Random blood glucose: a robust risk factor for type 2 diabetes.

ORIGINAL

ARTICLE

Random Blood Glucose: A Robust Risk Factor For Type 2 Diabetes Michael E. Bowen, MD, MPH,1,2, Lei Xuan, PhD2, Ildiko Lingvay, MD, MPH, MSCS,3,4, Ethan A. Halm, MD, MPH1,2 1

Division of General Internal Medicine, Department of Medicine, University of Texas Southwestern Medical Center. Dallas, TX. 2Division of Outcomes and Health Services Research, Department of Clinical Sciences, University of Texas Southwestern Medical Center. Dallas, TX. 3Division of Endocrinology, Department of Medicine, University of Texas Southwestern Medical Center. Dallas, TX. 4Department of Clinical Sciences, University of Texas Southwestern Medical Center. Dallas, TX.

Context: Although random blood glucose (RBG) values are common in clinical practice, the role of elevated RBG values as a risk factor for type 2 diabetes is not well described. Objective: To examine non-diagnostic, RBG values as a risk factor for type 2 diabetes Design: Cross-sectional study of NHANES participants (2005–2010) Participants: 13,792 non-fasting NHANES participants without diagnosed diabetes. Primary Outcome: Glycemic status (normal glycemia, undiagnosed prediabetes, or undiagnosed diabetes) using hemoglobin HbA1C as the criterion standard. Analysis: Multinomial logistic regression examined associations between diabetes risk factors and RBG values according to glycemic status. Associations between current US screening strategies and a hypothetical RBG screening strategy with undiagnosed diabetes were examined. Results: In unadjusted analyses, a single RBGⱖ100 mg/dL (5.6 mmol/L) was more strongly associated with undiagnosed diabetes than any single risk factor [Odds Ratio (95% CI) 31.2 (21.3 – 45.5)] and remained strongly associated with undiagnosed diabetes [20.4 (14.0 – 29.6)] after adjustment for traditional diabetes risk factors. Using RBG⬍100 mg/dL as a reference, the adjusted odds of undiagnosed diabetes increased significantly as RBG increased: RBG 100 –119 mg/dL [7.1 (4.4 –11.4)], RBG 120 –139 mg/dL [30.3 (20.0 – 46.0)], RBGⱖ140 mg/dL [256 (150.0 – 436.9)]. As a hypothetical screening strategy, an elevated RBG was more strongly associated with undiagnosed diabetes than current USPSTF guidelines (hypertension alone; p⬍0.0001) and similar to ADA guidelines (p⫽0.12). Conclusions: A single RBG ⱖ 100 mg/dL is more strongly associated with undiagnosed diabetes than traditional risk factors. Abnormal RBG values are a risk factor for diabetes and should be considered in screening guidelines.

T

ype 2 diabetes is a significant and costly public health epidemic that is projected to affect 1 in 3 US adults by 2050 (1). This projected increase is driven largely by the 86 million Americans with prediabetes, which if unrecognized and untreated will likely progress to frank diabetes (2, 3). In spite of well-established diabetes screening guidelines (4, 5), over 8 million people with diabetes and 80

million people with prediabetes remain undiagnosed or unaware of their condition (2, 6). Current US diabetes screening guidelines recommend targeted screening of high risk individuals; however, guidelines define diabetes risk differently. The US Preventative Services Task Force (USPSTF) guideline recommends screening only when an individual has a sustained

ISSN Print 0021-972X ISSN Online 1945-7197 Printed in U.S.A. Copyright © 2015 by the Endocrine Society Received November 14, 2014. Accepted February 2, 2015.

Abbreviations:

doi: 10.1210/jc.2014-4116

J Clin Endocrinol Metab

jcem.endojournals.org

The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 22 March 2015. at 14:41 For personal use only. No other uses without permission. . All rights reserved.

1

2

Random blood glucose as a diabetes risk factor

elevation in blood pressure (BP) (treated or untreated) ⬎135/80 (5). In contrast, the multifactorial American Diabetes Association (ADA) guideline (4) recommends screening for all individuals age 45 and older and individuals at any age if their BMI is ⱖ 25 kg/m2 and they have one additional risk factor including: nonwhite race, family history of diabetes, hypertension, dyslipidemia, history of cardiovascular disease, physical inactivity, polycystic ovarian syndrome, history of gestational diabetes, delivery of an infant weighing ⬎ 9 pounds, or other clinical conditions associated with insulin resistance. Screening is also recommended for individuals with prediabetes when HbA1C values, fasting glucose values, and 2-hour glucose values obtained via an oral glucose tolerance test (OGTT) are abnormal but fail to reach diagnostic thresholds (4). The International Diabetes Federation guidelines extend these recommendations even further to recommend screening for individuals with random, nondiagnostic glucose values between 100 and 199 mg/dL (5.6 – 11.0 mmol/ L), although the evidence supporting this recommendation is limited (7). Random blood glucose values are not included as a risk factor in US screening recommendations. In order to improve screening strategies and identification of individuals with undiagnosed diabetes and prediabetes, greater understanding of the risk factors for undiagnosed diabetes and prediabetes are needed. This study seeks to: 1) describe the prevalence diabetes risk factors in individuals with undiagnosed diabetes and prediabetes; 2) examine the association between random glucose values and undiagnosed diabetes and prediabetes; 3) explore the associations between ADA screening guidelines, USPSTF screening guidelines, and a hypothetical random glucose screening strategy with undiagnosed disease. Research Design and Methods We analyzed merged data from the 2005–2010 National Health and Nutrition Examination Surveys (NHANES). NHANES is a repeated, cross-sectional, stratified survey designed to be representative of the noninstitutionalized US population using a multistage probability sample. Participants complete an in-home interview for basic demographic and health information along with a scheduled visit to a mobile examination center for physical examination and laboratory testing (8). All participants gave written informed consent and the research ethics boards of the National Center for Health Statistics approved all protocols.

