The Prevalence and Identification of Risk Factors for ... - Diabetes Care

O R I G I N A L

A R T I C L E

T

The Prevalence and Identification of Risk Factors for NIDDM in Urban Africans in Cape Town, South Africa NAOMI S. LEVITT, MD JUDITH M. KATZENELLENBOGEN, MSC DEBORAH BRADSHAW, PHD MARGARET N. HOFFMAN, MBCHB DIP COMM MED FRANCOIS BONNICI, MMED FCP

OBJECTIVE — To determine the prevalence of NIDDM and associated risk factors in urban Africans in Cape Town, South Africa. RESEARCH DESIGN AND METHODS—With a three-stage, proportional, stratified, random cluster method, we sampled 1000 Africans, >30 yr of age, living in African residential areas in Cape Town. We assessed glucose tolerance with a 75-g oral glucose tolerance test, according to World Health Organization criteria, and obtained anthropometric and demographic data. RESULTS— The response rate was 79%. The prevalence of NIDDM was 8.0% (confidence interval 5.8-10.3%), age-adjusted to world population figures and that of impaired glucose tolerance, 7.0% (confidence interval 4.9-9.1%). Multivariate analysis indicated that increased age (odds ratio 4.18), upper-segment fat distribution (odds ratio 2.94), proportion of life spent in an urban area (odds ratio 2.32), and obesity (odds ratio 2.31) were significant independent risk factors for NIDDM. In contrast, sex, family history, alcohol intake, and physical activity were not independent risk factors. Only increased age (odds ratio 4.06) was a significant risk factor for impaired glucose tolerance. CONCLUSIONS— The prevalence of NIDDM in urban Africans in Cape Town, South Africa, is moderately high, and considerably higher than previous reports from Africa. The association of NIDDM with urbanization has important implications in view of the large-scale urbanization occurring in southern Africa.

FROM THE DEPARTMENTS OF MEDICINE, COMMUNITY MEDICINE, UNIVERSITY OF CAPE TOWN, CAPE TOWN; AND THE CENTER FOR EPIDEMIOLOGICAL RESEARCH FOR SOUTHERN AFRICA, MRC,

TYGERBERG,

CAPE, SOUTH AFRICA. ADDRESS CORRESPONDENCE AND REPRINT REQUESTS TO NAOMI S.

LEVITT, MD, DEPARTMENT OF

MEDICINE, UNIVERSITY OF CAPE TOWN MEDICAL SCHOOL, OBSERVATORY 7925, CAPE, SOUTH AFRICA. RECEIVED FOR PUBLICATION 14 JANUARY 1992

AND ACCEPTED IN REVISED FORM 5 NOVEMBER

1992.

N I D D M , NON-INSULIN-DEPENDENT DIABETES MELLITUS; I G T , IMPAIRED GLUCOSE TOLERANCE; O G T T , ORAL GLUCOSE TOLERANCE TEST; W H O , WORLD HEALTH ORGANIZATION; C I , CONFIDENCE INTERVAL; O R , ODDS RATIO; FPG,

FASTING PLASMA GLUCOSE; WHR,

DIABETES CARE, VOLUME 16, NUMBER 4, APRIL

WAIST-TO-HIP RATIO.

1993

he lack of epidemiological studies on diabetes in Africans in southern Africa has led to considerable speculation concerning its prevalence. The most recent studies, conducted two decades ago, reported prevalence rates of 2.9% in Mamelodi, Pretoria, and 3.6% in Guguletu, Cape Town, South Africa (1). Further north, a more recent study in Tanzania, using the current WHO diagnostic criteria for diabetes, reported a prevalence rate of 0.8% for diabetes in rural Africans, which rose to 1.1% when age-adjusted to the U.S. population (2). With the changing sociodemographic picture in South Africa, the local data clearly need to be updated. Notwithstanding legislation used to restrict the number of Africans in the urban areas from 1954-1986, the pattern of urbanization has been in accord with international trends. By 1900 an estimated 10,000 Africans lived in Cape Town, and after 1940 three segregated townships were established. For a number of decades no further housing facilities were developed and, increasingly, pockets of squatter areas mushroomed where people lived in an attempt to escape local control.

