1-hour postload plasma glucose improves T2D prediction
A recent study has shown that 1-hour postload plasma glucose (PG), alone or in combination with metabolic markers, strongly predicts future risk of type 2 diabetes (T2D), outperforms the 2-hour PG and is cheaper to measure than metabolites.
“Metabolic markers provide a robust prediction of future risk of T2D in combination with postload PG measurements and glycated haemoglobin (HbA1c),” researchers said. “Shortening the standard 75-g oral glucose tolerance test (OGTT) to 1 hour improves its predictive value and clinical usability.”
Researchers measured PG and serum insulin levels at 0, 30, 60 and 120 minutes during an OGTT in 543 participants in the Botnia Prospective Study. Of these, 146 progressed to T2D within a 10-year follow-up period. Combinations of variables were used to assess 1,527 predictive models for progression to T2D.
The 1-hour PG showed better performance than every individual marker except 30-minute PG or mannose, whose predictive performances were lower but not significantly worse. The 1-hour PG also outperformed HbA1c according to DeLong test p value but not false discovery rate. [J Clin Endocrinol Metab 2019;104:1131-1140]
Predictive models significantly improved when metabolic markers were combined with PG measurements and HbA1c. In addition, mannose was identified as a robust metabolic marker.
“We constructed machine learning predictive models based on combinations of PG and insulin values, other clinical risk factors, HbA1c, and metabolic markers that we previously reported, thereby evaluating the performance of 1,527 models,” researchers said. [Diabetologia 2017;60:1740-1750]
“Of all 1,527 predictive models evaluated in this study, 139 models outperformed 1-hour PG in terms of area under the receiver operating characteristic curve (q<0.05). All these models included at least one metabolite as a variable,” they added.
A predictive performance higher than 1-hour PG alone can be achieved by adding HbA1c alone, 1-hour PG alone, 1-hour PG and clinical risk factors, or HbA1c and a PG measurement apart from the metabolites. The 1-hour PG significantly improved predictive performance of all multivariate models of metabolites.
Furthermore, there was a good balance between sensitivity (0.75) and specificity (0.68) with 1-hour PG. Of the 139 predictive models that outperformed 1-hour PG, 74 showed greater sensitivity, 132 showed greater specificity, 68 showed both greater sensitivity and specificity, and 47 showed greater sensitivity, specificity, accuracy, and positive and negative predictive values than 1-hour PG.
Included in these 47 models were the multivariate model involving HbA1c and the panel of six metabolites (AHB, [Hyp3]-BK, mannose, α-tocopherol, 10:1 carnitine, and X-12063); the model including HbA1c, fasting plasma glucose and six metabolites; and 10 different combinations of HbA1c and 30-minute PG and five metabolites. Twenty-eight of these models included HbA1c, while six models included clinical risk factors.
“International diabetes organizations should consider this recommendation because 1-hour PG more accurately identifies patients with dysglycaemia, critical in context of the burgeoning global population with obesity and glucose disorders,” researchers said.