Registry Forge — Drug repurposing analysis

Synthetic demonstration cohort · 2026-05-12 21:51 UTC
Privacy: Patient identifiers pseudonymized · ages banded · dates given as time-from-cohort-entry only · k-anonymity threshold 2 applied to medication-level summaries.

Cohort overview

80Patients in motor neuron disease cohort 0Deaths observed 2Medications meeting min-5 threshold 1ATC classes represented
Cohort defined by ICD-10-CM G12.21 / G12.20 / G12.22 / G12.23 / G12.29 / ICD-9-CM 335.20-29 / SNOMED 86044005 / Mondo MONDO:0004976 / MONDO:0019056 in problems[]. Exposure criteria adapted from Reimer 2026.

ALS clinical-interest medications · always shown, regardless of min-exposed threshold · k-anonymized at k=2

The five medications below have particular relevance to ALS clinical care and research: the three FDA-approved disease-modifying therapies, the FDA-approved-for-PBA combination commonly used off-label in ALS care, and the briefly-approved combination withdrawn after confirmatory trial failure. Presence and year-range are reported regardless of whether the medication meets the standard min-exposed threshold; counts are k-anonymized (values below k= 2 reported as “< 2”).

Medication FDA status Cohort presence Unique patients Year range Records
Riluzole
Rilutek, Tiglutik, Exservan · ATC N07XX02
FDA-approved
FDA-approved for ALS (1995)
Not observed00
Edaravone
Radicava, Radicava ORS · ATC N07XX14
FDA-approved
FDA-approved for ALS (2017)
Not observed00
Tofersen
Qalsody · ATC N07XX17
FDA-approved
FDA-approved for SOD1 ALS (2023, accelerated)
Not observed00
Dextromethorphan-quinidine
Nuedexta · ATC N07XX59
Off-label-of-interest
FDA-approved for PBA; off-label-of-interest in ALS
Not observed00
Sodium phenylbutyrate-taurursodiol
Relyvrio, Albrioza · ATC —
Withdrawn
FDA-approved for ALS 2022; withdrawn 2024 after PHOENIX
Not observed00

Medication summary · grouped by ATC class · k=≥2 for "n_exposed"

Other nervous system drugs (motor neuron disease)

Medicationn exposedn unexposeddeaths (exp)deaths (unexp)median surv (exp)median surv (unexp)direction
edaravone
N07XX14
6740015.4y9.9y↑ exposed longer
riluzole
N07XX02
6515009.9y9.9y

Methodology and limitations

This module produces the cohort assembly and exposure assignment described in Reimer et al. Lancet Digital Health 2026. It does not run the propensity score matching or the Cox proportional hazards analysis; those steps require dedicated tools (lifelines, statsmodels) and per-cohort tuning of caliper widths, immortal-time bias correction, and covariate balance assessment. The cohort dataset emitted by this module (drug_repurposing_cohort.csv) is the input to those tools.

Limitations of this analysis at any stage: small-cohort statistical power, inadequate covariate adjustment for known and unknown confounders, indication bias (a patient prescribed a PDE5 inhibitor may by virtue of that indication be healthier than average), immortal-time bias if exposure timing is mishandled, accuracy of EHR-coded motor-neuron-disease diagnosis, and lack of dose-response information in the bundle. Treat outputs as hypothesis-generating; prospective randomized trials remain the gold standard.

Cure ID intake export: produces one row per exposed (patient, medication) for submission to the FDA / NCATS-NIH / Critical Path Institute CURE Drug Repurposing Collaboratory Treatment Registry. Banded fields and pseudonymized identifiers comply with the registry's PII-free expectations.