THIO-SAFE

Clinical Problem

Azathioprine remains useful across autoimmune disease, but preventable myelotoxicity still occurs when genotype data, baseline cytopenias, interacting drugs, and monitoring plans are not integrated at the bedside. THIO-SAFE makes those factors explicit before treatment starts or when toxicity risk is being reassessed.

Methodology

THIO-SAFE is a transparent 10-domain weighted score built around pharmacogenetic liability (TPMT and NUDT15), marrow reserve, drug interactions, planned dose intensity, comorbidity, and monitoring readiness. Each domain contributes a visible weighted component to a 0-100 composite score. The implementation runs 5,000 Monte Carlo simulations to generate an approximate 95% interval.

Executable Python code

#!/usr/bin/env python3
"""
THIO-SAFE: Thiopurine myelotoxicity risk stratification before or during
azathioprine therapy in rheumatic and autoimmune disease.

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

Clinical purpose:
Estimate near-term risk of clinically relevant thiopurine myelotoxicity when
azathioprine is being considered or monitored in autoimmune disease. The score
supports genotype review, safer starting-dose selection, CBC monitoring plans,
and medication reconciliation.

This score does NOT replace pharmacogenetic prescribing guidelines, direct CBC
review, infection workup, or specialist judgment.

Key references:
- Relling MV, Gardner EE, Sandborn WJ, et al. Clin Pharmacol Ther. 2011.
  DOI: 10.1038/clpt.2010.320
- Yang SK, Hong M, Baek J, et al. Nat Genet. 2014.
  DOI: 10.1038/ng.3060
- Moriyama T, Nishii R, Perez-Andreu V, et al. Nat Genet. 2016.
  DOI: 10.1038/ng.3508
"""

from __future__ import annotations

import random
from dataclasses import dataclass, field
from typing import List


@dataclass
class ThioPatient:
    tpmt_status: str = "normal"  # normal, intermediate, poor, unknown
    nudt15_status: str = "normal"  # normal, intermediate, poor, unknown
    baseline_wbc: float = 6.5  # x10^9/L
    egfr: int = 90
    allopurinol_or_febuxostat: bool = False
    planned_azathioprine_mg_per_kg: float = 1.5
    chronic_liver_disease: bool = False
    prior_drug_related_cytopenia: bool = False
    concomitant_marrow_suppressant: bool = False
    cbc_monitoring_in_place: bool = True


@dataclass
class DomainScore:
    name: str
    score: float
    weight: float
    weighted: float
    detail: str


@dataclass
class ThioSafeResult:
    composite_score: float
    risk_category: str
    recommendation: str
    monitoring_comment: str
    ci_lower: float
    ci_upper: float
    domains: List[dict]
    notes: List[str] = field(default_factory=list)


WEIGHTS = {
    "tpmt": 0.26,
    "nudt15": 0.24,
    "baseline_wbc": 0.12,
    "renal": 0.06,
    "xanthine_oxidase": 0.12,
    "dose": 0.08,
    "liver": 0.04,
    "cytopenia_history": 0.04,
    "marrow_suppressant": 0.08,
    "monitoring": 0.04,
}


def score_tpmt(status: str):
    status = status.lower().strip()
    mapping = {
        "normal": (0, "TPMT normal metabolizer or normal activity"),
        "intermediate": (58, "TPMT intermediate metabolizer: reduced thiopurine tolerance"),
        "poor": (96, "TPMT poor metabolizer: severe myelotoxicity concern"),
        "unknown": (24, "TPMT status unknown before thiopurine decision"),
    }
    return mapping.get(status, (24, f"Unrecognized TPMT status '{status}', treated as unknown"))


def score_nudt15(status: str):
    status = status.lower().strip()
    mapping = {
        "normal": (0, "NUDT15 normal metabolizer or normal activity"),
        "intermediate": (70, "NUDT15 intermediate metabolizer: meaningful leukopenia risk"),
        "poor": (98, "NUDT15 poor metabolizer: extreme intolerance risk"),
        "unknown": (26, "NUDT15 status unknown before thiopurine decision"),
    }
    return mapping.get(status, (26, f"Unrecognized NUDT15 status '{status}', treated as unknown"))


