Table 3.
Meta-analysis (univariate and multivariate) using random effects model comparing the response rate in the experimental arms between personalized and nonpersonalized registration trials. Unlike Table 2, data from both randomized and nonrandomized trials were included
| Characteristic | Response rate in the experimental arm | |
|---|---|---|
| No. of studies | RR % (95% CI) | |
| Personalized status | ||
| Personalized | 44 | 48 (42 to 55) |
| Nonpersonalized | 67 | 23 (20 to 27) |
| P (univariate) | <.001 | |
| P (meta-regression)* | <.001 | |
| Study design† | ||
| Randomized | 56 | 25 (21 to 31) |
| Nonrandomized | 55 | 40 (35 to 45) |
| P (univariate) | <.001 | |
| P (meta-regression)* | .14 | |
| Tumor type | ||
| Solid | 66 | 24 (20 to 28) |
| Hematologic | 45 | 46 (40 to 53) |
| P (univariate) | <.001 | |
| P (meta-regression)* | <.001 | |
| Treatment status | ||
| Naïve | 30 | 40 (33 to 49) |
| Pretreated | 81 | 29 (25 to 34) |
| P (univariate) | .01 | |
| P (meta-regression)* | <.001 | |
| Class of agent | ||
| Cytotoxic | 23 | 30 (24 to 36) |
| Targeted | 88 | 33 (28 to 38) |
| P (univariate) | .44 | |
| P (meta-regression)* | NA | |
| Single agent | ||
| Yes | 87 | 31(27 to 36) |
| No | 24 | 35 (28 to 43) |
| P (univariate) | .39 | |
| P (meta-regression)* | NA | |
| Patients per arm‡ | ||
| 6–178 | 56 | 38 (33 to 44) |
| 179–839 | 55 | 27 (22 to 32) |
| P (univariate) | .04 | |
| P (meta-regression)* | .19 | |
| Route | ||
| Oral | 53 | 26 (23 to 35) |
| Intravenous | 58 | 36 (31 to 41) |
| P (univariate) | .09 | |
| P (meta-regression)* | NA | |
| Approval date‡ | ||
| 09/1998-09/2007 | 53 | 35 (29 to 42) |
| 10/2007-06/2013 | 58 | 30 (25 to 35) |
| P (univariate) | .23 | |
| P (meta-regression)* | NA | |
* Meta-regression analysis was performed, including only factors considered statistically significant in univariate analysis; otherwise, analysis was not applicable for the characteristic. All P values were two-sided. CI = confidence interval; RR = rate ratio.
† Response rates (95% CI) in the experimental arm were 41% (30%-52%) for randomized personalized trials (n = 18 studies), 20% (15%-25%) for randomized nonpersonalized trials (n = 38), 54% (46%-61%) for nonrandomized personalized trials (n = 26), and 29% (26%-34%) for nonrandomized nonpersonalized trials (n = 29); all P values were less than .001 for comparison of personalized vs nonpersonalized trials within randomized and nonrandomized trials.
‡ Cutoff used was the median of distribution.