In December 2016, the PrIMeD (Precision Individualized Medicine for Diabetes) program was established to form an interdisciplinary effort to investigate detection of those at risk of T1D using genetic screening and employment of the artificial pancreas to control T1D and develop a cure using immunologic methods of beta cell regeneration. The strategic goal of this report was to conduct a pilot study that could lead to establishment of a statewide network to provide genetic screening for T1D.

Participant recruitment and baseline data collection

This study was approved by the University of Virginia Human Subjects Research Institutional Review Board, and a study ethicist was involved in the study design.

Children aged 2 to 16 years were offered saliva-based screening for genetic risk of T1D. Clinical research coordinators were stationed in the waiting rooms of pediatric and diabetes clinics that were selected for recruitment and study activities. Parent/guardians of children who were in clinics for well-child visits were approached for participation in the study and provided with a flyer describing the study with an educational section describing signs and symptoms of diabetes (Additional File 1: Fig S1) to increase awareness of the disease process in type 1 diabetes. Participation in the study was voluntary. Written informed consent/assent (Additional File 1: Document S1) was obtained on a tablet from the children’s parent/legal guardian in private space within each site.

A short baseline questionnaire (Additional File 1: Document S2) was obtained on a tablet to evaluate personal medical history and family history of health and diabetes, including contact information, family structure, self-reported ancestry, history of T1D and other autoimmune diseases, and if a family member had a prior diagnosis of diabetes (and evidence of hospitalization or diabetic ketoacidosis). The consent form and questionnaire were available in English and Spanish. The clinical research coordinator obtained a saliva sample (DNA Genotek Inc./Oragene•Discover|OGR-500) from each participant. Younger children, and those with difficulty producing enough saliva to spit, were offered an assisted collection that involved the clinical research coordinator using a small sponge-like swab to gently collect saliva from the participant’s inner cheek area. Unique barcodes were assigned to individuals to link consent forms, questionnaires, and saliva kits.

Clinical research coordinators hand-delivered the saliva specimens to the University of Virginia School of Medicine CAP/CLIA-certified Medical Genetics Laboratory where samples were logged and inventoried under CAP/CLIA protocols. DNA was extracted from the saliva samples using standard laboratory protocols, and batches were used as input to custom genotyping using a Fluidigm genotyping platform (BioMark HD Reader and Juno™ system). No saliva specimens were kept for long-term storage. The T1D-focused genotyping array was customized using 74 SNPs (Additional File 1: Table S1) to generate a T1D GRS [4, 17]. In T1D-associated regions, SNPs were selected using a stepwise conditional logistic regression model [3]. Lead variants independently associated with T1D were selected for the T1D-focused genotyping array. If the flanking sequence of a lead variant resulted in low-quality genotyping score, we replaced the variant with a proxy SNP in high linkage disequilibrium. Arrays were genotyped following the manufacturer’s (Fluidigm) protocol. The T1D GRS consists of 26 SNPs in the HLA region, three SNPs in IFIH1 and IL2RA, two SNPs in INS, DEXI, PTPN2, and TYK2, and one SNP in the remaining T1D risk regions [3]. The genetic data produced by the custom array was subjected to both laboratory and statistical genetic quality control (e.g., SNP missingness and call rates).

Genetic risk score (T1D GRS) classification

The genotyping panel produced raw data in batches of 96 samples that were analyzed using a KING [22] software script. Samples and SNPs that passed quality control (e.g., sample call rates ≥ 80%, SNP call rates ≥ 95%) were saved in a binary file, with the T1D GRS for each participant generated using PLINK [23] software. The T1D GRS was calculated by computing the sum of the T1D-associated risk alleles weighted by the effect size estimated from robust fine mapping data [3, 4]. The proposed threshold from the T1D GRS (≥ 5) corresponds to a modest sensitivity (0.446) with high specificity (0.955) as estimated from the T1DGC resource of 16,086 samples of European ancestry (6670 T1D cases). The prediction of T1D risk in the training set was high (AUC = 0.889) and represents a balance between missing fewer individuals at “true” high genetic risk versus being cost-effective (numbers requiring subsequent islet autoantibody testing).

Each participant was determined to be at “high genetic risk” or “not high genetic risk” based on a selected T1D GRS threshold (“high” defined by T1D GRS ≥ 5). In a previous test/validation case–control sample (data not published), this threshold provided ~ 85% sensitivity and ~ 80% specificity, roughly equating to a tenfold increased risk over population rate.

Genetic risk communication

After application of the T1D GRS to define “high genetic risk” versus “not high genetic risk”, a report for each participant was generated. Since genetic risk was viewed as a research test, and was not CAP/CLIA certified, the information was not entered into the participant’s electronic health record. Results of the T1D GRS were returned to parents/legal guardians of the study participants, with approaches dependent upon disease status of the participant (with or without diagnosed T1D) and the T1D GRS value (“high” versus “not high”).

For participants with a “high genetic risk,” letters were mailed following either a successful phone contact (Additional File 1: Document S3) or 3 failed attempts at phone contact by a study pediatrician or genetic counselor. The letter provided basic facts about genetic and environmental risk factors for T1D, clinical signs and symptoms of hyperglycemia, and an interpretation of the T1D GRS. In addition, the letter included statements that genetic risk is “not destiny” (those at “high” genetic risk may not develop T1D; those at “not high” genetic risk may yet develop the disease).

For those participants at “not high genetic risk,” letters were mailed without initiating a phone call (Additional File 1: Document S4). All participants had access to a study phone line and email for questions. In addition, a certified genetic counselor was available to address questions from individuals that might impact their willingness to participate in the study as well as to provide additional risk interpretation for parents/guardians and participants after receipt of genetic risk results.

Islet autoantibody testing

Parents/guardians of study participants eligible for islet autoantibody testing were identified, and a separate written informed consent/assent was obtained, blood collection (venipuncture) was scheduled, and specimens were shipped to the Clinical Immunology Laboratory at the Barbara Davis Center for Diabetes (BDC). The BDC measured islet autoantibodies to insulin (IAA), GAD65 (GADA), IA-2 (IA-2A), and ZnT8 (ZnT8A) using radio-binding assays [24]. In the 2020 IASP Workshop, sensitivity and specificity for IAA was 62% and 99%, for GADA 78% and 99%, for IA-2 72% and 100%, and for ZnT8 74% and 100%, respectively.

Islet autoantibody screening results were returned to PrIMeD using a secure HIPAA-compliant electronic portal. Parents/guardians of participants with either 0 or 1 autoantibodies detected received a mailed copy of their results with interpretation; results were considered “research” and not placed in the electronic health record. For those participants with no islet autoantibodies detected, repeat screening was recommended in 3 years; for those with one islet autoantibody detected, repeat screening was recommended in 1 year.

Clinically meaningful autoantibody screening results (≥ 2 islet autoantibodies detected) were reported to the parent/guardian by the PRiMeD pediatric endocrinologist, followed by a mailed copy of the results. A CLIA-certified report was generated and placed in the participant’s electronic health record, unless requested otherwise. For participants at “high genetic risk” with ≥ 2 islet autoantibodies, parents/guardians were presented with options for engagement in structured clinical monitoring, e.g., home fingerstick self-monitoring of blood glucose (SMBG), continuous glucose monitoring (CGM), or hemoglobin A1c (HbA1c) testing. Information on volunteering for participation in immune therapy trials (e.g., TrialNet) was provided.

Follow-up of study participants

At least 1 year after study entry, a follow-up survey was conducted by clinical research coordinators. The survey was administered by telephone or email (for those with 3 failed phone contact attempts). Email attempts included an individualized link to the survey in Qualtrics HSD®, a HIPAA-compliant web-based survey platform. The follow-up questionnaire (Additional File 1: Document S5) obtained updated participant contact information and health history (T1D and other autoimmune disease development) of the participants and of their relatives.

Statistical analyses

Statistical analyses using were performed R version 4.2.1, employing the lm() routine with the T1D GRS as the outcome variable and age as the predictor; covariates included sex, family history, and self-reported ancestry. In one secondary analysis, the T1D GRS was partitioned into those SNPs in the HLA region (in the human major histocompatibility complex, MHC) and those not in the MHC, to determine if source of genetic information (MHC versus not) had an association with age at onset. In a second analysis, the full T1D GRS was evaluated for association with early versus late age at onset (early ≤ 5 years). Significance was considered for P ≤ 0.05 in all analyses.

Study implementation

Three sites affiliated with the University of Virginia Health System (UVA) were chosen to initiate recruitment—UVA Birdsong Clinic, UVA Pediatrics/Battle Building, Charlottesville, VA; Northridge Pediatrics, UVA Medical Park, Charlottesville, VA; Orange Pediatrics, Orange Medical Center, Orange, VA. These sites represent a university clinic environment (Birdsong), a private clinic environment (Northridge), and a rural clinic environment (Orange). Clinical research coordinators were hired in February 2018, one for each clinic site, and started a month of training for interaction with potential participants and their parent/legal guardian. A 2-week “shadowing” of each clinic was used to ascertain clinic-specific patterns of activities and to establish the optimal space for privacy during the consent/assent process. Staff were trained for collection of data and specimens in a manner to ensure minimum impact on clinic patient flow. Active recruitment was initiated May 2018. Ultimately, 12 clinical sites (university, private-practice, and diabetes subspecialty) from eight practices were used for recruitment into the study.