Aims/hypothesis:The aim of this study was to identify shared and distinct metabolite profiles prospectively associated with nephropathy, retinopathy and neuropathy at 15 years’ follow-up among 1947 participants in the Diabetes Prevention Program Outcomes Study, the long-term follow-up of the Diabetes Prevention Program (DPP).
Methods:We applied bootstrapped LASSO to 353 annotated metabolites to identify metabolites associated with one or more complication. For these metabolite hits, we tested for an interaction with DPP treatment arm, and ran multivariable models for the pooled sample or within treatment group as appropriate.
Results:
At follow-up, 572 participants had one or more complication (
n
=277 nephropathy,
n
=194 retinopathy,
n
=212 neuropathy). Of 105 metabolites that predicted any complication, 74 predicted one, 27 predicted two, and four predicted all three. In a pooled analysis of 69 metabolites without treatment arm interactions, histidine predicted lower odds of nephropathy (OR 0.75; 95% CI 0.69, 0.88), and serine predicted lower odds of nephropathy (OR 0.69; 95% CI 0.58, 0.82) and neuropathy (OR 0.68; 95% CI 0.56, 0.84). Of 36 metabolites that interacted with treatment arm, higher
N
-carbamoyl-β-alanine predicted greater odds of nephropathy (OR 1.99; 95% CI 1.38, 2.99) and C22:0-sphingomyelin predicted lower odds of neuropathy (OR 0.54; 95% CI 0.37, 0.77) in the metformin arm. In the lifestyle intervention arm, quinolinic acid predicted greater odds of neuropathy (OR 1.64; 95% CI 1.24, 2.19). These estimates accounted for sex, race, baseline age, BMI and smoking, and time elapsed during follow-up. Further adjustment for HbA
1c
during follow-up, incident diabetes and eGFR did not change the results.
Conclusions/interpretation:The existence of distinct metabolite profiles associated with single microvascular complications highlights the importance of characterising pathophysiological mechanisms specific to each complication, in addition to studying shared mechanisms across multiple complications.