ALLO-SCAR: Transparent Allopurinol Severe Cutaneous Adverse Reaction Risk-Context Stratification Before or During Therapy

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

Abstract

Allopurinol is a first-line urate-lowering therapy, but its use is shadowed by rare and potentially fatal severe cutaneous adverse reactions (SCAR), including DRESS, Stevens-Johnson syndrome, and toxic epidermal necrolysis. The bedside challenge is not merely awareness that HLA-B58:01 is important. It is integrating genotype, kidney function, starting dose, co-medications, and early systemic findings into a defensible decision about whether allopurinol should be started, continued, or stopped urgently. We present ALLO-SCAR, an executable Python skill for transparent allopurinol SCAR risk-context stratification. The model combines HLA-B58:01 status, ancestry context, chronic kidney disease, allopurinol dose, diuretic exposure, cardiovascular comorbidity or hypertension, prior mild allopurinol rash, days since treatment initiation, and early warning features including fever, facial edema, mucosal involvement, eosinophilia, liver-enzyme elevation, and creatinine rise. Outputs include visible component scores, categorical risk classes, recommended actions, alerts, and explicit limitations. In demonstration scenarios, a low-dose planned initiation case is LOW risk, a chronic kidney disease patient started at 300 mg/day with early systemic warning signs is CONTRAINDICATED / CRITICAL, and an HLA-B*58:01-positive patient with mucosal involvement and acute kidney injury is CONTRAINDICATED / CRITICAL. ALLO-SCAR is intended as an auditable triage and prescribing-context aid rather than a validated probability calculator and does not replace direct clinical assessment for SCAR.

Keywords: allopurinol, SCAR, DRESS, Stevens-Johnson syndrome, toxic epidermal necrolysis, HLA-B*58:01, gout, pharmacogenomics, clinical decision support, RheumaAI, DeSci

1. Clinical problem

Allopurinol remains the anchor urate-lowering agent for gout because it is effective, inexpensive, and guideline-supported. Yet a rare subset of exposed patients develop severe hypersensitivity syndromes with cutaneous and systemic organ injury. These reactions are uncommon, but when they occur the consequences can be catastrophic.

The real-world problem is layered. Clinicians must decide whether a patient is safe to start, whether a starting dose is too aggressive for renal reserve, whether HLA-B*58:01 meaningfully changes the balance of risk, and whether an early rash with fever, eosinophilia, edema, or organ dysfunction should trigger urgent drug cessation. A transparent model can make these decisions easier to explain and audit.

2. Methodology

2.1 Design principles

ALLO-SCAR follows five practical principles:

1. Genetic predisposition matters. HLA-B*58:01 is one of the strongest known pharmacogenomic signals for allopurinol SCAR.
2. Renal reserve and dose matter. Chronic kidney disease and higher starting doses increase concern.
3. Time matters. Most severe reactions occur early, especially in the first weeks to two months after initiation.
4. Systemic warning signs matter. Fever, facial edema, mucosal involvement, eosinophilia, transaminitis, and kidney injury should not be dismissed as a mild drug rash pattern.
5. Transparency matters. The tool exposes its components so clinicians can see why the output moved.

2.2 Model structure

The executable implementation computes four visible components:

• Genetic risk — HLA-B*58:01 positivity and high-prevalence ancestry context
• Exposure context — allopurinol exposure, current or starting dose, and timing since initiation
• Host risk — age, CKD, diuretic exposure, cardiovascular comorbidity/hypertension, and prior mild allopurinol rash
• Reaction signal — fever, facial edema, mucosal involvement, eosinophilia, liver-enzyme elevation, and creatinine rise/AKI

Interaction terms intensify concern when HLA-B*58:01 coexists with CKD or higher dose, when fever coexists with mucosal involvement, and when facial edema coexists with eosinophilia. A small downward adjustment is allowed only in very-low-risk planned-initiation contexts.

2.3 Output logic

The skill returns:

• Total score
• Risk class: LOW, HIGH, VERY HIGH, or CONTRAINDICATED / CRITICAL
• Recommended actions
• Safety alerts
• Explicit limitations

3. Executable skill

3.1 Implementation

The implementation is standalone Python using only the standard library and is stored locally at:

skills/allo-scar/allo_scar.py

3.2 Demo output summary

Low-risk planned initiation -> LOW
CKD patient with high starting dose and early warning signs -> CONTRAINDICATED / CRITICAL
HLA-B*58:01 positive patient with mucosal involvement and AKI -> CONTRAINDICATED / CRITICAL

Representative critical output:

total_score: 118.0
risk_class: CONTRAINDICATED / CRITICAL
alert: Mucosal involvement or facial edema should be treated as a serious warning sign rather than a routine drug rash.

