Keith F. Rust

Prevalence of Diabetes and Impaired Fasting Glucose in Adults in the U.S. Population National Health and Nutrition Examination Survey 1999–2002

OBJECTIVE—The purpose of this study was to examine the prevalences of diagnosed and undiagnosed diabetes, and impaired fasting glucose (IFG) in U.S. adults during 1999–2002, and compare prevalences to those in 1988–1994. RESEARCH DESIGN AND METHODS—The National Health and Nutrition Examination Survey (NHANES) contains a probability sample of adults aged ≥20 years. In the NHANES 1999–2002, 4,761 adults were classified on glycemic status using standard criteria, based on an interview for diagnosed diabetes and fasting plasma glucose measured in a subsample. RESULTS—The crude prevalence of total diabetes in 1999–2002 was 9.3% (19.3 million, 2002 U.S. population), consisting of 6.5% diagnosed and 2.8% undiagnosed. An additional 26.0% had IFG, totaling 35.3% (73.3 million) with either diabetes or IFG. The prevalence of total diabetes rose with age, reaching 21.6% for those aged ≥65 years. The prevalence of diagnosed diabetes was twice as high in non-Hispanic blacks and Mexican Americans compared with non-Hispanic whites (both P < 0.00001), whereas the prevalence of undiagnosed diabetes was similar by race/ethnicity, adjusted for age and sex. The prevalence of diagnosed diabetes was similar by sex, but prevalences of undiagnosed diabetes and IFG were significantly higher in men. The crude prevalence of diagnosed diabetes rose significantly from 5.1% in 1988–1994 to 6.5% in 1999–2002, but the crude prevalences were stable for undiagnosed diabetes (from 2.7 to 2.8%) and IFG (from 24.7 to 26.0%). Results were similar after adjustment for age and sex. CONCLUSIONS—Although the prevalence of diagnosed diabetes has increased significantly over the last decade, the prevalences of undiagnosed diabetes and IFG have remained relatively stable. Minority groups remain disproportionately affected.

Full Accounting of Diabetes and Pre-Diabetes in the U.S. Population in 1988–1994 and 2005–2006

OBJECTIVE—We examined the prevalences of diagnosed diabetes, and undiagnosed diabetes and pre-diabetes using fasting and 2-h oral glucose tolerance test values, in the U.S. during 2005–2006. We then compared the prevalences of these conditions with those in 1988–1994. RESEARCH DESIGN AND METHODS—In 2005–2006, the National Health and Nutrition Examination Survey included a probability sample of 7,267 people aged ≥12 years. Participants were classified according to glycemic status by interview for diagnosed diabetes and by fasting and 2-h glucoses measured in subsamples. RESULTS—In 2005–2006, the crude prevalence of total diabetes in people aged ≥20 years was 12.9%, of which ∼40% was undiagnosed. In people aged ≥20 years, the crude prevalence of impaired fasting glucose was 25.7% and of impaired glucose tolerance was 13.8%, with almost 30% having either. Over 40% of individuals had diabetes or pre-diabetes. Almost one-third of the elderly had diabetes, and three-quarters had diabetes or pre-diabetes. Compared with non-Hispanic whites, age- and sex-standardized prevalence of diagnosed diabetes was approximately twice as high in non-Hispanic blacks (P < 0.0001) and Mexican Americans (P = 0.0001), whereas undiagnosed diabetes was not higher. Crude prevalence of diagnosed diabetes in people aged ≥20 years rose from 5.1% in 1988–1994 to 7.7% in 2005–2006 (P = 0.0001); this was significant after accounting for differences in age and sex, particularly in non-Hispanic blacks. Prevalences of undiagnosed diabetes and pre-diabetes were generally stable, although the proportion of total diabetes that was undiagnosed decreased in Mexican Americans. CONCLUSIONS—Over 40% of people aged ≥20 years have hyperglycemic conditions, and prevalence is higher in minorities. Diagnosed diabetes has increased over time, but other conditions have been relatively stable.