def score_baseline_wbc(wbc: float):
    if wbc >= 5.0:
        return 0, f"Baseline WBC {wbc:.1f} x10^9/L"
    if wbc >= 4.0:
        return 24, f"Baseline WBC {wbc:.1f} x10^9/L"
    if wbc >= 3.0:
        return 56, f"Baseline WBC {wbc:.1f} x10^9/L"
    return 84, f"Baseline WBC {wbc:.1f} x10^9/L"


def score_renal(egfr: int):
    if egfr >= 60:
        return 0, f"eGFR {egfr} mL/min/1.73m²"
    if egfr >= 45:
        return 18, f"eGFR {egfr} mL/min/1.73m²"
    if egfr >= 30:
        return 34, f"eGFR {egfr} mL/min/1.73m²"
    return 52, f"eGFR {egfr} mL/min/1.73m²"


def score_binary(flag: bool, if_true: int, yes: str, no: str):
    return (if_true, yes) if flag else (0, no)


def score_dose(mg_per_kg: float):
    if mg_per_kg < 1.0:
        return 0, f"Planned azathioprine dose {mg_per_kg:.2f} mg/kg/day"
    if mg_per_kg < 1.5:
        return 12, f"Planned azathioprine dose {mg_per_kg:.2f} mg/kg/day"
    if mg_per_kg < 2.0:
        return 26, f"Planned azathioprine dose {mg_per_kg:.2f} mg/kg/day"
    return 44, f"Planned azathioprine dose {mg_per_kg:.2f} mg/kg/day"


def score_monitoring(cbc_monitoring_in_place: bool):
    if cbc_monitoring_in_place:
        return -28, "CBC monitoring plan is documented"
    return 22, "No CBC monitoring plan documented"


def compute_thio_safe(patient: ThioPatient, n_simulations: int = 5000, seed: int = 42) -> ThioSafeResult:
    items = [
        ("tpmt", score_tpmt(patient.tpmt_status)),
        ("nudt15", score_nudt15(patient.nudt15_status)),
        ("baseline_wbc", score_baseline_wbc(patient.baseline_wbc)),
        ("renal", score_renal(patient.egfr)),
        ("xanthine_oxidase", score_binary(patient.allopurinol_or_febuxostat, 92, "Allopurinol/febuxostat co-therapy detected", "No xanthine oxidase inhibitor co-therapy")),
        ("dose", score_dose(patient.planned_azathioprine_mg_per_kg)),
        ("liver", score_binary(patient.chronic_liver_disease, 24, "Chronic liver disease present", "No chronic liver disease")),
        ("cytopenia_history", score_binary(patient.prior_drug_related_cytopenia, 42, "Prior drug-related cytopenia", "No prior drug-related cytopenia")),
        ("marrow_suppressant", score_binary(patient.concomitant_marrow_suppressant, 38, "Concomitant marrow-suppressive therapy present", "No concomitant marrow-suppressive therapy")),
        ("monitoring", score_monitoring(patient.cbc_monitoring_in_place)),
    ]

    domains: List[DomainScore] = []
    composite = 0.0
    for name, (raw, detail) in items:
        weight = WEIGHTS[name]
        weighted = raw * weight
        composite += weighted
        domains.append(DomainScore(name, round(raw, 1), weight, round(weighted, 1), detail))

    composite = round(max(0.0, min(composite, 100.0)), 1)