Study Population The study population consisted of nonpregnant adults age 18 and older who completed both the NHANES interview and MEC examination components between 2005 and 2010. We excluded participants with diagnosed diabetes or prediabetes. Participants responding “yes” when asked “Other than during pregnancy, have you ever been told by a doctor or health pro-


fessional that you have diabetes or sugar diabetes?” were considered to have diabetes and excluded from the analytic sample. Participants responding “yes” to the question: “Have you ever been told by a doctor or other health professional that you have any of the following: prediabetes, impaired fasting glucose, impaired glucose tolerance, borderline diabetes, or that your blood glucose is higher than normal but not high enough to be called diabetes or sugar diabetes?” were also excluded. Participants examined in a nonfasting state who had both hemoglobin HbA1C and random serum blood glucose (RBG) test results were included in analysis.

Measures Patient characteristics and diabetes risk factors including age, race, family history of diabetes, and family history of cardiovascular disease were obtained from questionnaire data. BMI was calculated from measured height and weight and classified as normal weight (BMI ⬍ 25 kg/m2); overweight (BMI 25–30 kg/ m2); or obese (BMI ⱖ 30 kg/m2). Participants self-reporting hypertension or high cholesterol diagnosed by a physician or other health professional were considered to have hypertension and high cholesterol. Polycystic ovarian syndrome, acanthosis nigricans, and other clinical conditions associated with insulin resistance were not included as a risk factor for diabetes because we were unable to identify these participants within NHANES. Gestational diabetes and delivery of an infant weighing ⬎ 9 pounds were not included as risk factors because data were not available in 2005–2006 NHANES. Participants meeting age, BMI, family history, race, hypertension, high cholesterol, or cardiovascular disease criteria were considered to satisfy the ADA screening guideline. We subsequently classified patients using the ADA (4) and USPSTF (5) screening guidelines by whether screening was indicated (yes/no). All questionnaire and examination components utilized in analysis were identical across the 3 survey cycles used. Serum RBG measurements were determined using the Beckman Oxygen electrode, glucose oxidase method. Between 2007 and 2012, one instrument change occurred [Beckman Synchron LX20 (2007) to the Beckman Unicel CxC800 Synchron (2008 – 2012)]. All measurements were performed by Collaborative Laboratory Services, LLC (9). HbA1C assays were conducted using high-performance liquid chromatography (HPLC) methods on instruments certified by the National Glycohemoglobin Standardization Program (9). All HbA1C results were subsequently standardized to the Diabetes Control and Complications Trial reference standard (10). Between 2007 and 2010, all HbA1C assays were performed at a single laboratory site (Minneapolis, MN). The instrument used to analyze HbA1C changed from the Tosoh HbA1C 2.2 Plus (2005–2007) to the Tosoh HbA1C G7 (2007–2010). In spite of this, HbA1C values were analyzed without corrections as recommended by the National Center for Health Statistics (11). We defined diabetes, prediabetes and dysglycemia using HbA1C as the criterion standard. Diabetes was defined as having an HbA1C ⱖ 6.5% (48 mmol/mol) and prediabetes as having an HbA1C 5.7– 6.4% (39 – 46 mmol/mol). Dysglycemia was defined as having an HbA1C ⱖ 5.7% (39 mmol/mol). Fasting glucose was not included in our diabetes definition because individuals examined in a fasting state did not have a RBG value available.


doi: 10.1210/jc.2014-4116


Statistical Analysis We calculated means and percentages of participant characteristics by glycemic status and compared values across groups using Rao-Scott ␹2 tests for categorical variables and F tests for continuous variables. We used multinomial logistic regression to examine associations between risk factors and undiagnosed diabetes and prediabetes. We further aggregated risk factors according to ADA (4) and USPSTF (5) diabetes screening guidelines to examine associations between these screening strategies and undiagnosed diabetes and prediabetes. We also examined unadjusted and covariate adjusted associations between a single RBG ⱖ 100 mg/dL (5.6 mmol/L) with undiagnosed diabetes and prediabetes. Models were adjusted for traditional diabetes risk factors including age, sex, race, BMI, hypertension, hyperlipidemia, cardiovascular disease, and family history of diabetes. The odds ratios of detecting undiagnosed diabetes and prediabetes using ADA, USPSTF, and RBG screening strategies were compared using Wald tests. Analyses were conducted in SAS version 9.3 (SAS Institute Inc., Cary, NC) and Stata/SE version 13.1 (StataCorp LP, College Station, TX) using survey procedures to account for sampling weights. Appropriate sampling weights were applied to account for unequal selection probabilities and nonresponse such that estimates are representative of the US noninstitutionalized civilian population (8). P values ⬍ 0.05 were considered statistically significant. This study was approved by the institutional review board (IRB) at the University of Texas Southwestern Medical Center at Dallas.

