Table 3. Comparison of quality-adjusted life years by trial arm, across all participants and stratified by prognostic group (multiple imputed data).
All participants (n = 1,868) | p-value | Minimal/mild (n = 1,357) | p-value | Moderate (n = 288) | p-value | Severe (n = 223) | p-value | |
---|---|---|---|---|---|---|---|---|
3 months | ||||||||
Mean QALYs (SE) 1 | ||||||||
Intervention arm | 0.147 (0.0006) | 0.164 (0.0008) | 0.112 (0.002) | 0.089 (0.002) | ||||
Control arm | 0.145 (0.0006) | 0.163 (0.0008) | 0.108 (0.001) | 0.085 (0.002) | ||||
Difference in mean QALYs between arms (95% CI)1 | 0.002 (-0.0002 to 0.003) | 0.09 | 0.0008 (-0.001 to 0.003) | 0.47 | 0.004 (-0.001 to 0.008) | 0.13 | 0.003 (-0.002 to 0.008) | 0.22 |
Sensitivity analysis2 | 0.001 (-0.0004 to 0.003) | 0.15 | 0.0007 (-0.001 to 0.003) | 0.52 | 0.003 (-0.001 to 0.008) | 0.16 | 0.003 (-0.002 to 0.008) | 0.19 |
Sensitivity analysis3 | 0.001 (-0.0009 to 0.003) | 0.27 | -0.0002 (-0.003 to 0.002) | 0.85 | 0.004 (-0.0008 to 0.009) | 0.10 | 0.004 (-0.001 to 0.010) | 0.17 |
12 months | ||||||||
Mean QALYs (SE) 1 | ||||||||
Intervention arm | 0.607 (0.004) | 0.669 (0.005) | 0.482 (0.011) | 0.392 (0.012) | ||||
Control arm | 0.596 (0.004) | 0.661 (0.004) | 0.465 (0.010) | 0.367 (0.011) | ||||
Difference in mean QALYs between arms (95% CI)1 | 0.011 (0.001 to 0.022) | 0.049 | 0.008 (-0.005 to 0.021) | 0.23 | 0.016 (-0.014 to 0.047) | 0.29 | 0.024 (-0.011 to 0.059) | 0.18 |
Sensitivity analysis2 | 0.010 (-0.001 to 0.021) | 0.09 | 0.00 (-0.006 to 0.020) | 0.29 | 0.016 (-0.015 to 0.046) | 0.32 | 0.024 (-0.010 to 0.059) | 0.16 |
Sensitivity analysis3 | 0.015 (0.00004 to 0.030) | 0.049 | 0.009 (-0.008 to 0.027) | 0.32 | 0.048 (0.002 to 0.095) | 0.04 | 0.013 (-0.034 to 0.059) | 0.59 |
Abbreviations: QALYs = quality-adjusted life years; SE = standard error; CI = confidence interval
1 Baseline mean and the difference between the mean for the intervention arm minus the mean for control arm estimated using a generalised linear model (family = Gaussian, link = identity) with random intercepts for individuals and adjusted for baseline AQoL-8D utility weight, baseline PHQ-9 score, general practice and prognostic group (the final covariate only applied to the analysis involving all participants);
2 Same as 1, adjusted for factors associated with non-response to the primary outcome measure, the PHQ-9 score, at 3 and 12 months (age, gender, highest level of education, current employment status, hold a health care card, long-term illness, live alone, number of times visited a psychiatrist or counsellor in past 12 months and current use of antidepressants)
3 Same as 1, for complete cases only (i.e., no multiple imputation of missing data)