Over the past decade and especially since the lapse in enforcement of laws designed to control population movement, the Cape Town African population has grown substantially. For the first time since 1968, housing has been built for families instead of hostels for migrant laborers. This has resulted in the rapid development of new townships, which offer accommodations ranging from formal housing to squatting in both organized serviced sites and informal areas without services. Current estimates put the population in the greater Cape Town area at 2.5 million, of which 600,000 are Africans (3), predominantly Xhosa speaking. By the year 2000, 80% of the country's African population will be urbanized (4). If, as demonstrated in other societies, urbanization is associated with a rise in

601

The prevalence of NIDDM in urban Africans

the prevalence of diabetes (5-9), this will make a significant demand on urban health-care resources. This study was designed to determine the prevalence of NIDDM and its associated risk factors in Africans in Cape Town. Risk factors included sex, age, degree of urbanization, physical activity, obesity, upper-segment fat distribution, family history of diabetes, and alcohol abuse. RESEARCH DESIGN AND METHODS — The study took place in the greater Cape Town area over a 6-mo period beginning in May 1990. The study population comprised Africans, >30 yr of age, living in townships allocated for African settlement in Cape Town. (The recently repealed Group Areas Act designated separate areas for different racial groups.) The housing types comprised tents, shacks, hostels, and formal dwellings. We used proportional, three- stage cluster sampling, with stratification by area and housing type to draw a sample from the community. A 1985 census, with subsequent revisions in 1988, was the basis for weighting the strata. In the first stage of sampling, a range of maps and aerial photographs, complemented by scouting, was used to divide each township into clusters of —40 households. Thus, 250 clusters were randomly selected. In the second stage, a neighborhood of four households was randomly selected from each cluster. After a household census was completed, a single individual per dwelling was randomly selected by the field worker. No replacements were allowed, and households with no reply were visited three times before being classified as nonresponders. We asked selected participants to fast overnight. In the morning they were taken to a central venue. All participants, except those with a validated diagnosis of diabetes, underwent a standard 75-g OGTT (glucose monohydrate with the glucose dissolved in 250 ml of water). Blood samples were drawn for glucose

602

estimations in the fasting state and 2 h after ingestion of glucose. Samples were kept on ice, spun within 6 h, and stored at — 20°C until they were measured with the glucose oxidase method on an autoanalyzer (Beckman, Fullerton, CA). Diabetes and IGT were diagnosed according to the 1985 WHO criteria (2-h plasma glucose > 11.1 mM for diabetes and 2-h plasma glucose > 7.8 and < 11.1 mM with FPG 2 drinks/day during the week and/or >8 drinks on the weekend); light to moderate (40% of life spent in an urban area; obesity, men BMI >27 kg/m2, women BMI >25 kg/m2(13); uppersegment fat distribution, WHR (based on upper quartiles), >0.92 in men, >0.84 in women; physical activity (two levels), strenuous versus minimal to moderate (in work and/or leisure); alcohol, nondrinkers versus drinkers. Population-attributable risk fractions were calculated for urbanization, obesity, and WHR with the cutoff points mentioned above(13). RESULTS

Quality of data Of the sample of 1000, 71 had to be excluded because of political violence, and 5 because they were under age. The analysis was based on 729 subjects, for a response rate of 79%. Of these, 70% were women and 30% were men. We obtained no demographic details of the households of 51 nonresponders. Of the 144 people who were selected but did not participate, 63% were men, reflecting a greater difficulty to motivate men to participate. Thus, of the section of the target sample whose sex was known (873), 62% were women and 38% were men. This is significantly different from

DIABETES CARE, VOLUME 16,

NUMBER 4, APRIL

1993

Levitt and

Table 1—Sociodemographic characteristics for the study population

N* AGE (YR)I

MEN

WOMEN

214 48.7 ± 14.7

503 43.6 ± 12.1

EDUCATION NONE

(%)

< 5 YR (%) 5-10 YR(%) 11-12 Y R ( % )

17 17 62 4

18 13 66 4

37 24 21 18

30 44 13 13

44 20 36

82 13 5

EMPLOYMENT EMPLOYED

(%)

UNEMPLOYED (FIT FOR WORK) PENSIONER OTHER

(%)

(%)

(%)

ALCOHOL NONDRINKER (%) LIGHT TO MODERATE (%) HEAVY (%)

*Sex not recorded in 12 subjects. tData are means ± SD.

the distribution found in the household census (55% women and 45% men) in this age range. Because of the overrepresentation of women in the sample, the statistical analyses were stratified on the basis of sex.

