TrialMetrics

• 

  Trial Metrics Dashboard

Inspiration

Clinical operations teams use fragmented tools — Veeva/Medidata for enrollment, SAS/R for statistics, and spreadsheets for budgets. There's no unified view that combines statistical analysis with forecasting and budget tracking. Teams make decisions in silos, leading to delayed trials and underpowered studies.

As a CFA Charterholder with 10+ years in quantitative research, I saw an opportunity to bring portfolio management discipline to clinical research.

What it does

TrialMetrics is a clinical trial analytics dashboard that unifies:

• Statistical Power Analysis — Non-central t-distribution calculations with interactive power curves
• Enrollment Forecasting — OLS regression with HAC-robust standard errors (Newey-West)
• Budget Tracking — Burn rate, runway, and efficiency metrics
• AI-Powered Summaries — GPT-4.1 generates executive briefings with risk assessment and recommendations

How I built it

• Frontend: Streamlit with Plotly visualizations
• Data: Live connection to ClinicalTrials.gov API
• Statistics: SciPy and Statsmodels for power analysis and forecasting
• AI: OpenAI GPT-4.1-nano for natural language summaries

Challenges I ran into

• ClinicalTrials.gov doesn't provide historical enrollment data, so I modeled enrollment trajectories based on trial timelines
• Balancing statistical rigor with user-friendly visualizations
• Ensuring the AI summaries are actionable, not generic

What I learned

• How to apply financial risk models to clinical trial assessment
• HAC-robust standard errors for time series with autocorrelation
• Integrating AI to synthesize complex metrics into executive summaries

What's next for TrialMetrics

• Integration with sponsor CTMS (Veeva, Medidata, Oracle)
• Multi-trial portfolio views
• Predictive alerts for enrollment slowdowns

Built With

• numpy
• openai
• pandas
• plotly
• python
• scipy
• statsmodels
• streamlit