Results A total of 13 792 participants met eligibility criteria. Among all participants aged ⱖ 18 years without diagnosed diabetes, 1.9% had undiagnosed diabetes and 20.2% had undiagnosed prediabetes by HbA1C criteria. Participant characteristics by glycemic status are shown in

Table 1.

Table 1. Overall, the prevalence of traditional diabetes risk factors is increased in individuals with undiagnosed prediabetes and diabetes. The prevalence of older age, nonwhite race, positive family history of diabetes, increasing BMI, and a diagnosis of hypertension, hyperlipidemia, and cardiovascular disease increases across the spectrum of dysglycemia from normal to prediabetes, to diabetes, with the highest prevalence of risk factors present in individuals with undiagnosed diabetes (Table 1). Men were more likely than women to have undiagnosed diabetes. While those with undiagnosed diabetes and prediabetes were more likely to be screened in the past 3 years compared to those with normoglycemia, only slightly more than half of those with undiagnosed diabetes (56.1%) and half (49.6%) of those with undiagnosed prediabetes reported having been screened for diabetes in the past 3 years. However, this does not appear to be for lack of physician visits because only 21.0% and 13.8% of patients with undiagnosed diabetes and prediabetes respectively did not have a doctor’s visit in the past 12 months. The mean HbA1C increased across the glycemic spectrum: no diabetes HbA1C⫽5.2% (33 mmol/mol); undiagnosed prediabetes HbA1C⫽5.9% (41 mmol/mol); and undiagnosed diabetes HbA1C⫽7.6% (60 mmol/mol) (Table 1). Translating these HbA1C values into estimated average glucose values over the preceding 3 months (12), participants with no diabetes had an estimated average glucose of 103 mg/dL (5.7 mmol/L) compared with 123 mg/dL (6.8 mmol/L) and 171 mg/dL (9.5 mmol/L) in participants with undiagnosed prediabetes and diabetes respectively. The mean value of a single RBG measure across groups

Participant Demographics and Risk Factors by Glycemic Status Characteristic

Mean age, years Female Race Non-Hispanic white Non-Hispanic black Mexican American Other Hispanic Other race Education⬍12 yr Married Uninsured Family history diabetes Mean BMI, kg/m2 Hypertension Hyperlipidemia Cardiovascular disease No doctors visit past 12 months Screened for diabetes past 3 yr Mean random blood glucose, mg/dL Mean HbA1C, % Mean HbA1C, mmol/mol

3

No Diabetes

Undiagnosed Pre-diabetes

Undiagnosed Diabetes

p- value

41.7 (0.3) 50.8 (0.5)

55.4 (0.4) 51.7 (1.1)

58.6 (0.9) 41.3 (3.0)

⬍0.001 0.009

72.9 (1.8) 8.8 (0.8) 8.3 (0.9) 4.3 (0.6) 5.7 (0.5) 15.3 (0.8) 53.7 (1.0) 21.4 (0.9) 30.9 (0.6) 27.3 (0.1) 20.6 (0.7) 22.1 (0.7) 3.8 (0.3) 19.1 (0.5) 35.9 (0.5) 89.9 (0.2) 5.2 (0.01) 33 (0.1)

65.0 (0.9) 15.6 (0.6) 8.4 (0.4) 4.5 (0.3) 6.5 (0.6) 24.2 (0.9) 58.1 (1.1) 18.5 (0.8) 40.0 (1.1) 30.2 (0.1) 40.5 (1.1) 39.0 (1.1) 9.6 (0.6) 13.8 (0.7) 49.6 (1.1) 99.1 (0.4) 5.9 (0.01) 41 (0.1)

57.0 (5.2) 17.0 (2.4) 14.0 (2.5) 4.8 (1.0) 7.3 (2.7) 29.6 (2.7) 56.0 (3.2) 22.8 (3.3) 46.1 (2.7) 34.2 (0.5) 49.6 (2.8) 37.1 (3.2) 14.8 (1.6) 21.0 (2.8) 56.1 (2.8) 156.0 (5.3) 7.6 (0.1) 60 (1.1)

⬍0.001 ⬍0.001 0.005 0.83 0.48 ⬍0.001 ⬍0.001 0.012 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001

Data presented as percent (SE.) unless otherwise noted. All estimates are weighted to US population estimates. Random glucose (mg/dL) ⫻ 0.5551 ⫽ mmol/liter


