Table 2. Modelling the association of CSF biomarkers on AD outcomes.
| Model | Ferritin* | tau/Aβ1–42 | ApoE | |||
|---|---|---|---|---|---|---|
| Cross-sectional cognition (MR) | β (s.e.) | P value | β (s.e.) | P value | β (s.e.) | P value |
| ADAS-Cog13† | 0.139 (0.050) | 0.006 | 0.104 (0.046) | 0.025 | −0.178 (0.049) | 0.0003 |
| RAVLT | −1.77 (0.559) | 0.0017 | NS | NS | 1.033 (0.564) | 0.0677 |
| Longitudinal cognition (MELM) | β (s.e.) | P value | β (s.e.) | P value | β (s.e.) | P value |
| ADAS-Cog13† | ||||||
| Main effect | 0.178 (0.051) | 0.0005 | 0.129 (0.049) | 0.009 | −0.180 (0.051) | 0.0004 |
| Interaction time | 0.0005 (0.016) | 0.977 | 0.081 (0.016) | 2.70 × 10−7 | −0.044 (0.016) | 0.006 |
| RAVLT | ||||||
| Main effect | −1.60 (0.63) | 0.012 | −0.847 (0.608) | 0.165 | 1.03 (0.63) | 0.104 |
| Interaction time | −0.035 (0.152) | 0.817 | −0.610 (0.150) | 4.85 × 10−5 | 0.279 (0.152) | 0.066 |
| MCI conversion to AD | Statistic‡ | P value | Statistic‡ | P value | Statistic‡ | P value |
| Cox (Hazard ratio) | 1.10 (1.01–1.19) | 0.030 | 1.53 (1.03–2.28) | 0.037 | 0.83 (0.73–0.95) | 0.008 |
| LR (Odds ratio) | 2.32 (1.86–2.90) | 8.001 × 10−15 | 1.45 (1.16–1.80) | 0.0001 | 0.38 (0.30–0.48) | 1.88 × 10−17 |
| Rate of MRI atrophy (MELM) | β (s.e.) | P value | β (s.e.) | P value | β (s.e.) | P value |
| Hippocampus | −18.33 (7.86) | 0.019 | −35.31 (7.79) | 6.81 × 10−6 | 21.38 (8.02) | 0.008 |
| Lateral ventricles§ | 0.007 (0.003) | 0.008 | 0.013 (0.002) | 4.19 × 10−8 | −0.009 (0.003) | 0.0002 |
Cox, Cox proportional hazard model; LR: logistic regression; MELM, mixed effects linear model; MR, multiple regression; NS, not significant. All models initially contained the variables: age, gender, BMI, APOE genotype, baseline diagnosis; the MRI models additionally included intracranial volume. Minimal models for the cognition models included baseline diagnosis, gender, years of education and the AD CSF biomarkers. Minimal model for the Cox proportional hazard model (Cox) contained only the AD CSF biomarkers. Minimal models for the MRI models contained age, gender, baseline diagnosis, years of education, APOE ɛ4 status and intracranial volume. All subjects with available data were included in the cross-sectional cognition models. Only CN and MCI subjects were included in modelling of longitudinal cognition because short follow up of AD subjects (Supplementary Table 3). Only subjects who were classed as MCI at baseline were included in the MCI conversion models. The MRI models contained subjects who were classed as cognitively normal or MCI at baseline. AD subjects at baseline were not included because of low numbers and lack of follow-up (Supplementary Tables 3).
*Ferritin values were log-transformed, excluding non-parametric Cox and LR models.
†The β-coefficient is for the square root of ADAS-Cog13.
‡The statistics for the conversion models were based on one interquartile range change for each analyte (ferritin: 3.3 ng ml−1, tau/Aβ1–42: 0.67 units; ApoE: 3.1 μg ml−1).
§For Lateral ventricles, the β-coefficient is for natural log of the ventricle volume.