Abstract
Background
Network metaanalysis (NMA), a generalization of conventional MA, allows for assessing the relative effectiveness of multiple interventions. Reporting bias is a major threat to the validity of MA and NMA. Numerous methods are available to assess the robustness of MA results to reporting bias. We aimed to extend such methods to NMA.
Methods
We introduced 2 adjustment models for Bayesian NMA. First, we extended a metaregression model that allows the effect size to depend on its standard error. Second, we used a selection model that estimates the propensity of trial results being published and in which trials with lower propensity are weighted up in the NMA model. Both models rely on the assumption that biases are exchangeable across the network. We applied the models to 2 networks of placebocontrolled trials of 12 antidepressants, with 74 trials in the US Food and Drug Administration (FDA) database but only 51 with published results. NMA and adjustment models were used to estimate the effects of the 12 drugs relative to placebo, the 66 effect sizes for all possible pairwise comparisons between drugs, probabilities of being the best drug and ranking of drugs. We compared the results from the 2 adjustment models applied to published data and NMAs of published data and NMAs of FDA data, considered as representing the totality of the data.
Results
Both adjustment models showed reduced estimated effects for the 12 drugs relative to the placebo as compared with NMA of published data. Pairwise effect sizes between drugs, probabilities of being the best drug and ranking of drugs were modified. Estimated drug effects relative to the placebo from both adjustment models were corrected (i.e., similar to those from NMA of FDA data) for some drugs but not others, which resulted in differences in pairwise effect sizes between drugs and ranking.
Conclusions
In this case study, adjustment models showed that NMA of published data was not robust to reporting bias and provided estimates closer to that of NMA of FDA data, although not optimal. The validity of such methods depends on the number of trials in the network and the assumption that conventional MAs in the network share a common mean bias mechanism.
Keywords:
Network metaanalysis; Publication bias; Smallstudy effectBackground
Network metaanalyses (NMAs) are increasingly being used to evaluate the best intervention among different existing interventions for a specific condition. The essence of the approach is that intervention A is compared with a comparator C, then intervention B with C, and adjusted indirect comparison allows for comparing A and B, despite the lack of any headtohead randomized trial comparing A and B. An NMA, or multipletreatments metaanalysis (MA), allows for synthesizing comparative evidence for multiple interventions by combining direct and indirect comparisons [13]. The purpose is to estimate effect sizes for all possible pairwise comparisons of interventions, although some comparisons have no available trial.
Reporting bias is a major threat to the validity of results of conventional systematic reviews or MAs [4,5]. Accounting for reporting biases in NMA is challenging, because unequal availability of findings across the network of evidence may jeopardize NMA validity [6,7]. We previously empirically assessed the impact of reporting bias on the results of NMAs of antidepressant trials and showed that it may bias estimates of treatment efficacy [8].
Numerous methods have been used as sensitivity analyses to assess the robustness of conventional MAs to publication bias and related smallstudy effects [920]. Modeling methods include regressionbased approaches and selection models. We extend these approaches to NMAs in the Bayesian framework.
Methods
First, we extended a metaregression model of the effect size on its standard error, recently described for MAs [21,22]. In this approach, the regression slope reflects the magnitude of the association of effect size and precision (ie, the “smallstudy effect”), and the intercept provides an adjusted pooled effect size (ie, the predicted effect size of a trial with infinite precision). Second, we introduced a selection model, which models the probability of a trial being selected and is taken into account with inverse weighting in the NMA. Both adjustment models rely on the assumption that biases are exchangeable across the network, ie, biases, if present, operate in a similar way in trials across the network. Third, we applied these adjustment models to datasets created from US Food and Drug Administration (FDA) reviews of antidepressant trials and from their matching publications. These datasets were shown to differ because of reporting bias [23]. We compared the results of the adjustment models applied to published data and standard NMA for published and for FDA data, the latter considered the reference standard.
Datasets used
