Comparison of the Effect of Plasma Glucose Concentrations on Microvascular Disease Between Pima Indian Youths and Adults

RESEARCH DESIGN AND METHODS

Members of the Gila River Indian Community have participated in a longitudinal study of diabetes and its complications since 1965. Every 2 years, residents of the community aged ≥5 years are asked to participate regardless of health status. Each participant undergoes a physical examination, which includes direct opthalmoscopy, a 75-g oral glucose tolerance test, and a spot urinalysis for assessment of creatinine and protein. The present analysis is restricted to subjects for whom both FPG and 2-h PG were measured. Because FPG was routinely collected beginning in 1975, the majority of exams included those from 1975 onward. The opthalmoscopy is performed without the examiner knowing the diabetes status of the participants. Pupils are dilated in participants aged ≥15 years. Retinopathy was defined as the presence of at least one microaneurysm, hemorrhage, or evidence of proliferative retinopathy. The technique and standards for diagnosing retinopathy have not changed over the period of the study. PG concentration was measured by the potassium ferricyanide method with an autoanalyzer or by the hexokinase method (in recent years). Subjects were asked to void at the beginning of the oral glucose tolerance test, and a urine specimen was collected 2 h later. Proteinuria was assessed using a dipstick test; throughout the study, all patients with levels of protein greater than or equal to trace on the urine dipstick had total urine protein measurement performed by the method of Shevky and Stafford (18). Urine creatinine was measured by the alkaline-picrate method. Nephropathy was defined by a protein-to-creatinine ratio ≥0.5 g.

Statistical analysis.

Our aim was to analyze the prognostic significance of hyperglycemia in youths. Therefore, we examined the relationship of glucose levels measured at age 5–19 years to the incidence of complications at age 25–39 years (youths). For comparison, we similarly examined the relationship of glucose in young adults aged 20–34 years to complications at age 40–54 years (adults). A baseline exam was defined for each group as the last exam within the initial age range, when FPG and 2-h PG were measured. A follow-up exam was defined as the last exam within the second age range, when an examination for retinopathy was performed and proteinuria was measured. Subjects with retinopathy or proteinuria at the baseline exam were excluded. Subjects were defined as having microvascular disease if they had retinopathy, nephropathy, or both at their last exam. Separate analyses for retinopathy and nephropathy were also performed. Analysis conducted with and without exclusion of subjects taking glucose-lowering medications revealed that excluded subjects came primarily from the extreme upper glucose categories in both groups, somewhat weakening the association between glucose and complications in both groups (data not shown). Because our goal was to examine the biological effect of glucose and because eliminating subjects on medication would eliminate only the most hyperglycemic subjects, we chose to include those subjects. Individuals were classified into categories based on glucose concentrations in relation to reference points that were equidistant on a logarithmic scale. Cumulative incidence of microvascular disease was taken as the proportion of subjects within each category who developed disease.

The differences in the predictive properties between the youths and adults were assessed by combining the groups in the logistic-regression model (using FPG and 2-h PG as continuous variables) and testing an interaction term.

The properties of specific levels of FPG and 2-h PG for predicting microvascular disease were determined by calculating sensitivity and specificity. Sensitivity was defined as the percentage of subjects with a microvascular endpoint whose baseline FPG or 2-h PG was greater than or equal to a specific cutoff. Specificity was defined as the percentage of subjects without a microvascular endpoint, whose baseline FPG or 2-h PG was below the cutoff. Receiver operating characteristics (ROC) curves were also generated for FPG and 2-h PG for each group.

ROC curves are plots of the sensitivity versus 1-specificity and are useful for determining the characteristics of a diagnostic test. The area under the ROC curve, which was calculated with the Wilcoxon statistic (19), quantifies the diagnostic accuracy of a test. ROC curve areas range from 0.5 to 1.0. An area of 0.5 implies no separation of the test values in the affected and unaffected groups. An area of 1.0 implies perfect separation between test values in the two groups. The area under the ROC curve indicates the strength of the association between the test variable and the outcome measure (19,20).

RESULTS

The youth group had 928 subjects, of whom 73 developed microvascular disease, and the adult group had 566 subjects, of whom 104 developed microvascular disease (Table 1). Median follow-up was slightly longer in the youth group.

