LEF-LUNG: Transparent Leflunomide-Associated Interstitial Lung Toxicity Risk Stratification in Rheumatic and Autoimmune Disease

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

Abstract

Leflunomide-associated interstitial lung toxicity is uncommon but clinically important because presentations can be abrupt, severe, and difficult to separate from rheumatoid arthritis-associated interstitial lung disease or pulmonary infection. The bedside problem is not merely whether the adverse event is rare. It is whether a susceptible patient with baseline interstitial lung disease, abnormal chest imaging, concurrent methotrexate exposure, or an early high-intensity loading strategy is being exposed without enough pulmonary caution. We present LEF-LUNG, an executable Python skill for transparent clinical risk stratification before or during leflunomide therapy in rheumatic and autoimmune disease. The model integrates pre-existing interstitial lung disease or pulmonary fibrosis, abnormal baseline imaging, age, smoking or chronic lung disease, low body weight, high-risk population context, planned or current leflunomide exposure, loading-dose strategy, early treatment window, prior methotrexate lung-toxicity context, concurrent methotrexate, new dyspnea, cough, fever, hypoxemia, and unresolved infection uncertainty. Outputs include visible component scores, categorical risk classes, safety alerts, recommended actions, and explicit limitations. In demonstration scenarios, a patient without pulmonary vulnerability is LOW risk, an older patient with baseline ILD and planned loading dose is VERY HIGH risk, and a symptomatic patient with dyspnea, fever, and desaturation during recent leflunomide exposure is VERY HIGH risk. LEF-LUNG is intended as an auditable triage tool rather than a diagnostic engine and does not replace imaging, pulmonary consultation, or urgent in-person assessment.

Keywords: leflunomide, interstitial lung disease, pneumonitis, rheumatoid arthritis, drug toxicity, pulmonary safety, clinical decision support, DeSci, RheumaAI

1. Clinical problem

Leflunomide remains an important disease-modifying option in rheumatoid arthritis and related inflammatory disease, particularly when methotrexate is insufficient or poorly tolerated. Yet lung toxicity has remained one of its most feared rare adverse effects. Reported cases suggest that risk is not evenly distributed. Patients with pre-existing interstitial lung disease, abnormal baseline imaging, early treatment exposure, loading-dose strategies, and certain population-level backgrounds appear more vulnerable to severe pulmonary injury.

This creates a recurring clinical dilemma. Respiratory symptoms after DMARD initiation are often nonspecific. Dyspnea, cough, fever, and hypoxemia may reflect infection, progression of rheumatoid lung disease, methotrexate-related toxicity, or leflunomide-associated injury. The cost of over-calling toxicity is treatment interruption; the cost of under-calling it may be respiratory failure.

LEF-LUNG was designed to make that risk structure explicit and auditable.

2. Methodology

2.1 Design principles

The score follows five defensible clinical principles:

1. Baseline pulmonary vulnerability matters most. Pre-existing ILD or pulmonary fibrosis should strongly influence any decision around leflunomide exposure.
2. Early high-intensity exposure matters. Loading doses and events arising within the first months of treatment deserve heightened concern.
3. Symptoms should be weighted by severity. Dyspnea and especially hypoxemia are stronger warning signals than isolated cough.
4. Competing diagnoses matter. Infection remains a major confounder and must be actively excluded.
5. This is a triage tool, not a diagnosis. The aim is to support safer escalation and monitoring, not to label a radiographic pattern definitively.

2.2 Model structure

The implementation computes four visible components:

• Host vulnerability — pre-existing ILD, baseline imaging abnormality, chronic lung disease or smoking, age, low body weight, and high-risk population context
• Exposure intensity — leflunomide exposure itself, loading-dose strategy, concurrent methotrexate, prior methotrexate lung-toxicity context, and early treatment timing
• Presentation severity — new dyspnea, cough, fever, and hypoxemia/desaturation
• Diagnostic uncertainty — unresolved pulmonary infection context

Interaction terms intensify concern when pre-existing ILD coexists with active leflunomide exposure, when loading-dose exposure occurs in the early treatment window, and when dyspnea and hypoxemia cluster together.

2.3 Output logic

The skill returns:

• Total score
• Risk class: LOW, INTERMEDIATE, HIGH, VERY HIGH
• 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/lef-lung/lef_lung.py

3.2 Demo output summary

RA patient without lung disease starting leflunomide without loading dose -> LOW
Older RA patient with baseline ILD and planned loading dose -> VERY HIGH
Recent leflunomide exposure with dyspnea, cough, fever, and desaturation -> VERY HIGH

Representative high-risk output:

total_score: 128.5
risk_class: VERY HIGH
alert: Hypoxemia or desaturation suggests potentially severe pulmonary involvement and lowers the threshold for urgent imaging.

4. Why this solves a real problem

Clinicians already know that leflunomide can cause lung injury, but that knowledge often remains diffuse and retrospective. The practical failure happens when risk factors are recognized one at a time but never integrated into a concrete triage threshold. LEF-LUNG turns that fragmented warning pattern into a visible framework that supports safer DMARD selection, clearer patient counseling, and earlier escalation to imaging or pulmonary review.

5. Limitations

1. This is an evidence-informed heuristic tool, not a validated incidence calculator.
2. It does not diagnose ILD subtype or distinguish drug toxicity from all competing pulmonary processes.
3. Drug toxicity, rheumatoid ILD progression, and pulmonary infection may overlap clinically and radiographically.
4. Population-level risk estimates vary across cohorts, especially where baseline ILD prevalence differs.
5. Definitive management requires individualized clinical assessment, imaging, oxygenation assessment, and medication review.

References

1. Suissa S, Bernatsky S, Hudson M. Antirheumatic drug use and the risk of hospitalization for interstitial lung disease in rheumatoid arthritis. Arthritis Rheum. 2006;54(5):1435-1439. DOI: 10.1002/art.21799
2. Sawada T, Inokuma S, Sato T, et al. Leflunomide-induced interstitial lung disease: prevalence and risk factors in Japanese patients with rheumatoid arthritis. Rheumatology (Oxford). 2009;48(9):1069-1072. DOI: 10.1093/rheumatology/kep163
3. Ju JH, Kim SI, Lee JH, et al. Risk of interstitial lung disease in patients with rheumatoid arthritis treated with leflunomide. Mod Rheumatol. 2007;17(5):414-417. DOI: 10.1007/s10165-007-0597-z
4. Roubille C, Haraoui B. Interstitial lung diseases induced or exacerbated by DMARDs and biologic agents in rheumatoid arthritis: a systematic literature review. Semin Arthritis Rheum. 2014;43(5):613-626. DOI: 10.1016/j.semarthrit.2013.09.005
5. Fraenkel L, Bathon JM, England BR, et al. 2021 American College of Rheumatology Guideline for the Treatment of Rheumatoid Arthritis. Arthritis Rheumatol. 2021;73(7):1108-1123. DOI: 10.1002/art.41752