UBC Faculty Research and Publications

Incorporating multiple interventions in meta-analysis: an evaluation of the mixed treatment comparison… O'Regan, Christopher; Ghement, Isabella; Eyawo, Oghenowede; Guyatt, Gordon H; Mills, Edward J Sep 21, 2009

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
52383-13063_2009_Article_374.pdf [ 593.63kB ]
Metadata
JSON: 52383-1.0132612.json
JSON-LD: 52383-1.0132612-ld.json
RDF/XML (Pretty): 52383-1.0132612-rdf.xml
RDF/JSON: 52383-1.0132612-rdf.json
Turtle: 52383-1.0132612-turtle.txt
N-Triples: 52383-1.0132612-rdf-ntriples.txt
Original Record: 52383-1.0132612-source.json
Full Text
52383-1.0132612-fulltext.txt
Citation
52383-1.0132612.ris

Full Text

ralssBioMed CentOpen AcceTrialsResearchIncorporating multiple interventions in meta-analysis: an evaluation of the mixed treatment comparison with the adjusted indirect comparisonChristopher O'Regan1, Isabella Ghement2, Oghenowede Eyawo3, Gordon H Guyatt4 and Edward J Mills*3,4,5Address: 1Department of Epidemiology, London School of Hygiene & Tropical Medicine, London, UK, 2Department of Statistics, University of British Columbia, Vancouver, British Columbia, Canada, 3Faculty of Health Sciences, Simon Fraser University, Burnaby, British Columbia, Canada, 4Department of Clinical Epidemiology & Biostatistics, McMaster University, Hamilton, Ontario, Canada and 5Faculty of Health Sciences, University of Ottawa, Ottawa, Ontario, CanadaEmail: Christopher O'Regan - christopher_oregan@merck.com; Isabella Ghement - isabella@ghement.ca; Oghenowede Eyawo - edeyawo@yahoo.com; Gordon H Guyatt - guyatt@mcmaster.ca; Edward J Mills* - emills@cfenet.ubc.ca* Corresponding author    AbstractBackground: Comparing the effectiveness of interventions is now a requirement for regulatoryapproval in several countries. It also aids in clinical and public health decision-making. However, inthe absence of head-to-head randomized trials (RCTs), determining the relative effectiveness ofinterventions is challenging. Several methodological options are now available. We aimed todetermine the comparative validity of the adjusted indirect comparisons of RCTs with the mixedtreatment comparison approach.Methods: Using systematic searching, we identified all meta-analyses evaluating more than 3interventions for a similar disease state with binary outcomes. We abstracted data on each clinicaltrial including population n and outcomes. We conducted fixed effects meta-analysis of eachintervention versus mutual comparator and then applied the adjusted indirect comparison. Weconducted a mixed treatment meta-analysis on all trials and compared the point estimates and 95%confidence/credible intervals (CIs/CrIs) to determine important differences.Results: We included data from 7 reviews that met our inclusion criteria, allowing a total of 51comparisons. According to the a priori consistency rule, we found 2 examples where the analyticcomparisons were statistically significant using the mixed treatment comparison over the adjustedindirect comparisons and 1 example where this was vice versa. We found 6 examples where thedirection of effect differed according to the indirect comparison method chosen and we found 9examples where the confidence intervals were importantly different between approaches.Conclusion: In most analyses, the adjusted indirect comparison yields estimates of relativeeffectiveness equal to the mixed treatment comparison. In less complex indirect comparisons,where all studies share a mutual comparator, both approaches yield similar benefits. Ascomparisons become more complex, the mixed treatment comparison may be favoured.Published: 21 September 2009Trials 2009, 10:86 doi:10.1186/1745-6215-10-86Received: 6 April 2009Accepted: 21 September 2009This article is available from: http://www.trialsjournal.com/content/10/1/86© 2009 O'Regan et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Page 1 of 12(page number not for citation purposes)Trials 2009, 10:86 http://www.trialsjournal.com/content/10/1/86BackgroundAcknowledging their enormous value for health interven-tion decision-making, clinicians, drug manufacturers, reg-ulatory agencies and the public are now requiring meta-analysis to identify the most effective intervention amonga range of alternatives.[1] As meta-analysis grows in pop-ularity, investigators have endeavoured to further enhanceits usefulness by proposing extensions meant to accom-modate a number of challenges. One important challengeis choosing from a number of potentially competing inter-ventions, not all of which have been subject to direct com-parison in properly conducted randomized trials; hereinreferred to as indirect comparisons.Until recently, meta-analysis addressed indirect compari-sons using flawed methods that examined only interven-tion groups and ignored control event rates.[2] In the lastdecades, methodological advances,[3] most notably, theadjusted indirect comparison, first reported in 1997,[4]and the mixed treatment comparison, first reported in2003, [5] have provided more sophisticated methods forquantitatively addressing indirect comparisons.The adjusted indirect comparison, first reported by Bucheret al.,[4] enables one to construct an indirect estimate ofthe relative effect of two interventions A and B, by usinginformation from randomized trials comparing each ofthese interventions against a common comparator C (e.g.,placebo or standard treatment). In this approach, directestimates of the relative effects of A versus C and B versusC, together with appropriate measures of uncertainty, areobtained using standard pairwise meta-analysis. Theseestimates are then appropriately combined to produce anindirect estimate of the relative effect of A versus B. A suit-able measure of uncertainty for the indirect estimate isalso produced.The multiple treatment approaches, based on developingmethods by several investigators,[6,7]most recently Luand Ades,[8] is a generalization of standard pairwisemeta-analysis for A versus B trials, to data structures thatinclude, for example, A versus B, B versus C, and A versusC trials. This approach, which can only be applied to con-nected networks of randomised trials, has two importantroles: (1) strengthening inference concerning the relativeefficacy of two treatments, by including both direct andindirect comparisons of these treatments, and (2) facilitat-ing simultaneous inference regarding all treatments, inorder to simultaneously compare, or even rank, thesetreatments.