The length of perimenopausal menstrual cycles increases later and to a greater degree than previously reported.

OBJECTIVE: To demonstrate that the perimenopausal increase in menstrual cycle length presented by Treloar et al. was biased by misidentified menopause dates, mean values classified by calendar year, and exclusion of menstrual cycles straddling two calendar years; and to use the revised data to investigate women's experiences of longer perimenopausal cycles.
DESIGN: Secondary analysis of prospectively collected menstrual cycle data.
SETTING: Center for Population and Health, Georgetown University. PATIENT(S): One hundred twenty white, college-educated, US women in the Tremin Research Program on Women's Health. INTERVENTION(S): None. MAIN OUTCOME MEASURE(S): Mean cycle length and time spent in >40-day cycles, by year before menopause. RESULT(S): Mean estimates for each of the 4 years before menopause were 30.48, 35.02, 45.15, and 80.22 days, respectively, compared with the original analysis: 33.60, 43.91, 55.87, and 54.58 days. In the year before menopause, the majority of women spent >or=75% of their time in cycles >40 days long. CONCLUSION(S): Treloar's estimates of mean cycle length were biased. Long cycles occurred throughout perimenopause, but the largest increase in mean cycle length did not occur until the final year before menopause. New estimates of the time spent in cycles >40 days may be useful clinically as well as epidemiologically for assessing menopausal onset and symptomatology.