To investigate the agreement between the 2- and 3-step methods for the detection of SMDs. Menstrual cycles (MCs, 98) of 59 athletes were assessed using the 2- and 3-step methods. Regular-length MCs (ie,≥21 and ≤35d) were classified as either having no SMD (luteal phase length ≥10d, midluteal progesterone concentration ≥16nmol·L-1, and being ovulatory) or having an SMD (eg,short luteal phase [<10d], inadequate luteal phase [midluteal progesterone concentration <16nmol·L-1], or being anovulatory). Method agreement was assessed using theMcNemar test and Cohen kappa (κ). Substantial agreement was observed between methods (κ = .72; 95% CI, .53-.91), but the 2-step method did not detect all MCs with an SMD, resulting in evidence of systematic bias (χ2 = 5.14; P = .023). The 2-step method detected 61.1% of MCs that had an SMD ([51.4, 70.8]), as verified using the 3-step method, and correctly identified 100% of MCs without an SMD. MCs classified as being disturbed using the 2-step method could be considered valid evidence of SMDs. However, MCs classified without SMDs do not definitively confirm their absence, due to the proven underdetection via the 2-step method.
Read full abstract