Statistical Methodology
Overtime, NJS has provided below expert statistical consulting for clients.
Oncology Trial Design
- Adaptive Group Sequential Design with application to seamless phase IIB and Phase III combination study
- Adaptive Multi-stage Adaptive Design toxicity monitoring critical value
- Adaptive Randomization
- Sample size re-estimation
- Bayesian Hierarchical Model
- IPE Method, RPSFT and IPCW
- Pick Winner Design (Biomarker Guided)
- Oncology Trial Design Simulator (sample size/power calculation and simulation)
- 3×3 Design Probabilities simulation and Bayesian CRM modeling/mTPI/BOIN
- Comparing dose-response shapes, e.g., linear, logistic, Emax, sigmoid Emax, exponential, etc
- Modified MSPRT for post-approval safety monitoring study
Trial Monitoring Tool
- Event Trial duration Projection (parametric, non-parametric, bootstrapping)
- Advanced Survival Analysis Models:
- Multiple methods for estimating true OS effect from treatment crossover data (Robin’s RPSFT, 2-dimension RPSFT, IPCW, etc.)
- Bootstrapping projecting Future Efficacy Distribution of P-values and Hazard ratios
- Conditional and Predictive Power Calculation
- Efficacy P-value vs. Time, hazard ratio vs. time (Landmark Analysis)
- Survival Model Distribution Validation and Testing
- Nonproportional hazard survival model
- PFS and OS joint distribution modeling.
Quality of Life Analysis for Oncology trials
- Fit pattern mixture model Qol data with non-ignoble missing (missing not at random, MNAR) due to dropout.
- Use the worst rank method to handle missing QoL scores due to informed dropouts as an alternative to the pattern mixture model.
Other
- Koch randomization based nonparametric methods for time to failure outcomes in multi-center trials
- Multiplicity adjusted confidence intervals for many to one comparison
- Overall multiplicity adjusted p-values
- Bayesian Statistical Prediction for Accelerated Submission (Device)
- Expectation–maximization (EM) algorithms
- Bayesian Classification and regression tree methods
- Statistical comparison using Virtual-Matched Historical Data
- Control using Virtual-Matched Historical Data
- Sequential Parallel Comparison Design (SPCD)
- Randomized Withdrawal Design
- Randomized Delayed Start Design