(P 229) DETERMINANTS OF DIAGNOSTIC INTERVAL IN EPITHELIOID HEMANGIOENDOTHELIOMA: A LINEAR AND QUANTILE REGRESSION ANALYSIS OF PATIENT-REPORTED DATA FROM THE SPAGN GLOBAL SARCOMA DIAGNOSIS PATHWAY SURVEY

Objective: Delayed diagnosis is a major challenge in managing epithelioid hemangioendothelioma (EHE), a rare vascular sarcoma with nonspecific symptoms. This study assessed demographic and clinical factors associated with diagnostic interval length in EHE patients.

Methods: Data were derived from the SPAGN Global Sarcoma Diagnosis Pathway Survey, including adults with EHE. Outcome was the log-transformed diagnosis interval, defined as time in days from onset of symptoms to current diagnosis. Associations between clinical and demographic variables and diagnostic interval were explored, using linear regression and quantile regression at the 25th, 50th, and 75th percentiles to capture effects across the distribution. Variable selection was AIC-based; missing data were imputed using chained equations. Bootstrapped 95% confidence intervals (R = 500) were calculated; coefficients were exponentiated for interpretability.

Results: A total of 50 EHE patients were included. The median diagnostic interval was 349 days (IQR 28-1,542). While no associations were statistically significant, trends suggested potentially meaningful effects. For instance, longer diagnostic intervals were seen in females (exp(β)=1.89, 95% CI: 0.88–4.06) and those with symptoms (exp(β)=1.72, 95% CI: 0.88–3.35), while older age trended toward shorter intervals (exp(β)=0.98 per year; 95% CI: 0.96–1.00).
Quantile regression revealed longer delays in 75th percentile analysis if symptoms were reported (exp(β) = 2.09; 95% CI: 1.12-3.92). Also, poor healthcare accessibility was significantly associated with shorter diagnostic intervals at the 50th (exp(β) = 0.43; 95% CI: 0.22-0.82) and 75th percentiles (exp(β) = 0.45; 95% CI: 0.25-0.83).

Conclusion: Delayed diagnosis in EHE remains substantial with a median diagnostic interval approaching 12 months. Although only a few variables reached statistical significance, the observed trends suggest an influence of various distribution-dependent factors. Quantile regression is important in uncovering associations masked in mean-based models, stressing the need for tailored diagnostics in ultra-rare cancers.