Assignment to menopausal status and estimation of age at menopause for women with missing or invalid data--a...

Gynaecologic surgeries and hormone medication increasingly interfere with the concept of determining menopausal status based on self-reported data about the last menstrual period-an approach frequently applied in epidemiological studies. This analysis compared different approaches of menopausal status determination, including a probabilistic approach that was developed to reduce misclassification of menopausal status.

Interview and questionnaire data relating to hormone status from about 16,000 German women were used to determine menopausal status using different approaches: menses based, age based, self-assessment based, and a probabilistic approach.

Applying strict menses-based criteria, 29.1% of the women were found to be with undetermined menopausal status. The probabilistic approach was applied to determine menopausal status and age at menopause for these women. The proposed approach uses the current age of the women with undetermined menopausal status, and the relative age-dependent frequencies of pre- or postmenopausal status among those with natural menopausal status to calculate weighting factors, which represent the probabilities of being pre- and postmenopausal of each woman previously undetermined. Correspondingly, the missing age at menopause was estimated with the same probabilistic approach. Applying various approaches to determine menopausal status, the ratio of pre- to postmenopausal women differed considerably, ranging from 1.03 to 1.61. In addition, considerable differences were observed with respect to the number of missing values for both menopausal status and age at menopause.

The probabilistic approach allowed to estimate menopausal status and age at menopause with the advantage that all available information from the entire study population and from the individual is used.