Objective To determine the accuracy of polymerase chain reaction (PCR) analysis of amniotic fluid for fetal toxoplasmosis according to clinical predictors of outcome and study centre.
Design Prospective cohort study.
Setting Nine European centres.
Population Women with suspected toxoplasma infection identified by prenatal screening.
Methods Logistic regression was used to examine the effects of gestational age at maternal seroconversion, treatment and timing of amniocentesis, on PCR accuracy, and to calculate the post-test probability of congenital toxoplasmosis.
Main outcome measures Infants had congenital toxoplasmosis if specific IgG persisted beyond 11.5 months. Uninfected infants had undetectable IgG in the absence of anti-toxoplasma treatment.
Results Of 593 PCR results, 64 were positive (57 confirmed infected), and 529 were negative (23 confirmed infected). The likelihood ratio for a positive PCR result decreased significantly with trimester at seroconversion, but did not change significantly for a negative result. Weak associations were detected between sensitivity and, inversely, with specificity, and gestational age at maternal seroconversion. There was no significant association between sensitivity and centre, type or duration of treatment, or timing of amniocentesis. Specificity differed significantly between centres (P < 0.001). The change in pre- to post-test probability of infection was maximal for a positive PCR after first trimester seroconversion, affecting 1% of women tested, and a negative PCR after third trimester seroconversion, affecting half the women tested.
Conclusions Prediction of the risk of congenital toxoplasmosis should combine estimates of test accuracy and maternal–fetal transmission, which take account of the gestational age at which the mother seroconverted. Local laboratory standards will affect the generalisability of these results.