Population Heterogeneity in Repeat Oral Glucose Tolerance Testing (OGTT)

Presentation Number: FP17-6
Date of Presentation: June 16th, 2013

Helen W Lau*1, Elliot M Landaw2 and Eli Ipp3
1Harbor-UCLA Medical Center, Torrance, CA, 2David Geffen School of Medicine at UCLA, Los Angeles, CA, 3Los Angeles Biomedical Research Institute at Harbor-UCLA Med Ctr, Torrance, CA


The OGTT is a traditional diagnostic test for diabetes and impaired glucose tolerance (IGT) whose usefulness has attracted recent controversy. Criticism has focused on short term irreproducibility in categorizing normal, impaired glucose tolerance (NGT, IGT) and diabetes (DM) due at least in part to glucose sampling variability at criteria boundaries coupled to regression to the mean (RTTM). In a small scale repeat study, we also noticed a possible intervention effect – subjects responded by changing habits prior to a second OGTT, even when asked not to, resulting in improved glucose tolerance. We therefore examined the effect of a repeat OGTT (OGTT-1 vs. OGTT-2) using a representative NHANES cohort of subjects having two tests no more than 1-6 weeks apart and also evaluated the literature to ascertain whether a bias could be found. Of 31 reported studies, only 4 studied similar short term reproducibility and provided a suitable resampling design and sufficient prevalence data to estimate population frequencies for categorization in two consecutive tests. This provided a total of 5 studies for computing a test statistic based on the ratio of the proportion who are IGT on OGTT-1 and NGT on OGTT-2 (IGT1-NGT2) to the proportion discordant in the reverse order (NGT1-IGT2). The null hypothesis of no intervention effect (and consistent with RTTM) is a discordance ratio of 1. The NHANES study (n=544 having 2 tests) revealed 13.4% were IGT-NGT discordant (8.1% IGT1-NGT2 vs 5.3% NGT1-IGT2), yielding a discordance ratio[95% CI] of 1.52[0.95-2.42] (p=0.08). The reported studies showed a prevalence of IGT-NGT discordance ranging from 6 to 27% and discordance ratios of 10.77[3.91-29.63] (n=212; p<0.001), 2.85[1.07-7.61] (n=498; p=0.036), 0.78[0.29-2.09] (n=60; p=0.618), 0.73[0.53-1.01] (n=1109; p=0.061). These results demonstrate significant heterogeneity in response to a repeat OGTT (p < 0.001). In summary, our analysis reveals a strikingly heterogeneous response to repeat OGTT, with 2 studies that are not distinguishable from RTTM and three in which a bias may exist towards improved outcome on OGTT-2 (IGT1-NGT2) that is greater than expected for a typical RTTM effect. We conclude that variability seen in repeat OGTT is likely more than random error in certain populations, where the OGTT may be also acting as an intervention. Such heterogeneity may influence interpretation of a short-term repeat OGTT if used for experimental or diagnostic purposes.


Nothing to Disclose: HWL, EML, EI