Skip to main content

Table 2 Estimates of the risk difference

From: Using Bayesian statistics to estimate the likelihood a new trial will demonstrate the efficacy of a new treatment

FirsFirst author

e date

Nb

Events

Nb

Events

Cumulative Evidence

Fictive Trial

    

Mean posterior estimates of outlier probabilities

Estimated risk differencea

(95% CrI)

Pr. risk diff. Above

Additional planned sample size

Predictive probability that proportion of failures in the next sample was in the treated compared to the control

Control

Treated

Control

Treated

5%

10%

5% less

10% less

Chareancholvanich

2013–03-02

40

3

40

1

0.0952

0.0476

−0.048

(−0.161; +0,059)

0.464

0.16

-

48.5%

25.6%

Victor

2013–04-26

64

18

61

15

0.2075

0.1650

−0.042

(−0.100; +0.242)

0.444

0.14

125

47.5%

24.0%

Roh

2013–08-03

48

5

42

5

0.1753

0.1517

−0.024

(−0.107; 0.059)

0.267

0.036

90

33.3%

9.8%

Hamilton

2013–08-06

26

8

26

9

0.1944

0.1813

−0.013

(−0.095; 0.069)

0.187

0.018

52

26.3%

6.6%

Boonen

2013–08-10

82

15

86

26

0.1908

0.2218

0.031

(−0.038; 0.101)

0.011

0.00012

168

5.4%

0.5%

Parratte

2013–08-15

20

2

20

4

0.1844

0.2202

0.036

(−0.031; 0.102)

0.0057

0.00003

40

3.8%

0.2%

Chotanaphuti

2013–09-04

40

5

40

2

0.1770

0.1987

0.022

(−0.039; 0.083)

0.010

0.00004

80

5.2%

0.3%

Woolson

2014–03-07

26

10

22

9

0.1925

0.2124

0.020

(−0.039; 0.081)

0.011

0.00005

48

5.6%

0.3%

Kotela

2014–06-28

46

14

49

24

0.2056

0.2474

0.042

(−0.016; 0.101)

0.001

0.000001

95

1.5%

<0.0%

Pfitzner

2014–07-16

30

13

60

11

0.2217

0.2388

0.017

(−0.039; 0.073)

0.009

0.00002

90

4.9%

0.2%

Yan

2014–09-14

30

13

30

8

0.2357

0.2406

0.005

(−0.050; 0.060)

0.024

0.00009

60

8.4%

0.4%

Abane

2015–01-09

67

22

59

19

0.2476

0.2495

0.002

(−0.050; 0.054)

0.025

0.00006

126

8.6%

0.4%

Molicnik

2015–03-04

19

4

19

0

0.2463

0.2410

−0.005

(−0.057; 0.045)

0.042

0.00013

38

11.2%

0.5%

  1. *Negative values favour the experimental treatment. Number of patients in the control (n. ctr) and experimental (n. exp) groups; number of events (ev. ctr and ev. exp); credibility interval (ctrCrI); probability (Pr)
  2. Estimation of the probabilities that the proportion of outliers in the experimental group is below that observed in the control group by 5 and 10% according to the accumulated evidence. Estimation of the Bayes predictive probability that the risk difference be of at least 5 and 10% in favour on the experimental group
Back to article page

BMC Medical Research Methodology

ISSN: 1471-2288

Contact us