OSPred tool

OSPred tool predicts overall survival hazard ratios from early clinical endpoints to quantify correlations between progression-free survival (PFS) measures and overall survival (OS) in non-small-cell lung cancer (NSCLC) clinical trial data.

Key Features:

• Predictive analysis: Predicts HR OS from user-specified early endpoint measures including HR PFS and odds ratios of PFS at 4 and 6 months (OR PFS4 and OR PFS6).
• Machine intelligence–based modeling: Leverages machine intelligence approaches combined with clinical trial data to map surrogate endpoints to OS.
• Comparative analysis by mechanism of action (MoA): Enables comparisons of investigational drugs against historical NSCLC studies stratified by specific mechanisms of action.
• Probability density output: Produces probability density charts that provide point predictions and confidence intervals for HR OS.
• Trial-level dataset integration: Operates on a trial-level dataset derived from published phase III reports to support data-driven inference.

Scientific Applications:

• Early go/no-go decision support: Informs early development decisions by estimating likely OS outcomes from surrogate endpoints in oncology drug development.
• Surrogate endpoint evaluation: Assesses the relationship between PFS-based metrics and OS to evaluate surrogate validity in NSCLC trials.
• Contextualizing new results: Places investigational drug outcomes in context with historical phase III NSCLC data and MoA-specific trends.
• Statistical interpretation: Provides point estimates and confidence intervals to quantify uncertainty in predicted HR OS for clinical planning.

Methodology:

Implemented in R Shiny using ggplot2 for visualization, metafor and boot for statistical analysis, mvtnorm for multivariate normal distributions, and dplyr for data manipulation; based on a trial-level dataset comprising 81 phase III reports covering 35 anticancer drugs with 156 observations in NSCLC and incorporating machine intelligence approaches with clinical data.