WCRF International and IASO Joint Conference 201316-17 April 2013Racial and ethnic differences inmeasures and effects of obesityProfessor TH Lam, JP, BBSMD, FFPH, FFOM, Hon FHKCCM, FHKAM, FRCPSchool of Public HealthThe University of Hong Kong
What is Obesity?• Simple yet complex• Epidemiologically: general (total, global) versuscentral - simplistic definitions and classification• Body weight, body fat and body fat percent (BF%)• BMI most commonly used for general obesity:only data available in most studies• Waist circumference (and WHR): centralobesity: data available only in more recentstudies• Different methods to measure body fat: all withlimitations
Racial/Ethnic Differences: BF%/BMIDeurenberg et al 2002 review• Indonesians (Malays, Chinese); SingaporeanChinese, Malays and Indians; Hong KongChinese• All Asians studied: higher BF% at a lower BMIcompared to Caucasians:Same BMI, Asian BF% 3-5% points higherSame BF%, Asian BMI 3-4% points lower• Different BF%/BMI ratio: body build, i.e. trunk-to-leg length ratio and slenderness, muscularity
BF%/BMI Ethnic Specific• Relationships between BF% and BMI differbetween ethnic groups• Not all “Asians” are equal: Chinese, Indonesiansof Malays ancestry and Thais• Not all “Chinese” are equal: New York, Beijingand Hong Kong(Deurenberg et al 2002)
Problems and Challenges• Data limited and most reports had small samplesize, uncertain representativeness, differentmethods/assumptions/equations for body fatassessment• Studies with different ethnic groups measured inthe same laboratory using the samemethodologies are needed.• Universal BMI cut-off points are not appropriate(Deurenberg et al 2002)
Some Comments• Obesity, adiposity, BF%, BMI, WC, etc are allcontinuous variables• Using cut-off points for obesity (general orcentral) are needed but would it add to ourunderstanding of the causes and effects ofobesity and its mechanisms?• Or would it generate more confusion?
Is this normal ?
Is this underweight ?
Is this central obesity ?
Is this central and general obesity ?
• Meta-analysis/systematic reviews on ethnicdifferences are based on studies withdifferent methods at different time periods indifferent places with different socio-economic developments. Many studieswere not specifically designed to compareethnic/racial differencesSome Comments
• Could ethnic/racial differences be explainedby extraneous factors: regional,socioeconomic developments (high, middleand low income countries, or regions withinthe same countries): gradual developments inmany decades versus recent rapiddevelopments; immigration, inter-generationaldifferencesSome Comments
Obesity Increases Risks of Many Diseases• Cardiovascular, type 2 diabetes, some cancer(see WCRF reports) and all-cause mortality(prospective evidence)• Obesity associated with many risk factors:dyslipidemia, high blood pressure (cross-sectionalassociations)• Obesity has many determinants: e.g. diet,sedentary living, physical inactivity, energybalance; socio-economic and obesity control(WCRF 2007: Body fatness, convincing-oesophagus, pancreas,colorectum, breast postmenopausal, endometrial, kidney)
Obesity Increases Risks of Many Diseases• Evidence predominantly from Westernpopulations, and mainly from BMI• Difficulties in separating the effect of obesity fromits associated risk factors• Adjustment of risk factors, treating them as“confounders” may not clarify the effects ofobesity and its interaction with other risk factors
BMI/WC/WHR and Cardiovascular RiskHuxley et al 2010 review• Most evidence from Caucasians• Type II diabetes (Vazquez et al 2007), meta-analysis of 32 cohort studies• BM/WC/WHR: RR of 1.87-1.88 for incidentdiabetes per standard deviation - similarassociations• Effect stronger in Caucasian than Asian forWHR but not BMI or WC
Pooled relative risk for BMI, WC and WHR with incidentdiabetes stratified by age, gender and geographical regionAbbreviations: F, female; M, maleAdapted from Vazquez et al., (2007)(Huxley et al 2010)
Obesity in Asia Collaboration(Huxley et al 2008)• Cross-sectional meta-analysis >263,000 subjects(73% Asian)• Except Caucasian men, central obesity morestrongly associated with prevalent diabetes thanBMIPer 0.5 SD: BMI - 20-30% prevalent ORWC and WHR - 40%• Hypertension:BMI/WC/WHR: Similar OR for both men andwomen per 0.5 SDStronger OR in Caucasian (40%) than non-Causasian men (30%)
(Huxley et al 2008)
(Huxley et al 2008)
Diabetes Epidemiology: Collaborative Analysisof Diagnostic Criteria in Asia Study(DECODA, 2008)• 16 cross-sectional studies• DM and BMI/WC/WHR: little differences• But a slightly stronger association withWeight/Height ratio in both men and women
DyslipidaemiaOAC (Barzi et al 2010)• Most comprehensive analysis• Total cholesterol, LDL and triglycerides andglobal/central obesity• Cross-sectional associations broadly similarbetween Asians and non-Asians• No single measure of body size was superiorfor discriminating dyslipidaemia• WHR of 0.8 in women and 0.9 in menapplicable across both regions fordiscriminating any form of dyslipidaemia (alsofor diabetes and hypertension, Huxley 2008)
