Professional Documents
Culture Documents
Diabetes in the Caribbean Wilks et al 1999-4
Diabetes in the Caribbean Wilks et al 1999-4
Abstract
*Tropical Metabolism Research Unit, University of Aims To characterize the prevalence of diabetes and associated risk attributes
the West Indies, Mona, Jamaica in the Jamaican population.
²Department of Preventive Medicine and Methods A random population sample was recruited by door-to-door
Epidemiology, Loyola University Stritch School of
Medicine, Maywood IL, USA
canvassing (n = 1303). A ®nal participation of 60% was achieved. Oral glucose
tolerance testing was conducted after an overnight fast and standard
³Clinical Epidemiology Unit, School of anthropometric and demographic data were collected.
Epidemiology and Health Sciences, University of
Manchester, Manchester, UK Results The prevalence of Type 2 diabetes mellitus was 9.8% (95%
con®dence interval (CI) 7.2±12.4) among men and 15.7% (95% CI 13.1±18.3)
Supported in part by grants from the US National
Institutes of Health, the European Commission among women with an overall prevalence of 13.4% (95% CI 11.5±15.2).
and the Wellcome Trust Impaired glucose tolerance was found among 12.3% of men and 14.7% of
women. The sex patterns were consistent with a fourfold excess of obesity in
Received 20 October 1998; revised 9 March 1999;
accepted 7 June 1999 women compared to men. The odds ratios for diabetes, fourth vs. ®rst
quartiles were 5.42 (95% CI 2.02±16.88) in men and 3.32 (95% CI 1.73±
6.63) in women for body mass index (BMI) and 17.39 (95% CI 3.86±78.27) in
men and 5.48 (95% CI 2.84±11.00) in women for WHR in a logistic model
controlling for age. The population attributes risk percentage, for diabetes, of
being overweight and having waist-to-hip ratio (WHR) greater than the
median (0.80) were 66% and 80%, respectively. The contribution of central
obesity, as characterized by WHR, was also signi®cant in sex-speci®c
multivariate models that included age and BMI. Prevalent hypertension and
family history of diabetes were likewise associated with increased odds of
having the disease.
Conclusions The prevalence of diabetes in Jamaica now exceeds that observed
among European-origin populations and re¯ects the emerging epidemic of
obesity. The excess risk for this population could not be attributed entirely to
relative weight. The pronounced sexual dimorphism in diabetes prevalence
most likely re¯ects the substantial excess of obesity among women compared
to men. Like many other island nations, Caribbean societies now appear to be
at substantial risk of diabetes.
Diabet. Med. 16, 875±883 (1999)
Keywords Caribbean, Jamaica, obesity, Type 2 diabetes mellitus
Abbreviations BMI, body mass index; FPG, fasting plasma glucose; IGT, im-
paired glucose tolerance; NGT, normal glucose tolerance; OGTT, oral glucose
tolerance test; OR, odds ratio; PAR%, population attributable risk percentage;
WHO, World Health Organization; WHR, waist-to-hip ratio
criteria as follows: FPG > 7.0 mmol/l (126 mg/dl) or 2-h post- association is one of cause and effect [22]. Therefore, PAR%
load plasma glucose concentration > 11.1 mmol/l (200 mg/dl) represents that proportion of disease in the total population that
or on treatment for diabetes (i.e. insulin or oral hypoglycaemic could be eliminated by removal of the harmful exposure [22].
agent or diet) [17]. Hypertension was de®ned as systolic For example, the PAR% for BMI could be interpreted as the
BP > 140 or diastolic BP > 90 or current use of anti-hyperten- proportion of diabetes that could be eliminated if individuals
sive medication, were encouraged and succeeded in reducing their BMI to below
25 kg/m2.
Statistical analyses
BMI, body mass index. Hypertension1: systolic BP > 140 or diastolic BP > 90 mmHg or anti-hypertension medication. Hypertension2: systolic BP
> 160 or diastolic BP > 95 or anti-hypertension medication.
²BMI > 25 kg/m2. ³BMI > 30 kg/m2.
***P-value < 0.0001 and *P-value < 0.05 for difference between two means or proportions (men vs. women).
Table 2 Mean (95% con®dence intervals) of selected variables strati®ed by glycaemic status and sex
Hypertension1: systolic BP > 140 or diastolic BP > 90 or anti-hypertension medication. *Differences between three means were assessed using
2
ANOVA and the c -test for categorical variables.
