The interaction between dysmorphology and epidemiology: methodologic issues of lumping and splitting.