Skip to main content

Table 2 Awareness and evaluation of the transitivity in the protocol and publication of the review

From: Low awareness of the transitivity assumption in complex networks of interventions: a systematic survey from 721 network meta-analyses

Characteristic

Levels

Total

(n=721)

Before

PRISMA-NMA

(n=361)

After

PRISMA-NMA

(n=360)

Odds ratio (95% CI)

n

%

n

%

n

%

Reporting and evaluating transitivity in the protocol

 [1] The study protocol is availablea

Yes

210

29.1

57

15.8

153

42.5

3.94 (2.79, 5.64)

No

511

70.9

304

84.2

207

57.5

 [2] If the protocol is ‘Available’ (210 SRs), the authors defined the transitivity assumption

Yes

31

14.8

9

15.8

22

14.4

0.90 (0.40, 2.17)

No

179

85.2

48

84.2

131

85.6

 [3] If the protocol is ‘Available’ (210 SRs), the authors mentioned that they planned to evaluate the transitivity assumption in the review

Yes

85

40.5

13

22.8

72

47.1

3.01 (1.54, 6.23)

No

125

59.5

44

77.2

81

52.9

Reporting and evaluating transitivity in the systematic review

 [4] The authors defined transitivity

Yes

218

30.2

130

36.0

88

24.4

0.57 (0.42, 0.79)

No

503

69.8

231

64.0

272

75.6

 [5] The authors explicitly stated in the methods that they planned to evaluate transitivity and reported the evaluation results

Yes

442

61.3

190

52.6

252

70.0

2.10 (1.55, 2.86)

No

279

38.7

171

47.4

108

30.0

 [6] The authors did not state in the methods any plans for transitivity evaluation, but evaluation results were found in the manuscript

Yes

30

4.2

19

5.3

11

3.1

0.57 (0.26, 1.19)

No

691

95.8

342

94.7

349

96.9

Acknowledging the implications of the transitivity evaluation

 [7] The authors conclude or imply the (im)plausibility of transitivity, or the difficulty to judge

Yes

265

36.8

128

35.5

137

38.1

1.12 (0.83, 1.51)

No

456

63.2

233

64.5

223

61.9

 [8] Among the reviews with a conclusion about transitivity (265 SRs), some authors explicitly refrained from NMAb

Yes

4

1.5

1

0.8

3

2.2

2.84 (0.36, 57.89)

No

261

98.5

127

99.2

134

97.8

 [9] Among the reviews with a conclusion about transitivity (265 SRs), implications were discussed or implied concerning at least one NMA parameterc

Yes

199

75.1

105

82.0

94

68.6

0.48 (0.27, 0.85)

No

66

24.9

23

18.0

43

31.4

Reporting the table of characteristics

 [10] A table of characteristics is providedd

Yes

680

94.6

336

93.6

344

95.6

1.47 (0.77, 2.88)

No

39

5.4

23

6.4

16

4.4

 [11] If a table of characteristics is provided (680 SRs), the structure of the table facilitates transitivity evaluatione

Yes

669

98.4

331

98.5

338

98.3

0.85 (0.24, 2.85)

No

11

1.6

5

1.5

6

1.7

 [12] Among the reviews with a proper table structure (669 SRs), there is at least one missing characteristic across the trials or comparisons

Yes

564

84.3

272

82.2

292

86.4

1.38 (0.91, 2.10)

No

105

15.7

59

17.8

46

13.6

  1. CI confidence interval, NMA network meta-analysis, PRISMA-NMA PRISMA extension statement for NMA, SR systematic review
  2. aA protocol is considered available when the systematic review reports a PROSPERO number, provides the protocol as supplementary material, or has published the protocol in the same or a different Journal; otherwise, a protocol is considered not available
  3. bThese authors judged transitivity to be questionable or difficult to judge due to limited data (e.g. few trials, low events, poor trial reporting, missing characteristics); hence, they decided to refrain from conducting network meta-analysis
  4. cThe network meta-analysis parameters include the summary treatment effects, intervention hierarchy measures, statistical heterogeneity, and inconsistency evidence (i.e. the difference between direct and indirect estimates and comparing pairwise meta-analysis with network meta-analysis treatment effects for the same comparisons)
  5. dThere was no access to the supplementary material of two eligible articles published before the PRISMA-NMA; hence, we could not extract any necessary information related to the table of characteristics. Only for the items related to ‘Reporting the table of characteristics’, we have restricted the articles published before PRISMA-NMA to those with access to their supplementary material, namely, 359 articles
  6. eTables of characteristics that facilitate transitivity evaluation include those with a trial-level structure (i.e. trials-by-characteristic table), trials grouped by comparison with characteristics at trial-level, or comparison-level summarised characteristics. Tables of characteristics that hinder transitivity evaluation include those with intervention-level summarised characteristics or descriptive statistics for each characteristic
Back to article page

BMC Medicine

ISSN: 1741-7015

Contact us