NS, not signi®cant. ²BMI > 25 kg/m2 and ³BMI > 30 kg/m2.
Table 3 Prevalence of impaired glucose tolerance (IGT) and diabetes (DIAB) by age and sex
25±34 134 3.0 2.2 197 8.6 2.5 331 6.3 2.4
35±44 113 4.4 6.2 192 13.0 10.4 305 9.8 8.8
45±54 105 17.1 13.3 182 17.6 18.7 287 17.4 16.7
55±64 87 14.9 21.8 126 15.1 30.2 213 15.0 26.8
65 + 81 29.6 9.9 86 25.6 30.2 167 27.5 20.4
Total 520 12.3 9.8 783 14.7 15.7 1303 13.7 13.4
and men with diabetes compared to their normoglycaemic increased with age with a slight downward trend for the
counterparts (21% vs. 16%, respectively; Table 2). last age group (65±74 years). The increase in the mean 2-h
Although mean FPG was the same for men and women level with age was much steeper than was FPG for both men
(5.5 6 2.3 mmol/l), women had signi®cantly higher 2-h and women.
post-glucose values (7.5 6 3.9 vs. 6.6 6 3.7 mmol/l; Based on the WHO criteria [15] the prevalence of
P < 0.001). When persons on insulin or oral hypoglycaemic diabetes was 9.8% (95% con®dence interval (CI) 7.2±
agents were excluded from the analysis (14 out of 78 people 12.4) for men and 15.7% (95% CI 13.1±18.3) for women
on insulin or oral agent), the 2-h post-load glucose level (Table 3), with an overall prevalence of 13.4% (95% CI
dropped to 6.2 6 2.8 mmol/l for men and 6.9 6 3.0 mmol/l 11.5±15.2). When age-standardized against the overall
for women. Mean FPG and 2-h post-load concentration population of Spanish Town the rates were 7.6% (95% CI
Figure 1 Prevalence of diabetes mellitus by quartiles of body mass index (BMI) among Jamaicans, by sex, c2 for trend P < 0.001. Inter-quartile
ranges for men (1st to 4th, respectively): 14.1±20.7; 20.8±22.8; 22.9±26.5; 26.6±48. Inter-quartile ranges for women (1st to 4th, respectively):
14.3±23.5; 23.6±27.5; 27.6±31.6; 31.7±56.2.
Table 4 Age-adjusted odds ratios and 95% con®dence intervals for diabetes by sex-speci®c body mass index (BMI) quartiles
Figure 2 Prevalence of diabetes mellitus by quartiles of waist-to-hip Ratio (WHR) among Jamaicans, by sex, c2 for trend P < 0.001. Inter-quartile
ranges for men (1st to 4th, respectively): 0.68±0.79; 0.80±0.83; 0.84±0.88; 0.89±1.07. Inter-quartile ranges for women (1st to 4th, respectively):
0.64±0.76; 0.77±0.79; 0.80±0.84; 0.85±1.35.
5.5±9.7), 12.0% (95% CI 10.0±14.0) and 10.2 (95% CI As expected, the prevalence of diabetes increased with
8.7±11.6) for men, women and the total population, age for both men and women, with the highest prevalence
respectively. When age-standardized against the standard occurring in the 55±64 years age group. This is maintained
world population of Segi, the rates were similar at 9.5% in the 65+ age group among women but falls sharply
(95% CI 7.1±12.0), 15.2% (95% CI 12.8±17.5) and among this oldest group of men. With the exception of the
12.8% (95% CI 11.1±14.5) for men, women and the total youngest age group (25±34 years), women had a higher
population, respectively. Seventeen individuals who re- prevalence of diabetes in all age groups. The distribution of
ported a history of diabetes and were not on any form of IGT followed similar patterns of increasing prevalence
treatment (i.e. oral, insulin or diet) were classi®ed with age. The sex-speci®c prevalence of IGT was 12.3% for
according to their FPG and OGTT values. Fifteen of the men and 14.7% for women, with a combined prevalence of
17 had normal blood glucose levels (i.e. both FPG and 13.7%. The IGT : DM ratio was greater among men at 1.3
OGTT values below 7.8 mmol/l) and the remaining two compared to 0.94 in women.
