Body mass index (BMI) is an important tool used by clinicians, epidemiologists and public health officials for the categorization of individuals based upon their relative weight. It has become the most commonly used measure of weight status due to its simplicity of calculation when collecting data for large population surveys.1 However, BMI is a measure of weight and height and does not directly measure adiposity, limiting its use for measuring levels of obesity.
Waist circumference has been shown to be a more accurate measure of body fat and therefore would offer an alternative to BMI.2 However, this does not mean that BMI should be discounted. It is important to understand how useful BMI is at estimating risk of health outcomes in comparison to waist circumference. This is important as self-reported data are easier to collect and inexpensive for large populations. More precise techniques for measuring obesity are not practical for large epidemiological studies or routine clinical usage. In this study, a comparison of BMI and waist circumference as measures of risk to multiple health outcomes is examined.
Individual-level data were taken from the Yorkshire Health Study (2010–12; n = 18 562, aged 16–85).3 Logistic regression models using BMI and waist circumference separately (both standardized using z-scores to improve their comparability) as explanatory variables against a series of chronic health conditions, illnesses or disabilities (separate outcomes variables). Unadjusted and adjusted models were produced, controlling for the following confounders of poor health: age, sex, ethnicity, deprivation (measured using the Indices of Deprivation 2010), smoking status, alcohol intake (units per week) and physical exercise levels. Data were self-reported.
Table 1 presents the results from the analysis. BMI and waist circumference were statistically significant predictors of multiple health outcomes, independent of known confounders. An increase in value of either measure results in a larger risk of an individual having a chronic health condition. Diabetes had the highest risk across both measures, with stroke and cancer less related to body size after controlling for known confounders. The analysis was repeated stratifying by age group. For adults (25–64) and the elderly (65+), the results were similar. However, for young adults (16–24), the results were mostly insignificant due to the decreased prevalence of health conditions in the young.
Table 1.
Results from unadjusted and adjusted logistic regression models explaining multiple health outcomes using separate models for body mass index and waist circumference (standardized using z-scores)
| Outcome |
BMI
|
Waist circumference
|
||
|---|---|---|---|---|
| Unadjusted | Adjusted | Unadjusted | Adjusted | |
| Fatigue | 1.422*** | 1.336*** | 1.507*** | 1.377*** |
| Pain | 1.533*** | 1.416*** | 1.587*** | 1.419*** |
| Insomnia | 1.251*** | 1.174*** | 1.232*** | 1.175*** |
| Anxiety | 1.196*** | 1.150*** | 1.154*** | 1.159*** |
| Depression | 1.362*** | 1.344*** | 1.339*** | 1.334*** |
| Diabetes | 1.841*** | 1.952*** | 2.238*** | 2.083*** |
| Breathing problems | 1.274*** | 1.198*** | 1.415*** | 1.264*** |
| High blood pressure | 1.660*** | 1.710*** | 1.804*** | 1.657*** |
| Heart disease | 1.350*** | 1.359*** | 1.691*** | 1.401*** |
| Osteoarthritis | 1.433*** | 1.431*** | 1.411*** | 1.337*** |
| Stroke | 1.163*** | 1.083 | 1.487*** | 1.156* |
| Cancer | 1.083* | 0.961 | 1.238*** | 1.011 |
| Any condition | 1.599*** | 1.382*** | 1.658*** | 1.374*** |
Models adjusted for age, sex, ethnicity, deprivation, smoking, alcohol intake and physical exercise.
*P < 0.05.
***P < 0.001.
Pearson's correlation coefficients were calculated for the unadjusted values of the odds ratios for both BMI and waist circumference (r = 0.866, P < 0.001) and the adjusted odds ratios (r = 0.965, P < 0.001). The correlation values show closer agreement once known confounders were controlled for, with odds ratio values being similar. This would suggest that there is little difference in the measures once known confounders are controlled for.
The analysis has indicated that BMI remains a useful measure for estimating risk of health outcomes in a large and representative sample. There was little difference between the measures once known confounders were controlled for. Different measures may be better for assessing individuals; however, BMI is still useful in a population setting and should not be discounted.
Conflict of interest
None declared.
Acknowledgements
This publication presents independent research by the National Institute for Health Research Collaboration for Leadership in Applied Health Research and Care for South Yorkshire (NIHR CLAHRC SY) a pilot which ended in 2013. Further details about the new NIHR CLAHRC Yorkshire and Humber can be found at www.clahrc-yh.nihr.ac.uk. The views and opinions expressed are those of the authors, and not necessarily those of the NHS, the NIHR or the Department of Health.
References
- 1. Shah NR, Braverman ER. Measuring adiposity in patients: the utility of body mass index (BMI), percent body fat, and leptin. PLoS ONE 2012;7:e33308. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Snijder MB, van Dam RM, Visser M, et al. What aspects of body fat are particularly hazardous and how do we measure them? Int J Epidemiol 2006;35:83–92. [DOI] [PubMed] [Google Scholar]
- 3. Green MA, Li J, Relton C, et al. Cohort profile: the Yorkshire Health Study. Int J Epidemiol. In press. http://ije.oxfordjournals.org/content/early/2014/07/09/ije.dyu121.full. [DOI] [PMC free article] [PubMed] [Google Scholar]