Table 2. Prognostic thresholds defining cohort-specific phenotypes.
| Embryos (No.)* | Cleavage Arrest (%)* | 8-cell embryo (No.)* | FSH (mIU/mL)* | Pregnancy-No. (%)† | No Pregnancy-No. (%) ‡ | Applicable Cases–No. (%) § | No. Trees ¶ | Reference Condition∥ | Odds Ratio | 95% Confidence Interval (C.I.) | |
| A1** | ≥6 | 177 (57.7) | 130 (42.3) | 307 (46.2) | 5 | ||||||
| A2 | <6 | 92 (25.7) | 266 (74.3) | 358 (53.8) | 5 | A1 | 3.9 | (2.8, 5.5) | |||
| B1 | ≥6 | ≤14.6 | 112 (70.4) | 47 (29.6) | 159 (23.9) | 4 | |||||
| B2 | ≥6 | >14.6 | 65 (43.9) | 83 (56.1) | 148 (22.3) | 4 | B1 | 3.0 | (1.9, 4.9) | ||
| B3 | ≥6 | 14.6–52.8 | 62 (47.3) | 69 (52.7) | 131 (19.7) | 1 | B1 | 2.6 | (1.6, 4.3) | ||
| B4 | ≥6 | ≤52.8 | 174 (60.0) | 116 (40.0) | 290 (43.6) | n/a | |||||
| B5 | ≥6 | >52.8 | 3 (17.6) | 14 (82.3) | 17 (2.6) | 1 | B1 | 10.6 | (3.2, 49.6) | ||
| B4 | 6.7 | (2.1, 30.9) | |||||||||
| C1 | ≥6 | ≥2 | 157 (63.6) | 90 (36.4) | 247 (37.1) | 1 | |||||
| C2 | ≥6 | <2 | 20 (33.3) | 40 (66.7) | 60 (9.2) | 1 | C1 | 3.5 | (1.9, 6.4) | ||
| D1 | ≥6 | >14.6 | ≥2 | 51 (53.1) | 45 (46.9) | 96 (14.4) | 1 | ||||
| D2 | ≥6 | >14.6 | <2 | 14 (26.9) | 38 (73.1) | 52 (7.8) | 1 | D1 | 3.0 | (1.5, 6.5) | |
| E1 | ≥6 | >14.6 | ≥2 | ≤4.6 | 14 (82.4) | 3 (17.6) | 17 (2.6) | 1 | |||
| E2 | ≥6 | >14.6 | ≥2 | >4.6 | 34 (46.9) | 37 (53.1) | 71 (12.2) | 1 | E1 | 4.8 | (1.4, 23.4) |
Cohort phenotypes defined by thresholds of non-redundant prognostic variables. Each set of conditions (A–E) use “AND” as the operator where more than one condition is listed.
No. of cases that satisfy the threshold conditions and have pregnancy outcome. This percentage is calculated by using the No. Applicable Cases as denominator. In general, conditions that discriminate between pregnancy and no pregnancy outcomes more highly are more robust and are expected to be more useful in both clinical management and translational research.
No. of cases that satisfy the threshold conditions and have no pregnancy outcome. This percentage is calculated by using the No. Applicable Cases as denominator.
The No. Applicable Cases is the total number of cases that satisfy the threshold conditions. This percentage is calculated by using the total number of cycles (665) as the denominator. In general, the larger the number of applicable cases, the more useful the set of conditions are for clinical management and counseling. However, for the purpose of translational research, conditions that define a smaller number of cases may have more specific correlates on a molecular level.
No. Trees shows the number of CART trees that utilize each set of conditions. There are a total of 5 trees. (See Supplemental Results.) Increased utilization indicates “usefulness” or “robustness” of that particular set of conditions.
Reference condition against which the Odds Ratio and 95% C.I. for having no pregnancy is calculated.
Conditions A–E are listed from most robust and “useful” to least “useful” based on: the number of trees that utilize each set of conditions, the number of applicable cases, and the odds ratio and 95% CI.