Quantifying heterogeneity in individual participant data meta-analysis with binary outcomes

Table 3 Sensitivity of heterogeneity measures to accounting for effect modification when prevalence of the outcome was 30%

			Two-stage approach		One-stage approach
τ ²	Number of studies	Strength of effect modification^a	\(\frac {I^{2}_{\text {emod}}}{I^{2}_{\text {crude}}}\)	\(\frac {R^{2}_{\text {emod}}}{R^{2}_{\text {crude}}}\)	\(\frac {I^{2}_{\text {emod}}}{I^{2}_{\text {crude}}}\)	\(\frac {R^{2}_{\text {emod}}}{R^{2}_{\text {crude}}}\)
0.5	15	Weak	0.17 (1.00)	0.89 (0.38)	1.00 (0.41)	0.91 (0.25)
1.0	15	Weak	0.01 (0.57)	0.60 (0.34)	0.99 (0.07)	0.81 (0.25)
1.5	15	Weak	0.06 (0.59)	0.50 (0.30)	0.98 (0.05)	0.75 (0.18)
0.5	15	Moderate	0.02 (1.00)	0.86 (0.39)	0.56 (0.68)	0.86 (0.34)
1.0	15	Moderate	0.08 (0.87)	0.68 (0.46)	0.82 (0.28)	0.98 (0.40)
1.5	15	Moderate	0.34 (0.77)	0.65 (0.39)	0.82 (0.25)	1.06 (0.46)
0.5	15	Strong	0.01 (1.00)	0.82 (0.42)	0.11 (0.23)	0.90 (0.37)
1.0	15	Strong	0.32 (1.00)	0.81 (0.43)	0.20 (0.28)	1.07 (0.57)
1.5	15	Strong	0.42 (0.93)	0.75 (0.46)	0.22 (0.29)	1.31 (0.86)
0.5	30	Weak	0.01 (1.00)	0.78 (0.39)	1.00 (0.27)	0.88 (0.19)
1.0	30	Weak	0.01 (0.36)	0.53 (0.22)	1.00 (0.04)	0.78 (0.16)
1.5	30	Weak	0.16 (0.49)	0.47 (0.22)	0.98 (0.03)	0.74 (0.14)
0.5	30	Moderate	0.01 (1.00)	0.77 (0.40)	0.59 (0.46)	0.86 (0.29)
1.0	30	Moderate	0.01 (0.54)	0.65 (0.30)	0.82 (0.19)	0.99 (0.33)
1.5	30	Moderate	0.19 (0.61)	0.59 (0.30)	0.82 (0.17)	1.07 (0.31)
0.5	30	Strong	0.01 (1.00)	0.79 (0.37)	0.09 (0.15)	0.95 (0.37)
1.0	30	Strong	0.01 (0.68)	0.70 (0.34)	0.16 (0.17)	1.10 (0.46)
1.5	30	Strong	0.33 (0.73)	0.71 (0.33)	0.16 (0.17)	1.23 (0.69)

Median (IQR) was presented
We present the ratios of the measure estimated from a model that ignored the effect modifier to one that included the effect modifier and an interaction term between it and the treatment status
^aEffect modification was classified as weak when β _w=1, β _xw=1, as moderate when β _w=1, β _xw=3, and as strong when β _w=2, β _xw=5

ISSN: 2046-4053