Evaluation of a Biplanar Vector-Based Diagnostic Model for Subthreshold Disease: In Silico Stress Testing of Composite Drift Score Performance Under Stochastic, 2D, and 3D Conditions

Traditional threshold-based diagnostics often miss early disease progression, particularly in subthreshold states. In a prior model, we proposed the Composite Drift Score (CDS), a vector-based index that quantifies directional drift from physiological norms. This metric uses two components: the Magnitude-to-Noise Ratio (MNR) and a Directional Emphasis Multiplier (DEM). This study stress-tests the CDS model under a range of biologically plausible and extreme conditions using in silico synthetic data. A total of 2000 simulated subjects were generated with varying prevalence, pause patterns, intra-subject variability, and progression speed. Two versions of CDS were evaluated: one using Coefficient of Repeatability (CR) and another using Mahalanobis distance (MD). Model performance was quantified using the area under the curve for lead time (AUC-LT), reflecting both timing and detection advantage. Results showed that CDS outperformed threshold-based methods in early detection, particularly under fragmented or physiological-range noisy conditions. The Mahalanobis-based variant demonstrated higher resilience under stress. The framework was also expanded to a three-variable version, preserving directional behavior and performance characteristics, suggesting dimensional scalability. These in silico findings indicate that the CDS model may warrant further investigation in clinical datasets.