The Obesity Paradox in Type 2 Diabetes Mellitus: Relationship of Body Mass Index to Prognosis: A Cohort Study

Costanzo et al (May issue) investigated the prospective relationship among obesity, CVD, and mortality in a large clinic-based cohort of people with type 2 diabetes. They found that patients whose baseline weight was below the normal range (BMI of 18.5-24.9 kg/m2) had a higher all-cause mortality risk, while being obese had not significant impact and being overweight even conferred a significant protection (HR=0.87, 0.79-0.95), particularly in people 60 years of age (1). The Authors have omitted to compare their findings to our earlier results from Casale Monferrato study (2-3) that support and strengthen their observations. The Casale Monferrato study is an on-going population-based study and analyses accounted for relevant risk factors, such as HbA1c, albumin excretion rate and apoA1/apoB, that were not included in the present analyses. In the first (1992-1992) survey (1,475 patients with a 10 year mean duration of type 2 diabetes, 25% with coronary heart disease and 972 deaths over a median 10 year follow-up), people with a BMI in the upper quartile (> 30.0 kg/m2) had an adjusted HRs of 0.75 (0.58-0.96) for all-cause and 0.67 (0.45-0.95) for cardiovascular mortality compared to people in the lower quartile of the BMI distribution (<24.2 kg/m2, 18 patients only with BMI < 18.5 Kg/m2). Results did not change after exclusion of smokers, patients with previous CHD or subjects who died within 2 years of follow-up and, of interest, this protective effect was observed exclusively in the elderly (age ?65 years). Therefore, both Costanzo and our results suggest that in the elderly the presence of overweight/obesity might confer an advantage on the mortality risk that requires further investigation. A modification effect by age was also confirmed in the second (2000) survey of the Casale Monferrato Study (2,272 patients with type 2 diabetes with a 5.4 year median follow-up). In this study, central obesity conferred an HR for cardiovascular mortality of 2.15 (1.18-3.92) in those aged < 70 years (median age of the cohort), which remained significant after multiple adjustments, including the negative confounding effect of NT-proBNP, whereas no association was evident in those aged 70 years and over.
1. Costanzo P, Cleland JG, Pellicori P, Clark AL, Hepburn D, Kilpatrick ES, Perrone-Filardi P, Zhang J, Atkin SL. The obesity paradox in type 2 diabetes mellitus: relationship of body mass index to prognosis: a cohort study. Ann Intern Med. 2015;162:610-8.
2. Perotto M, Panero F, Gruden G, Fornengo P, Lorenzati B, Barutta F, Ghezzo G, Amione C, Cavallo-Perin P, Bruno G. Obesity is associated with lower mortality risk in elderly diabetic subjects: the Casale Monferrato study. Acta Diabetol. 2013;50:563-8.
3. Bruno G, Barutta F, Landi A, Cavallo Perin P, Gruden G. The effect of age and NT-proBNP on the association of central obesity with 6-years cardiovascular mortality of middle-aged and elderly diabetic people: the population-based Casale Monferrato study. PLoS One 2014;9:e96076.

