Body Roundness Index is better predictor than BMI for clinical problems

 A recent study found that the Body Roundness Index (BRI) is a better predictor of adverse clinical effects than Body Mass Index (BMI), our usual surrogate marker for obesity-related adverse outcomes (see obesity body roundness index JAMA2024 in dropbox, or doi:10.1001/jamanetworkopen.2024.15051)

 

Details:

--32,995 US adults >20yo selected  from the National Health and Nutrition Examination Survey (NHANES) were evaluated for the relationship between Body Roundness Index (BRI) and all-cause mortality, with mortality data derived from the CDC website and linked to the NHANES database

    -- since 1999, NHANES conducted surveys every 2 years based on in-home interviews and study visits; 10 cycles from 1999-2018 were included. Nonpregnant individuals were selected to be representative of the noninstitutionalized civilian population

    -- BRI was calculated as: 364.2 − 365.5 × √ 1 − [waist circumference in centimeters / 2π]² / [0.5 × height in centimeters]² , according to the formula developed by Thomas et al

    -- in this study, BRI was partitioned into to five increasing quintiles

-- there were some changes in the demographics over the 2-year cycles from 1999-2018. The overall averages of demographics over these 5 quintiles:

    -- mean age 47, 50% female, 8% Mexican-American/11% non-Hispanic Black/70% non-Hispanic White, education level <9th grade 5%/9-11 grade 12%/high school graduate 25%/ some college 30%/college graduate 27% 

    -- PIR (poverty impact ratio, the ratio of family income to poverty threshold) <1 in 13%/at least 1 in 87%, cigarette smoking 44%, alcohol drinking 77%, family history CVD 11%, family history diabetes 42%

 

-- main outcome: the association between BRI and all-cause mortality

-- adjusted analyses controlled for age, sex, race and ethnicity, educational level, PIR, smoking status, drinking status and family history of CVD or diabetes

 

Results:

-- the mean BRI gradually increased from 4.80 (4.62-4.97) from 1999 to 5.62 (5.37-5.86) in 2018, an average biennial change of 0.95%; the increases were more dramatic in women, elderly individuals, and those self-identifying as Mexican-American (as in graphs below), but was also found by education (lower BRI in college grads), those with lower poverty income levels (ie, worse off financially), smoking, not drinking, or family history of CVD or diabetes:

 

 

 

-- mortality: after 10 years of follow-up, there were 3452 deaths from all causes (10.5% of the population), and this was a U-shaped curve after full statistical adjustment:

 

 

  

 -- comparing those with BRI of the middle quintile of 4.5-5.5 as the reference, those with BRI<3.4 had a 25% increased mortality risk, HR 1.25 (1.05-1.47), and those with BRI of 6.9 or greater had a 49% increased risk, HR 1.49 (1.31-1.70)

 

-- in comparison, BMI in the fully adjusted model did show a higher all-cause mortality but only in quintile 1 (44% increase, HR 1.44 (1.26-1.64)) and quintile 5 (31% increase, HR 1.31 (1.14-1.51)), whereas BRI was statistically significant also in quintile 4 with a 25% increased risk

 

Commentary:

-- as we know only too well, obesity is a global epidemic and getting worse:

    -- more than 1 billion people are obese

    -- obesity is one of the 5 top risk factors for mortality, with about 5 million deaths worldwide in 2019 likely attributable to obesity

-- in many (but not all) studies, obesity is defined as specific BMI ranges, BUT:

    -- It has been clear for many decades (ie back to the 1970s at least) that visceral adiposity is the main bad actor in terms of subsequent diseases:

        -- visceral adiposity (which is fat found around organs in the abdominal cavity), as opposed to subcutaneous adiposity, is associated with systemic inflammation, insulin resistance, metabolic syndrome, diabetes, cancer

            -- many comparative studies have found that visceral adiposity is much more highly associated with adverse clinical effects, with some studies finding no increased risk in those with just subcutaneous adiposity (though the BMIs of these groups may well be similar)

        -- the gold standard for assessing visceral adiposity is by imaging, eg by CT or MRI scanning

        -- but there are surrogate measures that provide reasonably accurate assessment of visceral adiposity, including:

            -- waist circumference (>35 inches for women or >40 inches for men)

                -- waist circumference has been validated by CT scanning as an accurate surrogate marker for visceral adiposity

            -- waist-to-hip ratio (>0.85 for women or >0.90 for men); people who are “pear-shaped” (with more fat tissue in the buttocks and thighs) typically having peripheral subcutaneous adiposity vs “apple-shaped” (typically with central visceral adiposity)

