How reliable are random plasma glucose levels (RPGs) compared to fasting blood glucose or A1c?

Early detection of diabetes can prevent other major complications from occurring, such as micro and macrovascular complications, yet many Americans go undiagnosed.  The purpose of this study was to evaluate if routine outpatient random blood glucose (RPG) levels can predict or diagnose patients with diabetes, which may lead to early preventive care for these patients.

The researchers in this study evaluated 942,446 patients without a diagnosis of diabetes at the U.S. Veterans hospital in a retrospective cohort study.  These participants had to have at least 3 or more random plasma glucose levels at baseline within a year and one or more visits with their primary care physician per year within a 5-year follow-up. The mean age of the participants was 63 years of age, 82.7% of them were white, and the mean BMI was 28.9 kg/m2.  The primary outcome for this study was the incidence of diabetes.  The incidence of diabetes was defined as having an outpatient prescription for antihyperglycemic medication(s) and a diagnostic code. 

Over the 5-year follow-up period, there were 94,599 patients that obtained a diagnosis of diabetes while the other 847,847 were not diagnosed with diabetes.  Patients in the diabetic and nondiabetic groups had comparable characteristics, except that the patients that had received a diagnosis of diabetes had a higher BMI (32 vs. 28 kg/m2) and higher random plasma glucose levels (150 vs. 107 kg/m2).  Most of the patients that received the diagnosis of diabetes were more likely to be black (P>0.0001; 18% vs. 15%).  Researchers used a receiver operating characteristic (ROC), which is a way to show the performance of two possible outcome classes (also known as a binary classifier).

The ROC curve for prediction was 0.701 in patients diagnosed with diabetes within the first year based on demographic factors.  The ROC curve increased to 0.708 when other factors such as systolic blood pressure, non-high-density lipoprotein cholesterol, and smoking were added.  ROC curve was significantly higher in the prediction of baseline RPG alone (ROC 0.878) when two or more levels were over given values.  When additional factors were added, as stated above, the ROC curve increased to 0.900 with elevated RPG levels. Obtaining two or more RPG ­­> 115 mg/dL has a specificity and sensitivity of 77 and 87 percent, respectively.  Whereas, having a two or more RPG > 130 mg/dL has specificity and sensitivity of 93 and 59 percent, respectively.  When predicting the diagnosis of diabetes in 3 or 5 years the ROC was reduced but still significantly elevated when using RPG levels alone (ROC 0.839 and 0.809, respectively). 

The conclusion of this study was patients should be tested outpatient using RPG to determine if further testing is needed, allow the preventive measures to be taken before the onset of diabetes, and for early diagnosis of diabetes.  Future studies should show how reliable RPG levels are in diagnosis when compared to fasting blood glucose (FBG) and hemoglobin A1c (HbA1c). 

Practice Pearls:

  • Random plasma glucose levels can aid in the diagnosis and prediction of diabetes in patients.
  • Many patients diagnosed with diabetes in this study had overweight or obesity,  and were mostly African American.
  • When additional factors such as systolic blood pressure, non-high-density lipoprotein cholesterol, and smoking were added, there was an increased risk of the participant to acquire diabetes.

“Random Plasma Glucose Levels Can Predict Diabetes Diagnosis.” PracticeUpdate, www.practiceupdate.com/news/25968/2/8?elsca1=emc_enews_daily-digest&elsca2=email&elsca3=practiceupdate_diab&elsca4=diabetes&elsca5=newsletter&rid=OTE0MTIxOTE4NTkS1&lid=10332481.

Mary K. Rhee, Yuk-Lam Ho, et al. “Random Plasma Glucose Predicts the Diagnosis of Diabetes.” PLOS ONE, Public Library of Science, journals.plos.org/plosone/article?id=10.1371/journal.pone.0219964#sec015.

Keri Hames, PharmD Candidate, Florida A&M University, College of Pharmacy & Pharmaceutical Sciences

SOURCE: http://www.diabetesincontrol.com/predicting-diabetes-diagnoses-using-random-plasma-glucose/