demonstrated impaired glucose tolerance. Prevalence of Based on the sex-speci®c empirical quartiles, there was a
diabetes increased to 11.5%, 17.0% and 14.8%, respec- consistent trend of increasing risk with increasing BMI for
tively, for men, women and all persons combined based on both men and women (Fig. 1; Table 4). Comparing all
the new American Diabetic Association (ADA) diagnostic other quartiles to the lowest quartile, there was a two to
criteria [17]. The mean FPG and 2-h post-load glucose ®vefold increased risk of developing diabetes as estimated
concentration for the additional 19 persons classi®ed as from age-adjusted odds ratios obtained from sex-speci®c
diabetic using the new ADA criteria were 7.2 6 0.2 (range logistic regression models. The absolute contribution of
7.0±7.7 mmol/l) and 8.7 6 1.8 (4.6±10.8 mmol/l) respec- BMI (>25 kg/m2) to the prevalence of diabetes as measured
tively. The prevalence of impaired fasting glucose (IFG) by the PAR% was 66% (95% CI 52±76).
using the new ADA diagnostic criteria was 5.1%. From this Similar analyses were performed using WHR for men
point on, all discussion refers to prevalence estimates based and women (Fig. 2, Table 5). As with BMI, the risk of
on the earlier WHO criteria. diabetes increased monotonically with WHR for both men
Table 5 Age-adjusted odds ratios and 95% con®dence intervals for diabetes by sex-speci®c waist-to-hip quartiles.
Table 6 Age adjusted Mantel±Haenszel odds ratio and 95% con®dence intervals by hypertension status and family history
Men Women
and women. In contrast, however, the increase in risk was indicate greater awareness of diabetes resulting in early
much greater for each subsequent quartile, with a 17-fold screening and diagnoses in the general population.
and ®vefold excess for men and women, respectively, for However, the proportion of newly diagnosed cases is still
the highest compared to the lowest quartile of WHR. The very high, an indication of a large reservoir of undetected
absolute contribution of WHR (> 0.8) to the prevalence of cases similar to that reported in the United States [26] and
diabetes as measured by the PAR% was 80% (95% CI 69± elsewhere [19]. Furthermore, among the 78 known cases
88). receiving pharmacological treatment (i.e. insulin and/or
The contributions of family history (i.e. any grand- oral hypoglycaemic agents) only 14 (18%) had values for
parent, parent or sibling) and of concurrent hypertension FPG below 7.8 mmol/l or 2-h post glucose load below
are displayed in Table 6. A logistic analysis comparing 11.1 mmol/l.
people with diabetes to people with normal tolerance The participation rate in this population survey was 60%.
indicated that family history of diabetes was associated While it is recognized that this relatively low rate may reduce
with approximately a fourfold increased risk among men the generalizability of the study ®ndings, it is however,
and a twofold increased risk among women. higher thanthe estimate of57.9% fromprevious population
Approximately half (82/174) of the diabetic people were survey in Jamaica [8]. Given that each participant was
also hypertensive. The OR from the logistic models expected to undergo an OGTT, the participation rate of
evaluating the association between hypertension and 60% at the population level will be considered a success by
diabetes status were 1.8 for men and 2.0 for women. The most investigators in this line of scienti®c endeavour.
corresponding risk values were 1.4 and 2.5 for men and Furthermore, studies that require venepuncture will gen-
women when hypertension was de®ned as systolic BP erally become less attractive in Jamaica because of the
> 160 or diastolic BP > 95 or currently taking anti- unfounded but widely held belief that one is at risk of
hypertension medication. The association of diabetes and acquired immune de®ciency syndrome (AIDS).
