There has been some controversy recently about how we should randomize twins (and triplets or more) in neonatal trials.
The question being whether randomizing the babies independently or randomizing them as a group is preferable. Those who wish to randomize as a group think it will help enrollment, without compromising the trial, those who think we should randomize individually are worried about the loss of power, and that there should at least be an analysis of such babies as a cluster. For some studies, especially in the preterm infant, a large proportion of eligible babies may be twins, so it can become an important issue.
Now fortunately someone has asked parents their opinion! And twins!
The authors from Rainbow Babies Hospital in Cleveland contacted about 200 parents of multiples, and over 300 multiples (97% twins) and asked them their preference. They were contacted from numerous sources, and quite a large proportion had actually been in an NICU (about half the parents and 1/3 of the multiples) and/or participated in research (about 19% of parents, and 9% of the multiples).
Most of the correspondents had a preference as to randomization strategy, about 70 to 80%, and most of those, particularly among the parents, preferred randomization to the same group (70 to 80% among parents and 60% among the multiples themselves), very few preferred being randomized into the alternate group (a strategy I didn’t talk about above but which has been proposed as a way of balancing groups), and a few more preferred independent randomization, (between 12 and 20%).
I think this is pretty conclusive, and very important. If parents and babies (or rather former babies) largely prefer being randomized into the same group, then we should probably in the future plan to randomize multiples as clusters, and analyze as clusters.
I gave a presentation at the 2012 PAS meeting about this very issue. In fact you can look at the presentation if you want on this website/blog, under presentations from our group, the 2012 Boston PAS meeting. My conclusion was the the impact of randomization method will really depend on the intra-multiple correlation. In a disease with a strong genetic component, and often similar outcomes between twins (such as BPD) the implications are different than for a disease such as severe IVH where there is much lower correlation between siblings. One of my slides says this :
If infants are randomized to the same group, i.e. as a cluster, then the analysis should take account of that
If there is little correlation between twins (eg severe IVH) then a strict cluster approach, such as the multiple outputation technique will reduce power, and risk type 2 error
If there is strong correlation between twins (eg BPD) then not correcting for clusters will, in general, make confidence intervals more narrow, and risk type 1 error
And even if you independently randomize twins, some of them will end up in the same group by chance ; you should probably, strictly, account for that in the analysis (particularly for diseases with a stronger genetic component), but even worst case scenarios show the total impact is minor.
One of my recommendations at the end of the presentation was that we should look at the impacts on families, I wasn’t exactly thinking of this kind of questionnaire, but I am very happy it was done.
The other question I have from a methodological point of view is this, now that we have this information, and, although clearly in favour of randomization as a cluster, it is far from unanimous:
What if we gave parents, at the time of randomization the option of having both their twins (or all the triplets) in the same group or of being randomized independently?
They could then be counted as either individuals or as a cluster in the randomization schedule, and analyzed as a cluster if randomized as a cluster. It would be a bit more complicated, but I can’t immediately think of any objection, and it would respect parental wishes, and show them how important their views are, at a critical time in the relationship of the researchers with the family.
Neonatal Research 