Evaluating outcomes of newborn screening programs.

Abstract

Newborn screening is a medical intervention. For every program, there should be evidence of its effect and effectiveness. The four questions to be addressed, very broadly, are: What is the effectiveness for case-finding (sensitivity, specificity, and positive predictive value)? What are the benefits of early detection versus clinical detection? What harm results from the program? Are the costs reasonably balanced in relation to benefits? Ideally, there would be randomized controlled trials (RCTs) of screening for each disorder. In practice, power calculations reveal that for very rare disorders this is not feasible. The numbers of screened and unscreened babies required would be huge, and trials would last for decades. There have only been RCTs of newborn screening for cystic fibrosis (birth prevalence 25-40 per 100,000 in Caucasians). No such trials were ever attempted for hypothyroidism, with a similar birth prevalence, and it may not now be ethical to mount one. Instead, lower orders of evidence must be used. Double-blind randomized controlled trials should be planned for not-so-rare disorders where possible. Where it is not feasible, careful planning and collection of data, plus the use of both historical controls and contemporaneous controls from other regions may have to suffice. To introduce programs with no plans for full evaluation is not best practice. Evaluation of outcomes of all kinds, not simply of case-finding, must be mandatory. Data for case-finding should be collected actively, with systematic searching for missed cases. Data about benefits need to be collected in well-planned long-term studies, although short-term benefits are also valuable. Good studies of harm, mainly from false positive results, are urgently needed. The problem of costs and benefits is difficult, and a "reasonable balance" rather than positive cost/benefit ratio seems desirable.