Statistically twins

Article Outline

• Copyright

Premature birth is common for twins and higher multiple births. Consequently, clinical studies of preterm infants must consider how to deal with twins, and there are three options: exclude twins, randomize them separately, or randomize the twins as a “cluster” of two to the same intervention group. Excluding twins means that 20-30% of the very low birth weight infants population will not be available for the trial. Parents seem to prefer that both infants receive the same randomization. Biologically, twins are not independent as they have shared genetics and maternal environment as fetuses. Thus, there is a problem with dealing with twins statistically, a situation addressed by Hibbs et al in this issue of The Journal. They report that many trials in neonatal medicine ignore the problem. Using one recent trial of inhaled nitric oxide to prevent bronchopulmonary dysplasia, they demonstrate the benefit of inhaled nitric oxide with twins analyzed as clusters and no benefit when the analysis assumed independence for each twin. The history of twins studies in clinical research is very rich, but twins can also complicate the interpretation and statistical analysis of trials when they make up a large percent of the randomized patients.