1. The trial protocol (Contents page), in the
supplementary data [1] states that the Randomization procedure is
presented in Section 5.2 on page 23; however Section 5.2 described
Laboratory procedures (and not Randomization). The trial protocol
published in the journal ‘Trials’ [2] also does not describe the
random sequence generation process; it only states: "A block
randomization procedure with variable block sizes will be used to
maximize unpredictability". Thus, as pointed out in the Commentary
[3], the procedure used to generate the randomization sequence is
‘unclear’.
2. The authors’ assertion: "computer-generated
block randomization done independently by the USA based study
statistician" is not found either in the Supplementary data [1] or
the published trial protocol [2]. The published paper [1] states
that the external statistician performed allocation concealment (and
not random sequence generation). The Commentary [3] already reported
that allocation concealment was adequate.
3. The supplementary data [1] (Contents page)
states that Blinding is described in Section 5.3 on page 23; however
this section is missing in the text. The published protocol [2]
states "TSB will be estimated using standard methods" without
commenting on blinding. Thus it does not appear that blinding of the
outcome assessor was done, hence it was described as ‘Inadequate’
[3]. Although many randomized trials cannot (and need not) include
blinding of outcome assessors, the importance in this trial has
already been highlighted previously [3].
4. The Supplementary data [1] and published
protocol [2] have two different sample size calculations. The former
describes a sample size of 124 infants (days of phototherapy not
mentioned), whereas the latter describes the sample size as 560
treatment days. Neither affects the assessment that there is lack of
clarity for information provided on the number of infants in the
safety analysis [3]. It is important to note that this criterion has
to be evaluated in trials for each outcome.
5. What can we learn (and apply) from this trial?
Filtered sunlight could be efficacious for mild(er) neonatal
jaundice (recall that the threshold was 3 mg/dL lower than standard
practice) and can be used if (i) intensive monitoring is
performed (as in the trial) and (ii) adequate backup
phototherapy units are available (as about 1 in 7 babies would
require phototherapy). Unfortunately, the trial does not explore
whether we can predict which babies will require phototherapy,
making it necessary to have back-up arrangements.