Figures 1 and 2 show the frequency distributions of baseline FPG and 2-h PG on a log scale plotted with the incidence (percent) of microvascular disease. The frequency distributions for glucose clearly differ between the groups, with much higher percentages of individuals in the youth group (aged 5–19 years) in the lower glucose categories. For FPG (Fig. 1) and 2-h PG (Fig. 2), the incidence of microvascular disease began to increase dramatically at similar points in each age group. For FPG, the increase occurred at FPG 7.34 mmol/l in the youths and at 7.53 mmol/l in the adults; for 2-h PG, the increase occurred at 9.97 mmol/l (youths) and 10.32 mmol/l (adults). The sustained increase in the percentage of subjects within these glucose groups who developed microvascular disease suggested an approximate threshold. Yet, the percentage of subjects developing microvascular disease who had baseline glucose concentrations above these apparent thresholds was very different in each group. In fact, among the subjects who developed microvascular disease, the percentage with baseline FPG ≥6.7 mmol/l or 2-h PG ≥9.0 mmol/l was twice as high in adults (64.4 and 76.0%, respectively) as in youths (23.2 and 31.5%, respectively). Therefore, in youths, only a minority of subjects who developed microvascular disease had relatively high (FPG ≥6.7 mmol/l and 2-h PG ≥9.0 mmol/l) baseline blood glucose.

When the age groups were combined in a logistic regression analysis including FPG, group (youth or adult), and an interaction term, there was no significant interaction, i.e., the slopes of the complication-FPG curves were not significantly different between age groups. Similarly, there was no significant interaction between group and 2-h PG. This suggested that a given difference in glucose values was associated with a similar increase in risk for development of microvascular disease in both groups. The absolute risk for development of complications was higher at each FPG and 2-h PG level for adults compared with youths, but not significantly. Odds ratios comparing adults to youths for the models containing FPG and 2-h PG were 1.3 (P = 0.13) and 1.1 (P = 0.75), respectively.

The diagnostic properties of FPG and 2-h PG are shown in Table 2. Sensitivity for FPG ≥7.0 mmol/l and 2-h PG ≥11.1 mmol/l was substantially lower for youths than for adults. Sensitivity improved at lower glucose thresholds (in this case FPG ≥5.9 mmol/l and 2-h PG ≥8.9 mmol/l) for youths without a substantial decrease in specificity, which was not the case for adults. Furthermore, ROC curve areas were lower for youths than for adults (Table 3), suggesting a weaker relationship between microvascular disease and baseline glucose. However, when nephropathy and retinopathy were analyzed separately, the area was lower for nephropathy but nearly equal for retinopathy in youths compared with adults.

CONCLUSIONS

The current guidelines for the diagnosis of diabetes are based on adult-population studies that have evaluated the relationship between glucose concentrations and microvascular disease (1). Because of the paucity of cases, similar information in youths has not been available. The ongoing biennial examinations of the Pima Indians have allowed us to collect follow-up data on a large number of individuals at various glucose levels during childhood and adolescence. We have shown that development of microvascular disease was related to levels of FPG and 2-h PG in youths. Results for the FPG were consistent with current ADA and WHO criteria; the results for the 2-h PG suggested a slightly lower diagnostic glucose level.

Based on these data, adult diagnostic guidelines for diabetes can be reasonably applied to youth. The similar slopes for FPG and 2-h PG in the logistic model suggest that youths are at increased risk of microvascular disease as glucose values increase, as in adults. However, our analysis suggested that the relationship between microvascular disease and glucose concentrations was different in youths than in adults. The ROC curve areas showed that microvascular disease was less well predicted by baseline glucose in youths than in adults. This difference between youths and adults is reflected in the low sensitivity in youths according to the 1997 ADA criteria and the difference in the percentage of complications defined by subjects with elevations in FPG and 2-h PG. The majority of the youths who developed microvascular disease still had glucose regulation in the impaired or normal ranges at baseline. Therefore, a large number of young at-risk subjects would not be identified by FPG or 2-h PG, unless “at-risk” was defined by a lower cutoff point or unless factors in addition to glucose were found to be of more or additive importance.

The differences between the two age groups in the predictive properties of glucose concentrations for subsequent microvascular disease may be largely caused by the differences in the frequency distributions of glucose at baseline (Figs. 1 and 2). The difference in the incidence of microvascular disease between the two groups at any given baseline glucose was modest. Therefore, in the younger subjects (youths), a greater percentage of those who developed microvascular disease may have come from the lower glucose categories simply because they made up a much larger percentage of the baseline population.

Another potential consequence of the differences in frequency distributions of glucose at baseline was that a greater proportion of younger individuals who developed microvascular disease developed proteinuria, a complication that may be less specific for hyperglycemia than retinopathy. In fact, when ROC curve analyses were conducted separately for each complication, retinopathy was as strongly related to glucose in youths as it was in adults, as demonstrated by the nearly equal ROC curve areas. The ROC curve area for nephropathy was lower in youths than in adults, suggesting that other factors specific for youth might be important in the development of nephropathy in this group. For instance, the hyperinsulinemia documented in young Pimas (21) might influence development of the kidney via interaction with the growth hormone axis (22, 23, 24).

In summary, despite the lower sensitivity for established diagnostic thresholds and the weaker association with microvascular disease, the incidence of microvascular disease increased at similar points in youths and adults in this population with a high prevalence of type 2 diabetes. In conclusion, it is reasonable to apply the adult diagnostic criteria for diabetes to children and adolescents.