171 Estradiol and other sex-hormone dynamics in menstrual migraine 8 Missing data for one month were imputed using data from the other month, with reasons for missingness documented. A missingness heatmap was generated and missing values were assumed to be missing at random (MAR), which was accounted for in the linear models used for analysis. For the primary analysis, linear regression was used to assess peak estradiol and estradiol decline as outcome variables, adjusting for BMI and age. Secondary analysis involved pathway analysis using the GlobalTest package (v5.60.0), allowing groups of variables to be evaluated for their association with migraine, while also adjusting for BMI and age.19 Variables were grouped by biological properties, including menstrual cycle hormones (progesterone, estradiol, LH, and FSH) and androgens (17-OHP, androstenedione, DHEA, DHEA-S, DHT, and testosterone). Individual hormones, predefined declines, and ratios were also included. The GlobalTest, sensitive to small differences and inter-variable associations, assessed whether any covariate within a group was linked to migraine.19 All analyses align with our predefined hypotheses and were not exploratory in nature. Two additional sensitivity analyses were conducted: one excluding women aged ≥45 years, and another excluding those aged ≥40 years. Plasma hormone levels were visualized using the ggplot2 package (v3.4.4). Graphs were created with log₂-transformed data, displaying means and 95% confidence intervals for each day. To improve interpretability, the y-axis was back-transformed to the original scale. RESULTS A total of 62 women (40 with migraine, 22 controls) completed the inclusion visit (eFigure 2). Thirteen participants were excluded for reasons including BMI >30, fewer than one migraine attack per month (migraine group), or failure to detect an LH surge during two months of testing. Five participants withdrew or were lost to follow-up. Ultimately, 49 participants (31 migraine, 18 controls) completed the first month of measurements. Four participants did not complete the second month due to time constraints, and one was excluded due to pregnancy. Upon data completion, one participant was excluded because plasma levels of estrogen deviated more than two standard deviations from the mean. Two participants (one with migraine, one control) were excluded due to low AMH and high FSH levels, indicating perimenopausal status. Table 1 presents baseline characteristics of the final study population, consisting of 46 participants: 28 women with migraine and 18 controls. All women had regular menstrual cycles with an average cycle length of 28 ± 3 days for the migraine group
RkJQdWJsaXNoZXIy MjY0ODMw