Sensitivity and Specificity issues regarding BMI thresholds
By Steven B. Halls, MD FRCPC
Many medical tests have issues with sensitivity and specificity. What do these words mean?
Sensitivity: is the ability to correctly detect disease.
Specificity: is the ability to avoid calling Normal things as disease.
A perfect medical test would have 100% sensitivity and 100% specificity. It would positively identify all the true cases of disease, and it would never mislabel anything normal as disease.
But very few tests are perfect. Using Body Mass Index as a medical test for labelling "overweightness" as disease, is a classic example of an imperfect test in many textbooks.
To achieve 100% sensitivity, you could propose that ANY BMI greater than 0 is "overweight", which would correctly find everyone who is truely overweight (100% sensitivity), but it would incorrectly mislabel everyone Normal as falsely overweight ( 0% specificity). Not good.
When a medical test is imperfect, scientists try to strike a balance between sensitivity and specificity. To do this, they plot sensitivity and 1-specificity on a graph, called a "ROC curve". (ROC means Receiver-Operator-Characteristic). Below is an example ROC curve chart.
The Red line is a better medical test than the Black line, because the curve of the Red line comes closer to the Green asterisk. To pick the best point along the ROC curve, scientists generally look for the shortest distance from the Green asterisk, to the Red line. In this case, the best point gives 0.81 sensitivity (81%) and 0.21 false positive fraction ( 79% specificity).
Sometimes the scientific "shortest distance" method doesn’t always give the "best" choice. Sometimes people want a test with HIGH sensitivity. For example, when the blood bank tests blood for the AIDS virus, they want a test with 100% sensitivity, even though some false-positive test results will occur. Conversely, before doing an autopsy, the pathologist tests for the "presence of death", requiring HIGH specificity, so as to avoid autopsying someone who isn’t dead.
In the case of "Overweightness" and "Obesity", some people, (usually thin people), believe that overweightness is a serious disease, and they want to persuade politicians to spend extra money on the problem. They may choose to set a BMI threshold fairly low, to increase apparent prevalence of the disease. But that isn’t a very popular view with the general public.
People get angry if they are labelled as overweight, when they are not. But overweight people don’t mind being called normal. As a matter of popular acceptance, I believe a Body Mass Index threshold should have higher specificity than sensitivity.
I have commented on several published scientific articles about body mass index and obesity. I don’t always agree with the authors. You can read my comments by clicking on the links below.
- Wellens RI, Roche AF, Khamis HJ, Jackson AS, Pollock ML, Siervogel RM.
Relationships between the Body Mass Index and body composition. Obes Res, Jan 1996;4(1):35-44
- Price RA, Sorensen TI, Stunkard AJ. Component distributions of body mass index defining moderate and extreme overweight in Danish women and men. Am J Epidemiol, Jul 1989;130(1):193-201
- Frankenfield DC, Rowe WA, Cooney RN, Smith JS, Becker D. Limits of body mass index to detect obesity and predict body composition. Nutrition, 2001 Jan;17(1):26-30
- Curtin F, Morabia A, Pichard C, Slosman DO. Body mass index compared to dual-energy x-ray absorptiometry: evidence for a spectrum bias. J Clin Epidemiol, Jul 1997;50(7):837-43
- Lazarus R, Baur L, Webb K, Blyth F. Body mass index in screening for adiposity in children and adolescents: systematic evaluation using receiver operating characteristic curves. Am J Clin Nutr, Apr 1996;63(4):500-6
- Sardinha LB, Teixeira PJ. Obesity screening in older women with the body mass index: a receiver operating characteristic (ROC) analysis. Science and Sports, Jul 2000; 15: 212-9
Back to the halls.md v2 criterion description.
- Vasconcelos Fde A, Cordeiro BA at al. Sensitivity and specificity of the body mass index for the diagnosis of overweight/obesity in elderly. (They say, the WHO BMI criteria are not good indicators of obesity for the elderly, and I agree.)