Background
Source | Study objectives, design and analysis | Reported relapse rates | |
---|---|---|---|
Rapaport et al. [1] | Escitalopram continuation treatment to prevent relapse; a multi-center, placebo controlled, randomized withdrawal study; 36-week randomized treatment; Kaplan-Meier estimate and log-rank test as primary statistical analysis | Escitalopram | 26.0%* (109) |
Placebo | 40.0%* (116) | ||
Keller et al. [2] | Long-term efficacy and tolerability of gepirone ER; a multi-center, placebo controlled, randomized withdrawal study; 40-44 weeks of randomized treatment; chi-square test; Kaplan-Meier estimate and log-rank test as Primary statistical analysis | Gepirone ER | 20.6% (26/126) |
Placebo | 28.2% (35/124) | ||
Kamijima et al.[3] | Efficacy, safety and tolerability of sertraline in the prevention of relapse; a multicenter, placebo controlled, randomized withdrawal study; 16-week randomized treatment; Kaplan-Meier estimate and log-rank test as primary statistical analysis | Sertraline | 08.5% (10/117) |
Placebo | 19.5% (23/118) | ||
Perahia et al. [4] | Efficacy, safety and tolerability of duloxetine in the prevention of relapse; A multi-center, placebo controlled, randomized withdrawal study; 26-week randomized treatment; Kaplan-Meier estimate, log-rank test as primary statistical analysis | Duloxetine | 17.4% (23/132) |
Placebo | 28.5% (39/137) | ||
Kocsis et al. [5] | Long-term efficacy and safety of venlafaxine ER in preventing recurrence; a multi-center, placebo controlled, randomized withdrawal study; 12 months randomized treatment; Kaplan-Meier estimate and log-rank test as primary statistical analysis | Venlafaxine ER | 23.1%* (129) |
Placebo | 42.0%* (129) |
Methods
Cox cure regression
The model
Parameter estimation
Cure survival CART
The model
The algorithm
Splitting
Pruning
Selection of a pruned sub-tree
Results
Study design and major results of a sertraline randomized withdrawal study
95% CI | |||||
---|---|---|---|---|---|
Standard logistic regression
| Estimates | SE | Lower | Upper | p-value |
Intercept | 1.686 | 1.687 | -1.621 | 4.992 | 0.3178 |
Treatment | 0.968 | 0.429 | 0.152 | 1.848 | 0.0239 |
Baseline HAM-D score at OP | 0.007 | 0.062 | -0.111 | 0.134 | 0.9096 |
Baleline HAM-D score at DP | -0.100 | 0.068 | -0.240 | 0.029 | 0.1443 |
Gender | -1.842 | 0.575 | -3.123 | -0.817 | 0.0014 |
Age | 0.026 | 0.020 | -0.013, | 0.068 | 0.2020 |
The number of episodes | 0.174 | 0.139 | -0.016 | 0.531 | 0.2111 |
Duration from the 1st episode | -0.002 | 0.003 | -0.008 | 0.005 | 0.5575 |
Duration of this episode | -0.002 | 0.022 | -0.041 | 0.046 | 0.9296 |
Interval from the previous episode | 0.004 | 0.007 | -0.008 | 0.019 | 0.5553 |
Complication | 0.572 | 0.413 | -0.241 | 1.388 | 0.1654 |
Maximum log-likelihood | -81.603 | ||||
AIC | 185.206 | ||||
95% CI
| |||||
Standard Cox regression
|
Estimates
|
SE
|
Lower
|
Upper
|
p-value
|
Treatment | -0.845 | 0.388 | -1.604 | -0.086 | 0.0293 |
Baseline HAM-D score at OP | 0.005 | 0.053 | -0.099 | 0.109 | 0.9222 |
Baseline HAM-D score at DP | 0.063 | 0.059 | -0.052 | 0.180 | 0.2847 |
Gender | 1.639 | 0.543 | 0.575 | 2.702 | 0.0025 |
Age | -0.023 | 0.018 | -0.058 | 0.012 | 0.1977 |
The number of episodes | -0.152 | 0.130 | -0.406 | 0.102 | 0.2399 |
Duration from the 1st episode | 0.002 | 0.003 | -0.004 | 0.007 | 0.5729 |
Duration of this episode | 0.005 | 0.019 | -0.033 | 0.042 | 0.8038 |
Interval from the previous episode | -0.003 | 0.006 | -0.014 | 0.009 | 0.6274 |
Complication | -0.518 | 0.358 | -1.221 | 0.184 | 0.1482 |
Maximum log-likelihood* | -180.917 | ||||
AIC | 381.834 |
Application of Cox cure regression and cure survival CART to sertraline data
95% CI | |||||
---|---|---|---|---|---|
Cured incidence
| Estimates | SE | Lower | Upper | p-value |
Intercept | 1.571 | 2.870 | -4.053 | 7.196 | 0.5840 |
Treatment | 1.177 | 0.429 | 0.335 | 2.018 | 0.0061 |
Baseline HAM-D score at DP | -0.122 | 0.071 | -0.262 | 0.018 | 0.0869 |
95% CI
| |||||
Uncured survivals
|
Estimates
|
SE
|
Lower
|
Upper
|
p-value
|
Gender | 1.953 | 0.535 | 0.905 | 3.001 | 0.0003 |
Complication | -0.967 | 0.418 | -1.798 | -0.159 | 0.0193 |
Maximum log-likelihood | -179.578 | ||||
AIC | 369.155 |
95% CI | |||||
---|---|---|---|---|---|
Standard logistic regression
| Estimates | SE | Lower | Upper | p-value |
Intercept | 2.865 | 0.848 | 1.203 | 4.527 | 0.0007 |
Treatment | 1.021 | 0.421 | 0.196 | 1.846 | 0.0153 |
Baseline HAM-D score at DP | -0.107 | 0.066 | -0.236 | 0.222 | 0.1036 |
Gender | -1.737 | 0.563 | -2.841 | -0.633 | 0.0021 |
The number of episodes | 0.133 | 0.114 | -0.091 | 0.357 | 0.2443 |
Complication | 0.585 | 0.407 | -0.213 | 1.384 | 0.1507 |
Maximum log-likelihood | -81.868 | ||||
AIC | 177.736 | ||||
95% CI
| |||||
Standard Cox regression
|
Estimates
|
SE
|
Lower
|
Upper
|
p-value
|
Treatment | -0.913 | 0.379 | -1.655 | -0.171 | 0.0159 |
Gender | 1.511 | 0.534 | 0.465 | 2.557 | 0.0046 |
The number of episodes | -0.131 | 0.109 | -0.345 | 0.083 | 0.2296 |
Maximum log-likelihood* | -183.520 | ||||
AIC | 373.040 |
95% CI | |||||
---|---|---|---|---|---|
Cured incidence
| Estimates | SE | Lower | Upper | p-value |
Intercept | 0.753 | 1.374 | -1.940 | 3.447 | 0.5835 |
Treatment | 30.954 | 10.001 | 11.353 | 50.555 | 0.0020 |
Baseline HAM-D score at DP | -0.233 | 0.646 | -1.449 | 1.033 | 0.7183 |
Treatment × Baseline HAM-D score at DP | -30.180 | 13.497 | -56.634 | -3.725 | 0.0254 |
95% CI
| |||||
Uncured survivals
|
Estimates
|
SE
|
Lower
|
Upper
|
p-value
|
Gender | 1.926 | 0.535 | 0.878 | 2.974 | 0.0003 |
Complication | -0.895 | 0.469 | -1.814 | 0.024 | 0.0563 |
Maximum log-likelihood | -177.499 | ||||
AIC | 366.998 |