Prevalence of Diabetes and High Risk for Diabetes Using A1C Criteria in the U.S. Population in 1988–2006

OBJECTIVE We examined prevalences of previously diagnosed diabetes and undiagnosed diabetes and high risk for diabetes using recently suggested A1C criteria in the U.S. during 2003–2006. We compared these prevalences to those in earlier surveys and those using glucose criteria. RESEARCH DESIGN AND METHODS In 2003–2006, the National Health and Nutrition Examination Survey included a probability sample of 14,611 individuals aged ≥12 years. Participants were classified on glycemic status by interview for diagnosed diabetes and by A1C, fasting, and 2-h glucose challenge values measured in subsamples. RESULTS Using A1C criteria, the crude prevalence of total diabetes in adults aged ≥20 years was 9.6% (20.4 million), of which 19.0% was undiagnosed (7.8% diagnosed, 1.8% undiagnosed using A1C ≥6.5%). Another 3.5% of adults (7.4 million) were at high risk for diabetes (A1C 6.0 to <6.5%). Prevalences were disproportionately high in the elderly. Age-/sex-standardized prevalence was more than two times higher in non-Hispanic blacks and Mexican Americans versus non-Hispanic whites for diagnosed, undiagnosed, and total diabetes (P < 0.003); standardized prevalence at high risk for diabetes was more than two times higher in non-Hispanic blacks versus non-Hispanic whites and Mexican Americans (P < 0.00001). Since 1988–1994, diagnosed diabetes generally increased, while the percent of diabetes that was undiagnosed and the percent at high risk of diabetes generally decreased. Using A1C criteria, prevalences of undiagnosed diabetes and high risk of diabetes were one-third that and one-tenth that, respectively, using glucose criteria. CONCLUSIONS Although A1C detects much lower prevalences than glucose criteria, hyperglycemic conditions remain high in the U.S., and elderly and minority groups are disproportionately affected.

The Prevalence of Meeting A1C, Blood Pressure, and LDL Goals Among People With Diabetes, 1988–2010

OBJECTIVE To determine the prevalence of people with diabetes who meet hemoglobin A1c (A1C), blood pressure (BP), and LDL cholesterol (ABC) recommendations and their current statin use, factors associated with goal achievement, and changes in the proportion achieving goals between 1988 and 2010. RESEARCH DESIGN AND METHODS Data were cross-sectional from the National Health and Nutrition Examination Surveys (NHANES) from 1988–1994, 1999–2002, 2003–2006, and 2007–2010. Participants were 4,926 adults aged ≥20 years who self-reported a previous diagnosis of diabetes and completed the household interview and physical examination (n = 1,558 for valid LDL levels). Main outcome measures were A1C, BP, and LDL cholesterol, in accordance with the American Diabetes Association recommendations, and current use of statins. RESULTS In 2007–2010, 52.5% of people with diabetes achieved A1C <7.0% (<53 mmol/mol), 51.1% achieved BP <130/80 mmHg, 56.2% achieved LDL <100 mg/dL, and 18.8% achieved all three ABCs. These levels of control were significant improvements from 1988 to 1994 (all P < 0.05). Statin use significantly increased between 1988–1994 (4.2%) and 2007–2010 (51.4%, P < 0.01). Compared with non-Hispanic whites, Mexican Americans were less likely to meet A1C and LDL goals (P < 0.03), and non-Hispanic blacks were less likely to meet BP and LDL goals (P < 0.02). Compared with non-Hispanic blacks, Mexican Americans were less likely to meet A1C goals (P < 0.01). Younger individuals were less likely to meet A1C and LDL goals. CONCLUSIONS Despite significant improvement during the past decade, achieving the ABC goals remains suboptimal among adults with diabetes, particularly in some minority groups. Substantial opportunity exists to further improve diabetes control and, thus, to reduce diabetes-related morbidity and mortality.

Population Inferences and Variance Estimation for NAEP Data

In the National Assessment of Educational Progress (NAEP), population inferences and variance estimation are based on a randomization-based perspective where the link between the observed data and the population quantities of interest is given by the distribution of potential values of estimates over repeated samples from the same population using the identical sample design. Because NAEP uses a complex sample design, many of the assumptions underlying traditional statistical analyses are violated, and, consequently, analysis procedures must be adjusted to appropriately handle the structure of the sample. In this article, we discuss the use of sampling weights in deriving population estimates and consider the effect of nonresponse and undercoverage on those estimates. We also discuss the estimation of sampling variability from complex sample surveys, concentrating on the jackknife repeated replication procedure-the variance estimation procedure used by NAEP-and address the use of a simple approximation to sampling variability. Finally, we discuss measures of the stability of variance estimates.