In their recent publication, Costanzo et al (1) have analysed data from 10,568 patients attending an NHS diabetes service in England. They have concluded that neither overweight (BMI 25-29.9) or obesity (BMI >30) were associated with an increased mortality risk compared to normal weight (BMI 18.5-24.9). They have also reviewed the studies on this subject to date and discussed their strengths and weaknesses, including our own study of 106,640 patients who were undergoing NHS care in Scotland.
We have two major areas of comment. The first is the inaccuracies in reporting the findings of our own study (2), which is the largest study published in this area. The authors describe it as using retrospective outcomes when it was prospective (using similar methods to their own study), claim we do not exclude underweight when we excluded all patients with a BMI <20 and fail to mention that our endpoints included cancer deaths, split into smoking and obesity related cancers, and respiratory deaths as well as vascular and all-cause mortality. We feel this is important to clarify given the very large size of our study, one of the few using contemporary data.
Our second area of comment is potential reasons for their finding of no increase in risk of mortality with increasing BMI. This is likely due to their choice of reference group including very low BMI; the inclusion of BMI 18.5-20 will have included a number of patients with low BMI due to chronic disease and these would be unusual patients in any type 2 diabetes cohort. Also the use of prevalent diabetes and therefore random BMI without any adjustment for factors such as medications or year of diagnosis will be biased by the effect of medications on weight and factors affecting choice of medications such as glycaemic control or drug availability. Our study used BMI at diagnosis to avoid this potential for bias.
There is no one study in this area which is perfect. Our own did not have data on co-morbid conditions but instead did a number of sensitivity analyses excluding early deaths and smokers. Other studies have used historical data or have insufficient sample sizes to explore cause of death. However those which are the most robust have generally agreed that all cause mortality is higher in patients who are obese and it is in these patients that weight management is a priority. It is probably time to move away from observational epidemiology in this area given the inability to remove all potential bias and the large numbers of studies published to date, and concentrate efforts on lifestyle intervention studies to improve outcomes for obese patients with type 2 diabetes.
(1) Costanzo P, Cleland JG, Pellicori P, Clark AL, Hepburn D, Kilpatrick ES, et al. The obesity paradox in type 2 diabetes mellitus: relationship of body mass index to prognosis: a cohort study. Ann Intern Med. 2015;162(9):610-8
(2) Logue J, Walker J, Leese G, Lindsay R, McKnight J, et al. The Association Between BMI Measured Within a Year After Diagnosis of Type 2 Diabetes and Mortality. Diabetes Care. 2013;36(4):887-93

From their analysis of 10,568 patients with prevalent type 2 diabetes, Dr. Costanzo and colleagues conclude that compared with the normal BMI range (18.5-24.9 kg/m2), obesity (BMI >30 kg/m2) was not associated with higher all-cause mortality, and overweight (25.0-29.9 kg/m2) was associated with a significant lower all-cause mortality risk.(1) A major limitation is their lack of information on participants’ specific causes of death, given the extreme heterogeneity in the all-cause mortality endpoint. In our analysis of 11,427 male and female US health care professionals,(2) we observed a direct linear relationship between BMI and death from cardiovascular causes. For cancer deaths, however, the highest risk was seen among those in the lowest BMI category; further, this was limited to ever smokers. This finding supports our conclusion that underlying chronic illnesses, such as cancer, and confounding by smoking status, are two major sources of bias in the relationship between body weight and mortality.

Removal of participants with illnesses diagnosed at baseline or multivariable adjustment for crudely assessed smoking measures are insufficient to fully account for these biases. Authors should have included multivariable adjusted hazard ratios simultaneously excluding smokers, participants with prevalent chronic disease, and deaths occurring early in follow-up to provide minimally biased estimates. The authors also could have separated the BMI categories further. Combining those with a BMI of 18.5 through 24.9 can be misleading if those with the lowest BMI are likely to be the sickest due to reasons other than diabetes.

The authors propose that normal weight patients with type 2 diabetes may have a more severe form of the disease that is less related to their body weight; however, given their study population is prevalent diabetics, we cannot know whether participants were overweight leading up to their diagnosis, and experienced subsequent weight loss due to other underlying illnesses, medications, or intentional lifestyle modifications. As we observed in our NEJM paper which used pre-diagnosis body weight, the lowest mortality risk was observed among those with BMI in the normal range when potential sources of recent weight loss were addressed. Typically, the “obesity paradox” tends to occur among chronically ill individuals who have lost weight and become frail and are also at an increased risk of death.

References
1. Costanzo P, Cleland JG, Pellicori P, Clark AL, Hepburn D, Kilpatrick ES, et al. The obesity paradox in type 2 diabetes mellitus: relationship of body mass index to prognosis: a cohort study. Ann Intern Med. 2015;162(9):610-8.
2. Tobias DK, Pan A, Jackson CL, O'Reilly EJ, Ding EL, Willett WC, et al. Body-mass index and mortality among adults with incident type 2 diabetes. N Engl J Med. 2014;370(3):233-44.