Table 5.
Comparison of primary endpoints reported in FDA summaries with corresponding publications and those published studies. Figures are numbers (percentage) unless stated otherwise
| All studies (n=152) | Pivotal studies* (n=127) | ||||||
|---|---|---|---|---|---|---|---|
| Summary | Published | Odds ratio (95% CI), P value† | Summary | Published | Odds ratio (95% CI), P value† | ||
| Type of endpoint analysis: | |||||||
| Non-inferiority | 11 (7) | 15 (10) | 1.65 (0.61 to 4.44), 0.32 | 9 (7) | 13 (10) | 1.84 (0.61 to 5.55), 0.28 | |
| Equivalence | 4 (3) | 3 (2) | —‡ | 4 (3) | 3 (2) | 0.70 (0.14 to 3.67), 0.68 | |
| Historical control | 13 (9) | 3 (2) | 0.10 (0.02 to 0.48), 0.004 | 13 (10) | 3 (2) | 0.10 (0.02 to 0.48), 0.004 | |
| Objective performance criteria | 47 (31) | 16 (11) | 0.15 (0.07 to 0.32), <0.001 | 43 (34) | 15 (12) | 0.15 (0.07 to 0.34), <0.001 | |
| Unknown | 77 (51) | 115 (76) | 4.01 (2.30 to 6.99), <0.001 | 58 (46) | 93 (73) | 4.30 (2.37 to 7.78), <0.001 | |
| Controls: | |||||||
| Prospective | 63 (41) | 63 (41) | 1.00 (0.50 to 1.99), 1.0 | 55 (43) | 55 (43) | 1.00 (0.43 to 2.30), 1.0 | |
| Retrospective | 20 (13) | 15 (10) | 0.64 (0.27 to 1.47), 0.29 | 19 (15) | 13 (10) | 0.54 (0.22 to 1.33), 0.18 | |
| None | 67 (44) | 68 (45) | 1.05 (0.57 to 1.94), 0.88 | 53 (42) | 54 (43) | 1.06 (0.54 to 2.11), 0.86 | |
| Unknown | 2 (1) | 6 (4) | 3.34 (0.63 to 17.66), 0.16 | 0 (0) | 5 (4) | —‡ | |
| Mean No of patients analyzed | 325.6 | 334.4 | 10.32 (−57.18 to 77.83), 0.77§ | 338.1 | 348.6 | 12.22 (−23.64 to 48.08), 0.51§ | |
| Direction of favorability of endpoint results: | |||||||
| Device | 94 (62) | 65 (43) | 0.44 (0.27 to 0.70), <0.001 | 86 (68) | 58 (46) | 0.37 (0.22 to 0.63), <0.001 | |
| Control | 8 (5) | 3 (2) | 0.35 (0.09 to 1.37), 0.13 | 8 (6) | 3 (2) | 0.35 (0.09 to 1.37), 0.13 | |
| Neither | 12 (8) | 14 (9) | 1.33 (0.56 to 3.15), 0.51 | 10 (8) | 11 (9) | 1.26 (0.49 to 3.26), 0.63 | |
| Unknown | 38 (25) | 70 (46) | 2.77 (1.65 to 4.64), <0.001 | 23 (18) | 55 (43) | 3.84 (2.10 to 7.05), <0.001 | |
*Study was classified as pivotal if it was only one included in summary, multicenter randomized controlled trial, or explicitly noted in summary as being pivotal study.
†Calculated by generalized linear mixed effects model with canonical logit link and random effect placed on intercept to account for heterogeneity across devices.
‡Too few events or counter-events for fitting of random effects model.
§Mean difference, (95% CI), P value calculated by general linear mixed effects model with random effect placed on intercept to account for device clustering.