University of British Columbia Vancouver, BC Canada Karim M. Khan University of Melbourne Melbourne NSW, Australia
De Souza et al. (1) have asserted that luteal phase abnormalities and decreased ovarian progesterone production do not affect "bone health." The investigators have done an excellent job of carefully documenting ovulation, attempting prospective monitoring of flow and menstrual cycles, measuring morphometric parameters, and observing activity through a daily exercise log (1). However, we believe that a better research design and larger sample size are necessary to support their conclusions.
Our prospective 1-yr study in ovulatory, normal-weight women of varying exercise habits documented that mean luteal phase length explained 22% of the variance in bone mineral density (BMD) change by quantitative computed tomography (QCT) (2). Further, QCT change significantly related to progesterone but not to estrogen (2). Because replication is essential for scientific knowledge, De Souzas study (1) is welcome. We believe, however, that four methodological issues explain the apparent discrepancies between their study (1) and ours (2):
1. De Souza et al. report no effect of luteal length on BMD by DXA measured at one time point and therefore confirm our cross-sectional assessments of BMD by QCT (2) and by DXA after 5 yr (3). Longitudinal research designs remove much of the individual variation in absolute BMD by looking at changes within individuals over time; this reduces variability, thus increasing statistical power to detect a real difference.
2. An important difference between our study and that of De Souza et al. is the type of bone assessed. DXA estimates the mineral in the whole bone (which includes both cortical and cancellous components), while QCT measures only cancellous bone within the vertebral body. In a recent prospective study in which measurements were made by both QCT and DXA in the same individuals, Friedlander et al. (4) found that a statistically significant (-3.0%) loss of cancellous bone using QCT corresponded to a nonsignificant change (-0.2%) in whole bone using DXA. This illustrates the increased responsiveness to change of cancellous bone measured by QCT.
3. Despite De Souzas assertion that there is no effect of short luteal phase cycles on BMD, in their study exercising women with short luteal phase lengths (1) had whole-body BMD 4% lower than in exercising, ovulatory women. They indicate that they had statistical power to detect a 1 SD (13%) difference, which is an effect size of 0.47.1Power guidelines for one-way ANOVA suggest that an effect size of 0.10 is small, of 0.25 is moderate, and 0.40 is large (5). In order to obtain 80% power to detect the 4% difference they report, De Souza et al. would have needed 222 participants. With the sample sizes they used, they had a 16% power to detect the 4% difference they observed.
It is regrettable that, in their Fig. 3, De Souza et al. report statistical tests with P > 0.05 as "significant correlations...." We feel it is inappropriate to state a nonsignificant result as a primary conclusion. We also believe it is a false dichotomy to argue that, if estrogen has an important BMD effect, progesterone cannot. Our work suggests that progesterone is synergistic with and complements the necessary actions of estrogen.
4. The validity of any data relating BMD change to changes in progesterone exposure depends on accurate identification of women with ovulatory disturbances. De Souza (1) argues that we misclassified ovulatory status by use of basal temperature (BBT). BBT as used in the past has been justifiably criticized: 1) because the interpretation of temperature shift has been subjective (by eye) and thus reproducibility and accuracy have been low; 2) because the date of the BBT rise follows ovulation, rather than preceding it, BBT rise is a poor predictor of the date of ovulation; 3) because a full cycle of temperatures is required before the date of temperature shift can be determined, it is a poor choice for infertility or contraceptive interventions; and 4) because life events such as late sleeping or illness alter the morning temperature.
For all of the above reasons we have developed a quantitative analysis system and attempted to eliminate the problems of the old BBT method (6). We use a computer algorithm (Maximina; J.C. Prior, Vancouver, BC), which has been validated against the day of the LH peak (6) and confirmed by assays of serum progesterone (2, and our unpublished data). Women in our study also recorded temperatures as a list along with a record of daily events that may have influenced the temperature reading including fever, poor sleep, waking early or late. Therefore, the method we use for basal temperature documentation of ovulation and luteal phase length is valid, accurate, noninvasive, inexpensive, and accessible to women. In physiological studies a quantitative BBT method with occasional hormonal confirmation is less burdensome for women than other available methods, and it allows for long-term, prospective data collection. Because it is clear that ovulatory characteristics do not remain constant over time (3, 7) and that they change cycle by cycle even within women (2, 3, 7, 8, 9), continuous or long-term monitoring is critical.
We agree with De Souza et al. that ideally, progesterone and estrogen should be directly assayed, rather than inferred from BBT-documented ovulatory status. We do not agree that the 3 months of intensive hormone sampling they performed gives a better picture of ovulatory status and experience than 12 months of continuous, quantitatively assessed BBT data.
In summary, we are grateful that colleagues such as B. L. Lasley, De Souza, Waller, and others have focused scientific attention on the variability of menstrual cycle and ovulatory characteristics and their influences on health and on BMD. However, we believe that the conclusive study should be adequately powered, should assess changes in BMD over several years utilizing both DXA and QCT technologies, and should document ovulatory and cycle characteristics continuously, or at least for six cycles per year. A method also needs to be developed that will allow accurate, noninvasive assessment of ovulation and hormonal characteristics that is acceptable to women for long-term data collection and that will allow epidemiological investigation of the relationships between ovulation and changes in cancellous and cortical bone.
Footnotes
Address correspondence to: Jerilynn C. Prior, Department of Medicine, The University of British Columbia, Vancouver Hospital and Health Sciences Center, 855 West 12th Avenue, Vancouver, BC Canada V5Z 1M9.
1 We estimated power using GPOWER (11) and a
pooled estimate of standard deviation of 0.115, calculated from Table
3. We assumed a one-way ANOVA design, with three groups of equal size.
The outlying value for the SEM in the sedentary ovulatory
group was assumed to be a typographical error; if this is not the case,
or if the groups have unequal sizes, then the required sample size
would be higher.
Received April 7, 1998.
References
HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |