Unfortunately studies like the new publication in Early Human Development don’t help anyone.
Chen L-C, Wu Y-C, Hsieh W-S, et al: The effect of in-hospital developmental care on neonatal morbidity, growth and development of preterm taiwanese infants: A randomized controlled trial. Early Human Development 2013, 89(5):301-306. This is a real pity. The authors have put a lot of time and resources into this trial, but there are so many problems with the publication that you can’t tell if they found anything. They randomized 178 VLBW infants into 3 groups.
That it is about all I can tell you for sure about this study. If we ask the simple questions that need to be reported in any publication regarding an RCT I have difficulty answering any of them
1. Registration. Was this trial registered? No information given, and I can’t find an entry on the meta-register of RCTs.
2. Eligibility. The authors state that the infants had to be ‘physiologically stable, i.e., no ventilator use, absence of apnea, bradycardia, or desaturation with or without oxygen use at post-menstrual age (PMA) of 36 weeks’ and ‘hospital discharge prior to PMA of 44 weeks’ but the babies were enrolled at an average PMA of age of 34.5 weeks, that is before they were eligible! Also they use BPD as an outcome criterion, and even use it in their sample size calculation, when by definition infants with BPD were not eligible.
3. Intervention. There were 3 groups, controls who got regular developmental care in the hospital, a clinic based intervention group, and a hospital based intervention group. I can’t really tell you what the intervention was, they say they got feeding support, massage and parent support and education. They also talk about family centered care, but there is no real details given for any of this bundle of interventions.
4. Outcomes. There is no primary outcome mentioned anywhere in the document. The table of results shows 18 outcome measures with significance testing on each. They have selected 2 which they say were significant, RoP grades 2 and 3, and desaturation during feeding. Now my gentle readers all know, I am sure, that this is horse dung. I don’t understand why the reviewers of this article didn’t know that. Clearly if you test 18 outcomes and for each one you use a significance level of 1 in 20, its pretty likely that you will find one to be significant.In fact the probability that at least one will be significant is 1-(0.95^18) (I think) which is 0.6 (ish). So in any study if you do 18 significance tests and use 0.05 as your threshold for significance, you will find something positive 60% of the time, even in a completely random set of numbers.
Also interesting is that more advanced RoP was an exclusion criterion. So they knew that most of the development of retinopathy occurs before they were even in the trial, but then they use it to show an effect of the intervention.
This study makes no sense! The authors note in the discussion that the study was unable to show an effect on duration of ventilation, when the infants were only eligible if they weren’t being ventilated.
How on earth did this get published? I went on the website of the journal ‘Early Human Development’ to see if their instructions for authors mention the CONSORT statement. That is a statement giving standards for the reporting of RCTs, which most quality journals have signed up to; they require conformity with its standards prior to accepting a report of an RCT. Any comparison of this report with CONSORT would be a clear and obvious fail.
The instructions for authors page is blank. Interesting.
But even the CONSORT statement doesn’t say that you can’t use outcome variables that are diagnosed before you enter the study, or that you can’t use eligibility criteria that occur after the babies are enrolled. I guess the CONSORT group never thought about that, maybe it should be in the next version?