Top panel, Annual rate of change in BMD of the lumbar spine in premenopausal (red bars), early perimenopausal (blue bars), late perimenopausal (yellow bars), and postmenopausal (green bars) African-American (n = 494), Caucasian (n = 944), Chinese (n = 221), and Japanese (n = 243) women from the full cohort. Rates of change were estimated from multivariable linear mixed models and adjusted for multiple covariates. Error bars represent 95% confidence limits. Comparisons were made between ethnic groups (P values for lumbar spine in pre-, early peri-, late peri-, and postmenopausal women, respectively): African-American vs. Caucasian, 0.65, 0.11, 0.10, and <0.001; African-American vs. Chinese, 0.031, <0.001, 0.006, and <0.001; African-American vs. Japanese, 0.87, <0.001, <0.001, and <0.001; Caucasian vs. Chinese, 0.004, 0.014, 0.06, and 0.003; Caucasian vs. Japanese, 0.58, <0.001, 0.009, and <0.001; Chinese vs. Japanese, 0.09, 0.43, 0.81, and 0.51. Bottom panel, Annual rate of change in BMD of the lumbar spine in premenopausal (red bars), early perimenopausal (blue bars), late perimenopausal (yellow bars), and postmenopausal (green bars) African-American (n = 198), Caucasian (n = 587), Chinese (n = 167), and Japanese (n = 181) women weighing between 50 and 78 kg. Rates of change were estimated from multivariable linear mixed models and adjusted for multiple covariates. Error bars represent 95% confidence limits. Comparisons were made between ethnic groups, with P values for lumbar spine in pre-, early peri-, late peri-, and postmenopausal women, respectively, as follows: African-American vs. Caucasian, 0.13, 0.24, 0.43, and 0.07; African-American vs. Chinese, 0.80, 0.007, 0.56, and 0.31; African-American vs. Japanese, 0.10, 0.010, 0.24, and 0.02; Caucasian vs. Chinese, 0.22, 0.041, 0.18, and 0.52; Caucasian vs. Japanese, 0.59, 0.06, 0.52, and 0.34; Chinese vs. Japanese, 0.16, 0.87, 0.11, and 0.17.

Annual rate of change in BMD of the lumbar spine in premenopausal (red bars), early perimenopausal (blue bars), late perimenopausal (yellow bars), and postmenopausal (green bars) women divided by tertiles of body weight in African-American (tertile 1, <73.4 kg; tertile 2, 73.4–91.5 kg; tertile 3, >91.5 kg), Caucasian (tertile 1, <64.0 kg; tertile 2, 64.0–78.1 kg; tertile 3, >78.1 kg), Chinese (tertile 1, <52.9 kg; tertile 2, 52.9–59.8 kg; tertile 3, >59.8 kg), and Japanese (tertile 1, <52.4 kg; tertile 2, 52.4–58.6 kg; tertile 3, >58.6 kg) women. Rates of change were estimated from multivariable linear mixed models and adjusted for multiple covariates. Error bars represent 95% confidence limits. Note that a relatively small number of Chinese and Japanese women in each weight tertile transitioned to late perimenopause or beyond (Chinese: tertile 1, n = 24; tertile 2, n = 24; tertile 3, n = 30; Japanese: tertile 1, n = 31; tertile 2, n = 31; tertile 3, n = 25) or became postmenopausal (Chinese: tertile 1, n = 22; tertile 2, n = 23; tertile 3, n = 24; Japanese: tertile 1, n = 24; tertile 2, n = 25; tertile 3, n = 22) during the follow-up period. Comparisons were made between tertile groups, with P values for African-Americans in pre-, early peri-, late peri-, and postmenopausal women, respectively, as follows: tertile 1 vs. 3, 0.008, 0.039, 0.30, and <0.001; tertile 2 vs. 3, 0.18, 0.88, 0.08, and <0.001; tertile 1 vs. 2, 0.31, 0.028, 0.57, and 0.78; with P values for Caucasians in pre-, early peri-, late peri-, and postmenopausal women, respectively: tertile 1 vs. 3, 0.012, <0.001, <0.001, and <0.001; tertile 2 vs. 3, 0.25, <0.001, <0.001, and <0.001; tertile 1 vs. 2, 0.21, 0.11, 0.95, and 0.29; with P values for Chinese in pre-, early peri-, late peri-, and postmenopausal women, respectively: tertile 1 vs. 3, 0.21, 0.015, 0.038, and <0.001; tertile 2 vs. 3, 0.56, 0.002, 0.020, and <0.001; tertile 1 vs. 2, 0.44, 0.57, <0.001, and 0.24; and with P values for Japanese in pre-, early peri-, late peri-, and postmenopausal women, respectively: tertile 1 vs. 3, 0.08, <0.001, 0.001, and 0.002; tertile 2 vs. 3, 0.18, 0.028, 0.002, and 0.003; tertile 1 vs. 2, 0.42, 0.14, 0.89, and 0.79.

Annual rate of change in BMD of the lumbar spine and total hip in premenopausal (red bars), early perimenopausal (blue bars), late perimenopausal (yellow bars), and postmenopausal (green bars) women (n = 1902). Rates of change were estimated from multivariable linear mixed models and adjusted for multiple covariates. Error bars represent 95% confidence limits. Comparisons were made across status categories: early peri- vs. premenopausal, P < 0.001 (spine) and P = 0.002 (hip); late peri- vs. early perimenopausal, P < 0.001 (spine) and P < 0.001 (hip); and post- vs. late perimenopausal, P = 0.002 (spine) and P < 0.001 (hip).

Annual rate of change in BMD of the lumbar spine (top panel) and total hip (bottom panel) in premenopausal (red bars), early perimenopausal (blue bars), late perimenopausal (yellow bars), and postmenopausal (green bars) women divided by tertiles of baseline body weight (tertile 1, < 60.7 kg; tertile 2, 60.7–77.3 kg; tertile 3, >77.3 kg) in the entire cohort. Rates of change were estimated from multivariable linear mixed models and adjusted for multiple covariates. Error bars represent 95% confidence limits. Comparisons were made between tertile groups, with P values for lumbar spine in pre-, early peri-, late peri-, and postmenopausal women, respectively, as follows: tertile 1 vs. 3, <0.001, <0.001, <0.001, and <0.001; tertile 2 vs. 3, 0.033, <0.001, <0.001, and <0.001; tertile 1 vs. 2, 0.13, 0.002, 0.39, and 0.044; and with P values for total hip in pre-, early peri-, late peri-, and postmenopausal women, respectively: tertile 1 vs. 3, 0.15, <0.001, 0.027, and <0.001; tertile 2 vs. 3, 0.51, 0.08, 0.09, and <0.001; tertile 1 vs. 2, 0.032, 0.017, 0.68, and 0.17.