4


increased from 89.9 mg/dL (5.0 mmol/L) in patients with no diabetes, to 99.1 mg/dL (5.5 mmol/L) in patients with undiagnosed prediabetes, to 156.0 mg/dL (8.7 mmol/L) in patients with undiagnosed diabetes. In univariate analyses, the risk factors most strongly associated [Odds Ratio (95% CI)] with undiagnosed diabetes were a single RBG ⱖ 100 mg/dL (5.6 mmol/L) [31.2 (21.3 – 45.5)], BMI ⱖ 25 kg/m2 [10.9 (6.8 – 17.5)], and age ⱖ 45 years [7.9 (5.7 – 10.9)]. Comorbid disease risk factors such as a history of cardiovascular disease, hypertension, or hyperlipidemia were more modestly associated with dysglycemia (Table 2). As a single risk factor, race and family history demonstrated the weakest associations with undiagnosed diabetes. Age ⱖ 45 years was the risk factor most strongly associated with undiagnosed prediabetes [OR (95% CI); 4.8 (4.2 – 5.4)] and undiagnosed dysglycemia [4.9 (4.4 – 5.6)]. A single RBG ⱖ 100 mg/dL (5.6 mmol/L) was more strongly associated with undiagnosed prediabetes [3.3 (3.0 – 3.8)] and undiagnosed dysglycemia [3.9 (3.5 – 4.4)] than all other risk factors examined except for age (Table 2). Table 3 shows a strong ‘dose response’ relationship between increasing RBG values and higher risk-adjusted odds of undiagnosed prediabetes, undiagnosed diabetes, and overall dysglycemia. For individuals with a single RBG 100 –119 mg/dL (5.6 – 6.6 mmol/L), the odds of undiagnosed diabetes were 7.1 (4.4 – 11.4). For glucose values between 120 –139 mg/dL (6.7–7.7 mmol/L) and ⱖ 140 mg/dL (7.8 mmol/L), the odds of undiagnosed diabetes were 30.3 (20.0 – 46.0) and 256 (150.0 – 436.9) respectively. In multivariate models, a single RBG ⱖ 100 mg/dL (5.6 mmol/L) was associated with a twenty-fold increased risk of undiagnosed diabetes [OR (95% CI); 20.4 (14.0 – 29.6)] and over double the odds of undiagnosed prediabetes [2.4 (2.1 – 2.7)] even after adjusting for age, sex, race, BMI, hypertension, hyperlipidemia, cardiovascular disease, and family history of diabetes. Overall, a single


RBG ⱖ 100 mg/dL (5.6 mmol/L) nearly tripled the riskadjusted odds of undiagnosed dysglycemia [2.8 (2.5 – 3.2)] (Table 4). As shown in Table 4, patients meeting USPSTF and ADA screening guidelines had increased odds of having undiagnosed prediabetes and undiagnosed diabetes. Meeting the single-factor USPSTF guideline criteria (presence of hypertension alone) nearly tripled the odds of prediabetes [OR (95% CI) 2.6 (2.4 –2.9)] and dysglycemia [2.7 (2.5–3.0)] and increased the odds of diabetes nearly 4-fold [3.8 (2.9 –5.0)]. Individuals meeting the multirisk factor ADA guidelines had approximately a ten-fold increase in the odds of prediabetes [9.8 (8.1–11.8)] and dysglycemia [10.6 (8.8 –12.7)]. Meeting ADA guidelines increased the odds of diabetes by over 50-fold [50.6 (17.7–144.5)]. A screening strategy based on a single RBG ⱖ 100 mg/dL (5.6 mmol/L) [OR (95% CI) 20.4 (14.0 – 29.6)] was much more strongly associated with undiagnosed diabetes than USPSTF guidelines (P ⬍ .0001) and not statistically different from the ADA screening guideline (P ⫽ .12). For detecting undiagnosed prediabetes, the single RBG strategy was similar to the USPSTF guideline (P ⫽ .32) but not as predictive as the ADA screening strategy (P ⬍ .0001).

Conclusions In a nationally representative sample of the US population without diagnosed diabetes, a single RBG ⱖ 100 mg/dL (5.6 mmol/L) is strongly associated with undiagnosed diabetes and demonstrates a robust dose response. In fact, a RBG ⱖ 100 mg/dL (5.6 mmol/L) was the single strongest predictor of undiagnosed diabetes outperforming all other traditional risk factors. As a simple strategy to detect undiagnosed diabetes, the association with undiagnosed disease for a single RBG ⱖ 100 mg/dL (5.6 mmol/L) is stronger than the USPSTF diabetes screening guidelines, which

Table 2. Univariate Associations Between Risk Factors and Undiagnosed Prediabetes, Diabetes, and Dysglycemia (Prediabetes ⫹ Diabetes) Odds Ratio (95% CI)

Characteristic

Undiagnosed Prediabetes


Age ⱖ 45 yr BMI ⱖ 25 kg/m2 Non-white race Family history diabetes Hypertension Hyperlipidemia Cardiovascular disease Random Glucose ⱖ 100 mg/dL

4.8 (4.2 – 5.4) 2.3 (2.0 – 2.6) 1.4 (1.3 – 1.7) 1.5 (1.4 – 1.6) 2.6 (2.4 – 2.9) 2.3 (2.0 – 2.6) 2.7 (2.2 – 3.3) 3.3 (3.0 – 3.8)

7.9 (5.7 – 10.9) 10.9 (6.8 – 17.5) 2.0 (1.4 – 2.9) 1.9 (1.5 – 2.4) 3.8 (2.9 – 5.0) 2.1 (1.6 – 2.8) 4.5 (3.4 – 5.9) 31.2 (21.3 – 45.5)

Dysglycemia 4.9 (4.4 – 5.6) 2.5 (2.2 – 2.8) 1.5 (1.3 – 1.7) 1.5 (1.4 – 1.7) 2.7 (2.5 – 3.0) 2.3 (2.0 – 2.6) 2.9 (2.4 – 3.4) 3.9 (3.5 – 4.4)

Results presented as odds ratios (95% CI). Weighted standard errors used to compute CIs. Random glucose (mg/dL) ⫻ 0.5551 ⫽ mmol/liter


doi: 10.1210/jc.2014-4116


5

Table 3. Dose Response Relationship Between Random Glucose and Undiagnosed Diabetes, Prediabetes, and Dysglycemia (Prediabetes ⫹ Diabetes) Adjusted Odds Ratio (95% CI)

Glucose Range



Random glucose ⬍ 100 mg/dL Random glucose 100 –119 mg/dL Random glucose 120 –139 mg/dL Random glucose ⱖ 140 mg/dL