Associations Between Trends in Race/Ethnicity, Aging, and Body Mass Index With Diabetes Prevalence in the United States: A Series of Cross-sectional Studies

Context The prevalence of diabetes has increased in the United States during the past several decades. The extent to which demographic changes and rising obesity rates have contributed to this increase is unknown. Contribution Data collected from nationally representative samples of men and women between 1976 and 2010 showed a greater increase in diabetes prevalence in men than in women. After adjustment for race/ethnicity, age, and body mass index, diabetes prevalence still increased in men but not women. Caution This study could not adjust for physical activity. Implication Body mass index is a substantial contributor to the observed increase in diabetes prevalence. The Editors The increasing prevalence of diabetes over the past few decades (13) has made it one of the most common and costly chronic disorders in the United States. The nationwide prevalence of diagnosed diabetes more than doubled between 1976 to 1980 and 1999 to 2004 (3), and the prevalence of diagnosed and undiagnosed diabetes combined increased by 33% between 1988 to 1994 and 2005 to 2010 (1). The increase in diabetes prevalence has coincided with an increase in certain risk factors for diabetes. The most important modifiable risk factors for diabetes are obesity and overweight. The prevalence of adult obesity, defined as a body mass index (BMI) of 30 kg/m2 or more, in national surveys was 22.3% from 1988 to 1994, 30.5% from 2000 to 2002, and 35.9% from 2009 to 2010, whereas the prevalence of overweight, defined as a BMI of 25.0 to 29.9 kg/m2, was 32.6%, 34.0%, and 33.3% during these respective periods (46). Racial and ethnic groups at increased risk for diabetes (3) make up a growing proportion of the population. Non-Hispanic black and Hispanic persons represented 11.7% and 6.4% of the U.S. population, respectively, in 1980, and their numbers grew to 12.6% and 16.3%, respectively, in 2010 (7, 8). Risk for diabetes increases with age (9). Furthermore, the U.S. population is aging; in recent years, the median age has increased from 30.0 years in 1980 to 37.2 years in 2010 (10). How much of the increased prevalence of diabetes is explained by an increase in these known risk factors and how much is due to other factors is unclear. We sought to determine the extent to which this increase is explained by changing distributions of race/ethnicity, age, and obesity prevalence in U.S. adults. To do so, we analyzed data from several National Health and Nutrition Examination Surveys (NHANES). Each survey was designed to be representative of the U.S. population. The surveys have been conducted in waves from 1976 to 1980 (NHANES II) and 1988 to 1994 (NHANES III), as well as the continuous NHANES from 1999 to 2010. We limited the study sample to adults aged 20 to 74 years for consistency across surveys. Methods Data Collection The NHANES is a series of stratified, multistage probability surveys designed to be representative of the U.S. noninstitutionalized civilian population (1113). Data were collected in an in-home interview and a subsequent visit to a mobile examination center. The response rate was 91% for the interview and 73% for the examination in NHANES II; respective response rates were 86% and 78% in NHANES III, 82% and 76% in 1999 to 2000, 84% and 80% in 2001 to 2002, 79% and 76% in 2003 to 2004, 80% and 77% in 2005 to 2006, 78% and 75% in 2007 to 2008, and 79% and 77% in 2009 to 2010. Each NHANES survey consisted of participants who were randomly selected to participate in a morning examination for which they were asked to fast or an afternoon or evening examination. We used data from the morning sessions, which were capable of independently producing national estimates. We excluded pregnant women from the analysis because pregnancy affects glucose and BMI measurements. Respective years and sample sizes from the NHANES series were 1976 to 1980 with 4343 participants, 1988 to 1994 with 7023 participants, 1999 to 2002 with 3848 participants, 2003 to 2006 with 3688 participants, and 2007 to 2010 with 5030 participants. For NHANES III and NHANES 19992010, all participants gave written informed consent and the research ethics boards of the National Center for Health Statistics approved all protocols. The National Center for Health Statistics did an internal human subjects review in NHANES II, which did not consist of institutional review board approval using current standards. A standardized questionnaire was used to collect demographic information, including age, race/ethnicity, and sex during an in-home interview, except race/ethnicity may have been determined on the basis of interviewer observation in NHANES II. Participants were also asked if they had ever been diagnosed with diabetes by a doctor (NHANES II and III) or a doctor or other health professional (NHANES 19992010). During the visit to the mobile examination center, height and weight were measured and BMI was calculated. A trained phlebotomist obtained a blood sample according to a standardized protocol, and fasting glucose was measured in plasma by a hexokinase method. Fasting glucose was measured by using an Abbott ABA-100 analyzer in NHANES II, and the interassay coefficient of variation was not reported. Respective equipment and coefficients of variation in the NHANES series were the Roche Cobas Mira chemistry system with 1.6% to 3.7% in NHANES III, Roche Cobas Mira chemistry system with 1.3% to 3.0% in NHANES 19992002, Roche Cobas Mira chemistry system or Roche/Hitachi 911 glucose analyzer with 1.3% to 2.3% in NHANES 20032006, and Roche Modular P chemistry analyzer with 0.8% to 2.6% in NHANES 20072010. Although NHANES III and NHANES 19992004 used the same equipment and methods to measure fasting glucose, NHANES 20052010 used different equipment and was therefore calibrated to the earlier data; calibration of NHANES II data was not possible (14, 15). Diabetes was defined as a self-reported previous diagnosis of diabetes or a fasting plasma glucose level of 7.0 mmol/L (126 mg/dL) or more. Although fasting plasma glucose level was consistently measured in all of the NHANES surveys, 2-hour plasma glucose and hemoglobin A1c levels were not. Statistical Analysis We calculated means or percentages of participant characteristics by survey year and diabetes status. We then used multivariable logistic regression models with diabetes as the outcome and terms for survey year, age, race/ethnicity, and BMI to find the adjusted prevalence of diabetes by survey period, for which we computed predicted margins (16). We fit an unadjusted logistic regression model with only a term for time (midpoint of survey yearthe results are denoted as unadjusted estimates); models including covariates of age, race/ethnicity, or BMI; and a model with all 3 variables. We compared these models to determine the extent to which covariates explained the trend in diabetes over time. All tests for trend were computed by including the midpoint year for each survey period as a continuous variable in regression models. We initially included sex and a sex-by-time interaction term in all models and found the interaction to be significant. Therefore, we conducted all analyses separately by sex. Appropriate sample weights were used to account for unequal probabilities of selection and nonresponse and thus provided estimates representative of the U.S. noninstitutionalized civilian population. Data were analyzed using SUDAAN (RTI International) and accounted for the NHANES-stratified, clustered sample design. Role of the Funding Source The Centers for Disease Control and Prevention (CDC) and the National Institutes of Diabetes and Digestive and Kidney Diseases funded NHANES and oversaw its conduct and reporting with regard to diabetes-related data. As employees or contractors of the Centers for Disease Control and Prevention or National Institutes of Diabetes and Digestive and Kidney Diseases, the authors had a direct role in the design and interpretation of the secondary analysis and the decision to submit the manuscript for publication. Results Participant characteristics by diabetes status are presented for men (Table 1) and women (Table 2). The no diabetes group includes persons with both normal fasting glucose levels (<5.6 mmol/L [<100 mg/dL]) and impaired fasting glucose levels (5.6 to 6.9 mmol/L [100 to 125 mg/dL]). Our study population was limited to persons aged 20 to 74 years; within this age range, a significant increase in mean age over time was seen only in women without diabetes. The percentage of non-Hispanic white men and women without diabetes significantly decreased over time, but the decrease was not significant in those with diabetes. Both the percentage of Mexican Americans and mean BMI significantly increased over time in all subgroups. Mean fasting plasma glucose levels significantly increased over time in men and women without diabetes but not in those with diabetes. Mean fasting plasma glucose levels were generally higher in men than in women among people with and without diabetes. Table 1. Baseline Characteristics of Men, by Diabetes Status and Survey Period, 19762010 Table 2. Baseline Characteristics of Women, by Diabetes Status and Survey Period, 19762010 The crude prevalence of diabetes increased between 1976 to 1980 and 2007 to 2010 for both men and women (both P for trends < 0.001) (Figure 1) (Supplement 1). After adjustment for age, race/ethnicity, and BMI, the prevalence of diabetes in men increased from 6.2% in 1976 to 1980 to 9.6% in 2007 to 2010 (P for trend < 0.001) (Figure 2). After identical adjustment, the prevalence of diabetes in women was 7.6% in 1976 to 1980 and 7.5% in 2007 to 2010 (P for trend = 0.69) (Figure 2). Figure 1. Crude and unadjusted prevalence of diabetes in men and women, 19762010. Points refer to crude prevalence, and lines refer to unadjusted prevalence (logistic regressionbased). Figure 2. Crude and adjusted prevalence of diabetes in men and women, 19762010. Adjustmen