Prevalence of glucose intolerance The crude prevalence of NIDDM was 6.3% and of IGT, 5.9%. When age adjusted to the world population, the prevalence was 8.0% (CI 5.8-10.3%) for NIDDM and 7.0% (CI 4.9-9.1%) for IGT. The prevalence of NIDDM among

Associates

participants, 30-65 yr of age, age adjusted to the world population, was 6.9% (CI 4.7-9.1%); in men it was 6.1% (CI 4.4-7.9%) and in women 7.4% (CI 4.610.1%). The prevalence of IGT in the 30- to 65-yr age-group, when ageadjusted to the world population, was 7% (CI 4.6-9.3%). The crude prevalence of NIDDM was similar in men (6.5%) and women (6.4%). The mean age of those with diabetes was 56 ± 12 yr, compared with 44 ± 13 yr in the nondiabetic participants. Of the 46 with diabetes, 24 had been diagnosed previously. Of these 24 known diabetic individuals with NIDDM, only 16 (66%) were seeing a medical practitioner or clinic for their diabetes. The mean FPG concentrations were 4.4 ± 1 . 6 mM for men and 4.4 ± 0.89 mM for women. A small increase in FPG and 2-h plasma glucose concentrations occurred with age (Table 3).

Univariate analysis of risk factors The prevalence of NIDDM increased with age in both men and women (P < 0.001), as shown in Fig. 1. In women the prevalence rose, particularly

Characteristics of respondents Tables 1 and 2 show the sociodemographic characteristics and risk-factor profiles of the participants. We noted a considerable degree of obesity, particularly among women who had spent a smaller proportion of their lives in the city. The migration pattern also differed by sex, with most of the men having arrived in the city in their second decade, presumably to seek jobs. In contrast, the women had come to the city in three waves, most during the latter part of their second decade, smaller numbers subsequently after 45 yr of age, and yet another group after 65 yr of age. The low education level of the sample is evident, as is the considerable degree of unemployment. Particularly noteworthy is the small proportion of participants who perform strenuous physical activity.

DIABETES CARE, VOLUME 16,

NUMBER 4, APRIL

Table 2—Prevalence of diabetes and IGT and a risk-factor profile for the study population MEN

WOMKN

6.5 14

6.4 32

6.0 13 24.2 ± 4.2 0.89 ± 0.07 4.6 27 (53)

5.9 30 30.9 ± 7.8 0.80 ± 0.07 8.0 34 (42)

DIABETES PREVALENCE

(%)

N

IGT PREVALENCE (%)

N BM1 (KG/M 2 )*

WHR* FAMILY HISTORY OF DIABETES (%) PERCENTAGE OF LIFE IN CITY, N 0 PHYSICAL ACTIVITY NONE

(%)

MINIMAL TO MODERATE STRENUOUS

(%)

(%)

39 40 21

44 54 1

*Data are means ± SD.

1993

603

The prevalence of NIDDM in urban Africans

Table 3—FPG and 2- h plasma glucose concentrations in relation to age AGE-GROUP

FPG

(MM)

N

2-H N

PLASMA GLUCOSE ( M M )

30-34 YR

35-44 YR

45-54 YR

55-64 YR

> 6 5 YR

4.2 ± 0.8 194 4.8 ± 1.4 190

4.3 ± 0.9 206 4.9 ± 1.4 205

4.5 ± 0.9 116 5.4 ± 1.6 116

4.6 ± 1.1 92 5.5 ± 1.7 92

4.9 ± 1.0 63 6.1 ± 1.7 62

Data are means ± SD. Diabetic patients were excluded. Information on precise age was missing from 12 subjects.

after 45 yr of age, whereas in men the rise occurred earlier, in those >35 yr of age. The crude prevalence of NIDDM increased with BMI and upper-segment fat distribution (Fig. 2). In those with lower-segment fat distribution the crude prevalence was 1.3% in the nonobese, compared with 4.8% in the obese. In participants with upper-segment fat distribution the nonobese had a crude prevalence of 10.9%, whereas the prevalence was 16% in the obese. These data indicate an additive effect of obesity and upper-segment fat distribution, which was consistent in men and women. Based on the Mantel-Haenszel \ 2 test, we found a significant relationship between NIDDM and urbanization (P = 0.004), with a marked rise evident in participants who spent > 40% of their lives in the city (Fig. 3). Only one of the participants who performed heavy physical activity had diabetes; we found no

statistically significant relationship between NIDDM and physical activity (P = 0.30). Nor did we find an association between NIDDM and family history (P = 0.627) or alcohol intake (P = 0.685).

