Spending Scarce Confirmatory PET Measurements: Target-Aligned Validation in A4/LEARN

Abstract:Anti-amyloid therapies and blood-based biomarkers are changing Alzheimer disease workups into a two-stage measurement workflow: screen broadly with cheaper information, then spend scarce confirmatory amyloid measurements where they support the decision that will be reported. Amyloid positron-emission tomography (PET) remains one such protocol measurement for amyloid burden, but PET slots, trial budgets, and payer-facing evidence packages are finite. This paper asks a deliberately operational question: when is simple transparent PET validation enough, and when is a fitted residual-uncertainty score worth the added complexity? For a weighted protocol target, the first-order value of validating subject i is the product of target influence and residual protocol uncertainty. Generic uncertainty sampling uses only the second factor and can spend PET measurements on subjects that are hard to predict but weak for the scientific, clinical, or commercial claim. We apply this rule to the A4/LEARN PET archive, treating observed PET as a design laboratory for scarce-confirmation studies. For the primary APOE4 carrier versus non-carrier contrast in Centiloid 24-or-higher PET positivity, simple APOE4-balanced validation recovers nearly all of the target-specific gain: at PET budget 200, the confidence-interval width ratio relative to random validation is 0.923 for APOE4 balancing and 0.914 for target-specific scoring, while generic uncertainty sampling is 0.980. Other targets behave differently: target-specific scoring gives larger gains for an age-slope analysis and for cutoff-indexed PET positivity. The practical message is simple: spend scarce protocol measurements according to the claim being validated, not only according to prediction uncertainty.

Applications (stat.AP); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Machine Learning (stat.ML)

arXiv:2608.22223 [stat.AP]

(or arXiv:2608.22223v1 [stat.AP] for this version)

https://doi.org/10.48550/arXiv.2608.22223 Focus to learn more arXiv-issued DOI via DataCite (pending registration)

README_CODE_RESULTS.md

README_ancillary.txt

docs/a4_analysis_ready_notes.md

docs/results_shareability_manifest_20260608.csv

requirements-core-20260608.txt

results/a4_pilot_validation_design_summary.csv

results/a4_table1_analysis_population.csv

results/a4_table2_contrast_design.csv

results/a4_threshold_process_multiplier_band_budget200.csv

results/figures/a4_apoe_exclusion_sensitivity_main.pdf

results/figures/a4_apoe_exclusion_sensitivity_main.png

results/figures/a4_apoe_exclusion_sensitivity_main.svg

results/figures/a4_contrast_sweep_punchline.pdf

results/figures/a4_contrast_sweep_punchline.png

results/figures/a4_contrast_sweep_punchline.svg

results/figures/a4_pilot_validation_design_main.pdf

results/figures/a4_pilot_validation_design_main.png

results/figures/a4_pilot_validation_design_main.svg

results/figures/a4_pilot_validation_design_main_cropped.pdf

results/figures/a4_sensitivity_main.pdf

results/figures/a4_sensitivity_main.png

results/figures/a4_sensitivity_main.svg

results/figures/a4_surrogate_roc.pdf

results/figures/a4_surrogate_roc.png

results/figures/a4_surrogate_roc.svg

results/figures/a4_threshold_process_main.pdf

results/figures/a4_threshold_process_main.png

results/figures/a4_threshold_process_main.svg

results/figures/a4_validation_design_ci_width.pdf

results/figures/a4_validation_design_ci_width.png

results/figures/a4_validation_design_ci_width.svg

results/figures/a4_validation_design_coverage.pdf

results/figures/a4_validation_design_coverage.png

results/figures/a4_validation_design_coverage.svg

results/figures/a4_validation_design_main.pdf

results/figures/a4_validation_design_main.png

results/figures/a4_validation_design_main.svg

results/figures/a4_validation_design_rmse.pdf

results/figures/a4_validation_design_rmse.png

results/figures/a4_validation_design_rmse.svg

results/figures/figure1_a4_main_punchline.pdf

results/figures/figure1_a4_main_punchline.png

results/figures/figure1_a4_main_punchline.svg

results/figures/figure1_a4_main_punchline_cropped.pdf

results/figures/figure1_recorded_probability_methodology.pdf

results/figures/figure1_recorded_probability_methodology.png

results/figures/figure1_recorded_probability_methodology.svg

results/figures/figure1_recorded_probability_methodology_cropped.pdf

results/figures/figure2_a4_pet_cutoff_sensitivity_main.pdf

results/figures/figure2_a4_pet_cutoff_sensitivity_main.png

results/figures/figure2_a4_pet_cutoff_sensitivity_main.svg

results/figures/figure2_a4_pet_cutoff_sensitivity_main_cropped.pdf

results/figures/figure2_a4_threshold_process_multiplier.pdf

results/figures/figure2_a4_threshold_process_multiplier.png

results/figures/figure2_a4_threshold_process_multiplier.svg

results/figures/figure2_a4_threshold_process_multiplier_cropped.pdf

results/figures/method_simulation_uncertainty_failure.pdf

results/figures/method_simulation_uncertainty_failure.png

results/figures/method_simulation_uncertainty_failure.svg

results/figures/method_simulation_uncertainty_failure_cropped.pdf

results/figures/pet_validation_reframed_design_ratios.pdf

results/figures/pet_validation_reframed_design_ratios.png

results/figures/pet_validation_reframed_design_ratios.svg

results/method_simulation_bias.csv

results/method_simulation_design_summary.csv

scripts/12_build_a4_analysis_ready.py

scripts/13_a4_surrogate_screen.py

scripts/14_a4_validation_design_prototype.py

scripts/15_a4_contrast_sweep.py

scripts/16_a4_pilot_validation_design.py

scripts/17_a4_sensitivity_sweep.py

scripts/18_a4_threshold_process.py

scripts/19_a4_tables_and_final_figures.py

scripts/30_method_simulations.py

scripts/33_a4_apoe_exclusion_sensitivity.py

scripts/34_a4_feature_timing_table.py

scripts/36_a4_empirical_audit_gates.py

scripts/37_a4_model_agnostic_sensitivity.py

scripts/38_editorial_reframing_empirical_summaries.py

scripts/39_editorial_reframing_p1_fix_summary.py

scripts/40_a4_apoe4_stratified_comparator.py

scripts/41_a4_broader_balanced_comparator_budget200.py

scripts/42_recorded_probability_methodology_figure.py

scripts/44_reframed_design_figure.py

scripts/manuscript_style.py

scripts/run_a4_main_outputs.sh

web_supplement_pet_validation.pdf

(82 additional files not shown) You must enabled JavaScript to view entire file list.

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.