A previous review by Turner et al. assessed the selective publication of antidepressant trials [23]. The authors identified all randomized placebocontrolled trials of 12 antidepressant drugs approved by the FDA and then publications matching these trials by searching literature databases and contacting trial sponsors. From the FDA database, the authors identified 74 trials, among which results for 23 trials were unpublished. The proportion of trials with unpublished results varied across drugs, from 0% for fluoxetine and paroxetine CR to 60% and 67% for sertraline and bupropion (Additional file 1: Appendix 1). These entire trials remained unpublished depending on the nature of the results. Moreover, in some journal articles, specific analyses were reported selectively and effect sizes differed from that in FDA reviews. The outcome was the change from baseline to followup in depression severity score. The measure of effect was a standardized mean difference (SMD). Separate MAs of FDA data showed decreased efficacy for all drugs as compared to published data, the decrease in effect size ranging from 10% and 11% for fluoxetine and paroxetine CR to 39% and 41% for mirtazapine and nefazodone (Additional file 1: Appendix 1). Figure 1 shows the funnel plots of published data. Visual inspection does not suggest stronger treatment benefit in small trials (ie, funnel plot asymmetry) for any of the 12 comparisons of each drug and placebo.
Additional file 1. Appendix 1. Summary effect sizes for the 12 comparisons of each antidepressant agent and placebo. Appendix 2. Winbugs codes. Appendix 3. Estimated parameters in the adjustment models applied to published data.
Format: DOCX Size: 19KB Download file
Figure 1. Contourenhanced funnel plots forthe antidepressant trials withpublished results. Each funnel plot is the scatter plot of the treatment effect estimates from individual trials against the associated standard errors; the vertical solid line represents the pooled estimate. In the absence of reporting bias, we might expect a symmetrical funnel plot. We may find the funnel plot is not symmetrical, ie does not resemble an inverted funnel, which may be due to reporting bias, however there are other possible sources of asymmetry. The contour lines represent perceived milestones of statistical significance (long dash p = 0.1; dash p = 0.05; short dash p = 0.01). If studies seem to be missing in areas of nonsignificance then asymmetry may be due to reporting bias rather than other factors.
Network metaanalysis
The standard model for NMA was formalized by Lu and Ades [2,24,25]. We assume that each trial i assessed treatments j and k among the T interventions in the network. Each trial provided an estimated intervention effect
size y_{ijk} of j over k and its variance v_{ijk}. We assume that y_{ijk} > 0 indicates superiority of j over k. Assuming normal likelihood and according to a randomeffects model, y_{ijk} ~ N(θ_{ijk}, v_{ijk}) and
Adjustment models
Metaregression model
We used a network metaregression model extending a regressionbased approach for adjusting for smallstudy effects in conventional MAs [21,22,2729]. This regressionbased approach takes into account a possible smallstudy effect by allowing the effect size to depend on a measure of its precision. Here, we assume a linear relationship between the effect size and its standard error and the model involves extrapolation beyond the observed data to a hypothetical study of infinite precision. The extended model for NMA is as follows:
Figure A in Additional file 2 shows a graphical representation of the model. In the regression equation, θ_{ijk} is the treatment effect adjusted for smallstudy effects underlying each randomized
comparison between treatments j and k; β_{jk} represents the potential smallstudy effect (ie, the slope associated with funnel
plot asymmetry for the randomized comparisons between treatments j and k). The model assumes that these comparisonspecific regression slopes follow a common
normal distribution, with mean slope β and common betweenslopes variance σ^{2}. This is equivalent to the assumption that comparisonspecific smallstudy biases
are exchangeable within the network. Since we assumed that y_{ijk} > 0 indicates superiority of j over k, β > 0 would mean an overall tendency for a smallstudy effect (ie, treatment contrasts
tend to be overestimated in smaller trials). Finally,
Additional file 2. Figures. Graphical representation of the adjustment models (A) regression model and (B) selection model. A solid arrow indicates a stochastic dependence and a hollow arrow indicates a logical function.
Format: PPT Size: 217KB Download file
This file can be viewed with: Microsoft PowerPoint Viewer
Selection model
We use a model that adjusts for publication bias using a weight function to represent the process of selection. The model includes an effect size model (ie, the standard NMA model that specifies what the distributions of the effect size estimates would be with no selection) and a selection model that specifies how these effect size distributions are modified by the process of selection [14,30]. We assume that the probability of selection depends on the standard error of the effect size, as a decreasing function of it. We adopt an approach based on a logistic selection model, as previously used in conventional MAs [18,31].
β_{0jk} ~ N(β_{0}, σ_{0}^{2}) and β_{1jk} ~ N(β_{1}, σ_{1}^{2})
Figure B in Additional file 2 shows a graphical representation of the model. In the logistic regression equation,
w_{i} represents the propensity of the trial results to be published, β_{0jk} sets the overall probability of observing a randomized comparison between treatments
j and k, and β_{1jk} controls how fast this probability evolves as the standard error increases. We expect
β_{1jk} to be negative, so trial results yielding larger standard errors have lower propensity
to be published. The model assumes exchangeability of the β_{0jk} and β_{1jk} coefficients within the network. By setting γ_{ijk} = θ_{ijk}/w_{i}, we define a simple scheme that weights up trial results with lower propensity of
being published so that they have a disproportionate influence in the NMA model. θ_{ijk} is the treatment contrast corrected for the selection process underlying each randomized
