10.07.2015 Views

Systematic review, meta-analysis and economic modelling of ...

Systematic review, meta-analysis and economic modelling of ...

Systematic review, meta-analysis and economic modelling of ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

DOI: 10.3310/hta17010 Health Technology Assessment 2013 Vol. 17 No. 1quality was assessed with each item scored as ‘yes’, ‘no’ or ‘unclear’. Further details on the modifiedversion <strong>of</strong> the QUADAS tool are provided in Appendix 2.The quality assessment for prognostic studies <strong>of</strong> biomarkers, exercise ECG <strong>and</strong> CTCA was conducted usingan adapted version <strong>of</strong> the framework described by Altman. 29 The assessment asked the following sevenquestions <strong>of</strong> each study:1. Sample <strong>of</strong> patients Are inclusion criteria defined?2. Sample <strong>of</strong> patients Are characteristics described (age <strong>and</strong> sex)?3. Outcome Is a MACE defined in the methods section?4. Outcome Is a MACE identification <strong>and</strong> definition independent <strong>of</strong> the index test?5. Outcome Is a MACE outcome recorded for at least 80% <strong>of</strong> the cohort from baseline episode?6. Analysis Was a multivariate <strong>analysis</strong> undertaken (were other variables, other than our variable <strong>of</strong>interest, included in the <strong>analysis</strong>)?7. Analysis Is troponin measured <strong>and</strong> included in the multivariate <strong>analysis</strong>, or is <strong>analysis</strong> stratified bytroponin or limited to those with a negative troponin?Questions 1 <strong>and</strong> 2 assessed adequacy <strong>of</strong> reporting. Question 3 aimed to determine whether or not theoutcome <strong>of</strong> interest (MACE) appeared to have been defined a priori by the researchers (i.e. in the methodssection rather than the results section). Question 4 aimed to determine whether or not a presentingdiagnosis (such as MI) that could have been associated with a positive index test was incorporated in thedefinition <strong>of</strong> MACEs. Question 5 assessed adequacy <strong>of</strong> follow-up. Although this was an inclusion criterionfor the <strong>review</strong>, 80% follow-up was not always clearly reported or achieved at all time points. Question 6assessed whether or not the study had explored beyond an association between the index test <strong>and</strong> MACEsto determine whether or not the biomarker added prognostic value beyond routine assessment. Question7 assessed whether this <strong>analysis</strong> was stratified by or adjusted for troponin, to determine whether thebiomarker added prognostic value to that provided by troponin.The methodological quality <strong>of</strong> each prognostic study in the <strong>review</strong> <strong>of</strong> biochemical markers was assessedby one <strong>review</strong>er (SG or CC) but checked by a second (SG or CC) using this modified version <strong>of</strong> the Altmancriteria. 29 The methodological quality <strong>of</strong> each included study in the <strong>review</strong> <strong>of</strong> CTCA <strong>and</strong> exercise ECG wasassessed by one <strong>review</strong>er (JL) but checked by a second (FM) using these same criteria. In all cases <strong>of</strong> doubtin either <strong>review</strong>, the principal investigator (SG) was consulted.Methods <strong>of</strong> data synthesisThe <strong>analysis</strong> was conducted using Bayesian Markov chain Monte Carlo simulation. In general, there areadvantages <strong>of</strong> the Bayesian approach over a Classical approach, including the ability to (1) analyse complexmodels exactly; (2) incorporate external evidence in addition to sample data; <strong>and</strong> (3) make probabilisticstatements about parameters. In particular, the approach allowed the direct use <strong>of</strong> a binomial likelihoodfor the sample data, including for studies with very small or zero counts; the ability to incorporateuncertainty in the estimate <strong>of</strong> the between-study st<strong>and</strong>ard deviation (SD), including in studies withrelatively few studies; 30 <strong>and</strong> the ability to generate probability distributions that represent parameteruncertainty about inputs to the <strong>economic</strong> model.The use <strong>of</strong> a r<strong>and</strong>om-effects model is motivated a priori by the assumption that the true sensitivities<strong>and</strong> specificities vary according to the study but that they arise from a common (bivariate) populationdistribution. Heterogeneity is common in <strong>meta</strong>-analyses <strong>of</strong> diagnostic test data <strong>and</strong> the results <strong>of</strong> theseanalyses are no exception. The pooled effects presented in the forest plots represent the means <strong>of</strong> thepopulation <strong>of</strong> sensitivities <strong>and</strong> specificities, <strong>and</strong> these are the parameters that are commonly presented asthe results <strong>of</strong> a <strong>meta</strong>-<strong>analysis</strong>. Also presented with the forest plots are predictive effects; these representthe range <strong>of</strong> estimates that we might expect to see in the population taking into account uncertaintyin both the estimate <strong>of</strong> the mean sensitivity <strong>and</strong> specificity <strong>and</strong> the uncertainty in the estimates <strong>of</strong> the© Queen’s Printer <strong>and</strong> Controller <strong>of</strong> HMSO 2013. This work was produced by Goodacre et al. under the terms <strong>of</strong> a commissioning contract issued by the Secretary <strong>of</strong> Statefor Health. This issue may be freely reproduced for the purposes <strong>of</strong> private research <strong>and</strong> study <strong>and</strong> extracts (or indeed, the full report) may be included in pr<strong>of</strong>essional journalsprovided that suitable acknowledgement is made <strong>and</strong> the reproduction is not associated with any form <strong>of</strong> advertising. Applications for commercial reproduction should beaddressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials <strong>and</strong> Studies Coordinating Centre, Alpha House, University <strong>of</strong> Southampton SciencePark, Southampton SO16 7NS, UK.13

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!