Table 4.
Effect of baseline PSQI measures on longitudinal change in CSF biomarkers
| Variables | log(CSF Aβ42) | log(CSF p-tau) | log(CSF t-tau) | |||
|---|---|---|---|---|---|---|
| β Coefficient (95% CI) | P-value | β Coefficient (95% CI) | P-value | β Coefficient (95% CI) | P-value | |
| Time × total PSQI score | 0.000 (−0.00002 0.000004) | 0.227 | 0.000 (−0.00001 0.00001) | 0.703 | 0.000 (−0.00001 0.00001) | 0.957 |
| Time × dichotomized PSQI score (ref. Total PSQI ≤5) | 0.000 (−0.0001 0.00003) | 0.295 | 0.000 (−0.00004 0.0001) | 0.534 | 0.000 (−0.00005 0.0001) | 0.868 |
| Time × sleep latency (ref. ≤ 15 min)a | ||||||
| 16–30 min | 0.000 (−0.0001 0.00005) | 0.471 | 0.000 (−0.00005 0.00007) | 0.749 | 0.000 (−0.0001 0.00003) | 0.291 |
| >30 min | 0.000 (−0.0001 0.0001) | 0.727 | 0.000 (−0.00007 0.0001) | 0.699 | 0.000 (−0.0001 0.00005) | 0.473 |
| Time × sleep duration (ref. > 7 h)b | ||||||
| 6–7 h | −0.0001 (−0.0001 0.00001) | 0.09 | 0.000 (−0.0001 0.00003) | 0.358 | 0.000 (−0.0001 0.0001) | 0.838 |
| <6 h | 0.000 (−0.0001 0.0001) | 0.57 | 0.000 (−0.0001 0.0001) | 0.868 | 0.000 (−0.0001 0.0001) | 0.902 |
| Time × sleep efficiency (ref. > 85%) | ||||||
| 75–84% | 0.000 (−0.0001 0.00003) | 0.269 | 0.000 (−0.00004 0.0001) | 0.435 | 0.000 (−0.0001 0.0001) | 0.76 |
| 65–74% | 0.000 (−0.0001 0.0001) | 0.529 | 0.000 (−0.0001 0.0001) | 0.828 | 0.000 (−0.0001 0.0001) | 0.988 |
| <65% | 0.000 (−0.0001 0.0001) | 0.687 | 0.000 (−0.0001 0.0001) | 0.48 | 0.000 (−0.0001 0.0001) | 0.808 |
| Time × sleep disturbance (ref. ≤ 9)c | ||||||
| 1–9 | −0.0002 (−0.0004−0.0001) | 0.006 | 0.000 (−0.0001 0.0001) | 0.771 | 0.000 (−0.0001 0.0001) | 0.712 |
| >9 | −0.0002 (−0.0004−0.0001) | 0.005 | 0.000 (−0.0001 0.0002) | 0.527 | 0.000 (−0.0001 0.0002) | 0.742 |
| Time × daytime dysfunction (ref. 0)d | ||||||
| ≥1 | −0.0001 (−0.00012 0.00001) | 0.125 | 0.000 (−0.00005 0.0001) | 0.888 | 0.000 (−0.0001 0.00005) | 0.959 |
ref.: Level of reference. aCategories corresponding to ‘31–60 min’ and ‘>60 min’ have been collapsed due to <20 observations in one category. bCategories corresponding to ‘5–6 h’ and ‘<5 h’ have been collapsed due to <20 observations in two categories. cCategories corresponding to scores ≤ 9 and > 9 have been collapsed due to <20 observations in two categories. dCategories corresponding to scores of ‘1–2’, ‘3–4’ and ‘5–6’ have been collapsed due to <20 observations in two categories. All models are adjusted by age, sex, APOE-ɛ4 carriership (and their interactions with time) and site of data collection (fixed effects). Additionally, models with log(CSF Aβ42) as outcome are adjusted by log(CSF p-tau), and models with log(CSF t-tau) or log(CSF p-tau) as outcomes, are adjusted by log(CSF Aβ42). A random intercept for each CSF biomarker and change over time (slope) are included as random effects.