comparison between treatments j and k. Finally,
Models estimation
We estimated 4 models: standard NMA model of published data, 2 adjustment models of
published data and a standard NMA model of FDA data. In each case, model estimation
involved Markov chain Monte Carlo methods with Gibbs sampling. Placebo was chosen
as the overall baseline treatment to compare all other treatments. Consequently, the
12 effects of drugs relative to placebo are the basic parameters. For 2 treatments
j and k, SMD_{jk} > 0 indicate that j is superior to k. In both the metaregression and selection models, we assumed that the active treatments
would always be favored by smallstudy bias as compared to placebo; consequently,
In the standard NMA model, we defined prior distributions for the basic parameters
Models comparison
We compared the results of the 2 adjustment models applied to published data and results of the standard NMA model applied to published data and the FDA data, the latter considered the reference standard. First, we compared posterior means and 95% credibility intervals for the 12 basic parameters and common variance, as well as for the 66 functional parameters (ie, all 12 × 11/2 = 66 possible pairwise comparisons of the 12 drugs). Second, we compared the rankings of the competing treatments. We assessed the probability that each treatment was best, then second best and third best, etc. We plotted the cumulative probabilities and computed the surface under the cumulative ranking (SUCRA) line for each treatment [34]. Third, to compare the different models applied to published data, we used the posterior mean of the residual deviance and the deviance information criteria [35].
Results
In the metaregression model applied to published data, the posterior mean slope β was 1.7 (95% credible interval −0.3–3.6), which suggests an overall tendency for a smallstudy effect in the network. The 12 regression slopes were similar, with posterior means ranging from 1.4 to 1.9. In the selection model applied to published data, the mean slope β_{1} was −10.0 (−18.0 – 2.50), so trials yielding larger standard errors tended overall to have lower propensity to be published. In both models, all estimates were subject to large uncertainty (Additional file 1: Appendix 3).
Table 1 shows the estimates of the 12 basic parameters between each drug and placebo according to the 4 models. As compared with the NMA of published data, both adjustment models of published data showed that the whole 12 estimated drug effects relative to placebo were reduced. For the metaregression model, the decrease in efficacy ranged from 48% for venlafaxine XR to 99% for fluoxetine. For the selection model, the decrease ranged from 13% for escitalopram to 26% for paroxetine. When considering the functional parameters (ie, the 66 possible pairwise comparisons between drugs), we found differences between the results of adjustment models and the standard NMA model applied to published trials (Figure 2). The median relative difference, in absolute value, between pairwise effect sizes from the regression model and the standard NMA model was 57.3% (25% – 75% percentile 30.3% – 97.6%); the median relative difference between the selection model and the standard NMA model was 29.2% (15.1% – 46.1%).
Table 1. Comparison of network metaanalysis(NMA)based estimates between the2 adjustment models appliedto published data andthe standard NMA modelapplied to US Foodand Drug Administration (FDA)data and to publisheddata
Figure 2. Difference plots of estimatesof pairwise comparisons ofthe 12 antidepressant agentsand placebo: regression modelof published data vs.standard network metaanalysis (NMA)model of published data(left panel); selection modelsof published vs. standardNMA model of publisheddata (right panel). The xaxes show the estimates from the standard NMA model applied to published data, the yaxes show the differences between the estimates from the adjustment (regression or selection) model of published data and the estimates from the standard NMA model of published data. Black dots are the 12 estimated drug effects relative to placebo; white dots are the 66 possible pairwise comparisons between the 12 drugs.
Figure 3 summarizes the probabilities of being the best antidepressant. Compared to the standard NMA of published data, adjustment models of published data yielded decreased probabilities of the drug being the best for paroxetine (from 41.5% to 20.7% with the regression model or 25.7% with the selection model) and mirtazapine (from 30.3% to 15.7% or 21.9%). They yielded increased probabilities of the drug being the best for venlafaxine (from 7.9% to 10.6% or 12.8%) and venlafaxine XR (from 14.1% to 21.0% or 23.5%).
Figure 3. Probabilities that each antidepressantdrug is the bestaccording to standard NMAof FDA data, regressionmodel, selection model orstandard NMA model ofpublished data.
Figure 4 shows cumulative probability plots and SUCRAs. For the standard NMA of published data, paroxetine and mirtazapine tied for first place and venlafaxine XR and venlafaxine tied for third. The selection model applied to published data yielded a slightly different ranking, with paroxetine, mirtazapine and venlafaxine XR tying for first and venlafaxine was fourth. In the regression model applied to published data, venlafaxine XR was first, venlafaxine and paroxetine tied for second and mirtazapine was fifth.
