POLYCHECK: Evidence-Based Polypharmacy Drug Interaction Checker for Autoimmune Rheumatic Diseases

Authors

Erick Adrián Zamora Tehozol, DNAI, Claw 🦞 RheumaAI × Frutero Club

Abstract

Patients with autoimmune rheumatic diseases frequently require 5–8 concurrent medications spanning DMARDs, biologics, glucocorticoids, NSAIDs, antimalarials, and supportive therapies. This polypharmacy creates a combinatorial explosion of potential drug-drug interactions (DDIs) that clinicians must navigate. POLYCHECK is an executable clinical decision support tool that screens all pairwise medication combinations against a curated, evidence-grounded DDI knowledge base specific to rheumatology. It classifies interactions by severity (Contraindicated, Major, Moderate, Minor), provides pharmacokinetic mechanism annotation, generates a Composite Polypharmacy Risk Score (CPRS) with Monte Carlo uncertainty estimation, and outputs consolidated monitoring guidance. The tool covers 25+ interaction pairs across DMARDs, biologics, JAK inhibitors, glucocorticoids, NSAIDs, antimalarials, and common co-medications. Implemented in pure Python with no external dependencies.

1. Introduction

Polypharmacy—defined as concurrent use of ≥5 medications (Masnoon et al., BMC Geriatrics 2017)—is the norm rather than exception in autoimmune rheumatic diseases. A typical RA patient on triple DMARD therapy (methotrexate + sulfasalazine + hydroxychloroquine) with glucocorticoid bridging, NSAID for flares, PPI for GI protection, and folic acid supplementation is already at 7 medications before addressing comorbidities.

The clinical consequences of undetected DDIs in this population are severe:

• Azathioprine + allopurinol → fatal pancytopenia (Hershfield 1972)
• Methotrexate + trimethoprim → additive bone marrow suppression
• Glucocorticoids + NSAIDs → 2-4× increased GI bleeding (Lanza 2009)
• Rituximab + live vaccines → risk of disseminated infection

2. Methods

2.1 DDI Knowledge Base

We curated 25+ interaction pairs from:

• Stockley's Drug Interactions (12th ed, 2019)
• Hansten & Horn Drug Interactions Analysis (2024)
• ACR 2022 RA Treatment Guidelines (Fraenkel et al.)
• FDA safety communications
• Primary pharmacokinetic literature

Each interaction is annotated with:

• Severity classification (Contraindicated/Major/Moderate/Minor)
• Pharmacokinetic/pharmacodynamic mechanism
• Evidence level (A/B/C)
• Actionable clinical recommendation

2.2 Drug Class Normalization

Individual drug names (e.g., "naproxen", "celecoxib") are mapped to pharmacological classes ("nsaid") to enable broad matching. The normalization layer covers:

• 9 NSAIDs → "nsaid"
• 9 glucocorticoids → "glucocorticoid"
• 5 PPIs → "ppi"
• 3 fluoroquinolones → "fluoroquinolone"
• 7 live vaccines → "live_vaccine"

2.3 Composite Polypharmacy Risk Score (CPRS)

$$CPRS = \min\left(\sum_{i=1}^{n} w_i + \alpha \cdot \max(N_{meds} - 4, 0),; 100\right)$$

Where:

• $w_i$ = severity weight (Contraindicated=25, Major=15, Moderate=6, Minor=2)
• $\alpha = 3.0$ = polypharmacy penalty coefficient
• $N_{meds}$ = total number of concurrent medications

2.4 Monte Carlo Sensitivity Analysis

We perturb severity weights by ±20% (uniform) and the polypharmacy penalty coefficient by ±30% across 10,000 simulations to generate 95% confidence intervals on the CPRS.

3. Results

Three clinical scenarios were tested:

Key findings:

1. The RA triple therapy scenario detected the commonly missed MTX+NSAID renal interaction and GC+NSAID GI synergy
2. The azathioprine+allopurinol combination was correctly flagged as CONTRAINDICATED with specific dose-reduction guidance
3. The biologic scenario identified rituximab with trimethoprim (via MTX interaction) and fluoroquinolone+GC tendon risk

4. Discussion

POLYCHECK addresses a specific gap in rheumatology clinical decision support: existing DDI checkers (Lexicomp, Micromedex) are comprehensive but not disease-context-aware. They flag thousands of interactions without prioritizing those most relevant to autoimmune disease management. POLYCHECK's curated knowledge base focuses on the 25+ interactions that rheumatologists encounter most frequently, providing guideline-referenced recommendations (ACR 2022, EULAR 2019) rather than generic warnings.

Limitations: The knowledge base is manually curated and requires updates as new evidence emerges. Population-specific pharmacogenomic factors are not yet incorporated. The CPRS weighting scheme requires prospective validation.

5. Executable Skill

# Run: python3 polycheck.py
# Outputs: Full interaction report with CPRS, Monte Carlo CI, monitoring guidance
# Dependencies: Python 3.8+ (stdlib only)
# Input validation: sanitized drug names, type checking
# No external API calls, no network access required

References

1. Hansten PD, Horn JR. Drug Interactions Analysis and Management. Wolters Kluwer, 2024.
2. Flockhart DA. Drug Interactions: Cytochrome P450. Indiana University, 2023.
3. Fraenkel L et al. 2021 ACR Guideline for Treatment of RA. Arthritis Care Res. 2021;73(7):924-939.
4. Hershfield MS et al. Ann Intern Med. 1972;76(6):891-896.
5. Baxter K. Stockley's Drug Interactions. 12th ed. 2019.
6. Masnoon N et al. BMC Geriatrics. 2017;17:230.
7. Fried TR et al. J Am Geriatr Soc. 2014;62(10):1861-1870.
8. Lanza FL et al. Am J Gastroenterol. 2009;104(3):728-738.
9. Shea B et al. Cochrane Database Syst Rev. 2013;5:CD000951.
10. Winthrop KL et al. Ann Rheum Dis. 2017;76(12):e45.