Table 5

Methodological assessment sample size calculations and trial justification in SW-CRTs, among those studies reporting a sample size calculation

All reports
N=45
1987–2012
N=18
2013–2014
N=27
Absolute difference (95% CI)p Value
Allowance for clustering
 Number (%)33 (73.3)11 (61.1)22 (81.5)20.4 (−6.5 to 47.2)0.130
Allowance for time effects
 Number (%)*15 (33.3)3 (16.7)12 (44.4)27.8 (2.3 to 53.2)0.063
Allowance for repeated measurements†*
 Number (%)3/24 (12.5)2/11 (18.2)1/13 (7.7)−10.5 (−37.5 to 16.5)0.576
Power methodology
 Hussey and Hughes14 (31.1)3 (16.7)11 (40.7)24.1 (−1.2 to 49.4)0.087
 Other, allowing for time effects*2 (4.4)0 (0)2 (7.4)7.4 (−2.5 to 17.3)0.509
 Other, not allowing for time effects14 (31.1)10 (55.6)4 (14.8)−40.7 (−67.3 to −14.2)0.004
 Not stated15 (33.3)5 (27.8)10 (37.0)9.3 (−18.3 to 36.8)0.519
Power methodology for additional features
 Transition periods0 (0)0 (0)0 (0)
 Interactions (eg, lag effects)0 (0)0 (0)0 (0)  
 Extended correlations*2 (4.4)2 (11.1)0 (0)−11.1 (−25.6 to 3.4)0.155
 Varying cluster size*3 (6.7)1 (5.6)2 (7.4)1.9 (−12.6 to 16.3)1.000
Variation in outcomes across clusters‡
 Reported using ICC20/33 (60.6)8/11 (72.7)12/22 (54.6)−18.2 (−51.7 to 15.4)0.314
 Reported using CV*10/33 (30.3)2/11 (18.2)8/22 (36.4)18.2 (−12.2 to 48.6)0.430
 Reported using DE*1/33 (3.0)1/11 (9.1)0/22 (0)−9.1 (−26.1 to 7.9)0.333
 Reported using between cluster variation*2/33 (6.1)0/11 (0)2/22 (9.1)9.1 (−2.9 to 21.1)0.542
  • Values are numbers (percentages) unless stated otherwise by year of publication.

  • p Value is for the comparison of 1987–1990 publications and 2013–2014 publications using a χ2 test for proportions or (*) using Fisher's exact test.

  • †Among those with a cohort design.

  • ‡As a percentage of studies for which some measure of variation was reported.

  • CV, coefficient of variation; DE, design effect; ICC, intracluster correlation; SW-CRT, stepped-wedge cluster randomised trial.