Figure 4. Cumulative ranking probability plotsfor the 12 antidepressantagents from the standardNMA model applied toFDA data (bold solidline) and published data(bold dotted line) andfrom the 2 adjustmentmodels applied to publisheddata (regression model inplain dashed line andselection model in plaindoubledashed line). On each plot, the xaxis shows possible ranks from r = 1 up to r = 13 and the yaxis shows the cumulative probabilities that the corresponding treatment is among the top r treatments. The closer the curve is to the upper left corner, the better the treatment. The surface under the cumulative ranking line is 1 when a treatment is the best and 0 when a treatment is the worst. FDA: standard NMA model applied to FDA data (bold plain line); Pub.: standard NMA model applied to published data (bold dash line); Reg.: regression model applied to published data (dash line); Sln.: selection model applied to published data (longdash shortdash line).
In adjustment models applied to published data, betweentrial heterogeneity and fit were comparable to those obtained with standard NMA of published data (Tables 1 and 2).
Table 2. Comparison of fit andcomplexity between the 2adjustment models and thestandard NMA model, allapplied to published data
The estimated drug effects relative to placebo from the regression and selection models were similar to those from the NMA of FDA data for some drugs (Table 1). There were differences when considering the 66 possible pairwise comparisons between drugs (Figure 5). Results also differed by models regarding the probability of being the best drug and the ranking of drugs. In the standard NMA of FDA data, the probability of being the best drug was 7.3% for mirtazapine, 33.9% for paroxetine, 19.3% for venlafaxine, and 25.7% for venlafaxine XR (Figure 3); paroxetine ranked first, and venlafaxine and venlafaxine XR tied for second (Figure 4).
Figure 5. Difference plots of estimatesof pairwise comparisons ofthe 12 antidepressant agentsand placebo: standard NMAmodel of published datavs. standard NMA modelof FDA data (upperpanel); regression model ofpublished data vs. standardNMA model of FDAdata (bottom left panel);selection model of publishedvs. standard NMA modelof FDA data (bottomright panel). The xaxes show the estimates from the standard NMA model applied to FDA data, the yaxes show the differences between the estimates from the adjustment (regression or selection) model of published data and the estimates from the standard NMA model of FDA data. Black dots are the 12 estimated drug effects relative to placebo; white dots are the 66 possible pairwise comparisons between the 12 drugs.
Discussion
We extended two adjustment methods for reporting bias from MAs to NMAs. The first method combined NMA and metaregression models, with effect sizes regressed against their precision. The second one combined the NMA model with a logistic selection model estimating the probability that a trial was published or selected in the network. The former method basically adjusts for funnel plot asymmetry or small study effects, which may arise from causes other than publication bias. The latter adjusts for publication bias (ie, the suppression of an entire trial depending on results). The two models borrow strength from other trials in the network with the assumption that biases operate in a similar way in trials across the domain.
In a specific network of placebocontrolled trials of antidepressants, based on data already described and published previously by Turner et al., comparing the results of adjustment models applied to published data and those of the standard NMA model applied to published data allowed for assessing the robustness of efficacy estimates and ranking to publication bias or related smallstudy effects. Both models showed a decrease in all basic parameters (ie, the 12 effect sizes of drugs relative to placebo). The 66 contrasts for all possible pairwise comparisons between drugs, the probabilities of being the best drug and the ranking were modified as well. The NMA of published data was not robust to publication bias and related smallstudy effects.
This specific dataset offered the opportunity to perform NMAs on both published and FDA data. The latter may be considered "an unbiased (but not the complete) body of evidence" for placebocontrolled trials of antidepressants [28]. The comparison of the results of the 2 models applied to published data and the standard NMA model applied to FDA data showed that the effect sizes of drugs relative to placebo were corrected for some but not all drugs. This observation led to differences in the 66 possible pairwise comparisons between drugs, the probabilities of being the best drug and the ranking. It suggests that the 2 models should not be considered optimal; that is, the objective is not to produce definitive estimates adjusted for publication bias and related smallstudy effects but rather to assess the robustness of results to the assumption of bias.
Similar approaches have been used by other authors. Network metaregression models fitted within a Bayesian framework were previously developed to assess the impact of novelty bias and risk of bias within trials [36,37]. Network metaregression to assess the impact of smallstudy effect was specifically used by Dias et al. in a reanalysis of a network of published headtohead randomized trials of selective serotonin reuptake inhibitors [38]. Along the line of the regressionbased approach of Moreno et al. in conventional MA, the authors introduced a measure of study size as a regression variable within the NMA model and identified a mean bias in pairwise effect sizes. More recently, Moreno et al. used a similar approach to adjust for smallstudy effects in several conventional MAs of similar interventions and outcomes and illustrated their