Some of the generally accepted criteria for screening programmes are inappr
opriate for newborn metabolic screening as they ignore the family dimension
and the importance of timely genetic information. Uncritical application o
f such criteria creates special difficulties for screening by tandem mass s
pectrometry, which can detect a range diseases with widely different natura
l histories and responsiveness to treatment. Further difficulties arise fro
m increasing demands for direct proof of the effects of screening on long-t
erm morbidity and mortality. The randomized controlled trial is held to be
the gold standard, but for ethical and practical reasons it will be impossi
ble to achieve for such relatively rare diseases. This approach also oversi
mplifies the complex matrix of costs and benefits of newborn metabolic scre
ening. A more workable approach could involve Bayesian synthesis, combining
quantitative performance data from carefully designed prospective pilot st
udies of screening with existing experience of the natural history, diagnos
is, and management of the individual disorders concerned.