavior. Participants in two studies were outpatient clinic or hospital patients (endocrinology and body mass reduction admits) and the rest were recruited from the general population. Eight studies used BMI as the primary and only obesity measure, while three studies also measured body fat. Four studies augmented BMI with other anthropometric measures and one study reported using four obesity measures including both direct and indirect. While no studies relied on self-reported height and weight data, three studies used cut-offs other than 25 kg/m2 and 30 kg/m2 to classify subjects into distinct groups. Only three studies had an overweight sample as well as obese. The median sample size was 40 (20–306). On average ~ 60% of the sample were OW/OB and the majority of reported mean BMI values were in the range of 35–40 kg/m2. One Australian study in particular, which used BMI and cut-off values of 23 kg/m2 and 27.5 kg/m2, had mean BMI of 30 kg/m2 for obese. All but three studies reported some significant obesity effect and two reported significant overweight effect. Authors of six studies, out of which five had used multiple anthropometric measures, included a mention of BMI’s limitation as a measure of obesity.

Issues related to balance, postural stability, and plantar pressure were discussed by 13 lab-based studies. Subjects were recruited from the general population in all but three studies, two of which had sampled from orthopedic rehabilitation and obesity treatment clinic patients. In one study, %BF was the primary obesity measurement used, but eight studies used BMI as the primary and only measure. Two other studies used both %BF and other anthropometric measures, and two studies used both BMI and WHR. It is noted that the two studies with the largest sample sizes used %BF measured by BIA. One included obesity clinic patients and the subjects in the other were part of a health promotion program for postmenopausal women. The median sample size among studies using only BMI was 40.5 (17–120) out of which on average 51(14) % were obese. No study in this sub-category included overweight subjects. Two studies, one testing Canadian and one testing Portuguese subjects reported using BMI cut-offs other than 25 and 30 kg/m2. With the exception of three studies, the average BMI reported for subjects was above > 35 kg/m2 and in four studies the mean BMI was > 40 kg/m2. In terms of significance of the obesity effect, two study reported no significant effect and four reported some but not all outcome measures to be significantly affected by obesity. Only two studies had a mention of inadequacy of BMI, and they both used BMI only.

The effects of obesity on functionality while performing specific occupationally-relevant tasks was investigated by nine lab-based studies. All but one study tested physical tasks such as the sit-to-stand movement, lifting, seated reach, cycling, and running. The remaining study focused on cognitive control. Three studies published by Italian authors tested hospital admits or recovering patients for body mass reduction or metabolic disorders. One study from Switzerland also recruited from individuals who were going to participate in a weight loss program at a hospital. Five studies relied on BMI only, while two added %BF measured by BIA and two added WC. The average sample size was 28.2 (SD = 14.5) and 60% of the included sample were obese. Only one study had an overweight group as well as obese. With the exception of two studies, the reported mean BMI for obese group was > 35 kg/m2. Five studies observed a significant effect of obesity on the performance of the specific tasks tested, while four reported no or partial effect. Three authors discussed how BMI is not the ideal obesity measure although only one used WC in addition.

Three studies discussed changes in physiological responses by obesity and the topics of four studies were not closely pertinent to the above mentioned subgroups. Details of these studies were reported in Table 2.

Discussion

Researchers worldwide have investigated the effect of obesity (sometimes including overweight) on a wide range of occupationally-relevant outcomes. Experts from diverse disciplines, including but not limited to, public health, medicine, health sciences and engineering, have contributed to our current understanding of the magnitude of an effect of obesity at work [37]. The diversity of scientific disciplines involved in obesity research has both advantages and disadvantages. It allows for more complex aspects of the obesity effect to be revealed by diverse methodologies. However, it increases the risk of misuse of methods due to unfamiliarity. In particular, the investigators’ understanding of obesity and the methods to measure it and classify individuals into distinct risk groups can affect the quality of the findings.

The present study focused primarily on examining the use of various obesity measurement methods and secondarily on sampling strategies. Two categories of publications were considered: those investigating the effect of obesity on occupational disease development or business outcomes and those studying how obesity alters task-level performance or functional capacity. As expected, studies in the first category had large sample sizes and were mostly public health studies, carried out by public health professionals. While the samples mostly consisted of participants from the general population or a certain occupation, the large sample sizes justified the use of BMI as the sole obesity measure in over 70% of these studies. It is noted that the vast majority of the publications in this category failed to report descriptive statistics regarding the obesity status of the obese group included in the sample. This could serve as a critical source of information for comparative analyses. The WHO expert consultation [38] suggests that wherever possible, researchers should use all BMI categories for reporting purposes, in order to facilitate international comparisons (i.e., 18.5, 20, 23, 25, 27.5, 30, 32.5 kg/m2, and in many populations, 35, 37.5, and 40 kg/m2).

Another issue with studies in the general category is in regard to abdominal obesity. It is often defined using waist circumference, especially in recent years, while waist-hip ratio was often used in earlier years. However, various cut points have been recommended over time, by different health organizations and across countries, and used across studies. Abdominal obesity is a major component of metabolic syndrome, a cluster of metabolic abnormalities that carry an increased risk of cardiovascular diseases and diabetes [39]. However, there is a subset of the obese population that are metabolically healthy and their inclusion in study samples can confound the results. Ortega et al., [40] studied a large cohort of 43,265 individuals and reported that when adjusting for fitness and other confounders, metabolically healthy but obese individuals had lower risk of all-cause mortality, non-fatal and fatal cardiovascular disease, and cancer mortality than their metabolically unhealthy obese counterparts. In their study, over 46% of the obese sample were metabolically healthy. From the reported exclusion criteria in the studies reported here, it cannot be decided whether obesity would have the same effects in the absence of other components of metabolic syndrome, particularly for outcomes such as healthcare cost, job disability, absenteeism, and presenteeism.

