diabetes risk: BMI less good predictor

 

Another study just came out finding that BMI is an inferior predictor of the development of type II diabetes (see BMI less good predictor for diabetes DiabReseClinPract2024 in dropbox or https://doi.org/10.1016/j.diabres.2024.111888)

 

Details:

-- 155,623 Japanese participants who had medical checkups when working at the Panasonic Corporation between 2008 and 2021, having had two consecutive health examinations during that interval, were evaluated for how several different anthropometric indices of adiposity were associated with subsequent development of type II diabetes

    -- 115,036 men and 40,587 women were evaluated

-- the anthropometric indices assessed:  

    -- BMI

    -- WC, waist circumference

    -- wBMI, waist circumference times the BMI (the waist-corrected BMI)

    -- ABSI , a-body shape index, the waist circumference divided by the BMI to the two thirds power times the square root of the height, a measure of the body shape (different from the body roundness index, see below)

    -- WHtR, the waist-to-height ratio

-- mean age 44, height 165 cm

-- BMI 22, waist circumference (WC) 80 cm, wBMI 20, ABSI .078, WHtR 0.485

-- blood pressure 121/76 for men, 113/70 for women, fasting plasma glucose 90

-- LDL 126 mg/dL in men and 115 in women, HDL 58 mg/dL in men and 70 in women, triglycerides 122 mg/dL in men and 73 in women; raised blood pressure 31% in men and 14% in women, raised triglycerides 26% in men and 6% in women, raised HDL in 10% of men and 9% of women

-- smoking none in 46% of men and 80% of women, past smoking in 16% of men and 7% of women, current smoking in 38% of men and 13% of women

-- alcohol consumption 32% of men and 5% of women, physical exercise 19% of men and 12% of women

    -- high blood pressure was defined as systolic at least 130 mmHg, diastolic of at least 85 mmHg or the use of antihypertensive medications; elevated triglycerides was defined as a level of at least 150 mg/dL or medications for dyslipidemia; low levels of HDL were defined as HDL less than 40mg/dL for men and 50 for women

    -- alcohol consumers were considered anyone drinking a minimum of 20 g of alcohol a day

    -- physical activity was defined as a minimum 30 minutes at least twice a week for a year or more

-- type II diabetes was defined as fasting plasma glucose levels of least 126 mg/dL or use of antidiabetic medications

-- the incidence of diabetes was monitored for 13 years from the individual's baseline to 2021

 

-- Main outcome was a relationship between each anthropometric index and the incidence of type II diabetes, adjusting for age, systolic blood pressure, triglyceride levels, HDL cholesterol levels, fasting plasma glucose levels, alcohol consumption, smoking status, and exercise habits

 

Results:

-- male participants were observed for 791,803 person-years, women were observed for 206,271 person-years

-- 8,005 men (10.11 per 1000 person-years) and 795 women (3.85 per 1000 person-years) developed type II diabetes

-- anthropometric indices: ABSI had a lower hazard ratio of developing diabetes as compared to the others (ie, it seemed to be protective), which were all largely between 1.4 and 1.5 (ie, 40-50% increases) with significant overlap of the confidence intervals

    -- overall, there was not much difference between men and women for all of the anthropometric indices and the development of diabetes, except for ABSI which was apparently protective of the development of diabetes for both men and women

 

-- their table 3 below shows the area under the curve (AUC) of the survival ROC curves for predicting the incidence of type II diabetes at 13 years by each of the anthropometric indices as well as the sensitivity, specificity, negative and positive predictive values, and negative and positive likelihood ratios by sex (note the WHtR was better than BMI for men, and BMI was actually as good as the others for women (this could be because women tend to have less abdominal obesity, and it is abdominal obesity that is associated with the metabolically terrible visceral obesity that leads to diabetes, so their BMI more reflects this more benign obesity: see below)

-- table 4 confirms that for men, BMI is pretty much the worst markers overall and WC, wBMI, and WHtR were the best, with WHtR being the most predictive (ABSI, was again in last place); and that for women, wBMI and WHtR were the best, but no significant difference from BMI

Commentary:

-- prior studies have found that all of these anthropometric indices are associated with type II diabetes incidence, with their differences as follows (for a review of these indicators and the risk of Type II diabetes, see https://www.bmj.com/content/376/bmj-2021-067516 ):

    -- BMI is a pretty nonspecific calculation, not differentiating the types of tissues that increase weight (fat vs muscle, etc), nor the location of the fat if present (abdominal fat is associated with worse prognoses than fat in the thigh region, ie apple shape vs pear shape, since the abdominal fat tends to go along with intra-abdominal visceral fat, which is associated with chronic inflammation, insulin resistance and the related diseases of metabolic syndrome and diabetes)

    -- WC (waist circumference), if measured accurately, does reflect intra-abdominal visceral fat as verified by CT and MRI (the gold standards for assessing visceral fat)

    -- ABSI is associated with increased mortality, and it can be substituted for WC in the metabolic syndrome criteria

