The sensitivity and specificity of each of five screening tests were estimated in each of three to ten countries by testing subjects drawn from the general populations of adults over thirty-four years of age. This permitted comparisons among countries and among the different tests (fasting, postprandial, and postglucose urine tests, and fasting and postprandial blood glucose values). Sensitivity and specificity of each test varied widely among populations. For example, the sensitivity of the two-hour urine glucose ranged from 17 per cent in Nicaragua to 100 per cent in East Pakistan. Apparently specificity and sensitivity of such tests are influenced by many factors including both the circumstances under which the tests are performed and the characteristics of the population tested. It is, therefore, not possible to predict prevalence rates reliably by extrapolating from the results of screening tests. However, we believe the data for specific populations on the sensitivity and specificity of various tests will provide a rough guide in predicting the cost-effectiveness of alternative approaches to case detection in those particular countries. For instance, these results suggest that roughly 56 per cent of the occult diabetics in Costa Rica in this age group would be detected by a two-hour urine glucose, but only about 41 per cent of those in whom this test was positive would prove to have diabetes. Even modest changes of criteria in defining either "diabetes" or "abnormality" of the screening results produced marked changes in rates of sensitivity and specificity. With few exceptions, tests which were more sensitive were, comparably, less specific, and the reverse was also true. Rates of "diabetes" were markedly influenced by modest changes in diagnostic criteria.
In each of four countries (Uruguay, Venezuela, Malaya and East Pakistan) where diets and other environmental factors differ greatly, the prevalence of diabetes as determined by impaired glucose tolerance was crudely estimated. Since all subjects received glucose loads, rates of prevalence are much higher than those obtainable by certain less sensitive standard methods. In the tested subjects over thirty years of age the prevalence of "diabetes" (two-hour venous blood glucose levels greater than 149 mg. per 100 ml.) was 6.9 per cent in Uruguay (6.8 per cent for males and 6.9 per cent for females). The prevalence of impaired tolerance in this age group in Venezuela was 7.3 per cent (4.5 per cent in males and 9.4 per cent in females), while in Malaya the rate was only 3.5 per cent (4.5 per cent in
males and 2.1 per cent in females). In East Pakistan impaired tolerance was present in only 1.5 per cent of this age group (1.2 per cent of males and 2.8 per cent of females). Comparable data are not available in the United States but with use of the technics employed abroad it was found that 17.2 per cent of volunteers in this age group in a Pennsylvania community had impaired tolerance. In East Pakistan, 83 per cent of calories were derived from carbohydrate. Comparable figures were 77 per cent for Malaya, 62 per cent for Venezuela and 53 per cent for Uruguay. In East Pakistan, only 7 per cent of the dietary calories were derived from fat; in Malaya, fat accounted for 21 per cent of dietary calories, in Venezuela, 24 per cent, and in Uruguay, 33 per cent. In East Pakistan only 29 per cent of dietary fat was animal fat. In Malaya, Venezuela, and Uruguay, comparable figures were 30, 35 and 62 per cent, respectively. In Uruguay, 34.4 per cent of the subjects were "obese" (30 per cent or more over "standard" weight), and in Venezuela 14.8 per cent were obese. In contrast none of the subjects from Malaya (566 persons), or East Pakistan (519 persons), was obese by these criteria. In Venezuela and Uruguay there was an association between the prevalence of diabetes and both parity and a history of large babies.
The prevalence of diabetes in Central America was somewhat greater than in East Pakistan and Malaya, less
than in Uruguay and Venezuela, and substantially less than in affluent societies such as the United States. Differences in prevalence among Central American countries were modest but probably significant in some instances. In all Central American countries diabetes was more common in females but this difference was probably attributable to the greater adiposity of the women. Age-matched populations from eleven different countries of three continents have now been tested using standardized methods. Prevalence of diabetes varied greatly, and differences were more related to environment than to race. These results support the hypothesis that environmental factors can increase or reduce prevalence by several-fold.