Articles
1392
www.thelancet.com
Vol 388 October 1, 2016
reported in accordance with the guidance from the
National Research Ethics Service,
which is a subdivision of
the National Patient Safety Agency.
Statistical analysis
We calculated that with 170 participants in each group,
the study would have 90% power at 5% signifi cance level
to detect a diff erence in mean EQ-5D score of 0·35 SD,
which represents an absolute change in score of 0·05
20
and is likely to be clinically important. We estimated that
the SD for intraocular pressure at 36
months would be
5 mm Hg and similar in the two randomised groups.
24–26
We therefore calculated that the study would have 90%
power at a 5% level of signifi cance to detect a mean
diff erence of 1·75 mm Hg. Additionally, with the
assumption that a maximum of 40%
of patients would
need glaucoma surgery, the power to detect a diff erence
of 15% in the need for glaucoma surgery would be 80%.
Thus, allowing for 15% loss to follow-up at 36 months,
we aimed to recruit 400 patients.
All the main analyses were based on intention to treat
(ITT) and were done at the end of the trial. Signifi cance
was set at 5% in the main analysis and 1% in the
subgroup analyses.
To assess the primary outcomes, we
used a repeated measures mixed eff ects model to analyse
the EQ-5D scores and intraocular pressure,
27
based on
the follow-up data obtained at 6, 12, 24, and 36 months.
Chia sẻ với bạn bè của bạn: