RESUMO
We analyzed two databases comprising repeated semen analyses of male patients visiting a fertility clinic in the Netherlands to assess intra- and interindividual variability for several semen parameters. We calculated reliability coefficients to estimate the attenuation of the slope if semen parameters are used to predict the probability of conception in regression models. Reliability was lowest for morphology (Population A: R = 0.48, Population B: R = 0.54), but somewhat better for motility (Population A: R = 0.66, Population B: R = 0.71) and count (Population A: R = 0.79, Population B: R = 0.81). Semen concentration had the largest reliability coefficients (Population A: R = 0.84, Population B: R = 0.84). Stimulations using empirical variance components indicated that using semen parameters as endpoints in case-referent studies might substantially underestimate measures of association. The results showed that most semen parameters were just moderately redundant (range: Kappa = 0.28 to Kappa = 0.45); only agreement between concentration and count was substantial (Population A: Kappa = 0.84, Population B: Kappa = 0.80). Hence, epidemiologic studies focusing on male infertility should take into account a variety of parameters.