                -- waist-to-hip ratio has also been validated by studies finding it is associated with low HDL levels and high insulin levels (https://www.nejm.org/doi/full/10.1056/NEJM199001253220404 , which references an older study showing correlation with CT scanning)

            -- waist-to-height ratio (a healthy ratio is <0.5)

                -- waist-to-height ratio has been validated, along with waist circumference, as a predictive value for diabetes outcomes, and both were superior to BMI; the waist-to-height ratio may have been a bit better than waist circumference in this study and has the advantage of being independent of age, sex, or ethnicity: https://www.cambridge.org/core/journals/nutrition-research-reviews/article/systematic-review-of-waisttoheight-ratio-as-a-screening-tool-for-the-prediction-of-cardiovascular-disease-and-diabetes-05-could-be-a-suitable-global-boundary-value/A65EC8CCE2A120C247F82C5074C24C7D

            -- a few studies have found that BRI has been very accurate in predicting cardiovascular risk, cardiometabolic risk, kidney disease, cancer, diabetes or metabolic syndrome, though waist-to-hip ratio and waist circumference were similarly accurate: eg,  https://onlinelibrary.wiley.com/doi/10.1111/obr.13023 or https://www.frontiersin.org/journals/nutrition/articles/10.3389/fnut.2022.801582/full

 

-- the AMA in June 2023 adopted the policy that BMI is an "an imperfect way to measure body fat in multiple groups given that it does not account for differences across race/ethnic groups, sexes, genders, and age-span. Given the report’s findings, the new policy supports AMA in educating physicians on the issues with BMI and alternative measures for diagnosing obesity." and the AMA comments that "BMI be used in conjunction with other valid measures of risk such as, but not limited to, measurements of visceral fat, body adiposity index, body composition, relative fat mass, waist circumference and genetic/metabolic factors."...  "the AMA also reinforced that relative body shape and composition is necessary when applying BMI as a measure of adiposity and that BMI should not be used as a sole criterion to deny appropriate insurance reimbursement". see https://www.ama-assn.org/press-center/press-releases/ama-adopts-new-policy-clarifying-role-bmi-measure-medicine

    -- BMI does have clear limitations. BMI was developed in 1832 by a mathematician (Adolphe Quetelet) to assess quantitatively the characteristics of a "normal man", which of course is really a "normal white man". 

    -- one concern about BMI is that it is elevated in those with increased muscle mass, not only in people with adiposity

    -- But, it should be noted that high BMI is still a pretty robust indicator (surrogate marker) of risk of outcomes associated with obesity, though inferior to the ones noted above that specifically reflect visceral obesity, with a few comments:

        -- this current study did find that BMI was almost as good as BRI as a risk factor

            -- all of the results for BMI in the study were located in the supplementary materia and not discussed in detail in the paper itself, other than commenting that “compared with BMI, BRI had narrower confidence intervals and higher sensitivity in estimating risk for all-cause mortality”. But to my inspection, there was not a dramatic difference. this article did find that BMI actually was associated with a significant increased risk of all-cause mortality in quintiles 1 and 5 in the fully adjusted model

            --  this may well reflect the fact that on a large population basis, BMI is a fairly good predictor of serious overweight leading to clinical problems, since about 40-45% of those classified as obese by BMI have visceral obesity, per https://diabetesjournals.org/care/article/32/3/481/27658/Patterns-of-Abdominal-Fat-DistributionThe 

            -- in addition, there may be confounding here: visceral adiposity seems to be pretty clearly associated with the adverse effects of chronic inflammation, metabolic syndrome, diabetes, etc as noted above. And much more so than superficial adiposity. But superficial adiposity is still related to many factors that in and of themselves seem to confer increased risk of these adverse outcomes, such as diet, exercise, some medications…. (see https://gmodestmedblogs.blogspot.com/2024/06/pitavastatin-decreases-pancreatic.html and https://gmodestmedblogs.blogspot.com/2023/12/colchicine-decreases-hip-and-knee.html for more info on this)

       -- as a perspective here, many of the surrogate markers we use regularly (eg A1c, LDL cholesterol, creatinine-based eGFR, etc) are quite flawed in terms of accurately reflecting the likelihood of adverse clinical outcomes: https://gmodestmedblogs.blogspot.com/2024/06/using-surrogate-markers-for-disease-are.html (which also comments on BMI as an inadequate surrogate marker)