    -- WHtR (waist-to-height ratio) is also useful for diagnosing metabolic syndrome

-- a prior study and subsequent blog found that body roundness index (BRI) was a much better predictor of mortality than BMI: https://gmodestmedblogs.blogspot.com/2024/09/body-roundness-index-is-better.html

    -- this study had a different mathematical definition of body shape, referred to as the body roundness index (RBI), instead of the ABSI (which did quite miserably in the current study)

    -- this blog also comments on other bad clinical outcomes, as well as diabetes, that track better with the body roundness index

    -- unlike the current study, there was no comparison to other measurements, such as waist circumference, or waist-to-hip ratio (a more difficult-to-reproduce ratio, since there is more variability to hip measurement), or waist-to-height ratio

-- another study compared different anthropometric measures, with Dual Energy X-ray Absorptiometry (DXA) being the gold standard for body fat distribution: BMI vs other measures Endocrine2022 in dropbox, or doi.org/10.1007/s12020-022-03030-x. They found that the WHtr for males and wBMI for females had the best correlation with fat mass percentage in the body in both those with “overfat” and “obesity” 

-- this current study found that the most useful indicators for high risk of diabetes varied between the sexes, with WHtR being most effective for men and both wBMI and WHtR for women (though BMI was also quite accurate in women, likely because BMI included a larger representation of the fat distribution in women that tends to be skewed away from the more dangerous abdominal visceral fat )

Limitations:

-- of note, the average BMI in this study was 22, which is pretty low. However, the BMI considered to reflect “obesity” in Asian patients is 25 or more, per the Japan Society for the Study of Obesity, still significantly higher than the 22 of this study; in terms of waist circumference in Japan, "obesity" is 85 cm in men and 90cm in women (the numbers in the US are 40 inches for men and 35 inches for women, which is 102 cm for men and 88 cm for women).

    -- for unclear reasons, this study used the cut-off values for BMI being 24.3 for men and 22.7 for women, but lower than the “obesity” definition of 25

    -- also, Japanese people have higher levels of abdominal visceral fat relative to abdominal subcutaneous fat than in Western folks

-- there was a lot of variability in my reviewing the reasonably extensive literature on the issue of the “best” anthropometric measurement to predict adverse outcomes. Pretty much all of them found BMI suffered in comparison with the others. Why is there such a difference in different studies? Likely because of large differences in the demographics, medical comorbidities, genetic predispositions, medications taken, etc in the different populations

-- as with many studies, several of the important comorbidities that have a role in diabetes development were not included (eg diet, exercise), and many that were included were binary (eg smoking) when there is a likelihood that the actual amount of smoking matters, the quitting date matters, passive exposure to smoking probably matters… 

 

So, this study reinforced a few things:

-- we have been using BMI as our marker of overweight categories, though this is pretty clearly not the best anthropometric measurement both by physiology (it is the one least reflective of the real bad actor: visceral fat), as well as clinical outcomes

-- but, as with all cultural issues (this one from the "medical culture"), we do continue to use suboptimal surrogate markers and even some medications because that's what everyone else around us is using

    -- for example, that also seems to be the case with markers of renal dysfunction: cystatin-C makes more sense physiologically (creatinine is influenced by so many meds, muscle mass/body composition, gender, age, etc; cystatin is affected by thyroid dysfunction, taking corticosteroids, inflammatory diseases, and high cell turnover states). several studies have found  that cystatin-C based eGFR calculations are better predictors of clinical outcomes than creatinine-based ones (eg, see https://gmodestmedblogs.blogspot.com/2023/12/cystatin-c-better-predictor-of-bad.html ). A meta-analysis of 11 generally diverse populations found that cystatin C alone or added  to creatine “strengthens the association between the eGFR and the risks of death and end-stage renal disease across diverse populations”: https://www.nejm.org/doi/full/10.1056/NEJMoa1214234.  And cystatin has been known to be more accurate than creatinine-based measurements for many decades. Yet, again, difficult to change our own culture...

    -- and there is a pretty reasonable argument that torsemide is likely superior to furosemide as a loop diuretic for heart failure: https://gmodestmedblogs.blogspot.com/2023/01/heart-failure-torsemide-vs-furosemide.html  

-- which anthropometric measurement to use is unclear, given the different outcomes in different studies. The choice for clinical care should probably be determined by the ease and reproducibility of the different ones, with a priority for the most predictive of adverse clinical outcomes. in this light, waist-to-hip ratio would be eliminated because of  the documented inconsistencies in assessing the hip circumference. Waist circumference by itself would be the easiest. The others require a calculation to be applicable, such as an automatic calculator in the electronic record system.

 

geoff    

If you would like to be on the regular email list for upcoming blogs, please contact me at gmodest@bidmc.harvard.edu

to get access to all of the blogs:  go to http://gmodestmedblogs.blogspot.com/ to see the blogs in reverse chronological order

or you can just click on the magnifying glass on top right, then type in a name in the search box and get all the blogs with that name in them