method using the dataset of Turner et al. [39]. Our approach differed in that we extended this metaregression approach to NMAs. We used the standard error of treatment effect estimate as the regressor. As well, we specified an additive betweentrial variance rather than a multiplicative overdispersion parameter. With the latter, the estimated multiplicative parameter may be < 1, which implies less heterogeneity than would be expected by chance alone. Selection model approaches have been considered recently. Chootrakool et al. introduced an approximated normal model based on empirical logodds ratio for NMAs within a frequentist framework and applied Copas selection models for some groups of trials in the network selected according to funnel plot asymmetry [40]. Mavridis et al. presented a Bayesian implementation of the Copas selection model extended to NMA and applied their method on the network of Turner et al. [41]. In the Copas selection model, the selection probability depends on both the estimates of the treatment effects and their standard errors. In the extension to NMA, an extra correlation parameter ρ, assumed equal for all comparisons, needs to be estimated. When applied to published data of the network of Turner et al., the selection model we proposed and the treatmentspecific selection model of Mavridis et al. yielded close results.
The 2 adjustment models rely on the assumption of exchangeability of selection processes
across the network; that is, biases, if present, operate in a similar way in trials
across the network. In this case study, all studies were, by construction, industrysponsored,
placebocontrolled trials registered with the FDA, and for all drugs, results of entire
studies remained unreported depending on the results [23]. Thus, the assumption of exchangeability of selection processes is plausible. More
generally, if we have no information to distinguish different reporting bias mechanisms
across the network, an exchangeable prior distribution is plausible, "ignorance implies
exchangeability" [42,43]. However, the assumption may not be tenable in other contexts in which reporting
biases may affect the network in an unbalanced way. It may operate differently in
placebocontrolled and headtohead trials [44], in older and more recent trials (because of trial registries), and for drug and
nondrug interventions [7]. In more complex networks involving headtohead trials, the 2 adjustment models
could be generalized to allow the expected publication bias or smallstudy bias for
activeactive trials to differ from that of the expected bias in trials comparing
active and inactive treatments [36]. In headtohead trials, the direction of bias is uncertain but assumptions in defining
The models we described have limitations. First, they would result in poor estimation of bias and effect sizes when the conventional MAs within the network include small numbers of trials [21]. Second, for the selection model, we specified the weight function. If the underlying assumptions (ie, a logistic link form and the chance of a trial being selected related to standard error) are wrong, the estimated selection model will be wrong. However, alternative weight functions (e.g., probit link) or conditioning (e.g., on the magnitude of effect size) could be considered. Finally, it was implemented with a weakly informative prior, which mainly suggested that the propensity for results to be published may decrease with increasing standard error. There is a risk that prior information overwhelms observed data, especially if the number of trials is low. Although they were somewhat arbitrarily set, our priors for the selection model parameters were in line with the values in previous studies using the Copas selection model [12,49]. Different patterns of selection bias could be tested, for instance, by considering various prior modes for p_{min} and p_{max}, the probabilities of publication when the standard error takes its minimum and maximum values across the network [15].
Conclusions
In conclusion, addressing publication bias and related smallstudy effects in NMAs was feasible in this case study. Validity may be conditioned by sufficient numbers of trials in the network and assuming that conventional MAs constituting the network share a common mean bias. Simulation analyses are required to determine under which condition such adjustment models are valid. Application of such adjustment models should be replicated on more complex networks, ideally representing the totality of the data as in Turner's, but our results confirm that authors and readers should interpret NMAs with caution when reporting bias has not been addressed.
Competing interest
The authors declared that they have no competing interest.
Authors' contributions
LT provided substantial contributions to conception and design, analysis and interpretation of data, drafted the article and revised it critically for important intellectual content. GC and PR provided substantial contributions to design and interpretation of data, and revised the article critically for important intellectual content. All authors read and approved the final manuscript.
Financial disclosure
Grant support was from the French Ministry of Health Programme Hospitalier de Recherche Clinique National (PHRC 2011 MIN0163) and European Union Seventh Framework Programme (FP7 – HEALTH.2011.4.12) under grant agreement n° 285453 (http://www.openproject.eu webcite). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Acknowledgments
The authors thank Laura Smales (BioMedEditing, Toronto, Canada) for editing the manuscript.
References

