Childhood Predictors of Young-Onset Type 2 Diabetes
Optimal prevention of young-onset type 2 diabetes requires identification of the early-life modifiable risk factors. We aimed to do this using longitudinal data in 1,604 5- to 19-year-old initially nondiabetic American Indians.
For type 2 diabetes prediction, we derived an optimally weighted, continuously distributed, standardized multivariate score (zMS) comprising commonly measured metabolic, anthropometric, and vascular traits (i.e., fasting and 2-h glucose, A1C, BMI, waist circumference, fasting insulin, HDL cholesterol, triglycerides, and blood pressures) and compared the predictive power for each feature against zMS.
In separate Cox proportional hazard models, adjusted for age, sex, and ethnicity, zMS and each of its component risk factors were associated with incident type 2 diabetes. Stepwise proportional hazards models selected fasting glucose, 2-h glucose, HDL cholesterol, and BMI as independent diabetes predictors; individually, these were weaker predictors than zMS (P < 0.01). However, a parsimonious summary score combining only these variables had predictive power similar to that of zMS (P = 0.33). Although intrauterine diabetes exposure or parental history of young-onset diabetes increased a child’s absolute risk of developing diabetes, the magnitude of the diabetes-risk relationships for zMS and the parsimonious score were similar irrespective of familial risk factors.
We have determined the relative value of the features of the metabolic syndrome in childhood for the prediction of subsequent type 2 diabetes. Our findings suggest that strategies targeting obesity, dysregulated glucose homeostasis, and low HDL cholesterol during childhood and adolescence may have the most success in preventing diabetes.
Excess global mortality attributable to type 2 diabetes in the year 2000 is estimated at one million deaths in developing nations and 1.9 million deaths in developed nations, or 2–8% of all deaths globally. Although conventionally type 2 diabetes has been considered a disease of adulthood, its occurrence during youth is increasingly common. Children exposed to diabetes in utero are at much greater risk of becoming obese and developing type 2 diabetes. Furthermore, individuals who develop type 2 diabetes in childhood and adolescence develop retinopathy and nephropathy at rates comparable with those of later-onset diabetes and by consequence have a substantial risk for microvascular complications at young ages. Thus, the prevention of young-onset type 2 diabetes presents a particularly important yet difficult challenge.
The optimal strategy for preventing any disease requires knowledge of its modifiable risk factors. In adult-onset diabetes, the risk factors include metabolic and anthropometric parameters such as hyperinsulinemia, impaired glucose regulation, obesity, and dyslipidemia, which when clustered together are often referred to as the metabolic syndrome. The strongest risk factors identified in adult Pima Indians are parental diabetes, BMI, insulin resistance, and impaired insulin secretion. Among Pima Indian children and adolescents free from diabetes, future type 2 diabetes is predicted by weight relative to height and concentrations of serum insulin and plasma glucose.
Few studies have explored the prospective relationships of risk factors related to the metabolic syndrome in childhood with subsequent type 2 diabetes, and little is known of the interdependence of these risk factors. Here, we explored the risk of young-onset type 2 diabetes associated with conventional anthropometric and metabolic traits in Pima Indian children and adolescents who had participated in a prospective cohort study. We also compared the ability of these risk factors, singly and in combination, to predict type 2 diabetes, and we examined the potential effect-modifying role of parental diabetes.
RESEARCH DESIGN AND METHODS
All residents of a geographic area of the Gila River Indian Community in Arizona ≥5 years old were invited to participate in a longitudinal study of diabetes. Participants selected for the present analysis consisted of 1,604 children and adolescents free from diabetes, the majority of whom are Pima or Tohono O’odham Indians, who had one or more research examination during childhood or adolescence (i.e., ≥5 and < 20 years old) and one or more follow-up examination at which diabetes was assessed. During the time frame of the present study, 2,828 nondiabetic children and adolescents were examined: 300 of these were excluded from the present analyses because data were missing for one of the predictor variables, and 924 were excluded because they did not have a follow-up examination. Those without follow-up tended to have been seen later in the study period than those with follow-up; they were also older and more likely to be male, but, accounting for these factors, there were no significant differences in metabolic variables. Participants ≥18 years old gave written informed consent, whereas for participants < 18 years old, written informed consent was given by a parent or guardian and written assent given by the participant. The Institutional Review Board of the National Institute of Diabetes and Digestive and Kidney Diseases approved the study.
In the present analyses, the extent to which variables measured at an initial baseline examination predict the incidence of future diabetes was assessed. For analysis of the whole group (age 5–19 years), the first examination at which an individual was nondiabetic and at which all relevant predictor variables had been measured was taken as the baseline examination. The period of follow-up extended from this baseline examination until the diagnosis of diabetes or until the last examination, whichever came first. In addition to analysis for the whole group, analyses were conducted in 5-year age bands (5–9, 10–14, and 15–19 years) based on age at the baseline examination. In these analyses, an individual’s first nondiabetic examination within the specified age range was taken as the baseline examination. Thus, individuals could appear in multiple age-groups if they had a nondiabetic examination when they were the relevant age and one or more subsequent examination. Parental diabetes was determined from examinations of both parents within the longitudinal study. For a child to be considered an offspring of a diabetic parent (ODP), one or more parent required a diagnosis of diabetes before age 30 years. For the children who were classified offspring of nondiabetic parents (ONDP), both parents required a nondiabetic examination at ≥30 years of age, irrespective of whether diabetes was diagnosed later in life. Individuals whose parents did not meet either set of criteria were considered of unknown status. Children were defined as having intrauterine exposure to diabetes (IED) if the mother had been diagnosed with diabetes before the child’s birth, regardless of the mother’s age at diagnosis. In eight individuals, IED occurred although the mother was >30 years old at onset of diabetes. Therefore, these children are classified as ONDP.
