Table 4.

Sample size calculations for a definite RCT with primary outcome of BMI SDs at 24 months under different assumptions

	80% power			90% power
	Minimum difference detectable			Minimum difference detectable
ICC	0.25	0.30	0.50	0.25	0.30	0.50
0.01	571	397	145	764	531	193
0.03	947	658	240	1267	880	320
0.06	1511	1050	383	2021	1404	511
0.10	2262	1572	573	3027	2103	765

All numbers are the total number of children required to be recruited, assuming 50 children per school and a loss to follow-up at 24 months of 20%.

The calculations take into account the reduction in variability associated with adjusting for baseline BMI SDs, conservatively assuming the correlation between baseline and 24 months data to be 0.8 (in our pilot study, this correlation was 0.93).

This is the sample size required to detect a true minimal difference in BMI SDs of 0.25, 0.3 or 0.5, assuming the SD is 1.3 (based on our pilot data).

BMI, body mass index; ICC, intraclass correlation coefficient; RCT, randomised controlled trial.