Table III.
Subgrouping and sensitivity analyses.
| Factor | Subgroup | Number of studies | Number of participants | Mean difference (95% CI) | P-value for heterogeneity, I2 |
|---|---|---|---|---|---|
| Main analysis | Overall analysis (no subgroups) | 7 | 1149 | 0.69 (−0.78 to 2.17) | 0.14, 36% |
| Menopausal status* | Premenopausal | 5 | 519 | 1.83 (0.25 to 3.40) | 0.85, 0% |
| Post-menopausal | 4 | 592 | −1.10 (−3.22 to 1.03) | 0.23, 30% | |
| Peri-menopausal | 1 | 16 | −0.37 (−6.12 to 5.38) | NR | |
| Baseline risk of breast cancer | Low risk | 4 | 715 | 0.97 (−0.77 to 2.71) | 0.72, 0% |
| Moderate risk | 2 | 241 | 3.22 (−0.18 to 6.63) | 0.62, 0% | |
| High risk | 2 | 194 | −0.24 (−3.47 to 2.99) | 0.07, 69% | |
| Isoflavone source | Isolated red clover isoflavones | 3 | 394 | 0.70 (−3.37 to 4.76) | 0.04, 69% |
| Isolated soy isoflavones | 1 | 30 | 2.12 (−2.42 to 6.66) | NR | |
| Soy foods | 2 | 559 | 0.54 (−1.71 to 2.79) | 0.34, 0% | |
| Soy protein | 2 | 166 | 1.54 (−0.83 to 3.92) | 0.85, 0% | |
| Any soy intervention | 5 | 755 | 1.14 (−0.40 to 2.68) | 0.82, 0% | |
| Dose, mg/day total isoflavones** | <50 | 3 | 394 | 0.70 (−3.37 to 4.76) | 0.04, 69% |
| 50 to <100 | 5 | 635 | 0.90 (−0.74 to 2.54) | 0.36, 8% | |
| 100+ | 1 | 243 | 1.30 (−3.38 to 5.98) | NR | |
| Duration** | 6 to <18 months | 7 | 1014 | −0.20 (−1.29 to 0.88) | 0.33, 13% |
| 18 to <30 months | 4 | 846 | 0.08 (−1.34 to 1.51) | 0.65, 0% | |
| 30+ months | 2 | 241 | 3.22 (−0.18 to 6.63) | 0.62, 0% | |
| Type of density assessment** | Visual assessment | 1 | 160 | 0.70 (−2.95 to 4.35) | NR |
| Semi-automated | 7 | 1023 | 0.67 (−1.01 to 2.36) | 0.11, 43% | |
| Fully automated | 2 | 277 | −0.22 (−1.57 to 1.13) | 0.34, 0% | |
| Sensitivity analyses | |||||
| SMD analysis including all breast density measures | Overall analysis | 8 | 1287 | 0.06 (−0.09 to 0.21) | 0.09, 39% |
| Premenopausal | 4 | 479 | 0.21 (0.04 to 0.38) | 0.95, 0% | |
| Post-menopausal | 5 | 730 | −0.07 (−0.33 to 0.18) | 0.05, 57% | |
| Peri-menopausal | 1 | 16 | −0.07 (−1.05 to 0.92) | NR | |
| Excluding studies fully funded by industry | Overall analysis | 7 | 1119 | 0.60 (−1.00 to 2.19) | 0.12, 41% |
| Premenopausal | 4 | 489 | 1.79 (0.11 to 3.47) | 0.72, 0% | |
| Post-menopausal | 5 | 730 | −0.07 (−0.33 to 0.18) | 0.05, 57% | |
| Peri-menopausal | 1 | 16 | −0.07 (−1.05 to 0.92) | NR | |
| Using Powles P-value data | Overall analysis (no subgroups) | 7 | 1149 | 0.43 (−0.91 to 1.77) | 0.24, 23% |
| Premenopausal | 5 | 519 | 1.63 (−0.03 to 3.30) | 0.94, 0% | |
| Post-menopausal | 4 | 592 | −0.89 (−3.04 to 1.26) | 0.18, 38% | |
| Peri-menopausal | 1 | 16 | −0.37 (−6.12 to 5.38) | NR | |
CI, confidence intervals; NR, Not relevant (assessment of heterogeneity is not relevant when only one study is included).
*As the menopausal status of some of the Atkinson study participants was not known the total numbers of participants are smaller when subgrouped by menopausal status than in the overall analysis.
**In some studies data were measured at more than one time point or using more than one technique, so numbers do not add up to total numbers of study participants. One study used two different dose levels and as these arms fell in separate dose subgroupings the full control group was used once in each subgroup, increasing the apparent number of participants.