[8]The adjusted indirect comparison and the mixed treat-ment comparison approach can be implemented throughThe basic assumptions underlying the adjusted indirectcomparison and mixed treatment comparison approachesare similar to but more complex than the assumptionsconcerning the standard meta-analysis approach. Just likestandard meta-analysis, both approaches rely on thehomogeneity assumption, which states that trials are suf-ficiently homogeneous to be quantitatively combined. Inaddition, both approaches require a similarity assump-tion - namely, that trials are similar for moderators of rel-ative treatment effect. The mixed treatment comparisonapproach also requires a consistency assumption, which isneeded to quantitatively combine direct and indirect evi-dence.[10]Both adjusted indirect comparison and mixed treatmentcomparison approaches to evaluating the relative impactof multiple alternative treatments have strengths andweaknesses.[11] The multiple treatment comparison usesboth direct and indirect evidence. The adjusted indirectmethod is comparatively simple and interpretable byusers, but requires that an intervention can only be com-pared with another intervention when they share amutual comparator (eg. placebo).[4] The mixed treatmentcomparison may be less intuitive as it can permit compar-isons when interventions do not share a comparator as itcreates a conceptual network[12,13] as well as borrowspower from trials that were not available for use in theadjusted indirect comparison approach.[14]Meta-analysts, agencies, and readers are now attemptingto gain further insight into the relative merits of the twoapproaches.[15] New US government initiatives to deter-mine the comparative effectiveness of interventionsrequire the use of indirect evidence, but do not provideguidance on what approach to use. Others, such as UK'sNational Institute for Clinical Excellence (NICE) provideadvice on the particular use of mixed treatment compari-sons and adjusted indirect comparisons.[15] To furtherelucidate the relative performance of the adjusted indirectcomparison and mixed treatment comparison methods,we applied both approaches to different comparativestudies that evaluated the effectiveness of multiple com-peting treatments for diverse health conditions. Ourobjective is to determine whether the adjusted indirectcomparison approach generates results comparable tothose produced by the mixed treatment comparisonapproach. We aim to determine if there are circumstanceswhere one method is preferable.MethodsEligibility CriteriaWe included systematic reviews of randomized clinical tri-als involving at least 4 different treatments (i.e., healthPage 2 of 12(page number not for citation purposes)a range of methods, including frequentist, Bayesian andvarious subspecies of each.[9]interventions used for treatment or prevention of thesame medical condition), as networks of three healthTrials 2009, 10:86 http://www.trialsjournal.com/content/10/1/86interventions have already received considerablestudy.[2,3,16] If a treatment consisted of several doses, weconsidered all doses to be equivalent. We also consideredno-treatment and placebo to be equivalent. Wheneverpresent, we excluded cluster randomized trials from thesesystematic review along with crossover trials and trialsreporting only continuous outcomes.Search StrategyWe (EM, OE) searched independently, in duplicate,PubMed from inception to January 2008 using the follow-ing search strategy: "network AND meta-analysis," "mixedtreatment AND meta-analysis," "indirect comparison,""indirect AND meta-analysis," and "mixed treatmentAND meta-analysis." Our search was limited to English-language articles. We supplemented our search strategyand findings from a review of network geometry of stud-ies[13] and from our own meta-analyses of multiple treat-ments (Perri D, O'Regan C, Cooper C, Nachega JB, Wu P,Tleyjeh I, Philips P, Mills EJ: Antifungal treatment for sys-temtic candida infectons: A mixed treatment compari-son meta-analysis. Unpublished).[17]Data AbstractionWe (EM, OE) abstracted independently, in duplicate,information addressing the systematic review aims,number of trials per comparison, number of individualswith each specific outcome and number of individualsrandomised to each intervention.Statistical analysesWe first plotted the geometric networks of comparisons tographically display what indirect comparisons our analy-ses aimed to assess.We conducted the mixed treatment comparisons usingfixed effects models similar to those introduced by Lu andAdes.[8] Although several definitions exist, we interpretthat the fixed effects approach assumes that there is a sin-gle true value underlying all the study results. That is,those studies would yield similar effects regardless of theparticular population enrolled, the intervention chosen,and the strategy for measuring the outcome of interest. Afixed effect model aims to estimate the common-trutheffect and the uncertainty around this estimate.[18] Weconsidered separate models for each outcome category(i.e., mortality, response) using approximately non-informed priors. We used these models as a basis forderiving the odds ratio [OR] for each treatment compari-son with 95% Credible Intervals (CrIs) - the Bayesianequivalent of a classical confidence interval.We estimated the posterior densities for all unknownage WinBUGS Version 1.4. Specifically, we simulated twoMCMC chains starting from different initial values ofselect unknown parameters. Each chain contained 20,000burn-in iterations followed by 20,000 update iterations.We assessed convergence by visualizing the histories ofthe chains against the iteration number; overlapping his-tories, that appeared to mix with each other, provided anindication of convergence. We based our inferences on the(convergence) posterior distributions of the relevantparameters. In particular, we estimated the OR for a giventreatment comparison by exponentiating the mean of theposterior distribution of the log OR, and constructed thecorresponding 95% CrI by exponentiating the 2.5th and97.5th percentiles of the posterior distribution of the logOR. Other parameters were estimated as means of corre-sponding posterior distributions.We measured the goodness of fit of our models to the databy calculating the residual deviance. Residual deviancewas defined as the difference between the deviance for thefitted model and the deviance for the saturated model,where the deviance uses the likelihood function to meas-ure the fit of the model to the data. Under the nullhypothesis that the model provides an adequate fit to thedata, the residual deviance is expected to have a meanequal to the number of unconstrained data points.For our relative effect sizes used in the adjusted indirectcomparison analyses, we used the same data as for themixed treatment comparison analyses. We conductedmultiple meta-analyses of head-to-head comparisons toobtain ORs and 95% Confidence Intervals [95% CIs]. Aswith the mixed treatment analyses, we applied the fixedeffects method. Once we obtained the summary estimatesof pooled head-to-head evaluations with CIs, we appliedthe adjusted indirect comparison approach.