Paternal age and Down's syndrome genotypes diagnosed prenatally: No association in New York State data

Summary An investigation of a paternal age effect independent of maternal age was undertaken for 98 cases of Down's syndrome genotypes diagnosed prenatally compared to 10,329 fetuses with normal genotype diagnosed prenatally in data reported to the New York State Chromosome Registry. The mean of the difference (delta) in paternal age of cases compared to those with normal genotypes after controlling for maternal age, was slightly negative,-0.27 with a 95% confidence interval of-1.59 to +1.06. A regression analysis was also done in which the data were first fit to an equation of the type lny=(bx+c) and then to the equation ln y=(bx+dz+c) where y = rate of Down's syndrome, x = maternal age, z = paternal age, and b, d, and c are parameters. This also revealed no evidence for a paternal age effect. The value of d (the paternal age coefficient) was in fact slightly negative,-0.0058, with an asymptotic 95% confidence interval of-0.0379 to +0.0263. Lastly, multiple applications of the Mantel-Haenszel test considering various boundaries in paternal age also revealed no statistically significant evidence for a paternal age effect independent of maternal age. These results are at variance with claims of others elsewhere of a very strong paternal age effect detected in studies at prenatal diagnoses. Five different hypotheses are suggested which may account for discrepancies among studies to date in findings on paternal age effects for Down's syndrome: (i) there are temporal, geographic, or ethnic variations in paternal age effects, (ii) there is no paternal age effect and statistical fluctuation accounts for all trends to date; (iii) methologic artifacts have obscured a paternal age effect in some studies which did not find a positive outcome; (iv) methodologic artifacts are responsible for the positive results in some studies to date; (v) there is a rather weak paternal age effect independent of maternal age in most if not all populations, but because of statistical fluctuation the results are significant only in some data sets. The results of all data sets to date which we have been able to analyze by one year intervals are consistent with a mean delta of +0.04 to +0.48 and in the value of d (the paternal age coefficient) of +0.006 to +0.017, and it appears the fifth hypothesis cannot be excluded. Projections based on this assumption are presented.