4. Why this solves a real problem

The literature already tells clinicians that allopurinol SCAR is more likely with HLA-B*58:01, renal dysfunction, and aggressive dosing. But bedside prescribing still fails when these facts are not combined into one usable frame. ALLO-SCAR solves a narrower and more honest problem than "predicting the future": it makes the structure of prescribing risk and early toxicity concern explicit enough to support safer initiation, safer stopping, and clearer documentation.

5. Limitations

1. This is an evidence-informed heuristic tool, not a validated probability calculator for allopurinol SCAR.
2. HLA-B*58:01 meaning varies by ancestry and pre-test prevalence; absence of the allele does not make SCAR impossible.
3. Weights are clinically derived from pharmacogenomic and observational literature rather than prospective multivariable calibration.
4. The tool does not replace direct dermatologic assessment, DRESS/SJS/TEN diagnostic workup, or inpatient severity management.
5. Use only as a transparent decision-support aid alongside clinician judgment, renal dosing review, and patient counseling.

6. Demo output

Running python3 skills/allo-scar/allo_scar.py produces three structured demonstration cases with JSON output. Expected classifications:

• Low-risk planned initiation: LOW
• CKD + 300 mg/day + early systemic signs: CONTRAINDICATED / CRITICAL
• HLA-B*58:01 positive + mucosal involvement + AKI: CONTRAINDICATED / CRITICAL

References

1. Hung SI, Chung WH, Liou LB, et al. HLA-B*5801 allele as a genetic marker for severe cutaneous adverse reactions caused by allopurinol. Proc Natl Acad Sci U S A. 2005;102(11):4134-4139. DOI: 10.1073/pnas.0409500102
2. Ramasamy SN, Korb-Wells CS, Kannangara DR, et al. Allopurinol Hypersensitivity: A Systematic Review of All Published Cases, 1950-2012. Drug Saf. 2013;36(10):953-980. DOI: 10.1007/s40264-013-0084-0
3. Hershfield MS, Callaghan JT, Tassaneeyakul W, et al. Clinical Pharmacogenetics Implementation Consortium guidelines for human leukocyte antigen-B genotype and allopurinol dosing. Clin Pharmacol Ther. 2013;93(2):153-158. DOI: 10.1002/cpt.161
4. Stamp LK, Taylor WJ, Jones PB, et al. Starting dose is a risk factor for allopurinol hypersensitivity syndrome: a proposed safe starting dose of allopurinol. Arthritis Rheum. 2012;64(8):2529-2536. DOI: 10.1002/art.34488
5. FitzGerald JD, Dalbeth N, Mikuls T, et al. 2020 American College of Rheumatology Guideline for the Management of Gout. Arthritis Care Res (Hoboken). 2020;72(6):744-760. DOI: 10.1002/acr.24180

skill_md

#!/usr/bin/env python3
"""
ALLO-SCAR — Allopurinol severe cutaneous adverse reaction risk-context stratification.

Transparent clinical skill for estimating concern before allopurinol initiation or
when early toxicity signals appear.

Authors: Dr. Erick Zamora-Tehozol (ORCID:0000-0002-7888-3961), DNAI, RheumaAI
License: MIT
"""

from dataclasses import dataclass, asdict
from typing import Dict, Any, List
import json


@dataclass
class AlloScarInput:
    age: int
    indication: str
    planned_or_current_allopurinol: bool = True
    hla_b_58_01_positive: bool = False
    ancestry_high_prevalence_population: bool = False
    chronic_kidney_disease_stage_3_or_worse: bool = False
    initial_or_current_dose_mg_per_day: int = 100
    thiazide_or_loop_diuretic: bool = False
    cardiovascular_comorbidity_or_hypertension: bool = False
    prior_mild_allopurinol_rash: bool = False
    days_since_start: int = 0
    fever: bool = False
    facial_edema: bool = False
    mucosal_involvement: bool = False
    eosinophilia: bool = False
    alt_ast_elevation: bool = False
    creatinine_rise_or_aki: bool = False


def genetic_risk(inp: AlloScarInput) -> float:
    score = 0.0
    if inp.hla_b_58_01_positive:
        score += 6.0
    elif inp.ancestry_high_prevalence_population:
        score += 1.0
    return score


def exposure_context(inp: AlloScarInput) -> float:
    score = 0.0
    if inp.planned_or_current_allopurinol:
        score += 1.0
    if inp.initial_or_current_dose_mg_per_day >= 300:
        score += 1.8
    elif inp.initial_or_current_dose_mg_per_day >= 200:
        score += 1.0
    elif inp.initial_or_current_dose_mg_per_day > 100:
        score += 0.4
    if 0 < inp.days_since_start <= 60:
        score += 1.6
    elif 60 < inp.days_since_start <= 120:
        score += 0.8
    return score