hypertension is signi®cant at both levels of de®nition in The pathophysiological signi®cance of the high rate of
women and in both comparisons, that is, diabetes vs. IGT observed in this population (12.3% for men and 14.7%
normal and diabetes and IGT vs. normal. Among men the for women) is not clear. Some investigators have suggested
association is not signi®cant in the diabetes vs. normal that it may indicate an increase in the public health impact of
comparison where the sample size is smaller. diabetes in the future [16,19,26,27]. It has also been
suggested that the ratio of IGT to Type 2 diabetes mellitus
may indicate the stage of an epidemic of glucose intolerance
Discussion
within a population [28]. The ratio in this study is similar to
Knowledge of the prevalence of diabetes and associated risk that observed in industrialized societies, an indication that
factors at the population level is essential for proper diabetes is not an emerging disease in Jamaica. Thus, by all
planning and implementation of public health policy aimed criteria, the problem of diabetes has reached epidemic
at prevention and control. Knowing the magnitude of the proportions in this and other communities in the Caribbean.
problem will also facilitate rational planning of research A caveat is, however, in order. Given the relative leanness
activities directed at understanding the aetiology of diabetes (meanBMI = 23.8 kg/m2)ofthemeninthiscommunity,IGT
and implementation of programmes to promote health. The may not necessarily mean deterioration of carbohydrate
present study used WHO standardized criteria [15] to de®ne metabolism since on repeat testing IGT often reverts to
diabetes and IGT status of a representative sample of men normal tolerance in lean and physically active populations
and women in Spanish Town, Jamaica. The prevalence of [29±32]. It may be that the impact of obesity on diabetes
diabetes for these participants with a mean age of 47 years prevalence has reached a plateau among women while it is
was 9.8% and 15.7% for men and women, with an overall still increasing among men. This question was not examined
prevalence of 13.4%. These estimates are similar to directly in the present study.
previously reported rates from Caribbean countries [2± There is limited evidence on the secular trends in the
8,23±25]. The adjustment using the standard world prevalence of diabetes in Jamaica. Florey et al. [33] reported
population of Segi made little difference to the prevalence aprevalenceof8.1%inbothmenandwomeninthesameage
estimates and this may be reassuring. There is however, a 2± group using criteria suggested by FitzGerald and Keen [34].
4% fall in prevalence when adjusted for the overall Diabetes was de®ned as a 2-h post-prandial glucose in
population of Spanish Town. As diabetes prevalence venous blood of greater than 6.7 mmol/l or greater than
increases with age this suggests that our sample was older 8.9 mmol/l at some other time point during a glucose
than the general population of Spanish Town. Among tolerance test. These criteria would include as diabetic some
persons classi®ed as diabetic in this study, 32% were newly of the subjects classi®ed in our study as having IGT. The
diagnosed cases, which is lower than the estimate of 48% prevalence of diabetes mellitus in this study at 9.8% and
from an earlier population survey conducted in Jamaica [8]. 15.7% among men and women, respectively, is by
The observed lower rate of newly diagnosed cases may comparison an underestimate. Using similar criteria the
increase in prevalence of diabetes over the last 25 years women (OR = 3.24 (95% CI 1.76±5.97); P < 0.0001),
would have been more marked than the 9.8% vs. 8.1% supporting the data of Perry et al. [39] and Colditz et al.
among men and the 15.7% vs. 8.1% among women [40]. As has been suggested [41], increases in the prevalence
suggested by this comparison. of obesity within a population often precede a rise in the
Similar to results from other Caribbean countries and the incidence of chronic diseases, most notably diabetes and
USA, there was a signi®cant rise in FPG with age. The oldest hypertension.
age group (65±74 years) showed a lower rate of rise relative While obesity, which undoubtedly has environmental as
to those who are younger. The prevalence of diabetes among well as genetic causes, contributes signi®cantly to the
men in this age group was also signi®cantly lower. Some of incidence of diabetes, it accounts for only a small proportion
the reasons that have been advanced for these observations of the variance in glucose tolerance. The signi®cant
include differential exposure by age group and decreased independent contribution of family history of diabetes
availability of susceptible persons in the non-diabetic observed here and in other studies [16,42,43] suggests the
population with age. Selective mortality prior to this age presence of genes that may increase susceptibility to
group may have accounted for this. A clear example of a diabetes. Another important risk factor that was not
cohort effect may be an increased energy intake and measured in the present study is the level of physical activity
declining physical activity for more recent generations [44,45]. Our research group is in the process of testing the
[16]. Cohort effects may also have a signi®cant impact on hypothesis that the difference in diabetes rate observed
known risk factors, including relative weight, as measured between Jamaican and Nigerian men, who have similar
byBMIandfatdistribution,measuredbyWHR.Therewasa levels of obesity, is a result of, in part, differences in physical
clear and consistent gradient between diabetes prevalence, activity and different body composition [46]. Preliminary
BMI and WHR for both men and women. Independent of ®ndings of a on-going study indicate that for a given BMI,
BMI, the OR for WHR was 1.6 (95% CI 1.4±1.8), lending Nigerian men have 50% lower levels of body fat. Parallel
supporttothenotionthatapartfromtheriskassociatedwith studies measuring resting metabolic rate and total energy
overall heaviness, fat distribution, especially around the expenditure using double-labelled water in Nigeria,
waist, adds to the risk of diabetes [35±37]. The PAR% was Jamaica and the US suggest group differences in physical
66% and 80%, respectively, for BMI and WHR, suggesting activity.