[4]For each systematic review, we determined if there wereimportant inconsistencies between the adjusted indirectcomparison and mixed treatment comparison approachesby comparing the 95% CrI produced by the formerapproach against the 95% CI produced by the latterapproach for the OR of each feasible treatment compari-son. We diagnosed inconsistency by assessing departuresfrom an a priori determined consistency rule stating thatthe lower and upper endpoints of the two types of inter-vals should not differ by more than 0.25 and 0.75, respec-tively, and the estimated ORs should not differ by morethan 0.5. EM and IG performed all statistical analyses.ResultsWe identified 44 potentially relevant systematic reviews ofthe effectiveness of multiple treatments for differentPage 3 of 12(page number not for citation purposes)model parameters using MCMC (Markov chain MonteCarlo) simulation, as implemented in the software pack-health conditions, including two of our own reviews thatwere ongoing during the search period (Perri D, O'ReganTrials 2009, 10:86 http://www.trialsjournal.com/content/10/1/86C, Cooper C, Nachega JB, Wu P, Tleyjeh I, Philips P, MillsEJ: Antifungal treatment for systemtic candida infec-tons: A mixed treatment comparison meta-analysis.Unpublished).[17] We narrowed down the scope of oursearch by excluding 13 reviews that incorporated fewerthan 4 treatments, 9 reviews that excluded eligible data forcomparisons, 3 reviews that did not create a network ofcomparisons, and 12 reviews that did not provide data onindividual outcomes in each study. In total, we includedseven systematic reviews in our analyses (Perri D, O'ReganC, Cooper C, Nachega JB, Wu P, Tleyjeh I, Philips P, MillsEJ: Antifungal treatment for systemtic candida infec-tons: A mixed treatment comparison meta-analysis.Unpublished) [4,17,19-22]with three different types ofnetwork structures: (I) star-network, having a commoncomparator and containing no loops (figures 1, 2, 3), (II)single-loop network (figures 4 and 5), containing onlyone loop, and (III) multi-loop network, containing two ormore loops (figures 6 and 7). All seven reviews were pub-lished between the years 1997 to present.Number of comparisonsThe seven systematic reviews retained in our analysesincluded between 4 and 8 treatments. Four reviews didnot have a no-treatment control intervention (Perri D,O'Regan C, Cooper C, Nachega JB, Wu P, Tleyjeh I, PhilipsP, Mills EJ: Antifungal treatment for systemtic candidainfectons: A mixed treatment comparison meta-analy- sis. Unpublished). [4,20,22] The number of trialsincluded in the seven systematic reviews ranged from 10to 29. Two reviews had insufficient mutual comparatorarms to allow the adjusted indirect comparison evalua-tion on each intervention (Perri D, O'Regan C, Cooper C,Star-network of evidence formed by the seven stent treat-ments on target lesion r vascularization event rates,toge her with information n the numb r of trials, umber of patie ts and number of e ts per (direct) tre tment com-r sonFi ur  1Star-network of evidence formed by the seven stent treatments on target lesion revascularization event rates, together with information on the number of trials, number of patients and number of events per (direct) treatment comparison. Each treatment is a node in the network. The links between nodes are trials or pairs of trial arms. The numbers along the link lines indicate 1 2 1 5 5 4          Comparison Number of Trials  per Comparison Number of Patients  per Comparison Number of Events per Comparison A vs. B 1 119 vs. 241 11 vs. 49 A vs. C 2 58 vs. 48 7 vs. 2 A vs. D 1 50 vs. 100 6 vs. 11 A vs. E 4 643 vs. 818 71 vs. 58 A vs. F 5 1,179 vs. 1,178 183 vs. 55 A vs. G 5 999 vs. 1,007 205 vs. 44  A B G C D E FA = BMS B = AES C = Polymeric EES D = MES E = Apolymeric PES F = Polymeric PES G = Polymeric SES Star-network of evidence formed by the treatments Placebo, Ketoprofen, Ibuprofen, Felbinac, Piroxicam, Indometh in and Other NSAID, together with inf rm tion on the num erof rials, num er of patients and number of eve ts per (direct) comparis nFigur 2Star-network of evidence formed by the treatments Placebo, Ketoprofen, Ibuprofen, Felbinac, Piroxicam, Indomethacin and Other NSAID, together with information on the number of trials, number of patients and number of events per (direct) compari-son.3 6 5 3          Direct Comparison Number of Trials  per Comparison Number of Patients  per Comparison Number of Events per Comparison A vs. B 6 260 vs. 258 101 vs. 203 A vs. C 5 186 vs. 183  67 vs. 112 A vs. D 3 203 vs. 210 57 vs. 62 A vs. E 3 194 vs. 283 85 vs. 184 A vs. F 3 110 vs. 197 39 vs. 95 A vs. G 9 377 vs. 397 166 vs. 292  B A E G9  F C D3 A = Placebo B = Ketoprofen C = Ibuprofen D = Felbinac E = Piroxicam F = Indomethacin G = Other NSAID Star-network of evidence formed by the four statin treat-ments and the placebo treatment in primary preven ion of cardiovascular mortality, together wi  inf rmatio on thenumber of rials, numbe  f pati nts and number of events pe  (di ect) co paris nFigur  3Star-network of evidence formed by the four statin treatments and the placebo treatment in primary prevention of cardiovascular mortality, together with information on the number of trials, number of 4 2 2 11      Comparison Number of Trials  per Comparison Number of Patients  per Comparison Number of Events per Comparison A vs. B 4 7,860 vs. 8,047 149 vs. 134 A vs. C 2 1,337 vs. 1,333 68 vs. 53 A vs. D 11 18,666 vs. 19,123  581 vs. 529 A vs. E 2 3,760 vs. 3,764 31 vs. 17  A B  C D EA = Placebo B = Atorvastatin C = Fluvastatin D = Pravastatin E = Lovastatin Page 4 of 12(page number not for citation purposes)the number of trials or pairs of trial arms for that link in the network.patients and number of events per (direct) compari-son.Trials 2009, 10:86 http://www.trialsjournal.com/content/10/1/86Nachega JB, Wu P, Tleyjeh I, Philips P, Mills EJ: Antifun-gal treatment for systemtic candida infectons: A mixedtreatment comparison meta-analysis. Unpublished).