A pilot study of a urinary IL-8 and IL-17-based formula for discriminating Hunner-type from non-Hunner-type interstitial cystitis.

OBJECTIVE

To determine whether urinary or plasma cytokines more effectively differentiate Hunner-type interstitial cystitis (HIC) from non-Hunner-type interstitial cystitis (NHIC), and to develop an interpretable discriminative formula based on the most informative features.

METHODS

Paired urine and plasma samples were collected from 47 patients with IC/BPS (15 HIC, 32 NHIC), and urine samples from 84 healthy controls were analyzed for baseline comparison. Twelve cytokines were quantified in each sample. Classification performance was evaluated using AutoGluon automated machine learning with stratified five-fold cross-validation across three feature sets: urine cytokines, plasma cytokines, and their combination. Symbolic regression with the QLattice algorithm, implemented within a nested cross-validation framework with seed averaging and a complexity constraint to reduce overfitting, was employed to generate a simplified discriminative formula.

RESULTS

The urine-based model outperformed both plasma-based and combined models (AUC 0.979 ± 0.028 vs. 0.741 ± 0.271 vs. 0.921 ± 0.087). A clear pathophysiological gradient was observed among healthy controls, NHIC, and HIC, with urinary IL-8 elevated 72-fold in HIC and 5-fold in NHIC compared to healthy controls. The QLattice-derived formula, incorporating uIL-17 and uIL-8, achieved a nested cross-validation accuracy of 90.5% (95% CI: 88.5%-92.4%) and an AUC of 0.948 (95% CI: 0.927-0.966).

CONCLUSION

Urinary cytokines, specifically IL-8 and IL-17, offer a more effective non-invasive approach for IC/BPS subtyping compared to plasma markers. The two-cytokine formula shows promise as an exploratory discriminative tool. However, external validation in larger, independent cohorts is necessary prior to clinical implementation.