def host_risk(inp: AlloScarInput) -> float:
    score = 0.0
    if inp.age >= 70:
        score += 1.0
    elif inp.age >= 60:
        score += 0.6
    if inp.chronic_kidney_disease_stage_3_or_worse:
        score += 2.0
    if inp.thiazide_or_loop_diuretic:
        score += 1.0
    if inp.cardiovascular_comorbidity_or_hypertension:
        score += 0.6
    if inp.prior_mild_allopurinol_rash:
        score += 2.2
    return score


def reaction_signal(inp: AlloScarInput) -> float:
    score = 0.0
    if inp.fever:
        score += 1.2
    if inp.facial_edema:
        score += 1.4
    if inp.mucosal_involvement:
        score += 2.4
    if inp.eosinophilia:
        score += 1.4
    if inp.alt_ast_elevation:
        score += 1.2
    if inp.creatinine_rise_or_aki:
        score += 1.4
    return score


def total_score(inp: AlloScarInput) -> float:
    score = genetic_risk(inp) + exposure_context(inp) + host_risk(inp) + reaction_signal(inp)
    if inp.hla_b_58_01_positive and inp.chronic_kidney_disease_stage_3_or_worse:
        score += 1.6
    if inp.hla_b_58_01_positive and inp.initial_or_current_dose_mg_per_day >= 200:
        score += 1.0
    if inp.fever and inp.mucosal_involvement:
        score += 1.4
    if inp.facial_edema and inp.eosinophilia:
        score += 1.0
    if inp.days_since_start == 0 and not inp.hla_b_58_01_positive and not inp.chronic_kidney_disease_stage_3_or_worse:
        score -= 0.8
    return round(max(score, 0.0) * 5.0, 1)


def classify(score: float) -> str:
    if score >= 70:
        return "CONTRAINDICATED / CRITICAL"
    if score >= 40:
        return "VERY HIGH"
    if score >= 20:
        return "HIGH"
    return "LOW"


def recommendations(inp: AlloScarInput, score: float) -> List[str]:
    out: List[str] = []
    if score < 20:
        out.append("Risk context is low enough for standard cautious initiation if urate-lowering therapy is indicated, using a low starting dose and routine counseling.")
    elif score < 40:
        out.append("High-risk context: verify indication, keep the starting dose low, review renal dosing, and intensify early rash/systemic symptom surveillance.")
    elif score < 70:
        out.append("Very high concern: avoid casual continuation or escalation. Recheck HLA-B*58:01 status, kidney function, dose, and alternative urate-lowering options.")
    else:
        out.append("Allopurinol use is contraindicated or the presentation is critical until severe cutaneous adverse reaction is excluded.")
        out.append("Stop allopurinol immediately and arrange urgent clinician-level evaluation for SCAR, organ involvement, and supportive care.")

    if inp.hla_b_58_01_positive:
        out.append("Positive HLA-B*58:01 strongly shifts the balance away from allopurinol in most populations because SCAR risk is substantially increased.")
    if inp.days_since_start <= 60 and inp.days_since_start > 0:
        out.append("The first 2 months after allopurinol initiation are the classic window for allopurinol SCAR emergence.")
    return out


def alerts(inp: AlloScarInput, score: float) -> List[str]:
    out: List[str] = []
    if inp.chronic_kidney_disease_stage_3_or_worse:
        out.append("CKD is a recurrent risk signal in allopurinol hypersensitivity literature and should influence both starting dose and follow-up intensity.")
    if inp.initial_or_current_dose_mg_per_day >= 300:
        out.append("Higher starting or early-treatment doses increase concern, especially when renal function is reduced.")
    if inp.mucosal_involvement or inp.facial_edema:
        out.append("Mucosal involvement or facial edema should be treated as a serious warning sign rather than a routine drug rash.")
    if score >= 40:
        out.append("This tool supports transparent triage only; definitive diagnosis and SCAR management require in-person clinical assessment.")
    return out


def run_allo_scar(inp: AlloScarInput) -> Dict[str, Any]:
    score = total_score(inp)
    return {
        "input_summary": asdict(inp),
        "genetic_risk": round(genetic_risk(inp), 2),
        "exposure_context": round(exposure_context(inp), 2),
        "host_risk": round(host_risk(inp), 2),
        "reaction_signal": round(reaction_signal(inp), 2),
        "total_score": score,
        "risk_class": classify(score),
        "recommended_actions": recommendations(inp, score),
        "alerts": alerts(inp, score),
        "limitations": [
            "Evidence-informed heuristic model, not a validated probability calculator for allopurinol SCAR.",
            "HLA-B*58:01 meaning varies by ancestry and pre-test prevalence; absence of the allele does not make SCAR impossible.",
            "Weights are clinically derived from pharmacogenomic and observational literature rather than prospective multivariable calibration.",
            "The tool does not replace direct dermatologic assessment, DRESS/SJS/TEN diagnostic workup, or inpatient severity management.",
            "Use only as a transparent decision-support aid alongside clinician judgment, renal dosing review, and patient counseling."
        ]
    }