Lumley T: Network metaanalysis for indirect treatment comparisons.
Stat Med 2002, 21(16):23132324. PubMed Abstract  Publisher Full Text

Lu G, Ades AE: Combination of direct and indirect evidence in mixed treatment comparisons.
Stat Med 2004, 23(20):31053124. PubMed Abstract  Publisher Full Text

Salanti G, Higgins JPT, Ades AE, Ioannidis JPA: Evaluation of networks of randomized trials.
Stat Methods Med Res 2008, 17(3):279301. PubMed Abstract  Publisher Full Text

Dwan K, Altman DG, Arnaiz JA, Bloom J, Chan AW, Cronin E, Decullier E, Easterbrook PJ, Von Elm E, Gamble C, et al.: Systematic review of the empirical evidence of study publication bias and outcome reporting bias.
PLoS One 2008, 3(8):e3081. PubMed Abstract  Publisher Full Text  PubMed Central Full Text

Song F, Parekh S, Hooper L, Loke YK, Ryder J, Sutton AJ, Hing C, Kwok CS, Pang C, Harvey I: Dissemination and publication of research findings: an updated review of related biases.
Health Technol Assess 2010, 14(8):193. PubMed Abstract  Publisher Full Text

Salanti G, Kavvoura FK, Ioannidis JP: Exploring the geometry of treatment networks.
Ann Intern Med 2008, 148(7):544553. PubMed Abstract  Publisher Full Text

Li T, Puhan MA, Vedula SS, Singh S, Dickersin K: Network metaanalysishighly attractive but more methodological research is needed.
BMC Med 2011, 9:79. PubMed Abstract  BioMed Central Full Text  PubMed Central Full Text

Trinquart L, Abbé A, Ravaud P: Impact of reporting bias in network metaanalysis of antidepressant placebocontrolled trials.
PLoS One 2012, 7(4):e35219. PubMed Abstract  Publisher Full Text  PubMed Central Full Text

Hedges LV: Modeling publication selection effects in metaanalysis.
Stat Sci 1992, 7(2):246255. Publisher Full Text

Silliman NP: Hierarchical selection models with applications in metaanalysis.