Multivariate analysis

We performed logistic regression analysis to assess the independent nature of the various risk factors for NIDDM. Table 4 shows the results of the backward selection logistic regression analysis. These data indicate, in descending order, age, upper-segment fat distribution, urbanization, and obesity are all independent risk factors for NIDDM in this population. The selected model did not indicate sex, physical activity, alcohol intake, or a family history of NIDDM, as risk factors. We also performed univariate and multivariate analyses of all risk factors for IGT. After diabetic cases were excluded, analysis indicated that age was the only significant independent risk factor for IGT, with an OR = 4.06 (CI 2.06-6.06; P = 0.0001). The population-attributable risk fractions were calculated for selected risk factors to demonstrate the potential reduction in disease occurrence that might ensue after changes in risk-factor levels. Population-attributable risk fractions for NIDDM were 40% for urbanization, 43% AGE (yr) for obesity, and 53% for upper-segment Figure 1—Prevalence of NIDDM by age in fat distribution. Because the analysis inmen O and women (E3) (P < 0.001). The dicates these are independent factors, if number of participants in each group is indiall three risk factors were removed, the cated.

604

prevalence of NIDDM would fall from 6.3 to 1%. CONCLUSIONS— This study demonstrates a moderate prevalence of NIDDM in Africans living in Cape Town. The age-adjusted rate of 8% is considerably higher than any previously reported in Africans either in South Africa (1) or the rest of Africa (2,15), but differences in methodology make comparisons difficult. One comparative study, which also used the WHO criteria, reported an ageadjusted prevalence of 1.1% in rural Africans in Tanzania (2). Three major differences between the two studies could account for the disparity: the age ranges studied (>30 yr of age in Cape Town and >15 yr of age in Tanzania), the greater degrees of obesity, and level of urbanization in the Cape Town study. We identify urbanization as a significant independent risk factor for NIDDM in this Cape Town (urban) sam-

Otmt

Lower etgment

Upptr sagmtnt

FAT DISTRIBUTION

Figure 2—Prevalence of NIDDM in relation to obesity and body fat distribution. The number of participants in each category is indicated.

DIABETES CAKE, VOLUME 16,

NUMBER 4, APRIL

1993

Levitt and Associates

city. The selection of >40% of life spent in the city to determine urbanization is supported by the prominent increase in the prevalence of NIDDM noted at this point. The factors contributing to the rise in prevalence of NIDDM with urbanization have not been defined clearly and may include either qualitative or quantitative alterations in diet, changes in the degree of physical activity, and stressrelated phenomena (5). Although we do not have dietary data available from this study, a considerable difference in the diet of rural and urban Zulu women has been reported. Albertse et al. (18) found the diet of rural Zulu women comprised 69% carbohydrate, 17% fat, 13% protein, and 37 g/day of fiber, compared with 50% carbohydrate, 31% fat, 16% protein, and 14 g/day of fiber in urban Zulu women. The significant alteration in diet constitutes an important factor that might result in a rise in the prevalence of NIDDM. Although several cross-sectional and prospective studies have suggested lack of physical activity is a risk factor for diabetes (19-21), our study did not confirm this relationship. Factors that could account for this disparity include: J) the relatively crude index of physical activity used, 2) the small proportion (7%) of the total sample that performed heavy physical activity, and 3) the imbalance in the number of men and women in the study sample and the relatively low level

8

I 6 20 - 39 40 - 59 60 - 79 SO - 100 Percentage of life spent in city

Figure 3—Age- and sex-adjusted prevalence of NIDDM in relation to urbanization expressed as quintiles of percentage stay in the city (P = 0.004). The number of participants in each Quintile is indicated.

pie, as have numerous other epidemiological studies (5-9). The definition and measurement of urbanization is under considerable debate (16). In this study we use the simple measures of length of time in the city and proportion of life spent in the city as markers of exposure to the urban environment. Interestingly, in 1960, Campbell (17) in Natal, South Africa, applied the rule of 20 yr to diabetes. He observed a peak incubation period of 18-22 yr of urban residence in a group of Zulus before diabetes developed. Our study also found a substantial rise in the prevalence of NIDDM after 20 yr of urban residence, but we prefer to use the proportion of life spent in the city as the exposure measure because of the complex interrelationship between age and duration of time spent in the

Table 4—Multivariate analysis of risk factors as predictors of diabetes based on backward selection logistic regression

AGEt SHAPE? URBANIZATIONS OBESITY||

OR (95% CD

P VALUE

4.18(2.48-9.52) 2.94(1.47-5.64) 2.32 (1.09-4.95) 2.31 (1.06-5.02)

0.0007 0.0017 0.0295 0.0351

"Overall r = 0.369 (13.6% of variation explained). tAge: £45 vs.