           -- So, this current article on measuring BRI reinforces the importance of using a surrogate obesity measurement that correlates with visceral fat, one that is much more clinically relevant than BMI: BRI (as well as waist circumference, or waist-to-hip ratio) have been well-documented to reflect CT-documented visceral obesity and therefore are more reasonable clinical surrogates to follow

-- the finding of a U-shaped curve for BRI and mortality is not commented on in the article, but is likely similar to the findings with BMI: individuals with low BMI include those who are chronically ill or have substance use disorders, including smokers who tend to be thinner than nonsmokers, as also found in the NHANES  study: https://www.sciencedirect.com/science/article/pii/S0002916523235479

 

Limitations:

-- the response rate for the NHANES surveys decreased over time: 76.62% of households interviewed and 69.83% of medical exams in 1999-2000, down to 48.24% for household interviews and 45.70% for medical exams in 2017-2018, potentially affecting the comparisons of the 2-year NHANES cycles over time (was there selection bias in who responded to the surveys over time??)

-- most of the demographics were binary (eg, either one smoked or drank vs not), not reflecting the huge differences in adverse effects from an occasional cigarette or drink to those with much higher volume), limiting our understanding of the role of these issues is in the relationship between BRI and mortality

-- one big limitation is that there was no information on how they incorporated important issues (drinking, smoking, PIR) into their model:

    -- did a single response to these questions at one time in the biennial NHANES assessment determine if the person was a smoker, etc (things do change over 20 years…)

        -- we do know that smoking/drinking/PIR are themselves associated with increased all-cause mortality. But on an individual basis, if a person smoked a few cigarettes at only one of the biennial assessments, their outcome attributable to smoking is likely hugely different than if 2 packs a day for the full 20 years, though perhaps they were both classified as “smokers”.  We really need individual-level granular data to know this

            -- ie, the statistically adjustments made in the article based on these groupings of smoking or drinking or even PIR are all potentially changing a lot over 20 years, and the statistical adjustment may be very misleading.

            -- this is a potential example of the ecological fallacy in statistics: assuming that group data (eg about smoking and BRI quintiles) is equivalent to data on the individuals tracked along these quintiles. A simple example to clarify "ecological fallacy": 2 towns have people with hypertension, but 1 town has much more salt consumption, suggesting that salt is the culprit.  But, perhaps on individual-based analysis, those not eating salt in the high-salt town actually had the hypertension. and perhaps they ate junk foods or didn't exercise or ???

            -- so, how did the BRI itself change on an individual basis?  We do have broad categories of change over the 5 time periods, but what were the specific individual changes that led to the observations in this study??

-- there is also a likely a dilutionary effect of assessing the relationship between BRI and all-cause mortality by including quite likely associated conditions (eg heart disease, diabetes, hyperlipidemia..) along with quite unlikely ones (eg car accident): ie, the actual association with specifically BRI-related mortality is likely understated

 

So, a few comments on this study

-- It was impressive that temporal trends of BRI increased over this 20 years of the study at a rate of about 1% biennially, and more so in women, those >65yo, and Mexican-American participants (the latter  group with a very high rate of obesity in the US)

-- there was a pretty clear relationship between BRI and mortality, though it was high also in the low BRI group but twice as high in the high BRI group. My guess is that this reflects the likelihood the low BRI group contained more individuals who had conditions associated with lower BRI, such as COPD, alcohol use disorder, other substance use disorders, psychiatric conditions, cancers, significant cardiovascular disease, etc. As per the Limitations section above, we are lacking in enough granular data to make sure the BRI/mortality data is robust

       -- it would have been very useful in this study to compare BRI to waist circumference, waist-to-hip ratio, and waist-to-height ratios to see if there was a significant difference. The BRI is based on weight and height, so is not so difficult to calculate (if one has an automatic computerized calculation), but it may well be that just the waist circumference is adequate. Though one advantage of the waist-to-height ratio is its independence of age, race and at least the several ethnicities studied, and it has the conceptual simplicity of “keep your waist circumference to less than half your height”

    -- but perhaps the most useful aspect of this study is that it raises yet again the importance of having some assessment of adiposity that focuses on visceral adiposity, since it has been known for decades to be related to chronic inflammation and the array of associated diseases (diabetes, cancer, etc as noted above).

    -- and perhaps insurance companies can be coerced into covering weight loss drugs if a more specific adiposity is measured that so clearly reflects large increases in morbidity and mortality????

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

  -- click on 3 parallel lines top left, if you want to see blogs by category, then click on "labels" and choose a category​

  -- 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

if you would the article, please email me.

please feel free to circulate this to others. also, if you send me their emails (gmodest@bidmc.harvard.edu), i can add them to the list