Larose DT, Dey DK: Modeling publication bias using weighted distributions in a Bayesian framework.
Comput Stat Data Anal 1998, 26:279302. Publisher Full Text

Copas J, Shi JQ: Metaanalysis, funnel plots and sensitivity analysis.
Biostatistics 2000, 1(3):247262. PubMed Abstract  Publisher Full Text

Duval S, Tweedie R: Trim and fill: a simple funnelplotbased method of testing and adjusting for publication bias in metaanalysis.
Biometrics 2000, 56(2):455463. PubMed Abstract  Publisher Full Text

Sutton AJ, Song F, Gilbody SM, Abrams KR: Modelling publication bias in metaanalysis: a review.
Stat Methods Med Res 2000, 9(5):421445. PubMed Abstract  Publisher Full Text

Copas JB, Shi JQ: A sensitivity analysis for publication bias in systematic reviews.
Stat Methods Med Res 2001, 10(4):251265. PubMed Abstract  Publisher Full Text

Preston C, Ashby D, Smyth R: Adjusting for publication bias: modelling the selection process.
J Eval Clin Pract 2004, 10(2):313322. PubMed Abstract  Publisher Full Text

Bowden J, Jackson D, Thompson SG: Modelling multiple sources of dissemination bias in metaanalysis.
Stat Med 2010, 29(7–8):945955. PubMed Abstract  Publisher Full Text

Carpenter J, Rucker G, Schwarzer G: Assessing the sensitivity of metaanalysis to selection bias: a multiple imputation approach.
Biometrics 2011, 67(3):10661072. PubMed Abstract  Publisher Full Text

Rucker G, Carpenter JR, Schwarzer G: Detecting and adjusting for smallstudy effects in metaanalysis.
Biom J 2011, 53(2):351368. PubMed Abstract  Publisher Full Text

Rufibach K: Selection models with monotone weight functions in meta analysis.
Biom J 2011, 53(4):689704. PubMed Abstract  Publisher Full Text

Moreno SG, Sutton AJ, Ades AE, Stanley TD, Abrams KR, Peters JL, Cooper NJ: Assessment of regressionbased methods to adjust for publication bias through a comprehensive simulation study.
BMC Med Res Methodol 2009, 9:2. PubMed Abstract  BioMed Central Full Text  PubMed Central Full Text

Moreno SG, Sutton AJ, Thompson JR, Ades AE: Abrams KR. Cooper NJ: A generalized weighting regressionderived metaanalysis estimator robust to smallstudy effects and heterogeneity. Stat Med; 2012.

Turner EH, Matthews AM, Linardatos E, Tell RA, Rosenthal R: Selective publication of antidepressant trials and its influence on apparent efficacy.
N Engl J Med 2008, 358(3):252260. PubMed Abstract  Publisher Full Text

Higgins JP, Whitehead A: Borrowing strength from external trials in a metaanalysis.
Stat Med 1996, 15(24):27332749. PubMed Abstract  Publisher Full Text

Lu G, Ades AE: Assessing evidence inconsistency in mixed treatment comparisons.

Lu G, Ades A: Modeling betweentrial variance structure in mixed treatment comparisons.
Biostatistics 2009, 10(4):792805. PubMed Abstract  Publisher Full Text

Stanley TD: Metaregression methods for detecting and estimating empirical effects in the presence of publication selection*.

Moreno SG, Sutton AJ, Turner EH, Abrams KR, Cooper NJ, Palmer TM, Ades AE: Novel methods to deal with publication biases: secondary analysis of antidepressant trials in the FDA trial registry database and related journal publications.
BMJ 2009, 339:b2981. PubMed Abstract  Publisher Full Text  PubMed Central Full Text

Rucker G, Schwarzer G, Carpenter JR, Binder H, Schumacher M: Treatmenteffect estimates adjusted for smallstudy effects via a limit metaanalysis.
Biostatistics 2011, 12(1):122142. PubMed Abstract  Publisher Full Text

Song F, Eastwood AJ, Gilbody S, Duley L, Sutton AJ: Publication and related biases.
Health Technol Assess 2000, 4(10):1115. PubMed Abstract  Publisher Full Text

Hedges LV: [Selection models and the file drawer problem]: comment.
Stat Sci 1988, 3(1):118120. Publisher Full Text

Bedrick EJ, Christensen R, Johnson W: A new perspective on priors for generalized linear models.

