Skip to main content

Table 1 Distribution of several characteristics across networks

From: Pattern-mixture model in network meta-analysis of binary missing outcome data: one-stage or two-stage approach?

Characteristic

Susceptible networks1

(n = 11)

Non-susceptible networks1

(n = 18)

All networks

(n = 29)

Total trials per network,

median (minimum, maximum)

21 (11, 104)

9 (4, 15)

13 (4, 104)

Trials per comparison

median (minimum, maximum)

2 (1, 13)

1 (1, 10)

1 (1, 13)

Degree of missing outcome data (%)

 Low

median (minimum, maximum)

0.03 (0.00, 0.57) [2]2

0.02 (0.00, 0.24) [10]

0.03 (0.00, 0.57) [12]

 Moderate and balanced

median (minimum, maximum)

0.12 (0.00, 0.62) [8]

0.09 (0.00, 0.37) [6]

0.11 (0.00, 0.62) [14]

 Moderate and unbalanced

median (minimum, maximum)

0.18 (0.00, 0.45) [1]

0.09 (0.03, 0.27) [1]

0.15 (0.00, 0.45) [2]

 Large and unbalanced

median (minimum, maximum)

–

0.30 (0.03, 0.87) [1]

0.30 (0.03, 0.87) [1]

Factors that affect within-trial normal approximation

 Trial sample size

median (minimum, maximum)

204 (12, 18,201)

364 (74, 8240)

262 (12, 18,201)

 Event risk

median (minimum, maximum)

0.58 (0.00, 1.00)

0.66 (0.12, 0.99)

0.60 (0.00, 1.00)

 Number of zero-cells

median (minimum, maximum)

1 (1, 4) [9]

–

1 (1, 4) [9]

  1. 1A network was ‘susceptible’ to within-trial normality approximation when there was at least one trial with a sample size less than 50 participants and/ or at least one trial-arm with observed event risk less than 5%
  2. 2Brackets indicate the number of networks with the studied characteristic