. 2017 Feb 6;27(10):2885–2905. doi: 10.1177/0962280216688033

Table 1.

Study information and percentage study weights toward the overall treatment effect from a one- and two-stage IPD random effects meta-analyses of 10 hypertension trials.

	Number of patients	Final SBP (mm Hg)		Treatment effect estimates in each trial	Weight values used to derive percentage study weights		Percentage study weights
Trial ID (i)	Control/ treatment	Control mean (sd)	Treatment mean (sd)	Treatment effect estimate^a ${\hat{θ}}_{i}$ , for each trial from first stage of two-stage IPD meta-analysis ( $var ({\hat{θ}}_{i})$ )	Two-stage model W_i	One-stage model (equation (18)) $W_{i} (\hat{β})_{21, 21}$	Two-stage model: Weights derived using equation (12)^b	One-stage model (equation (18)): Weights derived using equation (12)	One-stage model (equation (18)): Weights derived using number of patients^c
1	750/780	139.75 (17.81)	132.85 (16.68)	−6.66 (0.72)	0.095	0.096	11.05	11.06	5.35
2	199/150	179.89 (22.15)	165.06 (20.03)	−14.17 (4.73)	0.064	0.064	7.34	7.32	1.22
3	82/90	170.45 (26.91)	156.88 (21.26)	−12.88 (10.31)	0.043	0.043	4.99	4.99	0.60
4	2371/2427	138.54 (21.26)	130.09 (19.25)	−8.71 (0.30)	0.101	0.102	11.69	11.70	16.79
5	3445/3546	144.25 (17.58)	135.49 (16.32)	−8.70 (0.14)	0.103	0.104	11.93	11.95	24.46
6	1337/1314	164.58 (19.71)	153.99 (20.13)	−10.60 (0.58)	0.097	0.098	11.25	11.25	9.28
7	2371/2365	156.24 (20.12)	145.10 (19.05)	−11.36 (0.30)	0.101	0.102	11.68	11.70	16.57
8	131/137	189.11 (21.90)	171.46 (19.29)	−17.93 (5.82)	0.058	0.058	6.72	6.68	0.94
9	1139/1252	156.55 (16.86)	150.20 (15.84)	−6.55 (0.41)	0.099	0.100	11.50	11.52	8.37
10	2297/2398	165.24 (16.33)	154.87 (16.31)	−10.26 (0.20)	0.102	0.103	11.84	11.86	16.43
					Sum = var( $\hat{θ}$ ) = 0.87	Sum = var( $\hat{θ}$ ) = 0.87	100	100	100

The estimated between-study variance ( ${\hat{τ}}^{2}$ ) from REML was 7.13 in the one-stage analysis, and 7.18 in the two-stage analysis.

Treatment effect is the mean difference in final systolic blood pressure, after adjusting for baseline (i.e. using an ANCOVA model).

Equation (12) is equivalent to equation (7) here.

Kontopantelis and Reeves approach; θ is the summary treatment effect.