Reference 2.2 (1.9 – 2.5) 3.3 (2.6 – 4.2) 3.5 (2.2 – 5.5)

Reference 7.1 (4.4 – 11.4) 30.3 (20.0 – 46.0) 256.0 (150.0 – 436.9)

Dysglycemia Reference 2.3 (2.0 – 2.7) 3.8 (3.0 – 4.9) 8.4 (5.7 – 12.3)

All values adjusted for age, sex, race, BMI, hypertension, hyperlipidemia, cardiovascular disease, and family history of diabetes. Random glucose (mg/dL) ⫻ 0.5551 ⫽ mmol/liter

Table 4. Odds of Having Undiagnosed Diabetes, Prediabetes, or Dysglycemia (Prediabetes ⫹ Diabetes) According to Different Screening Strategies Odds Ratio (95% CI)

Screening Strategy



Dysglycemia

USPSTF Guidelines ADA Guidelines Random Glucose ⱖ 100 mg/dL*

2.6 (2.4 – 2.9) 9.8 (8.1 – 11.8) 2.4 (2.1 – 2.7)

3.8 (2.9 – 5.0) 50.6 (17.7 – 144.5) 20.4 (14.0 – 29.6)

2.7 (2.5 – 3.0) 10.6 (8.8 – 12.7) 2.8 (2.5 – 3.2)

All results presented as weighted odds ratios. USPSTF Guideline: diagnosed hypertension ADA Guideline: All individuals ageⱖ45 or those with BMIⱖ25 kg/m2 and family history of diabetes, non-white race, hypertension, high cholesterol, cardiovascular disease. *Adjusted for age, sex, race, BMI, hypertension, hyperlipidemia, cardiovascular disease, and family history of diabetes. Random glucose (mg/dL) ⫻ 0.5551 ⫽ mmol/liter

recommend screening based on the presence of hypertension alone (5), and similar to the multirisk factor, more complex ADA screening guidelines (4). Nondiagnostic, random glucose values should be considered in diabetes risk assessments and may improve detection of undiagnosed cases of diabetes and prediabetes. Among individuals with undiagnosed prediabetes and diabetes, traditional diabetes risk factors are common. In general, the prevalence of individual risk factors increases across the glycemic spectrum from no diabetes, to undiagnosed prediabetes and undiagnosed diabetes. Similar to other studies (13–17), we observed significant associations between age, BMI, hypertension, family history and undiagnosed disease. Although we found race and family history to be significant risk factors, their association with undiagnosed disease was more modest than other risk factors. Similar to a recent study finding that men have a higher prevalence of diagnosed diabetes than women (18), we found that men were also more likely to have undiagnosed diabetes. Although the vast majority of individuals with undiagnosed diabetes and prediabetes are insured and have visited a healthcare provider in the past 12 months, similar to other studies (19), we found that only half of eligible individuals reported being screened diabetes in the past 3 years. Given the burden of undiagnosed disease and the

prevalence of diabetes risk factors, missed opportunities for diabetes screening are common (15). Although the barriers to diabetes screening in clinical practice are poorly understood, the assessment of diabetes risk factors and ordering of diabetes screening is largely at the discretion of individual clinicians and may be an afterthought in timeconstrained clinic visits. Automated risk-assessment and clinical decision support can facilitate provider workflow and be effective tools to increase screening rates for cancer (20). However, automation and implementation of multifactorial diabetes screening guidelines in clinical practice is challenging and parsimonious risk assessment models are needed (21). The development of automated risk models capable of informing clinical decision support requires accurate data that are readily available. Identification and abstraction of traditional diabetes risk factors from the EMR is largely dependent on the data entry and coding of end users. However, laboratory data are readily available and electronically integrated into most EMRs. As such, both historic and prospective RBG values collected in routine clinical practice could provide a platform to develop automated diabetes risk assessment. Although prediabetes is a strong risk factor for diabetes (14, 22), it cannot inform the initial decision to screen because its diagnosis requires prior completion of a gold-standard diabetes screening test. As such,


6


prediabetes is very useful in targeting repeat screening to individuals at high risk for diabetes, but its usefulness as a risk factor is limited because most individuals with prediabetes remain undiagnosed and overall screening rates are suboptimal (19). Although the ADA does not recommend random glucose as a screening test for diabetes and does not provide guidance to interpret RBG values (4), random glucose is commonly used as an ‘opportunistic screen’ for diabetes in clinical practice (23, 24). However, abnormal RBGs are frequently overlooked and failure to pursue gold-standard diabetes testing in response to elevated RBG values is common (25). Although the International Diabetes Federation recommends that individuals with a random glucose value 100 –199 mg/dL (5.6 –11.0 mmol/L) undergo formal diabetes testing, this recommendation is absent from US guidelines. Our finding that a single RBG ⱖ 100 mg/dL (5.6 mmol/L) is more strongly associated with undiagnosed diabetes than any single traditional diabetes risk factor supports the consideration of random glucose as a diabetes risk factor. Our demonstration of a strong doseresponse relationship between the magnitude of the random glucose elevation and an increased risk for diabetes further supports this recommendation. Importantly, modest elevations in RBG values (100 –119 mg/dL; 5.6 – 6.6 mmol/L) that are commonly ignored in clinical practice (23, 25) were strongly associated with undiagnosed disease. Together, these findings indicate that nondiagnostic, random blood glucose values are an important indicator of dysglycemia that could play a critical role in screening and case identification strategies. The usefulness of random glucose values in the diagnosis of diabetes and prediabetes has a strong physiologic basis. Glucose homeostasis is a tightly regulated function of ␤ cell insulin production and insulin sensitivity (26) such that glucose concentrations are maintained in a narrow range (27). Disruptions in glucose homeostasis initially manifest as impaired glucose tolerance and increased glucose variability (26). In clinical practice, this is reflected as nondiagnostic RBG elevations and increased glucose variability. Given that random glucose testing accounts for over 95% of glucose testing in clinical practice, (23) opportunities to recognize random glucose abnormalities and incorporate them into risk assessment exist in routine clinical practice. Although random glucose is not currently included in US diabetes screening guidelines, our findings demonstrate the potential importance of random glucose values in screening strategies. Existing studies demonstrate that a single RBG ⱖ 125 mg/dL (6.9 mmol/L) is a good predictor of diabetes (28) and performs as well as more complicated models based on age, BMI, and race (29). We demonstrate


