ALLO-SAFE: Allopurinol Hypersensitivity Risk Stratification Before Urate-Lowering Therapy in Gout and Rheumatic Care

Authors: Dr. Erick Zamora-Tehozol, DNAI, RheumaAI
ORCID: 0000-0002-7888-3961

Abstract

Allopurinol is the preferred first-line urate-lowering therapy for most patients with gout, yet rare severe hypersensitivity reactions continue to produce preventable morbidity and mortality when genetic, renal, and prescribing risk factors are reviewed in isolation. We present ALLO-SAFE, a transparent 10-domain weighted clinical skill for relative risk stratification before or during very early allopurinol initiation. The model integrates HLA-B58:01 status, ancestry-linked pretest concern, chronic kidney disease, planned starting dose, thiazide exposure, prior rash history, age, chronic liver disease, urgency pressure to start therapy, and baseline monitoring readiness. The implementation is executable as standalone Python with no external dependencies and includes Monte Carlo uncertainty estimation. Demo scenarios separate genotype-negative conservative initiation (LOW), a chronic kidney disease patient with unknown genotype and non-conservative starting conditions (VERY HIGH), and an HLA-B58:01-positive patient with stacked risk factors (VERY HIGH). ALLO-SAFE is intended as an auditable bedside safety scaffold rather than an absolute event-probability engine. Limitations include evidence-informed weighting rather than prospective regression derivation, incomplete population portability, and the need to defer to formal genotype-guided prescribing and specialist review when severe prior reactions are suspected.

Clinical Problem

In gout care, allopurinol is common enough to become routine. That routine can obscure the fact that allopurinol hypersensitivity syndrome and severe cutaneous adverse reactions are uncommon but clinically devastating. Safety failures usually do not arise from one variable alone; they emerge when several modest risks stack together: kidney impairment, higher-than-conservative starting doses, thiazide co-prescription, absent genotype data, and logistical pressure to start therapy quickly.

Methodology

ALLO-SAFE uses ten transparent domains, each mapped to a weighted contribution in a 0-100 composite score. The weights were chosen to reflect relative bedside importance from published guideline and safety literature, with the heaviest contribution assigned to HLA-B*58:01. The model then performs 5,000 seeded Monte Carlo simulations with bounded jitter to communicate uncertainty instead of false precision.

Executable Skill

See skills/allo-safe/allo_safe.py and embedded code in SKILL.md.

Demo Summary

• Scenario 1: genotype-negative, preserved kidney function, 100 mg/day start → LOW
• Scenario 2: CKD + unknown genotype + thiazide + 200 mg/day + limited monitoring → VERY HIGH
• Scenario 3: HLA-B*58:01 positive + CKD + prior rash + 300 mg/day → VERY HIGH

Why this score exists

The purpose is not to compete with clinical judgment. It is to make the safety conversation explicit, auditable, and executable by an agent or clinician before allopurinol exposure becomes automatic.

Limitations

• Evidence-informed weighted model, not externally validated prospective prediction
• Relative stratification only; not a calibrated absolute risk estimate
• Not a substitute for HLA-B*58:01 testing where recommended
• Does not manage flare prophylaxis, urate targets, or febuxostat selection

References

1. FitzGerald JD, Dalbeth N, Mikuls T, et al. 2020 American College of Rheumatology Guideline for the Management of Gout. Arthritis Rheumatol. 2020;72(6):879-895. DOI: 10.1002/art.41247
2. Gonçalo M, Coutinho I, Teixeira V, et al. HLA-B*58:01 is a risk factor for allopurinol-induced DRESS and Stevens-Johnson syndrome/toxic epidermal necrolysis in a Portuguese population. Br J Dermatol. 2013;169(3):660-665. DOI: 10.1111/bjd.12389
3. Lu CY, Wang CW, Hui RCY, et al. Examining the use of allopurinol: Perspectives from recent drug injury relief applications. J Formos Med Assoc. 2019;118(1 Pt 1):206-214. DOI: 10.1016/j.jfma.2018.06.006