This article is from the Pregnancy Screening FAQ, by Lynn Gazis-Sax (firstname.lastname@example.org) with numerous contributions by others.
Amniocentesis is generally accurate. The most common error is a cell
culture failure, in which case the amniocentesis may need to be repeated.
This happens about 2% of the time (Scher and Dix). Other possible sources
of error are maternal cell contamination (so that cells from the mother
are examined instead of cells from the fetus), artifacts of the testing
process (such as pseudo-mosaicism), or mislabelling of the sample. (In
order to avoid these errors, labs will check a certain number of cells
before diagnosing an abnormality.) Better results are obtained from more
There are also some results which are inherently ambiguous, because we
don't really know the effect of the chromosomal abnormality detected.
Trisomy 21, or Down syndrome, is a well-defined syndrome (although,
even there, amniocentesis will not tell how severe a case of Down
syndrome the baby will have), but the results of sex chromosome
trisomies are less well known. Is XYY associated with more violent
behavior or isn't it? Some reports say yes, others no. The effects
of XXX are also unknown. For more discussion of these ambiguous
diagnoses, see Barbara Katz Rothman's book, _The Tentative Pregnancy_.