that RBG values as low as 100 mg/dL (5.6 mmol/L) confer a significant risk for diabetes and are associated with undiagnosed diabetes. Our findings suggest that a hypothetical screening strategy in which individuals with a single RBG ⱖ 100 mg/dL (5.6 mmol/L) undergo gold-standard diabetes screening and the ADA diabetes screening guidelines, which incorporate multiple risk factors, have similar associations with undiagnosed disease. Additionally, the random glucose strategy – even after adjustment for other risk factors – is more strongly associated with undiagnosed diabetes than the USPSTF guidelines, which likewise recommend screening based on a single risk factor. A major strength of this study is that the data are nationally representative of the noninstitutionalized, civilian, US population without diagnosed diabetes and prediabetes. However, there are several limitations. First, our analyses are limited to cross-sectional associations between a single RBG value and glycemic status. Because of this, we are unable to examine the development of diabetes over time. Second, individuals with diagnosed diabetes and prediabetes were excluded based on self-report; however, self-reported diabetes is both highly sensitive and specific for diagnosed diabetes (30). Third, because we sought to detect cases of undiagnosed diabetes and prediabetes, we were unable to examine prediabetes as a risk factor for undiagnosed diabetes. However, most individuals with prediabetes remain undiagnosed (2). Additionally, our estimates of the association between the ADA guideline and undiagnosed disease likely underestimate the true association given our inability to include all ADA screening criteria in our definitions. Fourth, our analyses utilize serum glucose values rather than plasma glucose values. However, analytic differences between serum and plasma glucose values are small and unlikely to exceed the day-to-day biologic variation of glucose (31). Thus, our findings reflect real-world practice where serum glucose values are routinely obtained as part of chemistry panels. Fifth, our analyses focus on community-dwelling NHANES participants and may not directly translate to medical practice if patients are acutely ill. Sixth, our definition of prediabetes and diabetes was based on HbA1C criteria alone because individuals examined in a fasting state did not have random glucose data available. As a result, our findings likely underestimate the burden of undiagnosed prediabetes and diabetes (32). In conclusion, our findings indicate that a single random glucose value ⱖ 100 mg/dL (5.6 mmol/L) is a robust risk factor for type 2 diabetes and should prompt clinicians to obtain gold standard diabetes screening tests to categorize glycemic status. Considering the large number of individuals in the US with undiagnosed diabetes and prediabetes (2), our findings have important implications


doi: 10.1210/jc.2014-4116

for screening and case finding strategies. First, our findings highlight the importance of RBG values ⱖ 100 mg/dL (5.6 mmol/L) as a risk factor for diabetes and provide evidence to support the inclusion of RBG values ⱖ 100 mg/dL as a risk factor in US diabetes screening guidelines. Second, our findings suggest that increased awareness of abnormal random glucose values and targeted diabetes testing may improve diabetes detection among individuals engaged in clinical care. Given the high frequency of glucose testing in clinical practice and the growing repository of computerized laboratory data within electronic medical records, the development of automated, glucose-driven risk assessment strategies may improve the screening and diagnosis of type 2 diabetes in clinical practice.

Acknowledgments This study was conducted using resources supported by the UT Southwestern Center for Patient-Centered Outcomes Research (AHRQ R24 HS022418), UT Southwestern National Center for Advancing Translational Sciences (UL1TR001105) and the Dedman Family Scholars in Clinical Care. Dr. Bowen was supported by the National Center for Advancing Translational Sciences of the NIH (KL2TR001103) and the NIH/NIDDK K23 DK104065. Drs. Halm and Xuan were supported in part by AHRQ R24 HS022418. Preliminary data from this study were presented at the 2014 Society of General Internal Medicine national meeting (San Diego, CA). M.B participated in the design, analysis, data interpretation, and drafting of the manuscript. L.X. participated in analysis, data interpretation, and manuscript revision. I.L. participated in the data interpretation and manuscript revision. E.H participated in the design, analysis, data interpretation, and manuscript revision. No potential conflicts of interest relevant to this article are reported. Address all correspondence and requests for reprints to: Corresponding Author and Reprint Requests: Michael E. Bowen MD, MPH, 5323 Harry Hines Blvd, E1.410C, Dallas, TX 75390 –9169, Telephone: 214 – 648-3233, Fax: 214 – 648-3232, Email: [email protected]. Disclosure Summary: The authors have nothing to disclose. Dr. Bowen was supported by the National Center for Advancing Translational Sciences of the NIH (KL2TR001103) and the NIH/NIDDK K23 DK104065. Drs. Halm and Xuan were supported in part by AHRQ R24 HS022418. This work was supported by Grant Support: This study was conducted using resources supported by the UT Southwestern Center for Patient-Centered Outcomes Research (AHRQ R24 HS022418), UT Southwestern National Center for Advancing Translational Sciences (UL1TR001105) and the Dedman Family Scholars in Clinical Care.