    rng = random.Random(seed)
    sims: List[float] = []
    for _ in range(n_simulations):
        noisy = ThioPatient(
            tpmt_status=patient.tpmt_status,
            nudt15_status=patient.nudt15_status,
            baseline_wbc=max(1.0, round(patient.baseline_wbc + rng.gauss(0, 0.25), 2)),
            egfr=max(10, int(round(patient.egfr + rng.gauss(0, 3)))) ,
            allopurinol_or_febuxostat=patient.allopurinol_or_febuxostat,
            planned_azathioprine_mg_per_kg=max(0.1, round(patient.planned_azathioprine_mg_per_kg + rng.gauss(0, 0.08), 2)),
            chronic_liver_disease=patient.chronic_liver_disease,
            prior_drug_related_cytopenia=patient.prior_drug_related_cytopenia,
            concomitant_marrow_suppressant=patient.concomitant_marrow_suppressant,
            cbc_monitoring_in_place=patient.cbc_monitoring_in_place if rng.random() > 0.01 else not patient.cbc_monitoring_in_place,
        )
        noisy_items = [
            ("tpmt", score_tpmt(noisy.tpmt_status)),
            ("nudt15", score_nudt15(noisy.nudt15_status)),
            ("baseline_wbc", score_baseline_wbc(noisy.baseline_wbc)),
            ("renal", score_renal(noisy.egfr)),
            ("xanthine_oxidase", score_binary(noisy.allopurinol_or_febuxostat, 92, "", "")),
            ("dose", score_dose(noisy.planned_azathioprine_mg_per_kg)),
            ("liver", score_binary(noisy.chronic_liver_disease, 24, "", "")),
            ("cytopenia_history", score_binary(noisy.prior_drug_related_cytopenia, 42, "", "")),
            ("marrow_suppressant", score_binary(noisy.concomitant_marrow_suppressant, 38, "", "")),
            ("monitoring", score_monitoring(noisy.cbc_monitoring_in_place)),
        ]
        total = sum(score * WEIGHTS[name] for name, (score, _) in noisy_items)
        sims.append(max(0.0, min(total, 100.0)))

    sims.sort()
    ci_lower = round(sims[int(0.025 * n_simulations)], 1)
    ci_upper = round(sims[int(0.975 * n_simulations)], 1)

    if composite < 10:
        category = "LOW"
        recommendation = "Standard azathioprine initiation is usually acceptable if clinically indicated; continue routine genotype-aware dosing and early CBC follow-up."
        monitoring_comment = "Check CBC and liver tests early after initiation and reinforce infection/bleeding precautions."
    elif composite < 25:
        category = "INTERMEDIATE"
        recommendation = "Use conservative dosing, verify genotype status if missing, and shorten the interval to first CBC review."
        monitoring_comment = "Reconcile interacting drugs and confirm that baseline counts are stable before escalation."
    elif composite < 40:
        category = "HIGH"
        recommendation = "Delay or reduce thiopurine exposure until genotype, interacting medications, and monitoring gaps are addressed."
        monitoring_comment = "This profile justifies intensified laboratory surveillance and low threshold to stop the drug if counts fall."
    else:
        category = "VERY HIGH"
        recommendation = "Avoid standard-dose azathioprine. Resolve interacting drug exposure, obtain pharmacogenetic guidance, and consider alternative therapy or drastic dose reduction only with specialist oversight."
        monitoring_comment = "Severe leukopenia risk is substantial enough that proactive prevention is safer than reactive monitoring alone."

    notes = [
        "This is a transparent bedside risk-stratification score, not an absolute probability model.",
        "TPMT and NUDT15 results should be interpreted using local pharmacogenetic reporting standards.",
        "Xanthine oxidase inhibitor co-therapy is weighted heavily because standard azathioprine doses can become dangerous in this setting.",
    ]
    if patient.tpmt_status == "unknown" or patient.nudt15_status == "unknown":
        notes.append("Unknown genotype status does not equal normal risk; it means preventable uncertainty remains.")

