Skip to main content

Table 3 Sample sizes for hypothetical randomised controlled trials with different analysis plans/outcome variable

From: Sample size considerations for trials using cerebral white matter hyperintensity progression as an intermediate outcome at 1 year after mild stroke: results of a prospective cohort study

Predictors included in RCT analysis

Outcome variable at 1-year follow-upa

Sample size (n per group)

Total sample size

Allowance for 20% dropout

Treatment group only (assumed difference 1 ml)

WMH volume

8508

17,016

21,270

Treatment group + baseline WMH volume (R 2 = 0.88)

WMH volume

987

1974

2468

Treatment group + age (R 2 = 0.19)

WMH volume

6872

13,744

17,180

Treatment group + baseline Fazekas score (R 2 = 0.69)

WMH volume

2628

5256

6570

Treatment group + all predictors used in multivariable analysis (R 2 = 0.90)

WMH volume

821

1642

2054

Treatment group + all predictors used in multivariable analysis (R 2 = 0.90) and 2-year follow-up

WMH volume

205

410

642b

Treatment group only (assumed OR 1.2)

Fazekas score

1499

2998

3748

Treatment group only (assumed OR 1.25, upper 95% CI)

Fazekas score

1001

2002

2504

Treatment group only (assumed OR 1.16, lower 95% CI)

Fazekas score

2262

4524

5656

Treatment group + baseline Fazekas (multiplierc 5.04)

Fazekas score

7554

15,108

18,886

Treatment group + baseline hypertension (multiplierc 1.00057)

Fazekas score

1500

3000

3750

Treatment group + baseline age (multiplierc 1.012)

Fazekas score

1515

3030

3788

Treatment group only (median increase 0.5)

Fazekas score

87,216

174,432

218,040

  1. RCT Randomised controlled trial, WMH White matter hyperintensity
  2. a Except for second multivariate sample size estimate, where 2-year follow-up was used
  3. b A 20% dropout per year was used for the 2-year follow-up (i.e., a total dropout rate of 36%)
  4. c The multiplier is the number by which the base sample size must be multiplied to allow for the inclusion of an additional predictor in the analysis without loss of power; the base sample size is that for an analysis with Fazekas score as outcome and treatment group as the only predictor