Validation of Prognostic Marker Tests: Statistical Lessons Learned From Regulatory Experience