Table 1 Summary of studies included in sensitivity analysis

Number of comparisons
Number of similar results
Number better in RCT
Number worse in RCT

Selection bias:

Low risk
6
6
—
—

No imbalance
22
19
1
2

Adjusted results
12
7
4
2

Partially controlled
18
15
2
1

Poorly controlled
33
30
3
—

Detection bias:

Low risk^*
78
66
8
3

Partially met, similarly measured
5
3
2
1

High risk
8
8
—
—

Exclusion bias:

No losses to follow up
38
32
4
2

1 person to 20% lost to follow up
33
30
2
1

> 20% loss to follow up
11
10
—
1

Unclear
9
4
4
1

Reasons for non-participation:

Refused
27
18
7
2

Refused because of preference
23
20
1
2

RCT versus preference trial
16
15
—
1

Not invited
8
8
—
—

Treated by non-participating clinicians
1
1
—
—

Different skills required for treatment:

Surgery and procedures
28
25
1
2

Drug treatment
22
17
4
1

Radiology
14
12
1
1

Usual care
9
8
1
—

Counselling and education
8
8
—
—

Watchful waiting
7
6
—
1

Different clinical area:

Oncology
28
23
3
2

Obstetrics and gynaecology
14
12
2
1

Cardiology
13
10
2
1

Other internal medicine
11
9
2
—

Psychology and drug misuse
9
9
—
—

Paediatrics
8
7
—
1

Respiration
2
2
—
—