[21] There were no three or greater-armed trials found inany of the seven systematic reviews.Analyses 1-3 (figures 1, 2, 3) represent star-shaped com-parisons whereby each intervention shares a mutual com-parator. Analysis 4 and 5 (figures 4 and 5) are networkswith a single loop demonstrating that multiple interven-tions have been compared, but do not necessarily have amutual comparator across treatment. Analyses 6 and 7(figures 6 and 7) are multi-loop comparisons wherebymore treatments exist that have not had mutual compara-tors.Analysis 1. Drug-eluting stents compared to bare-metal stents on target lesion revascularization event rates[22]We evaluated the impact of drug-eluting stents comparedto bare-metal stents on the outcome of target lesion revas-Single-loop network of evidence formed by the four antibi-otic and antiseptic treatments, togeth r with information on the number of trials, nu ber of patients and number of ev nts per (direct) treat ent omparisonFigure 4Single-loop network of evidence formed by the four antibiotic and antiseptic treatments, together with information on the number of trials, number of patients and number of events per (direct) treat-ment comparison.2  83  5         Direct Comparison Number of Trials  per Comparison Number of Patients  per Comparison Number of Events per Comparison A vs. B 2 99 vs. 98 80 vs. 35 B vs. C 8 322 vs. 356 60 vs. 98 B vs. D 3 114 vs. 149 45 vs. 117 C vs. D 5 176 vs. 157 73 vs. 108  A B D  CA = No Treatment B = Topical Quinolone Antibiotic C = Topical Non-Quinolone AntibioticD = Topical Antiseptic Single-loop network of evidence formed by five antifungal treatments, together with information on the number of tri-als, number of pati nts and number of events per (di ect) com risonFigure 5Single-loop network of evidence formed by five anti-fungal treatments, together with information on the number of trials, number of patients and number of 1 4 2 2          Direct Comparison Number of Trials  per Comparison Number of Patients  per Comparison Number of Events per Comparison A vs. B 4 232 vs. 245 28 vs. 28 A vs. C 2 81 vs. 110 9 vs. 6 A vs. D 1 41 vs. 43 5 vs. 3 A vs. E 2 88 vs. 49 4 vs. 4 B vs. C 1 91 vs. 97 7 vs. 12  B A  C D E1 A = Control B = Fluconazol C = Itraconazole D = Liposomal Amphotericin B E = Ketaconazole Multi-loop network of evidence formed by the four treat-ments f r pr vention of Pn umocystis carinii pne moni , toge her with inf rmation on the nu ber of trials, number of patients and number of events per (dir ct) reatmen  com-r sonFi ur  6Multi-loop network of evidence formed by the four treatments for prevention of Pneumocystis carinii pneumonia, together with information on the number of trials, number of patients and number of events per (direct) treatment comparison.2 2 6 3 9    Direct Comparison Number of Trials  per Comparison Number of Patients  per Comparison Number of Events per Comparison A vs. B 9 681 vs. 613 26 vs. 74 A vs. C 2 315 vs. 335 43 vs. 42 A vs. D 6 488 vs. 480 13 vs. 46 B vs. C 3 300 vs. 268 35 vs. 29 B vs. D 2 418 vs. 464 23 vs. 22 B A C D A = TMP-SXB = AP C = D D = D/P Multi-loop network of evidence formed by the eight antifun-gal treatme ts, t gether with information on th  number of trials, nu ber of patients and number of events per (direct)eatment compa isonFigure 7Multi-loop network of evidence formed by the eight antifungal treatments, together with information on the number of trials, number of patients and number 1 1 1 2    F   D    EB   C  A  H     G     Direct Comparison Number of Trials  per Comparison Number of Patients  per Comparison Number of Events per Comparison A vs. C 10 669 vs. 653 157 vs. 183 A vs. G 1 118 vs. 127 37 vs. 29 B vs. C 1 109 vs. 115 39 vs. 38 B vs. D 1 556 vs. 539 61 vs. 75 B vs. F 1 193 vs. 402 51 vs. 125 C vs. E 1 122 vs. 248 51 vs. 88 C vs. H 2 197 vs. 195 31 vs. 23 D vs. E 1 422 vs. 415 25 vs. 33 D vs. F 1 247 vs. 247 108 vs. 106 1 1 1 101 A = Fluconazole B = Caspofungin C = Amphotericin B Deoxycholate D = Amphotericin B Liposomal E = Voriconazole F = Micafungin G = Anidulafungin H = Itraconazole Page 5 of 12(page number not for citation purposes)events per (direct) treatment comparison. of events per (direct) treatment comparison.Trials 2009, 10:86 http://www.trialsjournal.com/content/10/1/86cularization event rates on the basis of 18 2-arm ran-domised trials comparing 7 different treatments. Figure 1displays the network of evidence available from these tri-als. Table 1 shows the results of the pairwise treatmentcomparisons when using direct, head-to-head data (inbold), the mixed treatment approach and the adjustedindirect comparison approach. In a single instance, themixed treatment comparison approach found a signifi-cant difference between the effects of two treatmentswhen the adjusted indirect comparison approach did not.According to the a priori consistency rule, the estimatedORs and associated uncertainty intervals were impor-tantly different between the two approaches for only fourpairwise treatment comparisons.Analysis 2. NSAIDS for acute pain[19]We evaluated the effects of 7 different interventions foracute pain from 29 trials that included 58 trial arms, for apossible 21 comparisons. See Figure 2 and Table 2. Wefound no important distinctions between the adjustedindirect comparison and mixed treatment comparisonapproaches.Analysis 3. Statins for the primary prevention of cardiovascular mortality[17]We evaluated the role of 4 statin interventions comparedto placebo/standard care for the prevention of cardiovas-cular mortality in primary prevention of cardiovasculardisease populations. See Figure 3 and Table 3. There were18 trials included, from 38 arms, allowing for a possible10 comparisons. We found no major discrepanciesbetween the two comparative approaches.Analysis 4. Topical treatment for treatment of ear discharge at 1 and 2 weeks [21]We evaluated the role of topical antibiotics for the preven-tion of ear discharge for patients with eardrum perfora-tions using 18 2-arm randomised trials comparing 4different treatments. Figure 4 displays the network of evi-dence available from these trials. The results of the 2 pair-wise treatment comparisons performed via the adjustedindirect comparison approach and 6 pair-wise treatmentcomparisons performed via the mixed treatment compar-ison approach are shown in Table 4. In one circumstance,the mixed treatment comparison approach found a statis-Table 1: Drug-eluting stents compared to bare-metal stents on revascularization status[22].Treatment Comparison Mixed Treatment Comparison Adjusted Indirect ComparisonOdds Ratio 95% Credible Interval Odds Ratio 95% Confidence IntervalAES vs. BMS 2.61 (1.32, 5.44) 2.51 (1.22, 5.56)Polymeric EES vs. BMS 0.31 (0.04, 1.63) 0.37 (0.06, 2.14)MES vs. BMS 0.93 (0.32, 2.88) 0.91 (0.28, 3.19)Apolymeric PES vs. BMS 0.64 (0.44, 0.93) 0.64 (0.44, 0.93)Polymeric PES vs. BMS 0.26 (0.19, 0.36) 0.27 (0.20, 0.37)Polymeric SES vs. BMS 0.17 (0.12, 0.24) 0.20 (0.13, 0.30)Polymeric EES vs. AES 0.12 (0.01, 0.73) 0.14 (0.02, 1.04)*MES vs. AES 0.36 (0.10, 1.34) 0.36 (0.09, 1.44)Apolymeric PES vs. AES 0.25 (0.11, 0.54) 0.25 (0.11, 1.57)Polymeric PES vs. AES 0.10 (0.05, 0.21) 0.10 (0.04, 0.23)Polymeric SES vs. AES 0.07 (0.03, 0.14) 0.07 (0.03, 0.18)MES vs. Polymeric EES 3.00 (0.40, 30.08) 2.45 (0.28, 21.48)***Apolymeric PES vs. Polymeric EES 2.06 (0.37, 16.49) 1.72 (0.26, 11.08)***Polymeric PES vs. Polymeric EES 0.85 (0.15, 6.79) 0.72 (0.11, 4.61)***Polymeric SES vs. Polymeric EES 0.56 (0.10, 4.53) 0.54 (0.08, 3.50)***Apolymeric PES vs. MES 0.69 (0.21, 2.16)) 0.70 (0.20, 2.42)Polymeric PES