On 2014 Sep 15, Andrea Messori commented:

Bayesian models implemented under Winbugs: can they be considered the new standard for conducting a network meta-analysis?

When multiple agents are available to treat the same disease condition, network meta-analysis of randomized controlled trials (RCTs) has the purpose of synthesising the available evidence. In the application of this technique, both direct and indirect comparisons are made between individual treatments. Direct comparisons are those for which a “real” randomised study is available while indirect comparisons are those for which no real trial has been conducted.

Initial approaches for conducting network meta-analyses were designed to separately evaluate each indirect comparison; hence, there were as many separate analyses as the number of indirect comparisons [1,3]. More recently, the Bayesian approach for network meta-analysis has emerged as the new standard in this area [4-7]. This method relies on a sort of “all-in-one” modelling in which a single model incorporates the evidence available from all clinical trials and estimates, through a unified approach, all comparative results for both direct and indirect comparisons.

In network meta-analysis, the phases of literature search and data extraction do not differ from those commonly employed for standard meta-analysis [5,6]. As regards the type of end-points, network meta-analysis favours binary end-points [4] as opposed to continuous ones. Hence, irrespective of whether the analysis is aimed at evaluating effectiveness or safety, the starting material for conducting a network meta-analysis is given by the rates of end-point occurrence (i.e. numerator and denominator) for each arm of each RCT.

In present times, statistical modelling for network meta-analysis recognises its new standard in these Bayesian approaches based on an “all-in-one” model and implemented in the WINBUGS environment [4,7]. This approach has found a wide acceptance also because the methods and the software, developed by an authoritative institution (the NICE), are freely available [7]. The code for these estimations is available as fixed-effect model and random-effect model.

The Winbugs Bayesian model employs a random sequence of chains, called a Markov chain Monte Carlo simulation. Each chain must be run for a length of time sufficient to allow model convergence (burn-in) before estimating posterior probabilities. Typically, the random-effect logistic regression model is created according to the binary outcome of subjects reaching the end-point concerned. Randomization within each study is preserved by specifying each arm in each study separately, thus accounting for the effect of the comparator. Results are generally presented as odds ratio or log odds ratio. Heterogeneity among studies can be accounted by applying meta-regression techniques and by consequently generating an index of heterogeneity [4,7].

As regards the software, these analyses can be conducted by using the software package WinBUGS 1.4.3 (Cambridge, United Kingdom) in combination with the meta-analysis code developed by the National Institute for Health and Care Excellence[7]. Odds-ratio is the typical outcome measure of this software; however, odds-ratios can be converted into risk ratios or risk differences by application of standard equations [8,9]

Sobieraj and co-workers [10] have published an article that reviews all previously published studies that have adopted the Bayesian approach.

In conclusion, the present literature clearly indicates that Bayesian network meta-analysis can be considered the new standard in this field.

References

1. Lumley T. Network meta-analysis for indirect treatment comparisons. Stat Med 2002, 30; 21(16): 2313-24

2. Bucher HC, Guyatt GH, Griffith LE, Walter SD. The results of direct and indirect treatment comparisons in meta-analysis of randomized controlled trials. J Clin Epidemiol 1997;50(6):683–91

3. Wells GA, Sultan SA, Chen L, Khan M, Coyle D. Indirect treatment comparison [computer program]. Version 1.0. Ottawa: Canadian Agency for Drugs and Technologies in Health; 2009. Software available at:http://www.cadth.ca/preview/en/resources/itc-user-guide

4. Greco T, Landoni G, Biondi-Zoccai G, D'Ascenzo F, Zangrillo A. A Bayesian network meta-analysis for binary outcome: how to do it. Stat Methods Med Res. 2013 Oct 28.

5. Hoaglin DC, Hawkins N, Jansen JP,. et al. Conducting Indirect-Treatment-Comparison and Network-Meta-Analysis Studies: Report of the ISPOR Task Force on Indirect Treatment Comparisons Good Research Practices—Part 2. Value Health 2011;14:429-37.

6. Jansen JP, Fleurence R, Devine B, et al. Interpreting Indirect Treatment Comparisons and Network Meta-Analysis for Health-Care Decision Making: Report of the ISPOR Task Force on Indirect Treatment Comparisons Good Research Practices: Part 1. Value Health 2011;14:417-28.

7. NICE Clinical Guidelines, No. 92. National Clinical Guideline Centre – Acute and Chronic Conditions (UK). London: Royal College of Physicians (UK); 2010. Available at http://www.ncbi.nlm.nih.gov/books/NBK116530/ Accessed 14 August 2014

8. Zhang J, Yu KF. What's the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA. 1998 Nov 18;280(19):1690-1.

9. ClinCalc Website. Odds-ratio to risk ratio. http://clincalc.com/Stats/ConvertOR.aspx

10. Sobieraj DM, Cappelleri JC, Baker WL, Phung OJ, White CM, Coleman CI. Methods used to conduct and report Bayesian mixed treatment comparisons published in the medical literature: a systematic review. BMJ Open. 2013 Jul 21;3(7). pii:e003111. doi: 10.1136/bmjopen-2013-003111. Print 2013.