The studies in the second category focused on task-level performance or functional capacity. There are three main points of discussion identified for these studies: 1) selection of obesity measurement(s) (e.g. BMI, WC, %BF) and the corresponding cut-points to distinguish obese from non-obese, 2) the study participants, both in terms of sample size and the population targeted (e.g. young adults, certain occupation groups, hospital admits), and 3) measurement considerations (e.g. site of measurement for WC). While these factors are all individually important, their interaction may also present a challenge to studies. For instance, when using BMI in a study with a small sample size, recruiting only young adults may be more problematic [41, 42] than when a larger group of older adults are classified based on BMI.

In this category, BMI was still the most frequently used obesity classification measure. Overall, the selection of an obesity measure should depend on the hypothesized mechanism by which the obesity effect would manifest. While obesity presents by both changes in anthropometry and metabolic function, acknowledgement of the considered causal pathway is advantageous to study rigor. Also, while obesity morphology may not be as crucial to the outcomes in the studies of the previous category, it is highly relevant to the dependent variables investigated by the studies in this category. In particular, balance and gait parameters are likely to be affected by the distribution of weight in the body, therefore not only obesity status, but also fat distribution needs to be taken into account. BMI by itself fails to do so, however other anthropometric measures such as WC and WHR are able to distinguish central obesity from lower body and general obesity. Across the 36 studies in the two aforementioned sub-categories, only 10 studies used additional anthropometric measures.

Caution should be made in the use of BMI in studies with small samples that include young adults. Statistically significant age dependencies have been reported in the relation between %BF and BMI, such that older adults have higher %BF compared with younger adults with comparable BMIs [25]. WHO expert consultation acknowledges the issue by stating that most studies show the relation between BMI and %BF to be dependent on age and sex, and also different across ethnic groups. Experts affirmed that Asian populations have different associations between BMI and percentage of body fat than do Western populations [38], however, due to lack of comprehensive data from all Asians, they suggested retaining WHO BMI cut-off points as international classifications. Using ethnic-specific cut-offs may come at the expense of consistency among studies. As such, we observed two studies from Portugal in the balance sub-category that participants were recruited from the general population, one using a BMI cut-off of 25.5 kg/m2 and the other using 30 kg/m2to distinguish obese from non-obese. Arbitrary grouping of subjects, not backed up by ethnic or other expected underlying differences, as was the case in these two studies, should also be minimized. Overall, it is alarming that only 20% of the studies in this category acknowledged the aforementioned shortcomings of BMI as the obesity measure.

The majority of studies in this category (~ 80%) were observational studies. To isolate an obesity effect, and in contrast to the majority of studies in the first category, subjects were selected such that they were mostly otherwise healthy. The representativeness of this group and the extent to which the findings based from them can be generalized to the overall obese population is concerning. This exclusion of obese with comorbidities from the study samples in this category and their possible inclusion in samples of the first category may contribute to the higher proportion of publications in the first category to report a significant obesity effect in comparison to the second category.

Another issue with the sample representativeness is including only severe obesity (classes II (BMI 35–39.9 kg/m2) and III (BMI ≥ 40 kg/m2). While this practice may be statistically sound and increase the likelihood of capturing the obesity effect, it again limits the generalizability of the result. For instance, in the United States the prevalence of obesity is estimated to be over 35% but less than 15% of the obese population (~ 5% of the total population) are categorized as class II and less than 7% (~ 2.5%) as class III [43].

There are considerations for proper use of each measurement as well. WC for instance is shown to be significantly different across sites of measurement, postures, phases of respiration, and time since last meal [31]. By following the existing measurement guidelines [44] studies are less prone to error and consistency across subjects and studies is also warranted. Also, practices such as having a single trained staff doing all the measurements when possible, keeping the measurement conditions homogenous across all subjects and using multiple measurements are beneficial for internal validity and worthy of report in research manuscripts.

Overall, we assessed obesity research in the occupational health field and showcased the practices of obesity measurement since 2000. The present study has many strengths, but also some limitations. While obesity has become a global epidemic, this review was limited to PubMed database as well as Google Scholar journal articles available in English, primarily due to the authors’ time and language proficiency constraints. Also, studies related to health promotion at work were excluded [45, 46]. Health risk assessment is a common part of these programs in which obesity status is commonly assessed as a health risk, however the topic of these studies were beyond the scope of this review. Moreover, although the effect of certain work types, such as shift work on the onset of obesity among workers is of importance and has been widely studied [47], this review focused on the obesity effect on occupationally-relevant outcomes.

Conclusion

Obesity is a serious global public health threat. In order to build up a comprehensive profile of its effects, it is crucial to have easy-to-use yet reliable measures that allow for classification of individuals into distinct risk groups. A large body of research has been conducted in the occupational health field regarding obesity. Use of indirect measures such as BMI may be justifiable in large scale public health studies due to their ease of use and low cost. However, due to limitations of these measures, cautious use of them is suggested as the sole obesity measure in small-scale observational studies.