What the cognitive outcome analyses found

Introduction

A reproducible calculation does not demonstrate that an index predicts a clinical outcome. This report summarizes two analyses of a frozen exposure-to-affinity construct against different cognitive measures and asks how much the current UnifiedACB research release can claim. ¹

The distinction between an internal signal and an independent external test is central. A positive association in one survey can change when a later dataset uses a different outcome, medication ascertainment, or population. The analysis must also be read in light of the historical 22-drug freeze; the current evidence audit retains 20 archived calculations and was not prospectively validated by either survey. ¹

Methods

The earlier NHANES 2011–2014 analysis examined cognitive test performance using a historical 22-drug freeze and survey-weighted adjustment. The later independent MEPS 2024 test prespecified cognitive limitation among adults aged 60 years or older and used a frozen Final-E1 index. These datasets differ in outcome definition and medication ascertainment. Neither analysis is a prospective test of the current 20 archived calculations. ^{1, 2, 3}

The NHANES primary strict analysis and an inclusive sensitivity analysis were distinguished from a restricted complete-routed-list analysis. The MEPS test specified its primary outcome before interpretation and reported effects per one survey-weighted standard deviation of the frozen index. The reporting also separated primary and age-restricted sensitivities and documented incomplete person-level medication coverage. STROBE reporting principles motivate explicit selection, missing-data, and sensitivity descriptions. ^{1, 5}

Neither dataset supplied complete concurrent medication administration, formulation, and patient-level dose for the present research model. Adjusted associations were examined in observational data and cannot establish that changing a medicine would change an individual's cognition. Outcome definitions and missing medication information limit direct comparison of the coefficients between surveys. ^{1, 2, 3}

Results

The NHANES primary strict Final-E1 analysis reported β = −0.0696 standard deviations of cognition per 1 U (95% CI, −0.1140 to −0.0253; P = .0031). A restricted complete-routed-list analysis (n = 1,935) did not show a clear association (strict P = .660). The earlier freeze is not the current 20-drug audit. ¹

The NHANES inclusive sensitivity estimate was β = −0.0769 (95% CI, −0.1120 to −0.0417). In the restricted complete-routed-list group, the inclusive analysis also lacked a clear association (P = .169). These sensitivity results show that exposure eligibility and list completeness materially affect interpretation of the earlier signal. ¹

The MEPS primary analysis reported an adjusted odds ratio of 1.009 per survey-weighted standard deviation of the frozen index (95% CI, 0.927–1.098; P = .835; n = 6,112). The age 65-and-older sensitivity was also null. Medication mapping covered 93.23% of non-equipment prescription rows as reproducibly mapped or named-staged, but person-level coverage remained incomplete. ¹

The age 65-and-older sensitivity produced an adjusted OR of 0.949 per standard deviation (95% CI, 0.849–1.061; P = .360). The mapping denominator included 119,434 non-equipment prescription rows. A row-level mapping rate should not be mistaken for complete exposure assessment for every participant. Secondary categorical or inclusive signals were not sufficiently stable to establish a cutoff. ¹

Discussion

The primary external result did not confirm the NHANES signal. The NHANES association is cross-sectional, sensitive to medication coverage and exclusions, and based on a different freeze. MEPS used another outcome and incomplete exposure ascertainment. Neither result establishes causality, an individual risk probability, superiority over established scales, or a validated clinical cutoff. The index was not retuned after the null MEPS result. ¹

A future dose- and time-aware study with actual medication administration and independent replication is needed before a clinical prediction claim can be considered. In the meantime, the website is an evidence-comparison resource. ^{1, 4}

The two studies should not be pooled as though they measured the same endpoint. A continuous standardized cognition result and a binary cognitive-limitation response have different interpretation, and their medication histories were collected differently. Unmeasured reasons for prescribing and baseline health may also confound an observed association. A null primary test does not mean all anticholinergic medicines are harmless; it means this frozen index did not demonstrate the proposed association with that specific external outcome under the specified analysis. ¹

The MEPS closeout reported stronger descriptive associations for several established scales in that dataset, but this comparison does not prove one scale's superiority across settings. A convincing next phase would freeze a version, assess dose and administration before an independently assessed outcome, quantify missingness, compare against established measures on common patients, and report null as well as positive findings. ^{1, 4}

References (AMA style)

1. Chaar M. Validation report: NHANES and MEPS evidence, frozen cohorts and limits. Unified ACB Tool; 2026. Published September 28, 2026. Source ↗
2. National Center for Health Statistics. NHANES 2013–2014: cognitive functioning (CFQ_H) data documentation. Centers for Disease Control and Prevention. Published 2017. Source ↗
3. Agency for Healthcare Research and Quality. MEPS HC-254A: 2024 prescribed medicines documentation. Published 2026. Source ↗
4. Chaar M. Step 20B MAR/delirium validation protocol. Unified ACB Tool; 2026. Published September 28, 2026. Source ↗
5. von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann Intern Med. 2007;147(8):573-577. doi:10.7326/0003-4819-147-8-200710160-00010 Source ↗