vs. MES 0.28 (0.09, 0.87) 0.29 (0.08, 1.00)Polymeric SES vs. MES 0.19 (0.06, 0.58 0.21 (0.06, 0.77)Polymeric PES vs. Apolymeric PES 0.41 (0.25, 0.67) 0.42 (0.26, 0.68)Polymeric SES vs. Apolymeric PES 0.27 (0.16, 0.45) 0.31 (0.17, 0.55)Polymeric SES vs. Polymeric PES 0.66 (0.41, 1.05) 0.74 (0.43, 1.25)Page 6 of 12(page number not for citation purposes)Bolded text denotes head-to-head meta-analysis evaluations. * Mixed treatment method identifies a significant effect, adjusted indirect comparison does not, adjusted indirect comparison identifies significant effect, mixed treatment comparison does not. ** Direction of effect differs between approaches, ***Important effect size differences.Trials 2009, 10:86 http://www.trialsjournal.com/content/10/1/86Page 7 of 12(page number not for citation purposes)Table 2: NSAIDS for acute pain[19].Treatment Comparison Mixed Treatment Comparison Adjusted Indirect ComparisonOdds Ratio 95% Credible Interval Odds Ratio 95% Confidence IntervalKetoprofen vs. Placebo 6.55 (4.35, 9.95) 6.06 (4.07, 9.04)Ibuprofen vs. Placebo 2.95 (1.92, 4.57) 2.70 (1.78, 4.09)Felbinac vs. Placebo 3.02 (2.01, 4.58) 2.91 (1.94, 4.39)Piroxicam vs. Placebo 2.75 (1.86, 4.08) 2.65 (1.80, 3.90)Indomethacin vs. Placebo 1.60 (0.99, 2.62) 1.58 (0.97, 2.57)Other NSAID vs. Placebo 3.74 (2.73, 5.13) 3.31 (2.46, 4.45Ibuprofen vs. Ketoprofen 0.45 (0.25, 0.81) 0.44 (0.25, 0.79)Felbinac vs. Ketoprofen 0.46 (0.26, 0.83) 0.48 (0.27, 0.84)Piroxicam vs. Ketoprofen 0.42 (0.24, 0.74) 0.43 (0.25, 0.76)Indomethacin vs. Ketoprofen 0.24 (0.13, 0.46) 0.26 (0.13, 0.48)Other NSAID vs. Ketoprofen 0.57 (0.34, 0.95) 0.54 (0.33, 0.89)Felbinac vs. Ibuprofen 1.02 (0.56, 1.86) 1.07 (0.60, 1.92)Piroxicam vs. Ibuprofen 0.93 (0.52, 1.67) 0.98 (0.55, 1.73)Indomethacin vs. Ibuprofen 0.54 (0.29, 1.04) 0.58 (0.30, 1.11)Other NSAID vs. Ibuprofen 1.27 (0.74, 2.15) 1.22 (0.73, 2.04)Piroxicam vs. Felbinac 0.91 (0.52, 1.60) 0.91 (0.51, 1.59)Indomethacin vs. Felbinac 0.53 (0.28, 1.01) 0.54 (0.28, 1.02)Other NSAID vs. Felbinac 1.24 (0.74, 2.08) 1.13 (0.68, 1.88)Indomethacin vs. Piroxicam 0.58 (0.31, 1.09) 0.59 (0.31, 1.11)Other NSAID vs. Piroxicam 1.36 (0.82, 2.25) 1.24 (0.76, 2.03)Other NSAID vs. Indomethacin 2.33 (1.31, 4.15) 2.09 (1.18, 3.70)Bolded text denotes head-to-head meta-analysis evaluations. * Mixed treatment method identifies a significant effect, adjusted indirect comparison does not, adjusted indirect comparison identifies significant effect, mixed treatment comparison does not. ** Direction of effect differs between approaches, ***Important effect size differences.Table 3: Statins for the prevention of cardiovascular mortality[17].Treatment Comparison Mixed Treatment Comparison Adjusted Indirect ComparisonOdds Ratio 95% Credible Interval Odds Ratio 95% Confidence IntervalsAtorvastatin vs. Placebo 0.88 (0.69, 1.11) 0.88 (0.70, 1.12)Fluvastatin vs. Placebo 0.77 (0.53, 1.11) 0.77 (0.53, 1.11)Pravastatin vs. Placebo 0.91 (0.80, 1.03) 0.91 (0.81, 1.02)Lovastatin vs. Placebo 0.67 (0.35, 1.24) 0.55 (0.31, 0.99)Fluvastatin vs. Atorvastatin 0.87 (0.56, 1.35) 0.87 (0.56, 1.35)Pravastatin vs. Atorvastatin 1.03 (0.79, 1.35) 1.03 (0.79, 1.33)Lovastatin vs. Atorvastatin 0.76 (0.38, 1.47) 0.62 (0.33, 1.15)Pravastatin vs. Fluvastatin 1.18 (0.80, 1.75) 1.18 (0.79, 1.74)Lovastatin vs. Fluvastatin 0.87 (0.41, 1.80) 0.71 (0.36, 1.41)Lovastatin vs. Pravastatin 0.74 (0.39, 1.38) 0.60 (0.33, 1.08)Bolded text denotes head-to-head meta-analysis evaluations. * Mixed treatment method identifies a significant effect, adjusted indirect comparison does not, adjusted indirect comparison identifies significant effect, mixed treatment comparison does not. ** Direction of effect differs between approaches, ***Important effect size differences.Trials 2009, 10:86 http://www.trialsjournal.com/content/10/1/86tically significant difference between the effects of twotreatments, when the adjusted indirect comparisonapproach did not.Analysis 5. Antifungal agents for preventing mortality in solid organ transplant recipients[20]We evaluated the role of antifungal agents for preventingmortality in solid organ transplant recipients on the basisof 10 2-arm randomised trials comparing 5 different treat-ments. The network of evidence for these trials is shownin Figure 5. The results for the 5 possible pair-wise treat-ment comparisons using the adjusted indirect compari-son approach and 10 comparisons using the mixedtreatment comparison are shown in Table 5. In a singlecase, the mixed treatment comparison approach found adifferent direction of effect than the adjusted indirectcomparison approach. The estimated ORs and associateduncertainty intervals produced by the two approacheswere importantly different for three pair-wise treatmentcomparisons.Analysis 6. Prophylactic treatments against pneumocystis carinii pneumonia and toxoplasma encephalitis in HIV-infected patients[4]We evaluated 4 different interventions from 22 trials with44 trial arms, allowing a possible 6 comparisons. See Fig-ure 6 and Table 6. In this example, the adjusted indirectcomparison was only required for one comparison butdiffered importantly from the mixed treatment method.Analysis 7. Antifungal agents for the prevention of mortality among patients with invasive candidemia(Perri D, O'Regan C, Cooper C, Nachega JB, Wu P, TleyjehI, Philips P, Mills EJ: Antifungal treatment for systemticcandida infectons: A mixed treatment comparison meta-analysis. Unpublished.)We evaluated the effectiveness of 8 different treatmentsfrom 19 trials, allowing 38 arms, for a possible 28 com-parisons. See Figure 7 and Table 7. For 9 comparisons wewere unable to conduct the adjusted indirect evaluation,as no suitable mutual comparator existed. The direction ofeffect differed between the two approaches in 4 studies. Inone circumstance, the adjusted indirect approach foundsignificant treatment effect while the mixed treatmentmethod did not.DiscussionOur paper presents important evidence on the relativeperformance of the adjusted indirect comparison andmixed treatment comparison approaches to evaluatingmultiple health interventions in the absence of sufficientdirect evidence.For the 3 star-networks considered in this paper, we foundthat both approaches led to similar results, as they coulduse all the available information in the data. In general,some slight difference may exist between the results pro-duced by the two approaches for this type of networksince the adjusted indirect comparison approach uses(approximate) normal likelihood while the mixed treat-ment comparison approach uses (exact) binomial likeli-hood. If one chooses to ignore such a slight difference, theadjusted indirect comparison approach is easier to use forstar-networks than the mixed treatment comparisonapproach.For the 2 single-loop networks included in this paper, wefound that