Table 1. Pooled odds ratios (ORs) and 95% confidence intervals (CIs) for high versus low soy isoflavone intake (as defined in each original study) associating with breast cancer risk, calculated from summary and stratified (by predefined factors) analyses.
| Group | No. of studies (cases) | Heterogeneity of ORs | OR (95%CI) | ||||
| χ2(dfΛ) | P * | I2(%) | |||||
| Premenopausal | |||||||
| Summary | 30(10888) | 107.33(34) | 0.000 | 68.3 | 0.74(0.64–0.85) | ||
| AsianΨ | 17(5466) | 36.32(17) | 0.004 | 53.2 | 0.59(0.48–0.69) | ||
| WesternΨ | 14(5422) | 35.51(16) | 0.003 | 54.9 | 0.90(0.77–1.04) | ||
| Soy isoflavone$/protein | 22(6710) | 98.44(26) | 0.000 | 73.6 | 0.76(0.62–0.89) | ||
| Soy bean/soy products(foods)# | 11(4702) | 29.87(10) | 0.001 | 66.5 | 0.64(0.49–0.80) | ||
| Earlier time@ | 12(5027) | 25.26(11) | 0.008 | 56.5 | 0.74(0.59–0.88) | ||
| Later time& | 18(5861) | 81.98(22) | 0.000 | 73.2 | 0.75(0.61–0.89) | ||
| Postmenopausal | |||||||
| Summary | 31(16705) | 223.33(35) | 0.000 | 84.3 | 0.75(0.63–0.86) | ||
| AsianΨ | 18(4581) | 93.28(18) | 0.000 | 80.7 | 0.59(0.44–0.74) | ||
| WesternΨ | 14(12124) | 25.20(16) | 0.066 | 36.5 | 0.92(0.83–1.00) | ||
| Soy isoflavone$/protein | 21(9341) | 205.48(25) | 0.000 | 87.8 | 0.73(0.58–0.88) | ||
| Soy bean/soy products(foods)# | 13(7727) | 138.77(12) | 0.000 | 91.4 | 0.72(0.48–0.97) | ||
| Earlier time@ | 10(5002) | 25.55(9) | 0.002 | 64.8 | 0.76(0.59–0.94) | ||
| Later time& | 21(11703) | 197.66(25) | 0.000 | 87.4 | 0.74(0.60–0.88) | ||
* P values (two-sided) were based on the Q test of heterogeneity.
Λ Degree of freedom (df) dose not equal N-1 because three studies’ data [8]–[10] were extracted as three or two independent studies.
@Before 2006.
&
Since 2006.
$
Soy isoflavone includes exposure measured by dietary isoflavone intake, plasma genistein concentration, or urinary isoflavone excretion.
#
Soy bean/soy products includes soybeans, soy, beans or tofu.
Ψ The study [10] was carried out in both Japan and Brazil by Motoki Iwasaki et al., its data was retained as three individual studies for analysis, hence larger total number of studies stratified by study region than summary studies.