the need for vigorous prevention efforts among even those in In summary, evidence presented here con®rms previous
the normal range of relative weight. reports that the prevalence of diabetes and associated risk
The epidemic of obesity in the Caribbean, especially factors, especially obesity, has reached epidemic propor-
amongwomen,isbeginning toreceivedeservedattention,as tionsin theCaribbean. Rates,atleast amongwomen,exceed
evidence by the recent conference on the origins and those for most European-origin populations [19]. While it is
consequences of obesity [38]. Using the National Institutes imperative to re®ne our understanding of the epidemiology
of Health consensus values (BMI > 27.8 kg/m2 for men and and genetics of diabetes, the implementation of coherent
27.3 kg/m2 women), Forrester et al. reported that females public health policy advocating preventive practices ±
have twice the rate of being overweight as males in Jamaica, including reduction in the level of obesity through nutri-
Barbados and St Lucia [9]. For women the rate of obesity, tional education and increased physical activity, an under-
de®ned as a BMI > 31.1 kg/m2 for men and 32.3 kg/m2 standing of family history and early detection programs for
women, was four times that of men in Jamaica (3.9 vs. high risk groups ± has become an urgent challenge.
15.6%). In Barbados, the rate of obesity for women was
three times that of men, with over 30% of the women
classi®ed as obese [9]. Similar high rates of overweight and References
obesity has been observed in the island of Trinidad [2]. The 1 Gulliford MC. Epidemiological transition in Trinidad and Tobago,
WHO classi®cation for obesity suggests that a BMI < 25 kg/ West Indies 1953±1992. Inter J Epidemiol 1996; 25: 357±365.
m2 is normal, however, two recent studies have shown that 2 Miller GJ, Maude GH, Beckles GLA. Incidence of hypertension and
BMI within the so called safe range are associated with non-insulin dependent diabetes mellitus and associated risk factors in
a rapidly developing Caribbean community: the St James survey,
diabetes [39,40]. In our data analysed by BMI quartiles
Trinidad. J Epidemiol Community Health 1996; 50: 497±504.
(Table 4), there is an increased risk of diabetes beginning 3 Alleyne SI, Cruickshank JK, Golding AL, Morrison EY. Mortality
with the second quartile (a lower limit of 20.8 kg/m2 for men from diabetes mellitus in Jamaica. Bull Pan Am Health Organ 1989;
and 23.6 kg/m2 for women). Although statistically signi®- 23: 306±314.
cant for women, the failure to achieve signi®cance for men 4 Hagley K. Chronic non-communicable diseases and their impact on
probably re¯ects the small sample size involved in the Caribbean women. West Indian Med J 1990; 39: 4±11.
5 Miller GJ, Kirkwood BP, Beckles GLA, Alexis SD, Carson DC, Byam
comparison between the second and ®rst quartile. When the
NT et al. Adult male all-cause, cardiovascular and cerebrovascular
data was dichotomised at a BMI of 23 kg/m2, individuals mortality in relation to ethnic groups, systolic blood pressure and
above this cut off were at an increased risk of diabetes in both blood glucose concentration in Trinidad, West Indies. Int J Epidemiol
men (OR = 2.84 (95% CI 1.47±5.47); P = 0.002) and 1988; 17: 62±69.