the adjusted indirect comparison and mixedtreatment comparison approached yielded comparableestimates of relative treatment effectiveness. However, thetwo approaches will be expected to yield different resultsfor general single-loop networks, simply because themixed treatment comparison approach is based on allavailable information in the data but the adjusted indirectcomparison approach is not.Table 4: Topical treatment for treatment of ear discharge at 1 and 2 weeks[21].Treatment Comparison Mixed Treatment Comparison Adjusted Indirect ComparisonOdds Ratio 95% Credible Interval Odds Ratio 95% Confidence IntervalTopical Quinolone Antibiotic vs. No Treatment 0.13 (0.06, 0.24) 0.08 (0.01, 0.51)Topical Non-Quinolone Antibiotic vs. No Treatment 0.21 (0.10, 0.44) 0.28 (0.09, 0.52)Topical Antiseptic vs. No Treatment 0.71 (0.32, 1.55) 0.61 (0.22, 1.22)Topical Non-Quinolone Antibiotic vs. Topical Quinolone Antibiotic1.67 (1.17, 2.31) 1.62 (0.92, 2.85)Topical Antiseptic vs. Topical Quinolone Antibiotic 5.64 (3.70, 8.70) 4.31 (1.34, 13.90)Topical Antiseptic vs. Topical Non-Quinolone Antibiotic 3.37 (2.25, 5.03) 3.02 (0.74, 12.29)*Page 8 of 12(page number not for citation purposes)Bolded text denotes head-to-head meta-analysis evaluations. * Mixed treatment method identifies a significant effect, adjusted indirect comparison does not, adjusted indirect comparison identifies significant effect, mixed treatment comparison does not. ** Direction of effect differs between approaches, ***Important effect size differences.Trials 2009, 10:86 http://www.trialsjournal.com/content/10/1/86Finally, we found that both the adjusted indirect compar-ison and the mixed treatment comparison approach pro-duced comparable estimates of relative treatmenteffectiveness for the two multi-loop networks consideredin this paper. As pointed out by one of the referees duringpeer-review, in general, the adjusted indirect comparisonapproach may be difficult, if not impossible, to apply forthis type of network. As an illustration, suppose we areinterested in the indirect estimate for the OR of the pair-wise comparison of treatments C and D in Figure 7, wherethere is no direct comparison between these two treat-ments. But, through the network of evidence, there arethree ways to perform the adjusted indirect comparison oftreatments C and D: (1) using comparisons C versus E andD versus E; (2) using comparisons C versus B and D versusB; (3) using comparisons C versus B, B versus F, and F ver-sus D. Clearly, these comparisons will lead to differentresults. One possible way to deal with this problem is toapply the adjusted indirect comparison approach threetimes to these data sets respectively and then combinethem together to get a pooled estimate. But, crucially,these three routes to the estimate of the OR of the pair-wise comparison of treatments C and D are not in thiscase statistically independent. As a result, the resultingestimates cannot be pooled by a simple weighted average.The mixed treatment comparisons approach, however,will combine this information simultaneously and pro-duce a coherent set of estimates for all the treatment con-trasts, based on all the data.The adjusted indirect comparison approach may be pre-ferred for star-networks, as it is typically easier to imple-Table 5: Antifungal agents for preventing mortality in solid organ transplant recipients[20].Treatment Comparison Mixed Treatment Comparison Adjusted Indirect ComparisonOdds Ratio 95% Credible Interval Odds Ratio 95% Confidence IntervalFluconazole vs. Control 0.81 (0.48, 1.37) 0.94 (0.54, 1.63)Itraconazole vs. Control 0.90 (0.41, 1.99) 0.49 (0.15, 1.56)Liposomal Amphotericin B vs. Control 0.50 (0.09, 2.33) 0.54 (0.10, 2.52)Ketoconazole vs. Control 1.83 (0.38, 8.93) 1.66 (0.41, 6.66)Intraconazole vs. Fluconazole 1.12 (0.52, 2.41) 1.69 (0.58, 5.33)Liposomal Amphotericin B vs. Fluconazole 0.62 (0.10, 3.17) 0.57 (0.09, 3.39)Ketoconazole vs. Fluconazole 2.27 (0.43, 12.09) 1.76 (0.39, 7.94)***Liposomal Amphotericin B vs. Itraconazole 0.55 (0.09, 3.11) 1.10 (0.14, 8.65)**, ***Ketoconazole vs. Itraconazole 2.03 (0.35, 11.87) 3.38 (0.54, 21.16)***Ketoconazole vs. Liposomal Amphotericine B 3.68 (0.41, 35.30) 3.07 (0.34, 27.48)***Bolded text denotes head-to-head meta-analysis evaluations. * Mixed treatment method identifies a significant effect, adjusted indirect comparison does not, adjusted indirect comparison identifies significant effect, mixed treatment comparison does not. ** Direction of effect differs between approaches, ***Important effect size differences.Table 6: Prophylactic treatments against pneumocystis carinii pneumonia and toxoplasma encephalitis in HIV-infected patients[4].Treatment Comparison Mixed Treatment Comparison Adjusted Indirect ComparisonOdds Ratio 95% Credible Interval Odds Ratio 95% Confidence IntervalAP vs. TMP-SMX 2.68 (1.90, 3.81) 3.19 (2.02, 5.03)D vs. TMP-SMX 1.38 (0.94, 2.04) 0.92 (0.58, 1.46)D/P vs. TMP-SMX 3.02 (1.92, 4.79) 3.22 (1.70, 6.10)D vs. AP 0.52 (0.34, 0.78) 0.90 (0.53, 1.52)D/P vs. AP 1.13 (0.71, 1.80) 0.86 (0.47, 1.57)D/P vs. D 2.19 (1.26, 3.82) 3.5 (1.59, 7.69)***Page 9 of 12(page number not for citation purposes)Bolded text denotes head-to-head meta-analysis evaluations. * Mixed treatment method identifies a significant effect, adjusted indirect comparison does not, adjusted indirect comparison identifies significant effect, mixed treatment comparison does not. ** Direction of effect differs between approaches, ***Important effect size differences.Trials 2009, 10:86 http://www.trialsjournal.com/content/10/1/86ment than the mixed treatment comparison approach andprovides similar results. For single-loop networks, onecould use either approach, though the results produced bythe two approaches might generally be different, reflectingthe fact that the mixed treatment comparison approachrelies on all of the information available in the data butthe adjusted indirect comparison approach does not. Formulti-loop networks, it might be difficult, if not impossi-ble, to implement the adjusted indirect comparisonapproach in some situations, rendering the mixed treat-There are strengths and limitations that should be consid-ered when interpreting this manuscript. Strengths includeour extensive searching of systematic reviews and inclu-sion of unpublished systematic reviews. It is possible thatwe missed systematic reviews that may have met ourinclusion criteria, however our searches were extensive,were supplemented with others' systematic reviews.