TABLE 1.

Sample Size Requirements to Detect a Treatment Effect Using ADL Total Versus IRT Scoring, as Estimated From the Training (HC Trial) and Validation (DHA Trial) Data Sets

Trial Length (mo)	N/arm				%-Reduction
	HC		DHA
	ADL	IRT	ADL	IRT	HC	DHA
12	486	452	490	442	7.1%	9.8%
18	322	274	341	282	15%	17.4%
24	267	214	291	228	19.8%	21.7%

Both N/arm and %-Reduction assume no dropout and equal allocation.

ADL indicates activity of daily living; DHA, the Alzheimer’s Disease Cooperative Study (ADCS) docosahexaenoic acid trial; HC, the ADCS homocysteine trial on which the IRT model was trained; IRT, item-response theory; N/arm, sample size per arm, to detect a 25% reduction in the mean rate of decline, with observations every 6 months, power=0.8, 2-tailed significance level=0.05; %-Reduction, the percentage reduction in required sample size for IRT score versus ADL total scoring.