GRAD: A Two-Stage Algorithm for Resolving Diagnostic Uncertainty in the Plasma p-tau217 Gray Zone

IntroductionPhosphorylated tau-217 (p-tau 217) is widely used as a plasma-based biomarker for Alzheimers Disease (AD) detection, demonstrating superior accuracy for detecting brain amyloid pathology. However, 30-50% of patients fall within an intermediate diagnostic "gray zone" where biomarker results are indeterminate, often decreasing physician confidence and requiring subsequent diagnostic workup. To address this, we developed a two-stage machine learning algorithm GRAD: Gatekeeper & Reflex for Alzheimers Disease to increase clinical confidence and reduce the AD health economic burden. MethodsWe initially analyzed 320 participants from the Alzheimers Disease Neuroimaging Initiative (ADNI) with plasma biomarkers and amyloid PET. We then built a two-stage machine learning classifier mimicking real clinical workflow where the stage 1 Gatekeeper used the gold-standard marker: p-tau217 with respective 25%/75% probability thresholds. The stage 2 Reflex step applied Random Forest multi-marker classification (p-tau 217, AB42/40, NFL, GFAP) for difficult-to-diagnose gray zone cases. To ensure statistical robustness, leave-one-out cross-validation with bootstrap confidence intervals was used. We externally validated the GRAD architecture on 1,644 A4 Study participants, with MRI enhancement analysis in 1,044 gray zone cases. To measure cost-effectiveness we compared our GRAD-staged testing to universal PET. ResultsThe models Gatekeeper resolved 55.6% of ADNI cases with 88.8% accuracy (NPV 91.8%, PPV 85.0%). The complete pipeline achieved AUC 0.867 (95% CI: 0.825-0.904), with 80.6% sensitivity, 80.0% specificity, LR+ 4.03, LR-0.24. For the difficult-to-diagnose gray zone cases, the Reflex machine learning model achieved AUC 0.755. In our A4 validation, the predictions correlated strongly with centiloid (r= 0.693). Expanding beyond plasma biomarkers, MRI integration improved gray zone classification from AUC 0.829 to 0.853 (p=0.014). The cost modeling analysis projected a 67% reduction in spending versus the current standard of universal PET. DiscussionOur clinically-staged diagnostic algorithm, GRAD, provides actionable classifications for the majority of patients while routing uncertain cases for additional workup. The GRAD framework offers a practical, cost-effective approach for implementing plasma biomarkers in clinical practice. Future iterations of this framework, with integration of novel biomarkers like MTBR-tau243 present a significant opportunity to alleviate the AD health-economic burden and eliminate expensive but unnecessary diagnostic measures. HighlightsO_LIGRAD: Two-stage "Gatekeeper + Reflex for Alzheimers Disease" algorithm resolves indeterminate plasma p-tau217 or gray zone patients with AUC of 0.755. C_LIO_LIOverall AUC of 0.867 (95% CI: 0.825-0.904) validated via leave-one-out cross-validation C_LIO_LIExternal validation in A4 Study demonstrates strong correlation with amyloid burden (r=0.693) C_LIO_LIMRI volumetric integration provides significant incremental value ({Delta}AUC=+0.025, p=0.014) C_LIO_LIProjected 67-71% cost reduction compared to universal PET screening C_LI Research in ContextO_ST_ABSLiterature ReviewC_ST_ABSWe searched PubMed, Google Scholar, and medRxiv databases for studies up to December 2025 that examined plasma p-tau217 diagnostic accuracy as well as "gray zone" management of patients. While several studies demonstrate area-under-the-curve (AUC) of >0.90, these studies largely compare cognitively normal individuals to those with established AD dementia with maximal biomarker separation [1-6]. The gray zone problem, affecting 30-50% of tested individuals, remains unaddressed in the vast majority of clinical implementation frameworks [7,8]. More recent work has established probability-based interpretation [9], but more cohesive algorithms for gray zone resolution through multi-marker integration remain rare if present. Furthermore, the health economic impacts of such resolution have not been fully established. InterpretationOur two-stage algorithm provides a workflow with clinical implementation potential, analogous to established laboratory medicine (i.e TSH with reflex free T4 testing). By first identifying high-confidence cases through univariate p-tau217 (55.6% resolution at 88.8% accuracy), and then applying multi-marker classification only to uncertain cases, we are able to achieve optimal resource utilization while simultaneously maintaining diagnostic accuracy. The finding that MRI usage provides statistically significant improvement ({Delta}AUC=+0.025) has practical implications given the fact that there is a reasonable level of MRI availability in clinical settings. Future DirectionsWhile this work accomplishes several key priorities, future work is required to validate them in diverse clinical populations. In addition, integration of other plasma markers (ex. MTBR-tau243), development of clinical decision support tools, reimbursement mechanisms, and longitudinal validation for treatment monitoring will be necessary to ensure the appropriate infrastructure exists to support providers and patients. Preliminary evidence suggests that %p-tau217 (the ratio of phosphorylated to total tau-217) and MTBR-tau243, a mass spectrometry-based marker of tau tangle pathology, may substantially improve gray zone classification by capturing complementary aspects of tau biology not reflected in absolute p-tau217 concentrations alone, which is a direction that future technical work should examine further.