[2,13]and were conducted in duplicate to minimize bias. Weapplied the fixed effects method for both the adjustedindirect comparison and mixed treatment comparisonTable 7: Antifungal agents for the prevention of mortality among patients with invasive candidemia (Perri D, O'Regan C, Cooper C, Nachega JB, Wu P, Tleyjeh I, Philips P, Mills EJ: Antifungal treatment for systemtic candida infectons: A mixed treatment comparison meta-analysis. Unpublished)Treatment Comparison Mixed Treatment Comparison Adjusted Indirect ComparisonOdds Ratio 95% Credible Interval Odds Ratio 95% Confidence IntervalCaspofungin vs. Fluconazole 1.01 (0.60, 1.71) 0.85 (0.44, 1.64)**Amphotericin B Deoxycholate vs. Fluconazole 1.26 (0.96, 1.65) 1.31 (0.99, 1.74)Amphotericin B Liposomal vs. Fluconazole 1.20 (0.70, 2.06) - -Voriconazole vs. Fluconazole 1.20 (0.75, 1.94) 0.58 (0.33, 0.99)*, ***Micafungin vs. Fluconazole 1.22 (0.68, 2.16) - -Anidulafungin vs. Fluconazole 0.64 (0.36, 1.14) 0.65 (0.35, 1.19)Itraconazole vs. Fluconazole 0.89 (0.46, 1.69 0.54 (0.27, 1.05)Amphotericin B Deoxycholate vs. Caspofungin 1.24 (0.79, 1.94) 1.12 (0.62, 1.12)Amphotericin B Liposomal vs. Caspofungin 1.19 (0.90, 1.57) 1.31 (0.90, 1.91)Voriconazole vs. Caspofungin 1.19 (0.75, 1.89) 0.67 (0.32, 1.43)**Micafungin vs. Caspofungin 1.20 (0.89, 1.62) 1.25 (0.84, 1.88)Anidulafungin vs. Caspofungin 0.64 (0.29, 1.38) - -Itraconazole vs. Caspofungin 0.87 (0.41, 1.84 0.63 (0.27, 1.47)Amphotericin B Liposomal vs. Amphotericin B Deoxycholate0.96 (0.60, 1.53) 1.80 (0.86, 3.76)**Voriconazole vs. Amphotericin B Deoxycholate 0.96 (0.65, 1.42) 0.76 (0.48, 1.22)Micafungin vs. Amphotericin B Deoxycholate 0.97 (0.58, 1.61) 0.95 (0.55, 1.64)Anidulafungin vs. Amphotericin B Deoxycholate 0.51 (0.27, 0.96) 0.49 (0.25, 0.97)Itraconazole vs. Amphotericin B Deoxycholate 0.70 (0.39, 1.27) 0.71 (0.39, 1.28)Voriconazole vs. Amphotericin B Liposomal 1.00 (0.64, 1.55) 1.37 (0.77, 2.45)Micafungin vs. Amphotericin B Liposomal 1.01 (0.75, 1.35) 0.96 (0.66, 1.40)Anidulafungin vs. Amphotericin B Liposomal 0.54 (0.24, 1.17) - -Itraconazole vs. Amphotericin B Liposomal 0.74 (0.34, 1.57) - -Micafungin vs. Voriconazole 1.01 (0.61, 1.67) - -Anidulafungin vs. Voriconazole 0.54 (0.25, 1.13) - -Itraconazole vs. Voriconazole 0.74 (0.36, 1.50) 0.93 (0.43, 1.98)Anidulafungin vs. Micafungin 0.53 (0.24, 1.19) - -Itraconazole vs. Micafungin 0.73 (0.33, 1.59) - -Itraconazole vs. Anidulafungin 1.37 (0.58, 3.26) - -Bolded text denotes head-to-head meta-analysis evaluations. * Mixed treatment method identifies a significant effect, adjusted indirect comparison does not, adjusted indirect comparison identifies significant effect, mixed treatment comparison does not. ** Direction of effect differs between approaches, ***Important effect size differences.Page 10 of 12(page number not for citation purposes)ment comparison as the preferred choice for this type ofnetwork.approaches. Our goodness of fit checks indicated that thefixed effects mixed treatment comparison approach wasTrials 2009, 10:86 http://www.trialsjournal.com/content/10/1/86sensible for nearly all of the seven systematic reviews. Fur-ther sensitivity analyses performed for this approach con-firmed the robustness of the overall conclusions to theexclusion of discrepant trials. It is possible that we wouldhave found slight differences if we had employed the ran-dom effects method. More often than not however, thesemethods yield comparable estimates of relative treatmenteffects.[18] Some have argued that the fixed effectsmethod should now be preferred over a random effectsmethod as it places a greater weight on larger studies, thusstudies may have reduced bias.[23] Finally, we were una-ble to compare the adjusted indirect comparisonapproach with the head-to-head evaluations as, in this setof systematic reviews, there was an insufficient number oftrials with more than one comparator.Salanti and others have discussed the merits and chal-lenges of the mixed treatment approach.[12,23,24] Themixed treatment comparison is a resource intensiveapproach to conducting analyses as it requires knowledgeof Bayesian principles and working abilities with Win-BUGS, a somewhat user-unfriendly software for thoseunfamiliar with it. However, the mixed treatmentapproach also provides interesting additional informa-tion that may be useful to some readers. Additional infor-mation includes probabilities of a ranking order of theeffectiveness of interventions. For the sake of clarity, wehaven't presented the probabilities associated with eachanalysis. Probabilities may be difficult to interpretthough, particularly when there are not clear differencesamongst them. A further additional source of informationis that this analysis provides indirect comparisons withoutrequiring a mutual comparator, a possible strength overthe adjusted indirect approach. However, we cannot knowwhether this estimate is reliable or similar to an adjustedindirect approach until further trials become available.Finally, some have argued that the mixed treatment com-parison is a 'black-box,' as it may be difficult or impossibleto determine where an analysis has gone incorrectly.[25]Future validations of the analytic manner performed inthis manuscript may yield insights into the transparencyof this method. Finally, no reporting guidelines exist forthe mixed treatment approach. A step forward may be thedevelopment of minimum reporting criteria for thisapproach.[11,12]For less complex analyses, such as star-shaped networks,the adjusted indirect comparison may be easier for meta-analysts to apply in their general practice. One of us (GG)was involved in the development of this approach.[4] Theadjusted indirect comparison is limited in more complexevaluations, as compared to the mixed treatment compar-ison, as it requires the utilization of a mutual comparatorcussed above, the validity of indirect comparisons withoutmutual comparators that are performed via the mixedtreatment comparison approach may be reasonably ques-tioned. The adjusted indirect comparison approachrequires the knowledge of standard meta-analysis tech-niques and working knowledge of programmable soft-ware such as R, S-Plus, Stata or SAS, so is arguably alsoresource intensive. A recent free download of a simplesoftware may make this approach accessible for non-stat-isticians.[25,26]There is also concern that both the adjusted indirect com-parison and mixed treatment comparison approaches willhave less power than the direct approach and may some-times lead to indeterminate results, in the form of wideuncertainty intervals for relative intervention effects.Inferences based on such findings may therefore be lim-ited. In addition, it is not clear yet how to interpret resultsthat differ substantially between the two approaches.Finally, although the choice of approach may differ onlymarginally in treatment effect estimates, the impact ofsmall differences may affect future analyses based onstudy findings, such as cost-effectiveness models. There isa clear need to evaluate whether one method may impor-tantly impact cost-effectiveness projections overanother.