Simple clinical measures may be better prognosticators than proposed diabetes subgroups
medwireNews: The recently proposed five diabetes subgroups are reproducible, but are more complicated and no better than simple clinical variables for predicting disease progression and treatment response, researchers have reported.
The team from the University of Exeter Medical School in the UK successfully reproduced the subgroups in 3802 participants of the ADOPT trial, using the same clustering method used by the researchers who first proposed the classifications.
The subgroups identified in ADOPT contained similar proportions of patients as they did in the original study, and had very similar clinical characteristics, except that they all had lower glycated hemoglobin (HbA1c) because they were undergoing treatment in a clinical trial setting rather than being drawn from clinical practice.
Addressing delegates at the 54th EASD Annual Meeting in Berlin, Germany, researcher John Dennis revealed that, within the ADOPT trial, glycemic progression was strongly related to diabetes subgroup; patients in the mild obesity-related and severe insulin-deficient diabetes subgroups had markedly faster HbA1c increases than those in the mild age-related and severe insulin-resistant diabetes subgroups. The smallest subgroup – severe autoimmune diabetes – was omitted for these analyses focusing on type 2 diabetes.
However, simple clinical variables did better. The combination of age at diagnosis, BMI, sex, and baseline HbA1c predicted 25% of the variability in glycemic progression – more than the 13% predicted by the four diabetes subgroups.
Similarly, progression to renal disease varied significantly among the subgroups, but so did baseline renal function and accounting for this abolished the significant differences between the groups.
So within this clinical trial dataset with standardized follow-up, variation in the risk for progression to renal disease “is entirely explainable […] simply by baseline renal function,” said Dennis.
Looking at the 1-year response to treatments, which in ADOPT were sulfonylureas, thiazolidinediones, and metformin, the researchers found variation between the four subgroups for the first two, but not the last. The amount of variability explained ranged from 10% to 15%.
But again, the clinical variables of age at diagnosis, BMI, sex, and baseline HbA1c better predicted response, with the amount of variability explained ranging from 26% to 35%.
However, speaking from the audience, Tiinamaija Tuomi (University of Helsinki, Finland), who is part of the group that produced the original five classifications paper, noted that they also tested them against routine clinical variables and did not find the latter to be so strongly predictive. She suggested this could reflect differences in the chosen study populations.
Responding to her comments, Dennis said that “the key point we’re trying to get across is that, yes, the subgroups do show differences, but really if you want to predict renal progression our analysis suggests that the best thing [you can do] is to just be really simple – use the [estimated glomerular filtration rate] when you have the patient in front of you rather than assign them to much more complex clusters.”
medwireNews is an independent medical news service provided by Springer Healthcare. © 2018 Springer Healthcare part of the Springer Nature group