Table 1

 
Figure 1

 
Figure 2

 

Table 1 Number of missing trials in meta-analyses estimated by trim and fill method and magnitude of change in outcome with random effects model
 

Study number*
No of trials
No estimated missing
Original pooled odds ratio (95% CI)
Adjusted odds ratio† (95% CI)
% reduction in outcome
% change in excess odds ratio
  
Random effects model
Fixed effects model
    
1
10
3
2
2.42 (1.57 to 3.74)
1.99 (1.32 to 3.00)
17.9
30.5
2
10
0
0
0.21 (0.08 to 0.53)
—
0
0
3
11
5
5
1.11 (0.89 to 1.40)
1.01 (0.82 to 1.26)
8.7
87.3
4
15
0
0
1.41 (0.77 to 2.57)
—
0.0
0.0
5
18
3
3
1.23 (1.03 to 1.45)
1.15 (0.98 to 1.36)
5.9
32.1
6
13
4
4
3.00 (2.37 to 3.80)
2.74 (2.13 to 3.53)
8.6
12.9
7
24
3
3
4.34 (2.85 to 6.61)
3.51 (2.24 to 5.52)
19.0
24.7
8
10
0
0
0.71 (0.39 to 1.27)
—
0
0
9
18
0
0
1.45 (1.14 to 1.86)
—
0
0
10
13
6
6
1.62 (1.17 to 2.24)
1.13 (0.79 to 1.63)
30.2
78.9
11
18
5
5
2.31 (1.95 to 2.75)
2.05 (1.70 to 2.47)
11.5
20.2
12
10
0
0
1.10 (0.61 to 1.99)
—
0
0
13
12
2
2
1.36 (1.01 to 1.82)
1.27 (0.94 to 1.71)
6.6
25.0
14
14
1
1
1.20 (0.88 to 1.62)
1.19 (0.88 to 1.61)
1.0
6.0
15
10
0
1
2.32 (1.97 to 2.73)
—
0.0
0.0
16
12
1
1
4.31 (2.53 to 7.35)
3.84 (2.23 to 6.62)
10.9
14.1
17
18
0
0
1.88 (1.57 to 2.24)
—
0
0
18
11
2
2
1.33 (0.94 to 1.89)
1.26 (0.90 to 1.77)
5.4
21.6
19
10
0
0
1.35 (0.92 to 1.98)
—
0
0
20
14
1
1
7.29 (4.29 to 12.42)
6.83 (3.99 to 11.70)
6.3
7.3
21
15
0
0
1.14 (0.88 to 1.48)
—
0
0
22
12
0
0
5.73 (2.89 to 11.38)
—
0
0
23
18
0
4
0.61 (0.35 to 1.07)
—
0
0
24
16
3
3
1.18 (0.81 to 1.74)
1.07 (0.70 to 1.63)
10.1
64.7
25
13
0
0
1.78 (0.67 to 4.71)
—
0
0
26
15
0
0
1.05 (0.88 to 1.26)
—
0
0
27
31
6
5
1.90 (1.57 to 2.31)
1.73 (1.40 to 2.13)
9.1
19.1
28
10
0
0
1.00 (0.69 to 1.43)
—
0
0
29
12
1
1
1.30 (1.05 to 1.61)
1.29 (1.04 to 1.60)
0.9
3.8
30
18
5
5
0.71 (0.41 to 1.23)
0.52 (0.32 to 0.86)
25.9
119.9
31
13
0
0
1.11 (0.87 to 1.41)
—
0
0
32
12
0
0
1.26 (0.90 to 1.77)
—
0
0
33
10
3
3
1.94 (1.29 to 2.93)
1.71 (1.15 to 2.53)
12.1
25.0
34
10
0
0
2.46 (1.72 to 3.50)
—
0
0
35
47
8
8
1.62 (1.46 to 1.80)
1.49 (1.32 to 1.68)
7.9
20.6
36
16
0
0
1.72 (1.40 to 2.12)
—
0
0
37
10
0
0
1.93 (1.52 to 2.47)
—
0
0
38
19
0
0
1.86 (1.41 to 2.12)
—
0
0
39
22
0
0
1.40 (1.03 to 1.90)
—
0
0
40
11
0
0
1.13 (0.76 to 1.66)
—
0
0
41
12
1
2
1.38 (1.10 to 1.73)
1.37 (1.09 to 1.73)
0.7
2.5
42
11
0
0
3.38 (1.75 to 6.51)
—
0
0
43
11
0
0
0.80 (0.63 to 1.00)
—
0
0
44
16
3
3
1.25 (1.06 to 1.49)
1.19 (1.00 to 1.41)
5.2
25.4
45
11
0
4
2.50 (1.74 to 3.60)
—
0
0
46
18
1
1
1.08 (0.97 to 1.21)
1.07 (0.95 to 1.20)
1.1
14.6
47
17
4
4
3.47 (2.12 to 5.67)
2.51 (1.48 to 4.26)
27.7
38.9
48
10
1
1
0.94 (0.42 to 2.11)
0.79 (0.35 to 1.80)
15.5
295.1

* Corresponding to numbers on funnel plots in figure 2.

† Adjusted by trim and fill method with random effects model.  
 

Figure 1

    
 

Illustration of trim and fill method on meta-analysis of effect of gangliosides on mortality from stroke. (Top) original funnel plot; (middle) plot with asymmetric trials trimmed and weighted mean of remaining studies calculated; (bottom) funnel is filled with imputed trials allowing a 95% confidence interval to be calculated for adjusted pooled estimate
   

Figure 2

    

Funnel plots of 48 meta-analyses included in assessment. Scales vary between datasets since analysis does not compare reviews