[11]ConclusionIn conclusion, both the mixed treatment comparisonapproach and the adjusted indirect comparison approachprovide compelling inferences about the relative effective-ness of interventions. In less complex indirect compari-sons, where a mutual comparator exists, the adjustedindirect comparison may be favourable due to its simplic-ity. In more complex models, the mixed treatment com-parison appears to offer benefits for comparisons thatother methods cannot.AbbreviationsAES: actinomycin-d-eluting stent; BMS: bare-metal stents,PES; EES: everolimus-eluting stent; MES: micophenolate-eluting stent; PES: paclitaxel-eluting stent; SES: sirolimus-eluting stent; RCT: randomized clinical trial; OR: oddsratio; CI: confidence interval; CrI: credible interval; D:Dapsone; D/P: dapsone/pyrimethamine; AP: aerosolizedpentamidine; TMP-SMX: trimethoprim-sulfamethoxa-zole.Competing interestsNone declared. COR has previously been employed byPfizer Ltd. and is currently employed by Merck, Sharpe &Dohme (MSD) Ltd. MSD had no role in the development,execution or publication of the paper. EM has consultedPage 11 of 12(page number not for citation purposes)when performing indirect comparisons. However, as dis- to Pfizer Ltd. and received unrestricted research grantsPublish with BioMed Central   and  every scientist can read your work free of charge"BioMed Central will be the most significant development for disseminating the results of biomedical research in our lifetime."Sir Paul Nurse, Cancer Research UKYour research papers will be:available free of charge to the entire biomedical communitypeer reviewed and published immediately upon acceptancecited in PubMed and archived on PubMed Central Trials 2009, 10:86 http://www.trialsjournal.com/content/10/1/86from Pfizer Ltd., GG has received unrestricted researchgrants from several for-profit companies. IG runs a statis-tical consulting firm.Authors' contributionsCOR, EM, IG and GG were responsible for the study con-cept. COR, EM and OE were responsible for the studysearches. COR, EM, EO, IG and GG were responsible forstudy extraction and analysis. COR, EM, IG and GG wereresponsible for study writing. COR, IG, EO, GG and EMapproved the submitted manuscript.References1. Guyatt G, Schunemann H, Cook D, Jaeschke R, Pauker S, Bucher H:Grades of recommendation for antithrombotic agents.  Chest2001, 119:3S-7S.2. Song F, Altman DG, Glenny AM, Deeks JJ: Validity of indirect com-parison for estimating efficacy of competing interventions:empirical evidence from published meta-analyses.  BMJ (Clini-cal research ed) 2003, 326:472.3. Glenny AM, Altman DG, Song F, Sakarovitch C, Deeks JJ, D'Amico R,Bradburn M, Eastwood AJ: Indirect comparisons of competinginterventions.  Health technology assessment (Winchester, England)2005, 9:1-134.4. Bucher HC, Guyatt GH, Griffith LE, Walter SD: The results ofdirect and indirect treatment comparisons in meta-analysisof randomized controlled trials.  J Clin Epidemiol 1997,50:683-691.5. Ades AE: A chain of evidence with mixed comparisons: mod-els for multi-parameter synthesis and consistency of evi-dence.  Stat Med 2003, 22:2995-3016.6. Hasselblad V: Meta-analysis of multitreatment studies.  MedDecis Making 1998, 18:37-43.7. Lumley T: Network meta-analysis for indirect treatment com-parisons.  Stat Med 2002, 21:2313-2324.8. Lu G, Ades AE: Combination of direct and indirect evidence inmixed treatment comparisons.  Stat Med 2004, 23:3105-3124.9. Salanti G, Marinho V, Higgins JP: A case study of multiple-treat-ments meta-analysis demonstrates that covariates should beconsidered.  J Clin Epidemiol 2009, 62:857-864.10. Song F, Loke YK, Walsh T, Glenny AM, Eastwood AJ, Altman DG:Methodological problems in the use of indirect comparisonsfor evaluating healthcare interventions: survey of publishedsystematic reviews.  BMJ (Clinical research ed) 2009, 338:b1147.11. Sutton A, Ades AE, Cooper N, Abrams K: Use of indirect andmixed treatment comparisons for technology assessment.PharmacoEconomics 2008, 26:753-767.12. Salanti G, Higgins JP, Ades AE, Ioannidis JP: Evaluation of networksof randomized trials.  Statistical methods in medical research 2008,17:279-301.13. Salanti G, Kavvoura FK, Ioannidis JP: Exploring the geometry oftreatment networks.  Annals of internal medicine 2008,148:544-553.14. Higgins JP, Whitehead A: Borrowing strength from external tri-als in a meta-analysis.  Stat Med 1996, 15:2733-2749.15. NICE: Updated guide to the methods of technology appraisal- June 2008.  2008 [http://www.nice.org.uk/media/B52/A7/TAMethodsGuideUpdatedJune2008.pdf].16. Song F, Glenny AM, Altman DG: Indirect comparison in evaluat-ing relative efficacy illustrated by antimicrobial prophylaxisin colorectal surgery.  Controlled clinical trials 2000, 21:488-497.17. Mills EJ, Rachlis B, Wu P, Devereaux PJ, Arora P, Perri D: Primaryprevention of cardiovascular mortality and events with sta-tin treatments: a network meta-analysis involving more than65,000 patients.  Journal of the American College of Cardiology 2008,52:1769-1781.18. Guyatt GH, Rennie D: Users Guides to the Medical Literature.JAMA Press, Chicago; 2007:556. 19. Mason L, Moore RA, Edwards JE, Derry S, McQuay HJ: Topical20. Playford EG, Webster AC, Sorell TC, Craig JC: Antifungal agentsfor preventing fungal infections in solid organ transplantrecipients.  Cochrane database of systematic reviews (Online)2004:CD004291.21. Macfadyen CA, Acuin JM, Gamble C: Topical antibiotics withoutsteroids for chronically discharging ears with underlying ear-drum perforations.  Cochrane database of systematic reviews (Online)2005:CD004618.22. Biondi-Zoccai GG, Agostini P, Abbate A, Testa L, Burzotta F, Lotri-onte M: Adjusted indirect comparison of intracoronary drug-eluting stents: evidence from a meta-analysis of randomizedbare-metal-stent-controlled trials.  Int J Cardiol 2005,100:119-123.23. Pocock SJ: Safety of drug-eluting stents: demystifying networkmeta-analysis.  Lancet 2007, 370:2099-2100.24. Caldwell DM, Ades AE, Higgins JP: Simultaneous comparison ofmultiple treatments: combining direct and indirect evi-dence.  BMJ (Clinical research ed) 2005, 331:897-900.25. Wells GA, Sultan SA, Chen L, Khan M, Coyle D: Indirect evidence:indirect treatment comparisons in meta-analysis.  Ottawa:Canadian Agency for Drugs and Technologies in Health; 2009. 26. Wells GA, Sultan SA, Chen L, Khan M, Coyle D: Indirect treat-ment comparison [computer program]. Version 1.0.  Ottawa:Canadian Agency for Drugs and Technologies in Health; 2009. yours — you keep the copyrightSubmit your manuscript here:http://www.biomedcentral.com/info/publishing_adv.aspBioMedcentralPage 12 of 12(page number not for citation purposes)NSAIDs for acute pain: a meta-analysis.  BMC family practice2004, 5:10.

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.52383.1-0132612/manifest

Comment

Related Items