Study focuses on score accuracy to improve prediction models for the better.
Identifying patients at risk for developing type 1 diabetes (T1D) can allow for prevention trials as well as study of the pathophysiology of preclinical diabetes. Current prediction markers don’t remain constant throughout the course of the disease since they are based on immune and metabolic markers. Genetic markers though are independent of time and need only to be administered once. Previous studies have shown a connection of the HLA linked single nucleotide polymorphisms (SNPs) for T1D, as well as developed a T1D genetic risk score, which incorporates these regions to identify type 1 diabetes in particular. This study looked at the prediction accuracy of the T1D genetic risk score (GRS) in predicting the progression of islet autoimmunity and development of clinical T1D.
The participants for this study were all part of TrialNet Pathway to Prevention, a national program that follows at-risk relatives of T1D patients to detect their progression to diabetes. The participants all had at least one positive islet autoantibody, genotyped using Illumina ImmunoChip. The first set of screenings were to detect autoantibodies to glutamic acid decarboxylase, insulin, and insulinoma-associated antigen 2. If any screen came back as positive, patients were further screened for autoantibodies to zinc transporter 8 and islet cell antibodies. Patients were monitored for type 1 diabetes by hemoglobin A1C, autoantibody tests, and oral glucose tolerance test. Type 1 diabetes was diagnosed using the previously mentioned tests.
To calculate the T1D GRS for the patients, the odds ratios for the top 30 SNP variants linked to T1D were drawn. The T1D genetic risk score that separated it from type 2 diabetes was 0.280. A 10-SNP score was also calculated using just the top 10 T1D linked SNPs. The ability to predict progression was tested for both 30-SNP and 10-SNP. Many tests were used for statistical analyses. The Kaplan-Meier test was used to analyze the time it took to progress to type 1 diabetes, or for a single positive autoantibody to become multiple. Cox proportional hazards models were used and appropriate adjustments were made for both of the previously stated progressions of time. Time-dependent AUC analysis was used to test the accuracy of prediction for the T1D GRS and other factors.
Of the 1,244 participants, 90% had a first-degree relative with T1D. Median follow-up time was 5.4 years [CI 5.0-5.8]. There were 291 participants who had initially been positive for one antibody, from which 157 became positive for multiple autoantibodies, and 55 developed T1D. Of the initially screened participants, 953 had multiple autoantibodies and 419 developed T1D.
The values for the 30-SNP T1D GRS were from 0.138 to 0.341 (threshold 0.280).
T1D GRS was shown to predict clinical type 1 diabetes in patients positive for islet autoantibodies. Univariant analysis showed the significance of GRS to predict T1D with HR 1.7 and CI 1.43-2.0 P<.0001. GRS can also be added to the current prediction model and still play a significant factor (HR 1.29 CI 1.06-1.56; P=0.009). This study also looked at how effective the T1D GRS was at predicting progression to multiple autoantibody positivity. An increase of T1D GRS levels by 0.050 means a 50% increased risk of developing multiple autoantibody positivity from single positivity (HR 1.49 CI 1.1-2.05 P= 0.015). The genetic risk score, which used 10 SNPs instead of 30 SPNs, had similar results for every progression.
This study did have limitations. The first being that the T1D GRS was tested in a specific population, autoantibody positive relatives of T1D patients. The second limitation of the study was that the participants were >80% non-Hispanic whites. The T1D GRS should be tested in other races and adjustment can possibly be made.
This study showed that the T1D GRS can independently predict islet autoimmunity and development of type 1 diabetes in relatives of people who have type 1 diabetes. Adding this to current models also improved accuracy of predication. This study aligns with previous studies on genetic predictors of T1D, so the data looks promising. The T1D GRS can adapt to population characteristics and then combines genetic information into one number to add to prediction models. In models using the DPT-1 risk score, a score of >7 means that there are metabolic disturbances that have developed, in this case the T1D GRS is not a good prediction tool.
However in cases where there are no metabolic disturbances and thus the DPT-1 risk score is <7, T1D GRS is an excellent prediction tool. Age is a protective factor for T1D with the threshold age of 35, however in a previous study, T1D GRS identified more adults progressing to diabetes than it did children. Age plays a key role with the progression to T1D, because, as studies are pointing to, genetic factors differ with age.
- The T1D genetic risks score is a good independent predictor for progression of disease in an otherwise healthy looking individual.
- If a patient has a known DPT-1 risk score of >7 then the T1D GRS will not be accurate because the patient is then showing metabolic signs that something is wrong.
- Age is a factor that still needs to be further studied. Even though it is a protective factor, the GRS finds many adults with T1D. Genetics leading to T1D slowly differ as we age.
Redondo, Maria J., et al. “A Type 1 Diabetes Genetic Risk Score Predicts Progression of Islet Autoimmunity and Development of Type 1 Diabetes in Individuals at Risk.” Diabetes Care, vol. 41, no. 9, 2018, pp. 1887–1894., doi:10.2337/dc18-0087.
Arsalan Hashmi, PharmD. Candidate, LECOM College of Pharmacy