Table 3.
Patient CDRS-R benefits stratified by model prediction.
| CBT vs. FLX | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Predicted benefit strata | Treated with CBT | Treated with FLX | Estimated benefit | |||||||
| M | SD | N | M | SD | N | M | Cohen’s D | p | Adj. p | |
| 0–25% | 35.1 | 11.0 | 22 | 34.0 | 12.3 | 28 | 1.1 | 0.1 | 0.74 | 0.74 |
| 25–50% | 35.5 | 8.6 | 20 | 36.7 | 10.8 | 23 | −1.2 | −0.1 | 0.68 | 0.74 |
| 50–75% | 39.4 | 11.1 | 27 | 37.6 | 16.4 | 21 | 1.8 | 0.1 | 0.66 | 0.74 |
| 75–100% | 56.2 | 15.1 | 21 | 39.3 | 11.2 | 25 | 16.9 | 0.9 | 0.00 | 0.00 |
| CBT vs. COMB | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Predicted benefit strata | Treated with CBT | Treated with COMB | Estimated benefit | |||||||
| M | SD | N | M | SD | N | M | Cohen’s D | p | Adj. p | |
| 0–25% | 35.8 | 11.3 | 20 | 30.7 | 9.9 | 21 | 5.0 | 0.3 | 0.14 | 0.41 |
| 25–50% | 36.9 | 11.1 | 22 | 34.8 | 13.1 | 24 | 2.1 | 0.1 | 0.57 | 0.74 |
| 50–75% | 39.3 | 9.2 | 29 | 30.8 | 7.9 | 24 | 8.4 | 0.7 | 0.00 | 0.00 |
| 75–100% | 55.8 | 17.4 | 19 | 36.8 | 14.6 | 26 | 19.0 | 0.8 | 0.00 | 0.00 |
| FLX vs. COMB | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Predicted benefit strata | Treated with FLX | Treated with COMB | Estimated benefit | |||||||
| M | SD | N | M | SD | N | M | Cohen’s D | p | Adj. p | |
| 0–25% | 39.2 | 13.8 | 25 | 35.6 | 11.2 | 18 | 3.6 | 0.2 | 0.37 | 0.65 |
| 25–50% | 34.5 | 9.2 | 30 | 33.0 | 9.9 | 24 | 1.6 | 0.1 | 0.55 | 0.74 |
| 50–75% | 36.1 | 13.0 | 26 | 32.0 | 10.8 | 28 | 4.1 | 0.2 | 0.21 | 0.50 |
| 75–100% | 38.4 | 16.1 | 16 | 34.0 | 15.2 | 25 | 4.4 | 0.2 | 0.38 | 0.65 |
Bold indicates a significant treatment benefit in predicted benefit strata.
CDRS-R children’s depression rating scale-revised, CBT cognitive behavioral therapy, FLX fluoxetine, COMB combination treatment.
CBT vs. FLX: Rows represent groups of patients that are predicted to benefit from FLX over CBT with different magnitudes (bottom 25%, 25–50%, 50–75%, top 25%). The estimated benefit from FLX compared to CBT within each stratum is computed as the difference in CDRS-R between the patients who were treated with CBT and those treated with FLX. The participants who were predicted to benefit the most (top 25%) were estimated to benefit significantly from FLX with on average 16.9 CDRS-R difference. Adj p = adjusted p-value.
CBT vs. COMB: Rows represent groups of patients that are predicted to benefit from COMB over CBT with different magnitudes (bottom 25%, 25–50%, 50–75%, top 25%). The estimated benefit from COMB compared to CBT within each group is computed as the difference in CDRS-R between the patients who were treated with CBT and those treated with COMB. The participants who were predicted to benefit more from COMB (top 50%) were estimated to benefit significantly from COMB. Adj p = adjusted p-value.
FLX vs. COMB: Rows represent groups of patients that are predicted to benefit from COMB over FLX with different magnitudes (bottom 25%, 25–50%, 50–75%, top 25%). The estimated benefit from COMB compared to FLX within each group is computed as the difference in CDRS-R between the patients who were treated with FLX and those treated with COMB. Adj p = adjusted p-value.