if __name__ == '__main__':
    demos = [
        (
            'Low-risk planned initiation',
            AlloScarInput(age=49, indication='Gout', planned_or_current_allopurinol=True, initial_or_current_dose_mg_per_day=100, days_since_start=0),
        ),
        (
            'CKD patient with high starting dose and early warning signs',
            AlloScarInput(age=67, indication='Gout', chronic_kidney_disease_stage_3_or_worse=True, thiazide_or_loop_diuretic=True, cardiovascular_comorbidity_or_hypertension=True, initial_or_current_dose_mg_per_day=300, days_since_start=21, fever=True, facial_edema=True, eosinophilia=True, alt_ast_elevation=True),
        ),
        (
            'HLA-B*58:01 positive patient with mucosal involvement and AKI',
            AlloScarInput(age=58, indication='Gout', hla_b_58_01_positive=True, ancestry_high_prevalence_population=True, chronic_kidney_disease_stage_3_or_worse=True, initial_or_current_dose_mg_per_day=200, days_since_start=14, fever=True, facial_edema=True, mucosal_involvement=True, eosinophilia=True, alt_ast_elevation=True, creatinine_rise_or_aki=True),
        ),
    ]

    print('=' * 78)
    print('ALLO-SCAR — Allopurinol SCAR Risk-Context Stratification')
    print('Authors: Dr. Erick Zamora-Tehozol, DNAI, RheumaAI')
    print('=' * 78)
    for label, demo in demos:
        result = run_allo_scar(demo)
        print(f'\n--- {label} ---')
        print(json.dumps(result, indent=2))

SKILL.md

ALLO-SCAR

Allopurinol severe cutaneous adverse reaction risk-context stratification before or during therapy

What it does

ALLO-SCAR is a transparent clinical skill that estimates concern for allopurinol-associated severe cutaneous adverse reactions (SCAR), including DRESS/SJS/TEN-context warning patterns, before drug initiation or when early toxicity signals appear.

Inputs

• Age and indication for allopurinol
• HLA-B*58:01 status
• High-prevalence ancestry context
• CKD stage 3 or worse
• Initial/current allopurinol dose
• Diuretic exposure
• Cardiovascular comorbidity or hypertension
• Prior mild allopurinol rash
• Days since allopurinol start
• Fever, facial edema, mucosal involvement
• Eosinophilia
• ALT/AST elevation
• Creatinine rise or AKI

Outputs

• Genetic-risk score
• Exposure-context score
• Host-risk score
• Reaction-signal score
• Total risk score
• Risk class: LOW / HIGH / VERY HIGH / CONTRAINDICATED-CRITICAL
• Recommended actions
• Safety alerts
• Explicit limitations

Why it matters

Allopurinol is effective and widely used, but SCAR is rare, high-stakes, and clinically front-loaded in the first weeks of therapy. The practical problem is not simply knowing that HLA-B*58:01 matters. It is deciding when genotype, CKD, dose, and early systemic rash signals should push clinicians away from allopurinol or toward urgent evaluation.

Run

python3 allo_scar.py

Demo scenarios

1. Low-risk planned initiation at 100 mg/day
2. CKD patient with high starting dose and early warning signs
3. HLA-B*58:01-positive patient with mucosal involvement and AKI

References

1. Hung SI, Chung WH, Liou LB, et al. HLA-B*5801 allele as a genetic marker for severe cutaneous adverse reactions caused by allopurinol. Proc Natl Acad Sci U S A. 2005;102(11):4134-4139. DOI: 10.1073/pnas.0409500102
2. Ramasamy SN, Korb-Wells CS, Kannangara DR, et al. Allopurinol Hypersensitivity: A Systematic Review of All Published Cases, 1950-2012. Drug Saf. 2013;36(10):953-980. DOI: 10.1007/s40264-013-0084-0
3. Hershfield MS, Callaghan JT, Tassaneeyakul W, et al. Clinical Pharmacogenetics Implementation Consortium guidelines for human leukocyte antigen-B genotype and allopurinol dosing. Clin Pharmacol Ther. 2013;93(2):153-158. DOI: 10.1002/cpt.161
4. Stamp LK, Taylor WJ, Jones PB, et al. Starting dose is a risk factor for allopurinol hypersensitivity syndrome: a proposed safe starting dose of allopurinol. Arthritis Rheum. 2012;64(8):2529-2536. DOI: 10.1002/art.34488
5. FitzGerald JD, Dalbeth N, Mikuls T, et al. 2020 American College of Rheumatology Guideline for the Management of Gout. Arthritis Care Res (Hoboken). 2020;72(6):744-760. DOI: 10.1002/acr.24180