Platform

The dosing platform built
for oncology PK complexity

DoseMind models each patient's pharmacokinetics individually, updates exposure forecasts after every lab draw, and generates dose recommendations your investigators can review, approve, and submit.

Core Engine

Two-compartment Bayesian estimation, not a population nomogram

Most TDM software applies a population-average formula adjusted for body weight or creatinine. DoseMind fits a two-compartment PK model to each patient's observed concentration-time data using MAP Bayesian estimation.

After each sample, the model updates Cl, Vd, and ke for that individual. When a patient's renal function shifts mid-cycle, the model adjusts immediately - not at the next protocol-scheduled assessment.

The result: AUC target attainment rates in our pilot cohort improved from 61% to 97.3% versus standard Calvert-formula dosing.

DoseMind PK curve fitting interface
How It Works

From enrollment to dose recommendation

01

Protocol Configuration

Your clinical pharmacologist enters the drug, target AUC or Cmax band, PK model parameters from the investigator's brochure, and allowable dose increments. This takes 40-90 minutes per protocol and is validated before any patient is enrolled.

02

Patient Onboarding

Baseline demographics - weight, height, serum creatinine, ALT/AST, and prior chemotherapy exposure - are imported from the EDC via FHIR R4. Population PK priors are automatically assigned. The first-cycle starting dose is computed and queued for investigator approval.

03

Cycle-by-Cycle Updating

As PK samples post in the LIMS, DoseMind retrieves concentration values via HL7 ORU messages, runs a Bayesian update, and publishes a revised AUC forecast within 38ms. The investigator portal reflects the updated estimate in real time with no manual data entry required.

04

Dose Decision and Signature

The treating physician opens the next-cycle recommendation, reviews the AUC projection with credible interval, the toxicity probability score, and alternative scenario comparisons. They approve or modify the dose and provide an electronic signature that satisfies 21 CFR Part 11 requirements.

Platform Capabilities

Every component serves a clinical function

Bayesian PK Engine

Implements a two-compartment model with first-order absorption and linear elimination. Model parameters can be fixed (dense sampling Phase I) or estimated via MAP Bayesian inference (sparse sampling Phase II/III).

Supported model structures: one-compartment IV, two-compartment IV, one-compartment oral, two-compartment oral. Additional structures available via NONMEM import.

LIMS Integration

Native connectors for LabVantage, STARLIMS, and LabWare via HL7 v2.5 ORU messages and REST API. Fallback manual entry with dual-entry verification for sites without electronic LIMS.

Concentration data is validated against reporting range limits before inclusion in the model fit. Out-of-range values are flagged and require investigator acknowledgment.

Toxicity Risk Scoring

Uses a logistic regression model trained on published oncology trial datasets to estimate probability of grade 3/4 hematologic and hepatic toxicity. Input features: projected AUC, prior-cycle nadir ANC, and hepatic function indices.

Risk scores are displayed as probability (e.g., "21% chance of grade 3+ ANC nadir"), not as traffic-light categories. Clinicians see the number, not an abstracted color.

Regulatory Documentation

Auto-generates dose rationale sections formatted for IND annual reports, NDA/BLA submission appendices, and DSMB briefing documents. Output in DOCX and PDF with tracked-changes support.

Audit trail exports satisfy FDA 21 CFR Part 11 and EU Annex 11 requirements. Each record includes user identity, timestamp, action type, and reason for change.

Multi-Site Data Architecture

Population PK model parameters are aggregated centrally; patient-level data remain within each site's secure tenant. Cross-site pooling for population model updates requires explicit data sharing agreement flags in the protocol configuration.

Role-based access control at site, investigator, CRC, and sponsor-monitor levels. Single sign-on via SAML 2.0 or OIDC.

Simulation and Scenario Analysis

Before committing to a dose, investigators can simulate the exposure trajectory at +10%, -10%, or custom dose levels. The simulation panel displays projected AUC, time above minimum effective concentration, and Cmax at each alternative.

Simulation results are logged but do not constitute a dose decision until an approval action is taken.

Technical Specifications

Built for clinical-grade performance

Performance

Bayesian MAP estimation completes in under 38ms median latency per patient update on standard cloud infrastructure. 99.4% availability SLA with transparent status page. Concurrent user capacity: 500+ simultaneous sessions without degradation.

Data Security

Data encrypted in transit (TLS 1.3) and at rest (AES-256). PHI de-identification workflow per HIPAA Safe Harbor. SOC 2 Type II audit completed Q4 2024. Penetration test by Synack Red Team, December 2024.

Interoperability

FHIR R4 resource support: Patient, Observation, MedicationRequest, MedicationAdministration, DiagnosticReport. EDC integrations: Medidata Rave, Veeva Vault CDMS, Oracle Clinical One. HL7 v2.5 ORU for lab results.

Regulatory Compliance

Validated software per GAMP 5 Category 4 (configurable software). Validation package includes IQ/OQ/PQ documentation, traceability matrix, and risk assessment. FDA 21 CFR Part 11 and EU Annex 11 compliant.

Run a protocol validation pilot

Send us your protocol's PK parameters and target exposure range. We will configure DoseMind and run a retrospective simulation on your historical patient data before any prospective use.

Talk to the Team