Table 7 Sensitivity analysis: use of inverse-probability weighting (IPW) to account for missing data.

From: Association of sleep duration in middle and old age with incidence of dementia

  N cases/N total Main analysis Analysis using IPW to account for missing data
HRa (95% CI) P valueb HRa (95% CI) P valueb
Sleep duration at age 50
 Short: ≤6 h 211/3149 1.22 (1.01–1.48) 0.04 1.21 (0.99–1.48) 0.06
 Normal: 7 h 219/3624 1 (ref.)   1 (ref.)  
 Long: ≥8 h 91/1186 1.25 (0.98–1.60) 0.07 1.28 (0.99–1.66) 0.06
Sleep duration at age 60
 Short: ≤6 h 192/2759 1.37 (1.10–1.72) 0.005 1.31 (1.03–1.66) 0.03
 Normal: 7 h 142/2988 1 (ref.)   1 (ref.)  
 Long: ≥8 h 75/1417 1.15 (0.87–1.52) 0.34 1.25 (0.93–1.66) 0.14
Sleep duration at age 70
 Short: ≤6 h 171/2429 1.24 (0.98–1.57) 0.10 1.14 (0.88–1.46) 0.32
 Normal: 7 h 131/2578 1 (ref.)   1 (ref.)  
 Long: ≥8 h 90/1509 1.15 (0.88–1.51) 0.60 1.14 (0.85–1.52) 0.39
Trajectories of sleep duration between age 50 and 70
 Persistent short 103/1358 1.30 (1.00–1.69) 0.048 1.24 (0.94–1.64) 0.13
 Persistent normal 141/2520 1 (ref.)   1 (ref.)  
 Persistent long 35/461 1.28 (0.88–1.85) 0.20 1.33 (0.90–1.96) 0.15
 Change from short to normal 61/1086 1.20 (0.89–1.63) 0.23 1.21 (0.88–1.69) 0.24
 Change from normal to long 47/946 1.02 (0.73–1.42) 0.90 1.02 (0.71–1.45) 0.93
 Change from normal to short 39/504 1.13 (0.79–1.62) 0.50 1.12 (0.75–1.68) 0.57
Accelerometer-assessed sleep duration
 Tertile 1: 1 h 16 min–6 h 13 min 53/1296 1.63 (1.04–2.57) 0.03 1.63 (1.03–2.59) 0.04
 Tertile 2: 6 h 14 min–7 h 0 min 31/1296 1 (ref.)   1 (ref.)  
 Tertile 3: 7 h 1 min–10 h 6 min 27/1296 0.78 (0.46–1.32) 0.36 0.82 (0.47–1.41) 0.82
  1. CI confidence intervals, HR hazard ratio, IPW inverse-probability weighting, h hour, min minute.
  2. aHR estimated from a Cox regression adjusted for age (timescale), sex, ethnicity, education, and marital status, alcohol consumption, physical activity, smoking status, and fruit and vegetable consumption, BMI, hypertension, diabetes, cardiovascular disease, GHQ depression, and CNS medications.
  3. bTwo-sided for HR in comparison with the reference (ref.) category, without adjustment for multiple comparisons.