Exclusion criteria for reference studies: Context is key

We read with interest the response ¹ to our letter ² on pregnancy-specific reference intervals and the significance of body mass index (BMI). We welcome the opportunity to address some of the key points.

‘The values used in determining reference intervals are always selected from disease-free subjects’.

The IFCC’s series on defining reference intervals recommends that ‘health is not a well-defined condition and has come to be regarded as relative’, and ‘…in selecting reference individuals, the criteria of health applied are dictated by the aim of the laboratory investigation. Thus, [reference intervals] may not always be healthy subjects’. ³ If a state of absolute health does not exist, a transparent approach is required for risk stratification in reference studies. Such an approach may include the use of risk factors as proxies. For example, a reference study for a novel cardiac marker should probably exclude those with a personal history of myocardial infarction, but it may also be reasonable to exclude those with a higher risk of unreported or undiagnosed disease, like obesity. In contrast, while not strictly ‘healthy’, such a reference population may include individuals with asthma or arthritis. Context is key.

‘Obesity is a biological disadvantage to the person and thus fulfils the definition of a disease’.

Older age, female sex and obesity predispose to the development of breast cancer, but neither of these is a disease. While being mindful of the aforementioned potential utility of proxy markers, caution should be taken to differentiate between diseases and their risk factors.

‘Just because more than 50% of women in Britain and Ireland are overweight or obese is not a reason to pretend all are “normal”. Interpreting tests in the context of BMI is indeed important if the interpretation recognises that obesity is an illness.’

Whether one chooses to classify obesity as a disease or not, the fact is that women with a higher BMI constitute a significant proportion of the pregnant population, and they need reliable diagnostic tests. If the reference interval for an analyte differs by BMI there are two options: 1) create ‘high’ and ‘low risk’ alternatives a priori or 2) recruit inclusively and partition properly. Ultimately, the desired outcome is the same: ensuring that clinical tests are interpreted safely and appropriately in the context of the target population. We strongly believe that exclusion based on BMI should be the exception, not the rule.