DNAI · STORM · Stochastic Therapeutic Outcome & Risk Model

Treatment decisions in rheumatology
carry real probability. Model it.

Biologic selection, escalation timing, refractory disease -- each decision is a probability estimate. STORM makes that estimate explicit, grounded in Dr. Erick Zamora's outcome data.

Request clinical access Read the brief
Proof of concept · In clinical evaluation · Dr. Erick Zamora
STORM · Modelado Terapéutico · Dr. Zamora En uso

Caso: AR refractaria

Miguel A. · 61 años · M · AR severa Falla secundaria a adalimumab (3 años). DAS28 5.8 actual. Comorbilidades: HTA, EPOC leve.

Historia de tratamiento

MTX 20 mg + SSZ 2 g → MTX + ADA (3 años) → Falla secundaria. Sin rituximab previo.

Laboratorio

DAS28

5.8

ACPA

+ 340 UI

FR

+ 128 UI

EPOC

Leve (FEV1 72%)

Análisis de opciones

Recomendado Rituximab + MTX Respuesta ACPA+ estimada 68-74% a 6 meses. EPOC leve: sin contraindicación. Menor riesgo infeccioso vs IL-6. Primera elección en seropositividad alta.
Segunda línea Tocilizumab IV + MTX Respuesta estimada 58-65% a 6 meses. Monitoreo lipídico requerido. EPOC: precaución con infecciones respiratorias. Considerar si RTX falla.
Considerar después Tofacitinib + MTX JAKi: eficaz en falla anti-TNF, pero EPOC leve aumenta riesgo infeccioso. Reservar para falla de biológicos.

Protocolo RTX sugerido

RTX 1000 mg IV días 1 y 15 + MTX 20 mg/sem mantenimiento

Pre-medicación: metilprednisolona 100 mg IV + paracetamol + antihistamínico

Vacunas al día antes de primera dosis: neumococo, influenza

DAS28 + labs a 16 semanas para evaluar respuesta

"Every treatment decision is already a probability estimate in the physician's head. STORM makes that estimate explicit and auditable."

Dr. Erick Zamora - Rheumatologist · Primary operator

6 of 7 products
in the DNAI
proof stack

Biologic selection in refractory disease is a high-stakes probability estimate with no clear protocol.

Decision complexity

Biologic class switching after failure has no single right answer.

Anti-TNF failure, ACPA status, comorbidities, prior exposure -- the interaction of these variables exceeds what unaided clinical judgment can compute in a consultation.

Comorbidity weighting

Pulmonary, cardiac, and infectious risk changes the calculus entirely.

A patient with COPD, CKD, or prior infections has a different risk-benefit calculation for each biologic class. STORM weights comorbidities into the outcome estimate.

Outcome uncertainty

Clinicians estimate response probability intuitively -- without showing their work.

STORM makes the estimate explicit and grounded in Dr. Erick's outcome data. The physician still decides; the model shows the probability behind the decision.

How STORM works.

01

Enter the patient's treatment history and comorbidities.

Prior biologics, response history, current disease activity, serologic status, comorbidities. STORM maps the patient against Dr. Erick's outcome data for similar clinical profiles.

02

Model outcome probability for each treatment option.

Bayesian response estimates at 3 and 6 months for each candidate biologic or DMARD strategy. Comorbidity-adjusted risk for infections, cardiovascular events, and treatment discontinuation.

03

Structured risk-benefit output for the treatment decision.

Response probability per option, ranked recommendation with clinical rationale, monitoring protocol for the chosen strategy, escalation criteria if response is inadequate at 3 months.

Modeling capabilities

What STORM models today.

Current capabilities

  • Bayesian response probability for anti-TNF, IL-6, IL-17, IL-23, RTX, and JAK inhibitors
  • ACPA and RF seropositivity weighting in biologic selection
  • Comorbidity-adjusted infection and cardiovascular risk estimation
  • Treatment escalation timing from disease activity trajectory
  • Refractory disease pathway modeling after first and second biologic failure
  • MTX comedication optimization for biologic response
  • Monitoring schedule generation for selected biologic
  • Response assessment criteria at 3 and 6 months per EULAR

Access status

Proof of concept in active clinical evaluation.

STORM is in clinical evaluation by Dr. Erick Zamora. Access available to rheumatologists managing patients with refractory disease or complex biologic selection decisions.

Request clinical access Read the DNAI brief first

Part of the DNAI proof stack

STORM LES AI RheumaAI RheumaScore ORVS Beach Science / clawRxiv BiobadamexAI

Common questions.

Is this a replacement for clinical judgment?

No. STORM makes the probability estimate that is already implicit in the physician's decision explicit and grounded in outcome data. The physician decides. STORM shows the probability behind the decision and flags comorbidity-specific risks the calculation should include.

What data does it use for probability estimates?

Dr. Erick Zamora's clinical outcome data from his rheumatology practice, combined with published EULAR and ACR response rates for each biologic class. The estimates are specific to the clinical profile, not generic trial averages.

Which treatment classes does it cover?

Anti-TNF (adalimumab, etanercept, infliximab, certolizumab, golimumab), IL-6 inhibitors (tocilizumab, sarilumab), anti-CD20 (rituximab), IL-17 inhibitors, JAK inhibitors (tofacitinib, baricitinib, upadacitinib), and abatacept. Coverage expands with Dr. Erick's outcome data.

Does it handle lupus as well as RA?

The current model is optimized for rheumatoid arthritis biologic selection. Lupus treatment modeling is a planned expansion using the SLEDAI-based outcome data in Dr. Erick's practice. LES AI handles lupus treatment protocol currently.