Measurements.
Participants attended the clinic after an overnight fast. They underwent a 75-g oral glucose tolerance test (OGTT) for assessment according to World Health Organization diagnostic criteria. The participant was classified with type 2 diabetes if the fasting plasma glucose concentration was >7.0 mmol/l, the 2-h plasma glucose concentration was >11.1mmol/l, or there was an existing clinical diagnosis (as assessed by medical chart review). Standard anthropometric data were collected by trained observers with participants in lightweight clothing and no shoes as previously described in detail. No measures of puberty were available. Metabolic variables were measured according to previously described methods
Statistical analysis.
Analyses were performed using SAS software (V8.02; SAS Institute, Cary, NC). Participant characteristics are presented as the arithmetic mean ± SD or as the median (25th–75th) percentiles if the variable was not of Gaussian distribution. If necessary to reduce skewness, data were approximately normalized via logarithmic transformation. Proportional hazards models were used to test the association between each childhood metabolic trait and development of young-onset diabetes (i.e., age of diagnosis < 30 years of age), with control for age at baseline examination, sex, and fraction of Pima heritage. The validity of the proportionality assumption was tested for each variable by including a time-dependent interaction term. To facilitate comparisons among variables, the hazard rate ratio (HRR) was calculated for a 1-SD difference for each of the variables. To compare each of the variables with an optimal prediction score (standardized multivariate score [zMS]), we fit a proportional hazards model that included all of the predictor variables (i.e., fasting and 2-h glucose, A1C, BMI, waist circumference, fasting insulin, HDL cholesterol, triglycerides, systolic blood pressure [SBP], and diastolic blood pressure [DBP]) in addition to age, sex, and heritage. The ability of zMS to predict diabetes was then compared with that for each of the individual variables by the likelihood ratio test that compares the full model with that containing only the individual variable, age, sex, and heritage. The HRR associated with zMS was computed for a standardized sum of the individual variables weighted by their regression coefficients in the zMS model. Parsimoniously predictive subsets of the variables were identified by a backward stepwise procedure. All potential predictor variables were included in the initial model, and the weakest predictor was sequentially eliminated until all remaining variables were retained at P
< 0.05; (age, sex, and heritage were not made available to the stepwise procedure and are, thus, forcibly included in all models). To assess the extent that this submodel captured the information contained in the fully optimized model containing all variables, these two models were compared by the likelihood ratio test. These procedures were undertaken in each age stratum and in the entire dataset containing all children (i.e., 5–19 years of age). For children with data for parental diabetes status, we calculated HRRs for each variable stratified by parental diabetes and tested for interactions between parental diabetes status and each variable.
REFERENCES
Roglic G, Unwin N, Bennett PH, Mathers C, Tuomilehto J, Nag S, Connolly V, King H: The burden of mortality attributable to diabetes: realistic estimates for the year 2000. Diabetes Care 28:2130–2135, 2005
American Diabetes Association: Type 2 diabetes in children and adolescents. Pediatrics 105:671–680, 2000
Pinhas-Hamiel O, Dolan LM, Daniels SR, Standiford D, Khoury PR, Zeitler P: Increased incidence of non-insulin-dependent diabetes mellitus among adolescents. J Pediatr 128:608–615, 1996
Neufeld ND, Raffel LJ, Landon C, Chen YD, Vadheim CM: Early presentation of type 2 diabetes in Mexican-American youth. Diabetes Care 21:80–86, 1998
Rosenbloom AL, Joe JR, Young RS, Winter WE: Emerging epidemic of type 2 diabetes in youth. Diabetes Care 22:345–354, 1999
Fagot-Campagna A, Saaddine JB, Flegal KM, Beckles GL: Diabetes, impaired fasting glucose, and elevated HbA1c in U.S. adolescents: the Third National Health and Nutrition Examination Survey. Diabetes Care 24:834–837, 2001
Dabelea D, Hanson RL, Bennett PH, Roumain J, Knowler WC, Pettitt DJ: Increasing prevalence of type II diabetes in American Indian children. Diabetologia 41:904–910, 1998
Pettitt DJ, Baird HR, Aleck KA, Bennett PH, Knowler WC: Excessive obesity in offspring of Pima Indian women with diabetes during pregnancy. N Engl J Med 308:242–245, 1983
Abbreviations: DBP, diastolic blood pressure; IED, intrauterine exposure to diabetes; ODP, offspring of a diabetic parent; OGTT, oral glucose tolerance test; ONDP, offspring of nondiabetic parents; ROC, receiver operating characteristic; ROC AUC, area under the ROC curve; SBP, systolic blood pressure; zMS, standardized multivariate score
Paul W. Franks, Robert L. Hanson, William C. Knowler, Carol Moffett, Gleebah Enos, Aniello M. Infante, Jonathan Krakoff, and Helen C. Looker
Diabetes Epidemiology and Clinical Research Section, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Phoenix, Arizona
Genetic Epidemiology and Clinical Research Group, Department of Public Health and Clinical Medicine, Division of Medicine, Umea* University Hospital, Umea*, Sweden
Address correspondence to Dr. Robert L. Hanson, Diabetes Epidemiology and Clinical Research Section, National Institute of Diabetes and Digestive and Kidney Diseases, 1550 E. Indian School Rd., Phoenix, AZ 85014. E-mail: .(JavaScript must be enabled to view this email address)
Published online August 24, 2007
Diabetes 56:2964-2972, 2007
DOI: 10.2337/db06-1639