Quantifying the utility of islet autoantibody levels in the prediction of type 1 diabetes in children

Abstract

Aims/hypothesis: The aim of this study was to explore the utility of islet autoantibody (IAb) levels for the prediction of type 1 diabetes in autoantibody-positive children.

Methods: Prospective cohort studies in Finland, Germany, Sweden and the USA followed 24,662 children at increased genetic or familial risk of developing islet autoimmunity and diabetes. For the 1403 who developed IAbs (523 of whom developed diabetes), levels of autoantibodies against insulin (IAA), glutamic acid decarboxylase (GADA) and insulinoma-associated antigen-2 (IA-2A) were harmonised for analysis. Diabetes prediction models using multivariate logistic regression with inverse probability censored weighting (IPCW) were trained using 10-fold cross-validation. Discriminative power for disease was estimated using the IPCW concordance index (C index) with 95% CI estimated via bootstrap.

Results: A baseline model with covariates for data source, sex, diabetes family history, HLA risk group and age at seroconversion with a 10-year follow-up period yielded a C index of 0.61 (95% CI 0.58, 0.63). The performance improved after adding the IAb positivity status for IAA, GADA and IA-2A at seroconversion: C index 0.72 (95% CI 0.71, 0.74). Using the IAb levels instead of positivity indicators resulted in even better performance: C index 0.76 (95% CI 0.74, 0.77). The predictive power was maintained when using the IAb levels alone: C index 0.76 (95% CI 0.75, 0.76). The prediction was better for shorter follow-up periods, with a C index of 0.82 (95% CI 0.81, 0.83) at 2 years, and remained reasonable for longer follow-up periods, with a C index of 0.76 (95% CI 0.75, 0.76) at 11 years. Inclusion of the results of a third IAb test added to the predictive power, and a suitable interval between seroconversion and the third test was approximately 1.5 years, with a C index of 0.78 (95% CI 0.77, 0.78) at 10 years follow-up.

Conclusions/interpretation: Consideration of quantitative patterns of IAb levels improved the predictive power for type 1 diabetes in IAb-positive children beyond qualitative IAb positivity status.

Keywords: Islet autoantibody levels; Machine learning; Risk prediction models; Type 1 diabetes.

Fig. 1

Type 1 diabetes prediction performance (IPCW concordance index [C index] with 95% CI) for various covariate sets. (a) Performance for a model using baseline covariates; a model using baseline covariates and IAb positivity indicators from both initial and confirmatory visits; and a model using baseline covariates and IAb levels from both initial and confirmatory visits. (b) Performance for a model using baseline covariates, models using IAb positivity indicators from the initial visit, the confirmatory visit and both visits, and models using IAb levels from the initial visit, the confirmatory visit and both visits. The prediction start time (‘time 0’) was the seroconversion confirmatory visit. The duration of the follow-up period was 10 years. IAbs include GADA, IA-2A and IAA

Fig. 2

Comparison of type 1 diabetes prediction performance (IPCW concordance index [C index] with 95% CI) for two models. The first model (blue) used only the most recent IAb levels at the prediction start time (‘time 0’); the second model (grey) added baseline covariates and IAb levels from the initial and confirmatory visits to the most recent IAb levels. (a) Performance for various follow-up periods (T) ranging from 1 to 15 years. The prediction start time (‘time 0’) was the seroconversion confirmatory visit. (b) Performance for various test intervals (W) ranging from 0.25 to 5 years (W=0 is the confirmatory visit). In this analysis, the prediction time point (‘time 0’) was the time of the third visit (confirmatory visit+W). The follow-up period starts from the prediction time point and was fixed at 10 years

Fig. 3

(a) Type 1 diabetes prediction performance (IPCW concordance index [C index]) for various follow-up periods (T), ranging from 1 to 15 years, along the horizontal axis, and various intervals from confirmatory visit to the next test (W), ranging from 0.25 to 5 years, along the vertical axis (W=0 is the confirmatory visit). In this analysis, the prediction time point (‘time 0’) was the time of the third visit (confirmatory visit+W). The follow-up period starts from the prediction time point. All prediction models used just three covariates: GADA, IA-2A and IAA levels from the third visit. Darker shading indicates better performance. A standalone version of the table can be found as ESM Table 1. (b) Type 1 diabetes prediction performance for various follow-up periods (T) ranging from 1 to 15 years, with the test interval W fixed at 1.5 years. (c) Type 1 diabetes prediction performance for various test intervals (W) ranging from 0.25 to 5 years (W=0 is the confirmatory visit), with follow-up period (T) fixed at 5 years. (d) ORs for developing type 1 diabetes for a 1-, 2-, 5- and 10-fold increases in the levels of GADA, IA-2A and IAA separately for a prediction time point of W=1.5 years and a follow-up period (T) of 5 years. Confirm, confirmatory; y, years