On 2014 Sep 15, Andrea Messori commented:

Bayesian models implemented under Winbugs: can they be considered the new standard for conducting a network meta-analysis?

When multiple agents are available to treat the same disease condition, network meta-analysis of randomized controlled trials (RCTs) has the purpose of synthesising the available evidence. In the application of this technique, both direct and indirect comparisons are made between individual treatments. Direct comparisons are those for which a “real” randomised study is available while indirect comparisons are those for which no real trial has been conducted.

Initial approaches for conducting network meta-analyses were designed to separately evaluate each indirect comparison; hence, there were as many separate analyses as the number of indirect comparisons [1,3]. More recently, the Bayesian approach for network meta-analysis has emerged as the new standard in this area [4-7]. This method relies on a sort of “all-in-one” modelling in which a single model incorporates the evidence available from all clinical trials and estimates, through a unified approach, all comparative results for both direct and indirect comparisons.

In network meta-analysis, the phases of literature search and data extraction do not differ from those commonly employed for standard meta-analysis [5,6]. As regards the type of end-points, network meta-analysis favours binary end-points [4] as opposed to continuous ones. Hence, irrespective of whether the analysis is aimed at evaluating effectiveness or safety, the starting material for conducting a network meta-analysis is given by the rates of end-point occurrence (i.e. numerator and denominator) for each arm of each RCT.

In present times, statistical modelling for network meta-analysis recognises its new standard in these Bayesian approaches based on an “all-in-one” model and implemented in the WINBUGS environment [4,7]. This approach has found a wide acceptance also because the methods and the software, developed by an authoritative institution (the NICE), are freely available [7]. The code for these estimations is available as fixed-effect model and random-effect model.

The Winbugs Bayesian model employs a random sequence of chains, called a Markov chain Monte Carlo simulation. Each chain must be run for a length of time sufficient to allow model convergence (burn-in) before estimating posterior probabilities. Typically, the random-effect logistic regression model is created according to the binary outcome of subjects reaching the end-point concerned. Randomization within each study is preserved by specifying each arm in each study separately, thus accounting for the effect of the comparator. Results are generally presented as odds ratio or log odds ratio. Heterogeneity among studies can be accounted by applying meta-regression techniques and by consequently generating an index of heterogeneity [4,7].

As regards the software, these analyses can be conducted by using the software package WinBUGS 1.4.3 (Cambridge, United Kingdom) in combination with the meta-analysis code developed by the National Institute for Health and Care Excellence[7]. Odds-ratio is the typical outcome measure of this software; however, odds-ratios can be converted into risk ratios or risk differences by application of standard equations [8,9]

Sobieraj and co-workers [10] have published an article that reviews all previously published studies that have adopted the Bayesian approach.

In conclusion, the present literature clearly indicates that Bayesian network meta-analysis can be considered the new standard in this field.

References

1. Lumley T. Network meta-analysis for indirect treatment comparisons. Stat Med 2002, 30; 21(16): 2313-24

2. Bucher HC, Guyatt GH, Griffith LE, Walter SD. The results of direct and indirect treatment comparisons in meta-analysis of randomized controlled trials. J Clin Epidemiol 1997;50(6):683–91

3. Wells GA, Sultan SA, Chen L, Khan M, Coyle D. Indirect treatment comparison [computer program]. Version 1.0. Ottawa: Canadian Agency for Drugs and Technologies in Health; 2009. Software available at:http://www.cadth.ca/preview/en/resources/itc-user-guide

4. Greco T, Landoni G, Biondi-Zoccai G, D'Ascenzo F, Zangrillo A. A Bayesian network meta-analysis for binary outcome: how to do it. Stat Methods Med Res. 2013 Oct 28.

5. Hoaglin DC, Hawkins N, Jansen JP,. et al. Conducting Indirect-Treatment-Comparison and Network-Meta-Analysis Studies: Report of the ISPOR Task Force on Indirect Treatment Comparisons Good Research Practices—Part 2. Value Health 2011;14:429-37.

6. Jansen JP, Fleurence R, Devine B, et al. Interpreting Indirect Treatment Comparisons and Network Meta-Analysis for Health-Care Decision Making: Report of the ISPOR Task Force on Indirect Treatment Comparisons Good Research Practices: Part 1. Value Health 2011;14:417-28.

7. NICE Clinical Guidelines, No. 92. National Clinical Guideline Centre – Acute and Chronic Conditions (UK). London: Royal College of Physicians (UK); 2010. Available at http://www.ncbi.nlm.nih.gov/books/NBK116530/ Accessed 14 August 2014

8. Zhang J, Yu KF. What's the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA. 1998 Nov 18;280(19):1690-1.

9. ClinCalc Website. Odds-ratio to risk ratio. http://clincalc.com/Stats/ConvertOR.aspx

10. Sobieraj DM, Cappelleri JC, Baker WL, Phung OJ, White CM, Coleman CI. Methods used to conduct and report Bayesian mixed treatment comparisons published in the medical literature: a systematic review. BMJ Open. 2013 Jul 21;3(7). pii:e003111. doi: 10.1136/bmjopen-2013-003111. Print 2013.