6 Poon-King T, Henry MV, Rampersad F. Prevalence and natural Indians. In Eschwege E ed. Advances in Diabetes Epidemiology,
history of diabetes in Trinidad. Lancet 1968; 1: 155±160. INSERM Symposium no. 22. Amsterdam: Elsevier Biomedical Press,
7 Eldemire D, Hagley K. Diabetes mellitus in the Jamaican elderly. West 1982; 65±74.
Indian Med J 1996; 45: 82±84. 28 KeenH,JarretRJ,McCartneyP.Theten-yearfollow-upoftheBedford
8 RagoobirsinghD, Lewis-Fuller E,Morrison E.The Jamaican Diabetes Survey (1962±1972): glucose tolerance and diabetes. Diabetologia
Survey. A protocol for the Caribbean. Diabetes Care 1995; 18: 1277± 1982; 22: 73±78.
1279. 29 Dowse G, Zimmer P, King H. Relationship between prevalence of
9 Forrester T, Wilks R, Bennett F, McFarlane-Anderson, McGee D, impaired glucose tolerance and NIDDM in a population. Diabetes
Cooper R et al. 1996 Obesity in the Caribbean. In The Origins and Care 1991; 14: 968±974.
Consequences of Obesity, Ciba Foundation Symposium 201. 30 McLarty DG, Swai ABM, Kitange HM et al. Prevalence of diabetes
Chichester: John Wiley & Sons, 1996; 17±31. and impaired glucose tolerance in rural Tanzania. Lancet 1989; 1:
10 Lemeshow S, Robinson D. Surveys to measure programme coverage 871±875.
and impact: a review of the methodology used by the expanded 31 Swai ABM, McLarty DG, Kitange HM et al. Study in Tanzania of
programme onimmunization.WorldHealth StatQuart 1985;38:65± impaired glucose tolerance: Methodological myth? Diabetes 1991;
75. 40: 516±520.
11 Henderson RH, Sundaresan T. Cluster sampling to assess immuniza- 32 RiccardiG,Vaccaro O,RivelleseA,Pignalosa S,TutinoL,ManciniM.
tion coverage: a review of experience with a simpli®ed sampling Reproducibility of the new diagnostic criteria for impaired glucose
method. Bull WHO 1982; 60: 253±260. tolerance. Am J Epidemiol 1985; 121: 422±429.
12 Levy P, Yu ESH, Liu WT, Wong S-C, Zhang M-Y, Katzman R. Single- 33 Florey C, du V, McDonald H, McDonald J, Miall WE. The prevalence
stage cluster sampling with a telescopic respondent rule: a variation of diabetes in a rural population of Jamaican adults. Int J Epidemiol
motivated by a survey of 0dementia in elderly residents of Shanghai. 1972; 1: 157±166.
Statist Med 1989; 8: 1537±1544. 34 FitzGerald MG, Keen H. Diagnostic classi®cation of diabetes. Br Med
13 AtamanSL,CooperRS, RotimiCN,McGee D,Osotimehin B,KadiriS J 1964; i: 1568.
et al. Standardization of blood pressure measurement in an interna- 35 Ohlson LO, Larsson B, Svardsudd K, Welin L, Eriksson H,
tional comparative study. J Clin Epidemiol 1996; 49: 869±877.
Wilhelmsen L et al. The in¯uence of body fat distribution on the
14 Cooper RS,RotimiCN, AtamanSL,McGee D,Osotimehin B,KadiriS
incidence of diabetes mellitus. 13.5 years of follow-up of the
et al. Prevalence of hypertension in seven populations of African
participants in The Study of Men Born in 1913. Diabetes 1985; 34:
origin. Am J Public Health 1997; 87: 160±169.
1055±1058.
15 World Health Organization. WHO Expert Committee on Diabetes
36 Folsom AR, Gurke GI, Ballew C, Jacobs DRJ, Haskell WL, Donahue
Mellitus Second Report (Technical Report Series Number 626).