Wu Y, Shih WJ, Moore DF: Elicitation of a beta prior for Bayesian inference in clinical trials.
Biom J 2008, 50(2):212223. PubMed Abstract  Publisher Full Text

Salanti G, Ades AE, Ioannidis JP: Graphical methods and numerical summaries for presenting results from multipletreatment metaanalysis: an overview and tutorial.
J Clin Epidemiol 2010, 64(2):163171. PubMed Abstract  Publisher Full Text

Spiegelhalter DJ, Best NG, Carlin BP, Van Der Linde A: Bayesian measures of model complexity and fit.
J R Stat Soc Ser B Stat Methodol 2002, 64(4):583639. Publisher Full Text

Dias S, Welton NJ, Marinho VCC, Salanti G, Higgins JPT, Ades AE: Estimation and adjustment of bias in randomized evidence by using mixed treatment comparison metaanalysis.
J R Stat Soc Ser A Stat Soc 2010, 173(3):613629. Publisher Full Text

Salanti G, Dias S, Welton NJ, Ades AE, Golfinopoulos V, Kyrgiou M, Mauri D, Ioannidis JP: Evaluating novel agent effects in multipletreatments metaregression.
Stat Med 2010, 29(23):23692383. PubMed Abstract  Publisher Full Text

Dias S, Welton NJ, Ades AE: Study designs to detect sponsorship and other biases in systematic reviews.
J Clin Epidemiol 2010, 63(6):587588. PubMed Abstract  Publisher Full Text

Moreno SG, Sutton AJ, Ades AE, Cooper NJ, Abrams KR: Adjusting for publication biases across similar interventions performed well when compared with gold standard data.
J Clin Epidemiol 2011, 64(11):12301241. PubMed Abstract  Publisher Full Text

Chootrakool H, Shi JQ, Yue R: Metaanalysis and sensitivity analysis for multiarm trials with selection bias.
Stat Med 2011, 30(11):11831198. PubMed Abstract  Publisher Full Text

Mavridis D, Sutton A, Cipriani A, Salanti G: A fully Bayesian application of the Copas selection model for publication bias extended to network metaanalysis.
Stat Med 2012.
[Epub ahead of print].
Publisher Full Text 
McCandless LC, Gustafson P, Levy AR, Richardson S: Hierarchical priors for bias parameters in Bayesian sensitivity analysis for unmeasured confounding.
Stat Med 2012, 31(4):383396. PubMed Abstract  Publisher Full Text

Gelman A, Carlin JB, Stern HS, Rubin DB: Bayesian Data Analysis. 2nd edition. New York: Chapman Hall/CRC; 2004.

Rising K, Bacchetti P, Bero L: Reporting bias in drug trials submitted to the food and drug administration: review of publication and presentation.
PLoS Med 2008, 5(11):e217.
discussion e217.
PubMed Abstract  Publisher Full Text  PubMed Central Full Text 
Lathyris DN, Patsopoulos NA, Salanti G, Ioannidis JP: Industry sponsorship and selection of comparators in randomized clinical trials.
Eur J Clin Invest 2010, 40(2):172182. PubMed Abstract  Publisher Full Text

Lexchin J, Bero LA, Djulbegovic B, Clark O: Pharmaceutical industry sponsorship and research outcome and quality: systematic review.
BMJ 2003, 326(7400):11671170. PubMed Abstract  Publisher Full Text  PubMed Central Full Text

Bero L, Oostvogel F, Bacchetti P, Lee K: Factors associated with findings of published trials of drugdrug comparisons: why some statins appear more efficacious than others.
PLoS Med 2007, 4(6):e184. PubMed Abstract  Publisher Full Text  PubMed Central Full Text

Chalmers I, Matthews R: What are the implications of optimism bias in clinical research?
Lancet 2006, 367(9509):449450. PubMed Abstract  Publisher Full Text

Rucker G, Schwarzer G, Carpenter J: Arcsine test for publication bias in metaanalyses with binary outcomes.
Stat Med 2008, 27(5):746763. PubMed Abstract  Publisher Full Text
Prepublication history
The prepublication history for this paper can be accessed here: