ORF18-02 - A Three-Variable Diagnosis-Stage Screening Model to Differentiate Wolfram Syndrome from Type 1 Diabetes: A Single-Centre Study with Decision Curve Analysis

Background
Wolfram syndrome (DIDMOAD) is a rare progressive neurodegenerative disorder caused by WFS1 pathogenic variants, with an estimated prevalence of 1 in 770,000. Diabetes mellitus is typically the earliest manifestation, yet at onset it is clinically indistinguishable from type 1 diabetes (T1D), leading to diagnostic delays that preclude timely surveillance and family counselling. No validated clinical screening tool exists to guide referral for targeted genetic testing at the time of diabetes presentation, particularly in resource-constrained settings where molecular diagnosis remains inaccessible.
Methods
A single-centre retrospective–prospective study enrolled 37 paediatric and young adult patients (18 with clinically established Wolfram syndrome; 19 T1D controls; age ≤25 years). Three predictors available at diabetes onset were selected a priori: parental consanguinity, age at diabetes diagnosis, and baseline estimated glomerular filtration rate (eGFR, Bedside Schwartz equation). A Firth bias-reduced logistic regression model was developed to address small-sample bias and quasi-complete separation. Internal validation used 1,000 bootstrap resamples. Clinical utility was quantified via decision curve analysis.
Results
Groups differed significantly in parental consanguinity (61.1% vs 0%; p< 0.001), age at diabetes onset (median 5.0 vs 8.4 years; p=0.031), and baseline eGFR (81 vs 124 mL/min/1.73 m²; p=0.004). The three-variable model demonstrated excellent discrimination (apparent AUC 0.974; bootstrap-corrected AUC 0.969, 95% CI 0.936–0.980) and acceptable calibration (Brier score 0.051). At a sensitivity-prioritised threshold of 0.51, sensitivity was 0.94 and specificity 1.00. Decision curve analysis showed net benefit exceeding both "test all" and "test none" strategies across clinically relevant threshold probabilities. When consanguinity was excluded, discrimination fell substantially (AUC 0.815), underscoring its dominant role in autosomal recessive risk stratification.
Conclusions
This parsimonious model — requiring only three routinely available variables — identifies Wolfram syndrome at diabetes onset with high sensitivity and may serve as a practical, low-cost triage tool for targeted genetic testing. Its performance is particularly relevant for populations with elevated consanguinity rates, where the burden of rare autosomal recessive conditions is highest. External validation with molecular confirmation is required before clinical implementation.

*Unless otherwise noted, all abstracts presented at ENDO must not be released to the press or the public until the date and time of presentation. For oral presentations, the abstracts are embargoed until the session begins. The Endocrine Society reserves the right to lift the embargo on specific abstracts that are selected for promotion prior to or during ENDO.*