Table 5

Estimated sample sizes for two sided and one sided testing, based on the paired placebo event rates, calculated for five hypothetical situations: drugs working according to the RRR model (two), the ARR model (two), and a mixed model (one)

		RRR constant				ARR constant				Mixed model
	Placebo event rate(%)	RRR = 50%		RRR = 75%		ARR = 15%		ARR = 25%		Intervention event rate = 5%
Patient group	Placebo event rate(%)	2 sided	1 sided	2 sided	1 sided	2 sided	1 sided	2 sided	1 sided	2 sided	1 sided
Early RA, disease duration <2 years; late RA, disease duration ⩾2 years; RRR, relative risk reduction = (1−(negative event rate intervention group – negative event rate placebo group)×100); ARR, attributive risk reduction = (negative event rate placebo group – negative event rate intervention group); event rate, percentage of patients with progression of joint damage larger than smallest detectable change (SDC) based on the paired Sharp/van der Heijde change-score.
*ARR 15% or 25% not feasible owing to a placebo event rate smaller then 15%, estimated sample size based on the maximal ARR (9%); †ARR 25% not feasible owing to a placebo event rate smaller then 25%, estimated sample size based on the maximal ARR (20% and 17%).
Early RA
Regardless of baseline damage	9	463	365	172	135	76*	60*	76*	60*	151	119
Medium + high baseline damage, score ⩾4	20	191	151	71	56	75	59	31†	24†	42	33
High baseline damage, score ⩾21	36	93	73	34	27	138	109	25	20	18	14

Late RA
Regardless of baseline damage	17	232	183	86	68	59	46	38†	30†	54	43
Medium + high baseline damage, score ⩾4	20	200	157	74	58	71	56	32†	26†	45	35
High baseline damage, score ⩾21	25	150	118	55	44	97	77	24	19	31	25