Precision and Accuracy of Defining Obesity

The fifth item on the  disease definition modification checklist developed by the Guidelines International Network (G-I-N) Preventing Overdiagnosis Working Group published in JAMA Internal Medicine,  deals with issue of precision, accuracy, and reproducibility.

Obviously, any definition of obesity that requires clinical assessment and clinical judgement will not have the precision, accuracy, or reproducibility of simply measuring height and weight.

Thus, if we define obesity as the presence of abnormal or excess body fat that impairs health, we will necessarily have to deal with the issue of assessing health, which is not something that you can simply measure by stepping on a scale.

Rather, because abnormal or excess fat can affect virtually every organ system as well as psycho-social well-being, we are going to be faced with a rather complex system of diagnosing who has obesity and who hasn’t.

In fact, as the authors of the checklist point out,

“…an appropriate gold standard will rarely be available and therefore, traditional measures of diagnostic test accuracy, such as sensitivity and specificity, will generally not be appropriate.”

Both repeatability (agreement in identical conditions) as well as reproducibility (agreement across comparable conditions) may result from biological variability, analytical variability, and clinical judgement.

The only way to test the reproducibility and precision will be to evaluate the use of the new definition in clinical practice and ultimately determining whether or not clinicians can reasonably agree on who has the condition and who doesn’t.

While this may seem daunting to non-clinicians, let us remember that in clinical practice many diagnoses are dependent on clinical evaluations and clinical judgement, whereby experienced clinicians or specialists may perform better than the novice or the non-specialist (a good example is psychiatry, but there are countless other examples).

Moreover, there will always be grey areas in “borderline” cases, where examiners may disagree on the exact result and only time will tell, who is right.

Welcome to the messy world of clinical practice.

Just because BMI is simpler, more precisely measured, and more reproducible, does not make it a better measure of diagnosing whether or not someone actually has a disease.

After all it only makes sense that it will take a complex definition to diagnose a complex disease.

@DrSharma
Edmonton, AB