Ethnic Differences in the Association Between Body Mass Index and Hypertension

1
Ethnic Differences in the Association Between
Body Mass Index and Hypertension

• Colin Bell, Linda Adair, Barry Popkin
• Department of Nutrition, The University
• of North Carolina at Chapel Hill
• Department of Community Health,
• University of Auckland

2
Does a BMI of 25 kg/m2 mean the same thing in
different populations?

• BMI is only an approximate measure of body
  fatness
• Good evidence exists that some populations have
  different levels of body fat at similar BMIs
• Asian smaller frames, higher body fat than
  Caucasians (Deurenberg et al, Int J Obesity
  1998,1999)
• Polynesians larger frames, more lean body mass,
  lower body fat than Caucasians (Swinburn et al,
  Int J Obesity 1999)

3
If body fat differs, do obesity related
co-morbidities also differ?

• In Hong Kong Chinese, Ko et al observed increased
  prevalence of type-2 diabetes, hypertension,
  dislipidaemia and albuminuria at a BMI of 23
  kg/m2 (Ko et al, Int J Obesity 1999)
• In Polynesian populations serum lipids tend to be
  lower than for Caucasians in spite of higher BMIs
  (Bell et al, NZ Med J 2001, Scragg et al, NZ Med
  J 1993)
• However
• direct comparisons are needed in a variety of
  ethnic groups

4
Objective

• To determine whether there are ethnic differences
  in the association between BMI and hypertension
  in men and women aged 30 - 65 years

Ethnic groups

• 3,423 Chinese men and women (CHNS 1997)
• 1,929 Filipino women (CLHNS 1998)
• 7,957 non-Hispanic Whites, non-Hispanic Blacks,
  Mexican Americans (NHANES III 1988 - 1994)

5
Methodology

• Pooled cross-sectional data from three surveys
• Outcome Hypertension
• SBP ? 140 mm Hg , DBP ? 90 mm Hg, or on
  anti-hypertension medication
• Including those on medication biased the result
  towards the null or had no effect (see following
  figure)
• Main explanatory variable BMI
• Confounders Age. Physical activity, smoking and
  alcohol consumption were not major confounders.
  (see following figure)

6
We included pre-diagnoased individuals to boost
cell size and because their inclusion biased the
results towards the null
7
Physical activity, smoking status and alcohol
consumption had a minimal effect on the
association between hypertension and BMI in all
ethnic groups eg NHBlack women
8
Compared to US ethnic groups, Chinese men women
were less hypertensive Filipino women had
similar levels of hypertension to NHWhites
9
Compared to US ethnic groups, Chinese men women
Filipino women were less overweight (BMI ? 25
kg/m2)
10
Chinese men had higher odds of prevalent
hypertension, adjusted for age, than NH-Whites at
every category of BMI, including 23-24.9 kg/m2
11
Including waist circumference attenuated the
association for both Chinese NHWhite men but
the ethnic differences remained
12
The age-adjusted odds of prevalent hypertension
for Chinese and Filipino women were similar to
those for NH-Whites at low levels of BMI
13
Two problems can arise when using odds ratios in
this context

• Odds ratios are dependent on a reasonable number
  of subjects in the reference category
• Interpretation can be misleading because the
  analysis assumes that the underlying risk (or in
  this case prevalence) between the ethnic groups
  is the same

14
Subject numbers were sufficient but hypertension
prevalence differed markedly by ethnic group in
the referent BMI category (BMI 18.5-22.9 kg/m2)
15
To overcome this problem, we used prevalence
difference figures. A steeper slope was observed
at low levels of BMI for Chinese men (10.8? )
compared to NHWhite men (1.8 ?)
16
There was also some evidence of a steeper slope
at low levels of BMI for Chinese women (7.6 ? )
compared to NHWhite women (4.3 ? ). Filipino
women showed a 10.3? between the categories
23.0-24.9 and 25.0-26.9 kg/m2
17
Current WHO weight status recommendations for
Asia and the USA
18
Should lower definitions of overweight and
obesity be used for Asian populations?

• We have shown some evidence that the association
  between hypertension and BMI may be stronger in
  Chinese compared to NHWhites
• There was no evidence of a stronger association
  for Filipino women, however, a higher baseline
  prevalence may justify a lower cut-off
• To fully justify lower cut-offs we need
  longitudinal studies, data on all co-morbidities,
  consensus on appropriate methodology and more
  specific definitions of ethnicity

19
The value of ethnic-specific BMI cut-offs?

• At the clinical level
• In countries such as the USA, with considerable
  ethnic diversity, physicians would be better able
  to identify individuals at risk of obesity
  related co-morbidities
• At an international level
• At this level, the utility of a weight
  classification system is in the ability to
  compare populations and monitor changes overtime
  therefore there is no advantage in having
  ethnic-specific cut-offs