May 4, 2015

expert reaction to study investigating body weight in type 2 diabetes, and mortality

A relationship between BMI of patients with type 2 diabetes and prognosis was investigated in a study published in the journal Annals of Internal Medicine, where the authors reported that those who were overweight or obese were more likely to be hospitalised for cardiovascular events, but that those who were overweight (but not obese) had a lower mortality risk.

Prof. Nick Finer, Honorary Professor, National Centre for Cardiovascular Prevention and Outcomes, UCL Institute of Cardiovascular Science, UCL, said:

“The so called obesity paradox rears its head again. Over the past few years, a number of epidemiological papers have suggested that there may be a health advantage from being overweight compared to a ‘normal’ weight. In this study, Costanzo and colleagues report that in Hull, compared to being a normal weight, patients with diabetes survive longer if overweight but not obese, despite being more likely to have a cardiovascular event requiring hospitalisation. There are however major problems with this sort of study that make the conclusions insecure and it would be unhelpful for people with diabetes to take home the message that it is OK to be overweight.

“Body Mass Index (BMI: weight/height x height) is known to be an imperfect measure of body fat mass. Defining people by this measure alone will lump together individuals with substantial differences in nutritional status, disability, disease, and mortality risk into a particular BMI category.  Furthermore, the apparent reduced all-cause mortality risk in the overweight people could be an artifact. There is a u-shaped relationship of BMI and mortality such that people with a BMI between 18.5 and 22 have higher mortality than those with a BMI between 22 and 25. So putting both of these together into a single group of BMI 18.5-25, means that the higher mortality in the 18.5-22 BMI range will inevitably drive an increased mortality rate for the  ‘normal’ weight (or comparator) group as a whole. This would then favour the overweight groups of BMI 25-30. Waist circumference is a better predictor of health risks than BMI over the ‘normal’ and ‘overweight’ ranges. Its use is recommended in all guidelines, so the lack of data on waist circumference is a significant limitation.

“The study was on a mixture of patients attending the Hull clinic between 1995 and 2005. We know nothing about their treatment before attending, whether the clinic population represents the wider population with diabetes (many of whom will have been solely treated in primary care), or whether at presentation and inclusion to the study their weight was stable, falling or increasing.

“There are other serious confounding issues. Weight loss is a feature of poorly controlled diabetes, so those with a low BMI may have a lower body weight because they are sicker, and thus at greater risk than the overweight individuals. To reduce the risks of thinner patients being more ill, the authors excluded patients with chronic lung or kidney disease, which further limits the generalisability of the findings, but these are not the only chronic illnesses or lifestyles associated with weight loss. One example is that smokers have lower body weights and increased risks of cardiovascular disease and mortality. The study adjusted for smoking status, but not the number or duration of cigarettes consumed. Another example relates to heart failure, where previous reports have suggested an advantage of being overweight if you have this condition, although perhaps surprisingly this was not seen in this study. The ‘obesity paradox’ is probably best explained by the fact that people may be thinner because they are ill (so making it appear safer to be overweight). This effect has been termed ‘reverse causality’ – i.e. it is illness driving the lower body weight rather than the other way around. The technique of excluding people who died within two years of inclusion in the study (only done as a sensitivity analysis) may be inadequate to adjust for this ‘reverse causality’ since chronic illness that may exert its effects over a much longer time course.

“The study had no information of medication use in the subjects and it is quite possible that those with type 2 diabetes were more intensively treated for their blood glucose, blood pressure and cholesterol (risk factors for cardiovascular disease) if they were overweight. Since the overweight had a higher rate of hospitalisation for cardiovascular events one might speculate that they were then more carefully and intensively treated for their risk factors – i.e. a bias towards better care reducing mortality in the overweight. Another consideration is that thinner people with diabetes may have different types of diabetes from the overweight and obese, and may be more prone to hypoglycaemia (low blood glucose) since they are likely to remain more sensitive to insulin. Hypoglycaemia is known to be dangerous and increase mortality – a further bias to the finding of increased mortality in the leaner group.

“As the authors state, the study tells us nothing about the manifold benefits of weight loss in the overweight and obese demonstrated from a growing body or research. While the authors suggest appropriate caution in the clinical implications of their findings, the study, which has significant limitations, could add to the sometimes confusing messages received by the public. I would not want to see any headlines suggesting it is healthy to be fat, as this is not what the entirety of current evidence shows. Losing weight and getting fitter are key elements of diabetes care.”

‘The obesity paradox in type 2 diabetes mellitus: relationship of Body Mass Index to prognosis’ by Pierluigi Costanzo et al. published in Annals of Internal Medicine on Monday 4 May 2015.

Declared interests

Prof. Nick Finer is also Consultant Endocrinologist and Bariatric Physician at University College Hospitals, UCLH Centre for Weight Loss, Metabolic and Endocrine Surgery.