Baseline Serum Metabolites as Predictors of Teplizumab Response in Individuals With Type 1 Diabetes.
Elizabeth R Flammer, Lauren E Higdon, Srinath Sanda et al.
✦ AI-curated · Sources linked
In 30 seconds
This analysis examined whether baseline serum metabolites could predict response to teplizumab in 41 people newly diagnosed with type 1 diabetes who participated in the randomized AbATE trial. Response was defined as a <45% decline in baseline C-peptide at 2 years. Semitargeted metabolomics identified 15 significantly different metabolites: responders had higher levels of several TCA-cycle, amino-acid-derivative, and microbial metabolites, while nonresponders had higher glutamate and acylcarnitines. A random forest classifier achieved 0.769 accuracy and an area under the receiver operating characteristic curve of 0.881 in the test dataset.
Key findings
- Fifteen baseline metabolites differed significantly between teplizumab responders and nonresponders (P < 0.05).
- Responders had higher TCA-cycle metabolites, amino-acid derivatives, and microbial metabolites; nonresponders had higher glutamate and acylcarnitines.
- The random forest classifier achieved 0.769 accuracy and an area under the receiver operating characteristic curve of 0.881 in the test data set.
Why it matters
Response to teplizumab varies, and a baseline metabolic profile may help distinguish people more likely to preserve C-peptide after treatment. If validated, such profiles could support more individualized treatment decisions and inform prediction research for other immunotherapies in type 1 diabetes. The reported classifier performance is promising but does not establish clinical utility.
What to keep in mind
The analysis included 41 participants, and the authors state that further validation is needed. Classifier performance was evaluated using a split of this dataset, so external predictive performance is not established.