7

References 1. Boyle JP, Thompson TJ, Gregg EW, Barker LE, Williamson DF. Projection of the year 2050 burden of diabetes in the US adult population: dynamic modeling of incidence, mortality, and prediabetes prevalence. Population health metrics. 2010;8:29. 2. Centers for Disease Control and Prevention. National Diabetes Statistics Report: Estimates of Diabetes and Its Burden in the United States. 2014; http://www.cdc.gov/diabetes/pubs/statsreport14/national-diabetes-report-web.pdf. Accessed 30 October 2014. 3. Knowler WC, Fowler SE, Hamman RF, Christophi CA, Hoffman HJ, Brenneman AT, Brown-Friday JO, Goldberg R, Venditti E, Nathan DM. 10-year follow-up of diabetes incidence and weight loss in the Diabetes Prevention Program Outcomes Study. Lancet. 2009; 374:1677–1686. 4. American Diabetes Association. Standards of medical care in diabetes–2014. Diabetes Care. 2014;37 Suppl 1:S14 – 80. 5. U. S. Preventive Services Task Force. Screening for type 2 diabetes mellitus in adults: U.S. Preventive Services Task Force recommendation statement. Ann Intern Med. 2008;148:846 – 854. 6. Geiss LS, James C, Gregg EW, Albright A, Williamson DF, Cowie CC. Diabetes risk reduction behaviors among U.S. adults with prediabetes. Am J Prev Med. 2010;38:403– 409. 7. Colagiuri S. Global Guideline for Type 2 Diabetes. International Diabetes Federation Clinical Guidelines Task Force;2012. 8. Johnson C, Paulose-Ram R, CL O. National Health and Nutrition Examination Survey: Analytic guidelines, 1999 –2010. National Center for Health Statistics Vital Health Stat 2013; 2. 9. Centers for Disease Control and Prevention. National health and Nutrition Examination Survey [Internet]. Questionnaires, datasets, and related documentation. http://www.cdc.gov/nchs/nhanes/ nhanes questionnaires.htm. Accessed 30 October 2014. 10. Steffes M, Cleary P, Goldstein D, Little R, Wiedmeyer HM, Rohlfing C, England J, Bucksa J, Nowicki M. Hemoglobin A1c measurements over nearly two decades: sustaining comparable values throughout the Diabetes Control and Complications Trial and the Epidemiology of Diabetes Interventions and Complications study. Clinical chemistry. 2005;51:753–758. 11. Centers for Disease Control and Prevention. Updated advisory for NHANES Hemoglobin A1c (glycohemoglobin) data. 2012; http:// www.cdc.gov/nchs/data/nhanes/A1c webnotice.pdf. Accessed 11 November 2013. 12. Nathan DM, Kuenen J, Borg R, Zheng H, Schoenfeld D, Heine RJ. Group Ac-DAGS. Translating the A1C assay into estimated average glucose values. Diabetes Care. 2008;31:1473–1478. 13. Bang H, Edwards AM, Bomback AS, Ballantyne CM, Brillon D, Callahan MA, Teutsch SM, Mushlin AI, Kern LM. Development and validation of a patient self-assessment score for diabetes risk. Ann Intern Med. 2009;151:775–783. 14. Lindstrom J, Tuomilehto J. The diabetes risk score: a practical tool to predict type 2 diabetes risk. Diabetes Care. 2003;26:725–731. 15. Dall TM, Narayan KM, Gillespie KB, Gallo PD, Blanchard TD, Solcan M, O’Grady M, Quick WW. Detecting type 2 diabetes and prediabetes among asymptomatic adults in the United States: modeling American Diabetes Association versus US Preventive Services Task Force diabetes screening guidelines. Population health metrics. 2014;12:12. 16. Cowie CC, Harris MI, Eberhardt MS. Frequency and determinants of screening for diabetes in the U.S. Diabetes Care. 1994;17:1158 – 1163. 17. Edelman D, Edwards LJ, Olsen MK, Dudley TK, Harris AC, Blackwell DK, Oddone EZ. Screening for diabetes in an outpatient clinic population. J Gen Intern Med. 2002;17:23–28. 18. Menke A, Rust KF, Fradkin J, Cheng YJ, Cowie CC. Associations Between Trends in Race/Ethnicity, Aging, and Body Mass Index With Diabetes Prevalence in the United States: A Series of Crosssectional Studies. Ann Intern Med. 2014;161:328 –335. 19. Nogrady B. Only half of at-risk adults being screened for diabetes.


8

20.

21.

22.

23.

24. 25. 26.

27.


