Table 2. Estimated geometric mean and geometric s.d. E2 in cases and controls for each study, and the estimation methods used.
|
Estimated geometric mean (s.d.) E2
|
|||
|---|---|---|---|
| Study | Cases | Controls | Method used to estimate s.d. loge(E2) in cases and controls |
| Wysowski et al (1987) | 84.8 (1.91) | 101.5 (1.91) | Pooled s.d. across cases and controls from all other studies |
| Helzlsouer et al (1994) | 44.7 (3.00) | 47.9 (NA) | Two quantiles in cases (median in cases plus a further quantile calculated from no. of cases above and below control median) on a Q–Q plot used to estimate s.d. E2 in cases. s.d. loge(E2) calculated from mean and s.d. E2, assuming a normal distribution for loge(E2) |
| Rosenberg et al (1994) | 137.0 (2.19) | 138.4 (2.01) | s.d. loge(E2) in cases and controls reported |
| Thomas et al (1997) a | 86.0 (1.72) | 77.0 (1.72) | Geometric mean E2 in cases and controls, with 95% CIs, reported |
| Kabuto et al (2000) | NA | 87.4 (2.03) | s.d. log10(E2) in controls reported |
| Kaaks et al (2005) a | 86.3 (1.97) | 80.5 (2.03) | s.d. loge(E2) calculated from mean and s.d. E2, assuming a normal distribution for loge(E2), separately for cases and controls |
| Eliassen et al (2006) (follicular) | 49.4 (1.73) | 43.8 (1.82) | Quartiles plus median, 12.5th and 87.5th percentiles E2 on a Q–Q plot used to estimate mean and s.d. E2. s.d. loge(E2) calculated from mean and s.d. E2, assuming a normal distribution for loge(E2), separately for cases and controls |
| Eliassen et al (2006) (luteal) | 120.3 (1.43) | 117.9 (1.55) | As above, separately for luteal E2 |
Abbreviations: CIs=confidence intervals; NA=not applicable; Q–Q=quantile–quantile plot.
a
Thomas et al and Kaaks et al report the mean E2 in pmol l−1. These were converted to pg ml−1 by multiplying by 0.272.