Table 2.

Summary of results highlighting the impact of right-truncation in every tercile (denoted by T = 1 and 2) on the calculations of statistical power over a range of settings where statistical power was initially between 80% and 85% when no truncation was present (T = ∞) for multiple cohorts recruited over the trial period. The pre-specified between-cluster variance, ${\sigma }_c^{\mathrm{ }2},$ is the value of the between-cluster variance inputted for simulation of the trial data, and the average empirical between-cluster variance is the mean between-cluster variance estimated from the simulated datasets for each combination of λ and ${\sigma }_c^{\mathrm{ }2}$. The average empirical ICC is the mean intracluster correlation computed as described by Stryhn et al.²⁶ from the simulated datasets

						Statistical power (%)
	Number of clusters per tercile	Cluster sizes	Pre-specified between-cluster variance (${\sigma }_{\mathit{c}}^{\mathbf{}2}$)	Average empirical cluster variance	Average empirical ICC	T = ∞ per tercile	T = 2 per tercile	T = 1 per tercile
Annual event rate(λ) = 2.7	25	10	0.050	0.041	0.378	82.5	70.8	44.1
	30	30	0.100	0.092	0.592	82.4	78.5	70.9
	55	20	0.200	0.189	0.751	82.7	76.0	70.9
	80	30	0.300	0.294	0.823	82.6	77.7	75.1
	110	25	0.400	0.392	0.861	82.9	81.8	80.5