Table 3. Logistic Regression Analysis Assessing Factors Associated With Change in the Composite Management End Point.
| Effect | OR (95% CI) | P Valuea |
|---|---|---|
| Interceptb | 0.48 (0.28-0.80) | .006 |
| Main effects | ||
| Age (10 y) | 1.09 (1.02-1.17) | .008 |
| Sex (1 = female, 0 = male) | 1.04 (0.95-1.13) | .40 |
| Education (1 = college or advanced degree, 0 = other) | 0.92 (0.84-1.01) | .08 |
| Pre-PET Alzheimer disease drugs (1 = yes, 0 = no) | 0.93 (0.85-1.02) | .12 |
| Amyloid PET result (1 = positive, 0 = negative) | 3.54 (2.96-4.23) | <.001 |
| Primary pre-PET etiologic diagnosis (1 = Alzheimer disease, 0 = other) | 1.30 (1.12-1.51) | <.001 |
| Level of impairment (1 = Dementia, 0 = MCI) | 1.57 (1.36-1.82) | <.001 |
| Interactionsc | ||
| PET result × primary pre-PET etiological diagnosis Alzheimer disease | 0.60 (0.49-0.73) | <.001 |
| PET result × level of impairment | 0.54 (0.45-0.64) | <.001 |
Abbreviations: MCI, mild cognitive impairment; OR, odds ratio; PET, positron emission tomography.
a
Results considered significant at P < .0036 after Bonferroni correction for multiple comparisons.
b
Analysis set (n = 11 400) did not include 9 cases with uninterpretable amyloid PET results.
c
The 3-way interaction of amyloid PET result × primary pre-PET etiologic diagnosis × level of impairment and the 2-way interaction of primary pre-PET etiologic diagnosis × level of impairment were not significant (P = .91 and P = .89, respectively). The estimated Akaike Information Criterion for the full model, including these 2 interactions, was 50 474.5 and for the reduced, final model was 50 469.5.