Differential enrollment in twin registries: its effect on prevalence and concordance rates and estimates of genetic parameters