Some Comments• Cross-sectional associations - causationuncertain• Data on central obesity - more recent andscarce• Data from different countries/regions at differenttime periods• Variations of obesity and lipid measurement andresults, and heterogeneity in the associationswithin and between Asian and non-Asianpopulations
Asia Pacific Cohort Studies Collaboration(APCSC 2006)Obesity Indices and CVD Risk• >40 cohort studies: Asia and ANZ• 33 cohorts (310,000 people): BMI andCV events• 6 cohorts (45,998): waist and hipcircumference - 601 CHD and 346stroke events
• One SD associated with excess riskCHD(%) Stroke(%)Association No associationBMI 17(7-27) 3 (-9-16)WC 27(14-40) 5(-9-20)HC 10(1-20) 0(-11-13)WHR 36(21-52) 9(-8-28)Association: stronger inaged <65, men, non-Asians strongest for WCand WHR; weakest HCNo associationacross age, sex,region(Huxley et al 2010)
Some Comments• APCSC - a large individual data meta-analysis small number of cohorts (6/44) and eventsfor comparing different obesity indices andregional ethnic differences• Data mainly from ANZ; Asian cohorts had shortfollow up (mean 3.3 years or less; confoundedby pre-existing diseases)• Broad similarity: (a) different indices and CV riskand risk factors; (b) regions/ethnicity• Regional differences are NOT racial/ethnicdifferences
Obesity and Cancer: Ethnic Differences?APCSC 2010• BMI and cancer mortality• 39 cohorts, 424,519 people (77% Asian)• 4,872 cancer death from 401,215(excluding FU <3y)
Obesity and Cancer: Ethnic Differences?APCSC 2010• Increased risk (HR (95% CI)) in BMI ≥30 vs BMI18.5-24.9– all-causes (excluding lungand upper aerodigestive) 1.21(1.09-1.36)– colon 1.50(1.13-1.99)– rectum 1.68(1.06-2.67)– breast(≥60y) 1.63(1.13-2.35)– ovary 2.62(1.57-4.37)– cervix 4.21(1.89-9.39)– prostate 1.45(0.97-2.19)– leukaemia 1.66(1.03-2.68)(Parr et al 2010)
• No regional differences in HR for cancerand BMI except oropharynx and larynx:inverse in ANZ, absent in Asia• Asian data: mainly Japan• Insufficient data on WC, WHR• Test of regional interaction(heterogeneity): low stat. power(Parr et al 2010)
Age- and smoking-adjusted hazard ratios with 95% CI formortality from colorectal cancer by region after left censoringat 3 years according to categories of body-mass index(Parr et al 2010)
Obesity and Colorectal Cancer Meta-analysis(Moghaddan et al 2007)• 31 studies (23 cohort, 8 case control), 70,000events• BMI: 7%(4-10%) per 2 kg/m2WC: 4%(2-5%) per 2 cm• Association with general obesity: smallerstrength than previously reported• No ethnic differences but very few studies fromAsia• 2 Japanese and 1 Korean cohort studies: BMIassociation. No data on WC.
All-cause Mortality and BMI: Meta-analysis(Flegal et al 2013)• 97 studies, 2.88 million people, >270,000deaths• HR vs BMI : 18.5 to <25overweight (25-<30) 0.94(0.91-0.96)grade 1 obesity (30-<35) 0.95(0.88-1.01)grade 2,3 obesity (≥35) 1.29(1.18-1.41)obesity (1-3) (≥30) 1.18(1.12-1.25)• Overweight: Lower all-cause mortality
Some Comments• Baseline BMI to predict outcomes does notaccount for– Obesity/weight status (duration) and relatedfactors before and after baseline– Age of subjects and latency period– Duration of follow up - too short– Confounders: some could be determinants oroutcomes of obesity– Different socio-economic developments indifferent countries/regions at different timeperiods
Can we learn fromthe Stages ofEpidemic of Tobacco?
Four Stages of the Tobacco Epidemic(http://apps.who.int/bookorders/anglais/detart1.jsp?sesslan=1&codlan=1&codcol=76&codcch=22)
Stages of Tobacco Epidemic SE• Stage 1 to 4 for different regions/countries, by sex• A large gap of several decades betweenpeak of tobacco consumption and thepeak of tobacco-induced deaths• Full impact of adverse outcomes onlyobserved recently in the West (US/UK);not yet in LMIC (Asia)• Effective tobacco control decliningconsumption and mortality in the West
Stages of Epidemic of Obesity• The West: Early stage, Stage 2(a) Rising, high obesity level;(b) Rising mortality• LMIC: Stage 1(a) Early rise of obesity;(b) No or early rise of mortality• Same for men and women?• Need to interpret results and studies taking intoaccount stage developments• Inappropriate to pool results from different stages• Results observed in the past and now: Under-estimate the full impacts in the West; grosslyunder-estimate in LMIC
Life Course Studies of Obesity• Obesity/overweight can start from earlychildhood to late adulthood• Trajectory of obesity/overweight; extent andduration of obesity• Factors affecting changes in obesity/overweight(such as illnesses, efforts to reduce weight inhealthy people)• Reverse causation• Risk reversal of weight/obesity reduction• Inter-generational effect: parental obesity,pregnancy weight status on offsprings• Different effects on different diseases
is needed for theStages of Epidemic Life CourseObesity Research (SELCOR)Growing andComplex Global Epidemic of Obesityand forDifferent populations, regions,races and ethnicity