# E79 index from routine biomarkers stratifies rapid MCI-to-Alzheimer's conversion: AUC 0.891 in 607 ADNI patients

- Published: 2026-11-13 09:27 UTC
- Section: Life & Health › Medicine
- Tags: E79 Alzheimer index, MCI to Alzheimer's conversion, rapid progression prediction, spectral decomposition biomarkers, eigenmodes, ADNI cohort, negative predictive value, logistic regression, PCA comparison, anti-amyloid therapy candidacy
- URL: https://realtech.media/news/rt-26d1745f238e984f4e9e8d18_e79_index_from/

**Researchers built a logistic regression index on two eigenmodes of ten routine biomarkers. In the ADNI cohort it separated 24-month converters with 95.4% negative predictive value; external validation is pending.**

![E79 index from routine biomarkers stratifies rapid MCI-to-Alzheimer's conversion: AUC 0.891 in 607 ADNI patients](https://realtech.media/media/2026/11/5621ee8f3eedce08.jpg)
*Bobjgalindo, Wikimedia Commons, CC BY-SA 4.0*

A team led by John D. Mayfield, on behalf of the Alzheimer's Disease Neuroimaging Initiative (ADNI), described the E79 Alzheimer index, a tool that stratifies patients with mild cognitive impairment (MCI) by risk of rapid progression to Alzheimer's disease within 24 months. The work was published in Alzheimer's Research & Therapy and used data from 607 MCI participants. The index reached an AUC of 0.891 (bootstrap 95% CI 0.863–0.916) and uses only baseline clinical data, without genetic testing or advanced imaging [1].

Identifying rapid progressors matters for anti-amyloid therapy candidacy, monitoring intensity and prognosis. According to the authors, existing machine learning approaches report high AUC for any-conversion prediction but conflate slow and rapid progressors, which limits their clinical use [1].

## Spectral decomposition of ten biomarkers

The input panel contained ten routine biomarkers: cognitive, functional, structural MRI and cerebrospinal fluid measures. The authors computed a 10×10 Pearson correlation matrix and applied eigendecomposition to it. This revealed two disease-sensitive eigenmodes: E7, reflecting cognitive-functional instability, and E9, reflecting structural-pathological burden [1].

Individual biomarker profiles were projected onto the MCI-derived eigenvectors to obtain subject-level E7 and E9 scores. These scores were the inputs to a logistic regression model [1].

## Validation and comparison

On a stratified hold-out test set of 122 participants the AUC was 0.862; 5-fold cross-validation gave 0.880. At a fixed 90% sensitivity, the model showed 95.4% negative predictive value and 67.4% specificity [1].

Between the high-risk and low-risk groups the authors report an 80-fold risk enrichment: 47.9% of the high-risk group converted within 24 months versus 0.6% of the low-risk group [1].

Using DeLong's method with Bonferroni correction, E79 was statistically equivalent to logistic regression on all raw biomarkers (AUC 0.862 vs. 0.859, p = 1.000). It was significantly better than PCA compression of the same dimensionality (0.862 vs. 0.840, p = 0.004). Unlike PCA, the approach keeps one-to-one attribution to individual biomarkers [1].

The main limitation is derivation in a single ADNI cohort. External validation in ADNI4 is underway as follow-up matures through 2026–2027 [1].

## Sources

1. [Alzheimer s Research & Therapy](https://doi.org/10.1186/s13195-026-02126-0)
2. [bioengineer.org](https://bioengineer.org/multimodal-spectral-index-accurately-stratifies-rapid-conversion-to-alzheimers-disease)

---

Source: RealTech, realtech.media — https://realtech.media/news/rt-26d1745f238e984f4e9e8d18_e79_index_from/