RP et al. Relation of body fatness and its distribution to cardiovascular
Geneva: World Health Organization, 1980.
risk factors in young blacks and whites: the role of insulin. Am J
16 Knowler WC, Pettitt DJ, Saad MF, Bennett PH. Diabetes mellitus in
Epidemiol 1989; 130: 911±924.
the Pima Indians: incidence, risk factors and pathogenesis. Diab
37 HaffnerSM,SternMP,HazudaHP,PughJ,Patterson,JK.Upper-body
Metab Rev 1990; 6: 1±27.
and centralized adiposity measure different aspects of regional body-
17 TheExpertCommitteeontheDiagnosisandClassi®cationofDiabetes
fat distribution? Relationship to non-insulin-dependent diabetes
Mellitus. Report of the Expert Committee on the Diagnosis and
mellitus, lipids, and lipoproteins. Diabetes 1987; 36: 43±51.
Classi®cation of Diabetes Mellitus. Diabetes Care 1997; 20: 1183±
38 Chadwick DJ, Cardew G, eds. The Origins and Consequences of
1197.
Obesity. Ciba Foundation Symposium 201. Chichester: JohnWiley&
18 SAS User's Guide. Statistics, Version 6. Cary, NC: SAS Institute Inc.,
Sons, 1996.
1989.
39 Perry IJ, Wannamethee SG, Walker MK, Thomson AG, Whincup PH,
19 King H, Rewers M. Global estimates for prevalence of diabetes
mellitusandimpairedglucosetoleranceinadults.DiabetesCare1993; Shaper AG. Prospective study of risk factors for the development of
16: 157±177. non-insulin dependent diabetes in middle-aged British men. Br Med J
20 Segi M. Trends in cancer mortality for selected sites in 24 countries 1995; 310: 560±564.
(1950±59). Sendai, Japan: Tohuku University School of Medicine, 40 Colditz GA,WilletWC, Rotnitzky A, Manson JE. Weightgain as arisk
1963. factor for clinical diabetes mellitus. Ann Int Med 1995; 122: 481±486.
21 Greenland S, Drescher K. Maximum likelihood estimation of the 41 PoulterNR,KhawKT,HopwoodBE,MugambiM,PeartWS,SeverPS
attributable fraction from logistic models. Biometrics 1993; 49: 865± et al. Determinants of blood pressure changes due to urbanization: a
872. longitudinal study. J Hypertens 1985; 3: 5375±5377.
22 Hennekins CH, Buring JE. Epidemiology in Medicine. Boston: Little, 42 Zimmet PZ. Kelly West lecture 1991: challenges in diabetes
Brown and Company, 1987; 90±93. epidemiology ± from West to the rest. Diabetes Care 1992; 15: 232±
23 Foster C, Rotimi C, Fraser H, Sundarum C, Liao Y, Gibson E et al. 252.
Hypertension,diabetesandobesityinBarbados:®ndingsfromarecent 43 Knowler WC, Pettitt DJ, Savage PJ, Bennett PH. Diabetes incidence in
population-based survey. Ethnicity Dis 1993; 3: 404±412. Pima Indians: contribution of obesity and parental diabetes. Am J
24 Morrison EY. Diabetes mellitus in Jamaica. West Indian Med J 1983; Epidemiol 1981; 113: 144±156.
32: 199±200. 44 Helmrich SP, Ragland DR, Leung RW, Paffenbarger RSJ. Physical
25 Beckles G. The control of diabetes mellitus in the Caribbean activity and reduced occurrence of non-insulin-dependent diabetes
community. PAHO/WHO manual containing Proceedings of a mellitus. N Engl J Med 1991; 325: 147±152.
Workshop on Standardization of Management of Hypertension 45 Manson JE, Nathan DM, Krolewski SA, Stampfer MJ, Willett WC,
Held in St Lucia. Washington, DC: PAHO/WHO, 1986; 38±53. Hennekens CH. A prospective study of exercise and incidence of
26 Harris MI, Hadden WC, Knowler WC, Bennett PH. Prevalence of diabetes among US male physicians. JAMA 1992; 268: 63±67.
diabetes and impaired glucose tolerance and plasma glucose levels in 46 Luke A, Durazo-Arvizu R, Rotimi C, Prewitt TE, Forrester T, Wilks R
the US population aged 20±74 yr. Diabetes 1987; 36: 523±534. et al. Relation between body mass index and body fat in black
27 Bennett PH, Knowler WC, Pettitt DJ, Carraher MJ, Vasquez B. population samples from Nigeria, Jamaica, and the United States. Am
Longitudinal studies of the development of diabetes in the Pima J Epidemiol 1997; 145: 620±628.