Determination of luteal phase length by quantitative basal temperature methods: validation against the midcycle LH peak

Basal temperature data are known to provide unreliable assessments of luteal phase length when they are evaluated by qualitative, visual-pattern methods. This study of 24 cycles in 24 women compared the serum LH peak day with the luteal phase onset day determined by three quantitative basal temperature methods: a) a new computerized least mean square method developed by the authors; b) the mean temperature method reported by Vollman; and c) a computerized version of the World Health Organization cumulative sum method of Royston. The luteal phase onset day determined by the three quantitative basal temperature methods, (a, b, and c) correlated well with the midcycle LH peak (r = 0.879, 0.891, and 0.791, respectively, all p less than 0.001). The cumulative sum method, however, was only able to analyze 19/24 cycles. The mean delay between the LH peak day and the luteal phase onset day determined by thermal shift was 2.4 +/- 1.5, 2.7 +/- 1.4, and 4.1 +/- 2.0 d (mean +/- SD), respectively. The mean temperature method, but not the other two methods, showed an increasing delay between the LH peak day and the thermal shift day with longer follicular phase lengths. Rectal and oral temperature data from the same cycle give identical luteal onset days when analyzed by the least mean square and mean temperature methods, but discrepant days by the cumulative sum analysis. The least mean square technique is a reliable and precise method for population documentation of luteal phase lengths.