Abstract
Semen parameters are variable within individuals, but it is unclear whether 1 semen sample could represent a man's long-term average values in epidemiologic studies. Between 2005 and 2014, a total of 329 men from a fertility clinic in Boston, Massachusetts, provided 768 semen samples as part of the Environment and Reproductive Health (EARTH) Study. Total sperm count, sperm concentration, morphology, motility, and ejaculate volume were assessed. We used linear mixed models to compare values from men's first semen samples with their long-term averages and to calculate intraclass correlation coefficients for each parameter. We calculated positive predictive values (PPVs) and negative predictive values (NPVs) by comparing agreement in classification according to World Health Organization reference limits. There were no differences in mean semen parameters between men's first samples and the remaining replicates. Intraclass correlation coefficients ranged from 0.61 for morphology to 0.75 for concentration, indicating consistently greater between-man variability than within-man variability. Nevertheless, using 1 sample alone resulted in high NPVs but low PPVs (range, 43%-91%). The average of 2 samples was needed to achieve high PPVs (range, 86%-100%) and NPVs (range, 91%-100%). We conclude that 1 semen sample may suffice for studies aimed at identifying average differences in semen quality between individuals. Studies aimed at classifying men based on World Health Organization reference limits may benefit from collection of 2 or more samples.