    return ThioSafeResult(
        composite_score=composite,
        risk_category=category,
        recommendation=recommendation,
        monitoring_comment=monitoring_comment,
        ci_lower=ci_lower,
        ci_upper=ci_upper,
        domains=[d.__dict__ for d in domains],
        notes=notes,
    )


def print_case(label: str, patient: ThioPatient):
    result = compute_thio_safe(patient)
    print(f"\n=== {label} ===")
    print(f"Composite score: {result.composite_score}/100")
    print(f"Risk category: {result.risk_category}")
    print(f"95% CI: [{result.ci_lower}, {result.ci_upper}]")
    print(f"Recommendation: {result.recommendation}")
    print(f"Monitoring: {result.monitoring_comment}")
    if result.notes:
        print("Notes:")
        for note in result.notes:
            print(f"- {note}")
    print("Top domains:")
    for domain in sorted(result.domains, key=lambda d: d['weighted'], reverse=True)[:5]:
        print(f"- {domain['name']}: raw {domain['score']} × w {domain['weight']} = {domain['weighted']} ({domain['detail']})")


if __name__ == "__main__":
    print_case(
        "Scenario 1 — Genotype-normal patient starting conservative azathioprine",
        ThioPatient(
            tpmt_status="normal",
            nudt15_status="normal",
            baseline_wbc=6.4,
            egfr=92,
            allopurinol_or_febuxostat=False,
            planned_azathioprine_mg_per_kg=1.2,
            chronic_liver_disease=False,
            prior_drug_related_cytopenia=False,
            concomitant_marrow_suppressant=False,
            cbc_monitoring_in_place=True,
        ),
    )
    print_case(
        "Scenario 2 — Intermediate-metabolizer profile with borderline counts and marrow-suppressive co-therapy",
        ThioPatient(
            tpmt_status="intermediate",
            nudt15_status="normal",
            baseline_wbc=3.8,
            egfr=58,
            allopurinol_or_febuxostat=False,
            planned_azathioprine_mg_per_kg=1.7,
            chronic_liver_disease=False,
            prior_drug_related_cytopenia=False,
            concomitant_marrow_suppressant=True,
            cbc_monitoring_in_place=True,
        ),
    )
    print_case(
        "Scenario 3 — NUDT15-deficient patient with allopurinol exposure and no monitoring plan",
        ThioPatient(
            tpmt_status="normal",
            nudt15_status="poor",
            baseline_wbc=3.1,
            egfr=42,
            allopurinol_or_febuxostat=True,
            planned_azathioprine_mg_per_kg=2.0,
            chronic_liver_disease=True,
            prior_drug_related_cytopenia=True,
            concomitant_marrow_suppressant=False,
            cbc_monitoring_in_place=False,
        ),
    )

Demo output

=== Scenario 1 — Genotype-normal patient starting conservative azathioprine ===
Composite score: 0.0/100
Risk category: LOW

=== Scenario 2 — Intermediate-metabolizer profile with borderline counts and marrow-suppressive co-therapy ===
Composite score: 26.9/100
Risk category: HIGH

=== Scenario 3 — NUDT15-deficient patient with allopurinol exposure and no monitoring plan ===
Composite score: 50.4/100
Risk category: VERY HIGH

Why this score exists

In outpatient autoimmune care, azathioprine may be started while genotype reports, medication interactions, and CBC timing are reviewed by different people at different times. THIO-SAFE packages those risks into one auditable view to improve safer prescribing and documentation.

Limitations

• Evidence-informed weighted model, not a prospectively validated regression model
• Relative risk-stratification only; not an absolute probability engine
• Does not diagnose infection, marrow failure, or other non-thiopurine causes of cytopenia
• Must not replace CPIC guidance or urgent clinical review when severe cytopenia is suspected

References

1. Relling MV, Gardner EE, Sandborn WJ, et al. Clinical Pharmacogenetics Implementation Consortium guidelines for thiopurine methyltransferase genotype and thiopurine dosing. Clin Pharmacol Ther. 2011;89(3):387-391. DOI: 10.1038/clpt.2010.320
2. Yang SK, Hong M, Baek J, et al. A common missense variant in NUDT15 confers susceptibility to thiopurine-induced leukopenia. Nat Genet. 2014;46(9):1017-1020. DOI: 10.1038/ng.3060
3. Moriyama T, Nishii R, Perez-Andreu V, et al. NUDT15 polymorphisms alter thiopurine metabolism and hematopoietic toxicity. Nat Genet. 2016;48(4):367-373. DOI: 10.1038/ng.3508