Differences in survival between black and white patients with diabetic end-stage renal disease.

OBJECTIVE To evaluate whether the longer survival of blacks with diabetic end-stage renal disease (ESRD) relative to whites is due to racial differences in type of diabetes, comorbidity at ESRD onset, and ESRD treatment modality and to examine whether survival differences between blacks and whites occur only in certain population subgroups. RESEARCH DESIGN AND METHODS The Michigan Kidney Registry was used to ascertain all blacks and whites (n = 594) with diabetic ESRD in southeastern Michigan, with ESRD onset at age <65 years during 1974–1983. Patients were followed through 1988. Medical records were abstracted for type of diabetes, comorbidity at ESRD onset, and other factors. RESULTS Median survival among insulin-dependent diabetes mellitus patients was 27 months in blacks and 17 months in whites, and among non-insulin-dependent diabetes mellitus patients was 30 months in blacks and 16 months in whites. After adjustment for confounding factors by Cox proportional hazards analysis, the death rate was 45% lower in blacks than in whites on dialysis (relative death rate [RDR] = 0.55,95% confidence interval [CI] = 0.44–0.69), but was similar in blacks and whites with a renal transplant (RDR = 0.99, 95% CI = 0.64–1.52). Compared with dialysis, transplantation was associated with lower mortality in both races (whites, RDR = 0.50, 95% CI = 0.36–0.70; blacks, RDR = 0.89, 95% CI = 0.60–1.34), although the effect was not statistically significant in blacks. Racial differences in survival did not vary by type of diabetes or any additional factor. CONCLUSIONS Survival after ESRD onset is longer in blacks than in whites treated with dialysis, even after adjusting for comorbidity and other factors that affect survival. Survival does not differ by race among transplant patients.

Sampling and Weighting in the National Assessment.

This chapter describes procedures for obtaining the National Assessment of Educational Progress (NAEP) student samples used in the national and state assessments and for deriving survey weights for use in the analysis of the survey data. Following the description of general procedures, more detailed discussion is included about several issues that relate to the procedures used. In some cases, these involve procedures that NAEP is actively reviewing and investigating, with a view toward implementing improvements in the future. In other cases, the procedures, although well established in NAEP, involve technical aspects with interesting features not fully described in the available technical reports. Probability sampling techniques are used to select the student samples for the national assessment and for the Trial State Assessment program. The sample designs for these two types of samples depend on each other only in the implementation of procedures to minimize the overlap of the samples of participating schools. For both the national assessment and the Trial State Assessment, the goal of the sample design is to obtain samples of students from which estimates of subpopulation characteristics can be obtained with reasonably high precision while satisfying economic and operational constraints. We will describe the procedures used to obtain the student samples used in the national and state assessments and to derive survey weights for use in the analysis of the survey data. Additionally, we will discuss several

Mitigating case mix factors by choice of glycemic control performance measure threshold.

OBJECTIVE—Performance measures are tools for assessing quality of care but may be influenced by patient factors. We investigated how currently endorsed performance measures for glycemic control in diabetes may be influenced by case mix composition. We assessed differences in A1C performance measure threshold attainment by case mix factors for A1C >9% and examined how lowering the threshold to A1C >8% or >7% changed these differences. RESEARCH DESIGN AND METHODS—Using data from the 1999–2002 National Health and Nutrition Examination Survey for 843 adults self-reporting diabetes, we computed the mean difference in A1C threshold attainment of >9, >8, and >7% by various case mix factors. The mean difference is the average percentage point difference in threshold attainment for population groups compared with that for the overall population. RESULTS—Diabetes medication was the only factor for which the difference in threshold attainment increased at lower thresholds, with mean differences of 5.7 percentage points at A1C >9% (reference), 10.1 percentage points at A1C >8% (P < 0.05), and 14.1 percentage points at A1C >7% (P < 0.001). CONCLUSIONS—As 87% of U.S. adults have A1C <9%, a performance measure threshold of >9% will not drive major improvements in glycemic control. Lower thresholds do not exacerbate differences in threshold attainment for most factors. Reporting by diabetes medication use may compensate for heterogeneous case mix when a performance measure threshold of A1C >8% or lower is used.

Chapter 5: Population Inferences and Variance Estimation for NAEP Data:

In the National Assessment of Educational Progress (NAEP), population inferences and variance estimation are based on a randomization-based perspective where the link between the observed data and the population quantities of interest is given by the distribution of potential values of estimates over repeated samples from the same population using the identical sample design. Because NAEP uses a complex sample design, many of the assumptions underlying traditional statistical analyses are violated, and, consequently, analysis procedures must be adjusted to appropriately handle the structure of the sample. In this article, we discuss the use of sampling weights in deriving population estimates and consider the effect of nonresponse and undercoverage on those estimates. We also discuss the estimation of sampling variability from complex sample surveys, concentrating on the jackknife repeated replication procedure—the variance estimation procedure used by NAEP—and address the use of a simple approximation to sampling variability. Finally, we discuss measures of the stability of variance estimates.