Clinical Endocrinology News 2013; http://www.clinicalendocrinologynews.com/news/top-news/single-article/only-half-of-at-riskadults-being-screened-for-diabetes/ d23d462bfa6c020ddf93a82cd08c3d2e.html. Baron RC, Melillo S, Rimer BK, Coates RJ, Kerner J, Habarta N, Chattopadhyay S, Sabatino SA, Elder R, Leeks KJ, Task Force on Community Preventive S. Intervention to increase recommendation and delivery of screening for breast, cervical, and colorectal cancers by healthcare providers a systematic review of provider reminders. Am J Prev Med. 2010;38:110 –117. Noble D, Mathur R, Dent T, Meads C, Greenhalgh T. Risk models and scores for type 2 diabetes: systematic review. BMJ. 2011;343: d7163. Tabak AG, Herder C, Rathmann W, Brunner EJ, Kivimaki M. Prediabetes: a high-risk state for diabetes development. Lancet. 2012; 379:2279 –2290. Ealovega MW, Tabaei BP, Brandle M, Burke R, Herman WH. Opportunistic screening for diabetes in routine clinical practice. Diabetes Care. 2004;27:9 –12. Engelgau MM, Narayan KM, Herman WH. Screening for type 2 diabetes. Diabetes Care. 2000;23:1563–1580. Edelman D. Outpatient diagnostic errors: unrecognized hyperglycemia. Effective clinical practice : ECP. 2002;5:11–16. Kahn SE, Cooper ME, Del Prato S. Pathophysiology and treatment of type 2 diabetes: perspectives on the past, present, and future. Lancet. 2014;383:1068 –1083. Kahn SE, Prigeon RL, McCulloch DK, Boyko EJ, Bergman RN,


28.

29.

30.

31.

32.

Schwartz MW, Neifing JL, Ward WK, Beard JC, Palmer JP, et al. Quantification of the relationship between insulin sensitivity and beta-cell function in human subjects. Evidence for a hyperbolic function. Diabetes. 1993;42:1663–1672. Ziemer DC, Kolm P, Foster JK, Weintraub WS, Vaccarino V, Rhee MK, Varughese RM, Tsui CW, Koch DD, Twombly JG, Narayan KM, Phillips LS. Random plasma glucose in serendipitous screening for glucose intolerance: screening for impaired glucose tolerance study 2. J Gen Intern Med. 2008;23:528 –535. Ziemer DC, Kolm P, Weintraub WS, Vaccarino V, Rhee MK, Caudle JM, Irving JM, Koch DD, Narayan KM, Phillips LS. Age, BMI, and race are less important than random plasma glucose in identifying risk of glucose intolerance: the Screening for Impaired Glucose Tolerance Study (SIGT 5). Diabetes Care. 2008;31:884 – 886. Saydah SH, Geiss LS, Tierney E, Benjamin SM, Engelgau M, Brancati F. Review of the performance of methods to identify diabetes cases among vital statistics, administrative, and survey data. Ann Epidemiol. 2004;14:507–516. Sacks DB, Arnold M, Bakris GL, Bruns DE, Horvath AR, Kirkman MS, Lernmark A, Metzger BE, Nathan DM. Guidelines and recommendations for laboratory analysis in the diagnosis and management of diabetes mellitus. Clinical chemistry. 2011;57:e1– e47. Cowie CC, Rust KF, Byrd-Holt DD, Gregg EW, Ford ES, Geiss LS, Bainbridge KE, Fradkin JE. Prevalence of diabetes and high risk for diabetes using A1C criteria in the U.S. population in 1988 –2006. Diabetes Care. 2010;33:562–568.


Self-monitoring of blood glucose for type 2 diabetes 2.

Air pollution as a risk factor for type 2 diabetes.

Meat consumption as a risk factor for type 2 diabetes.

Self-monitoring of blood glucose in type 2 diabetes.

Should blood glucose strips be used in type 2 diabetes?

Self-reported discrimination, diabetes distress, and continuous blood glucose in women with type 2 diabetes.

Qualitative study of telemonitoring of blood glucose and blood pressure in type 2 diabetes.

Type 2 diabetes: an independent risk factor for tuberculosis: a nationwide population-based study.

Type 2 diabetes: a risk factor for liver mortality and complications in hepatitis B cirrhosis patients.

Sequence variants in SLC16A11 are a common risk factor for type 2 diabetes in Mexico.

Hypovitaminosis D: a novel risk factor for coronary heart disease in type 2 diabetes?

Obstructive sleep apnea as a risk factor for type 2 diabetes mellitus.

Vitamin D as a risk factor for premature atherosclerosis in patients with type 2 diabetes.

Type 2 diabetes as a risk factor for cognitive impairment: current insights.

Glucose Metabolism in High-Risk Subjects for Type 2 Diabetes Carrying the rs7903146 TCF7L2 Gene Variant.

Personalized glucose forecasting for type 2 diabetes using data assimilation.

A risk assessment model for type 2 diabetes in Chinese.

Cardiovascular Risk and Sodium-Glucose Cotransporter 2 Inhibition in Type 2 Diabetes.

Random blood glucose may be used to assess long-term glycaemic control among patients with type 2 diabetes mellitus in a rural African clinical setting.

Canagliflozin, a sodium glucose co-transporter 2 inhibitor, for the management of type 2 diabetes.

Hypertension is an independent risk factor for type 2 diabetes: the Korean genome and epidemiology study.

Recurrent subthreshold depression in type 2 diabetes: an important risk factor for poor health outcomes.

[Are high glucose levels a risk factor for dementia? High blood glucose levels increase the risk of dementia--not only in diabetes].

Evaluation of mean platelet volume in patients with type 2 diabetes mellitus and blood glucose regulation: a marker for atherosclerosis?