13.07.2015 Views

Evaluating non-randomised intervention studies - NIHR Health ...

Evaluating non-randomised intervention studies - NIHR Health ...

Evaluating non-randomised intervention studies - NIHR Health ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

162Review Method of incorporating Results of quality investigationquality into synthesisMayo-Smith, 1997 241 Not consideredMaziak, 1998 192 Qualitative Quality issues were discussed alongside the results of each study. A lack of rigorously controlled trials wasidentified and the reviewed evidence provides insufficient information to permit a reasonable conclusion to bemadeMcAweeney, 1997 325 Not consideredMcCusker, 1998 328 Quantitative Results stated to be similar for quality-adjusted and unadjusted effect measures. Only quality-adjusted resultsQuality weightingpresentedMelchart, 1994 93 Qualitative Methodological issues and results discussed. Reviewer’s estimate of strength of evidence provided by each studywas presented. Inconclusive or limited evidence of no efficacy was provided by 21 comparisons and limited orgood evidence of efficacy was provided by 10 comparisonsMiller, 1995 87 Quantitative No significant relation between these codes and methodological quality score was found (r = 0.06). QualityCorrelation analysis of quality score and score was also found to be unrelated to mean problem severity and to estimated cost of delivering the‘outcome logic scores’ (measure of the treatmentstrength of support for treatment efficacy)Morin, 1994 330 Quantitative Design quality was negatively correlated with ES for sleep onset latency (r = –0.24, df = 91, p < 0.03) only,Correlations between outcome and quality indicating that lower quality <strong>studies</strong> produced larger ESsof study design were calculatedMullen, 1985 275 Quantitative No impact on ES foundRegression analysis to examine effect ofstudy design and type of comparison groupMullen, 1992 274 Qualitative Some narrative discussion of methodological characteristics. Pre-/post-test <strong>studies</strong> were excluded from the metaanalysisowing to methodological limitations. Further implications of methodological flaws not discussed in any detailMulrow, 1986 197 Qualitative Some quality issues were discussed alongside study results. Authors conclude that methodological limitations donot allow many definitive conclusions but that some new and useful information was obtainedMurtagh, 1995 332 Quantitative Most effect were <strong>non</strong>-significant – only the source of participants, <strong>non</strong>-use of drugs and placebo treatmentDesign variables used as inputs into a contributed significantly to the prediction of the dependent variableregression to quantify impact on effect sizes(including source of sample, study design, useof objective validation)Naylor, 1987 334 Quantitative When all <strong>studies</strong> were combined (3 RCTs and 1 NRS), quality weighting abolished the significant treatment effectSensitivity analysis conducted using weighting for both overall survival from acute episode and overall survival to discharge from hospital. (Inclusion of the NRSby quality score. Weighted and unweighted led to larger treatment effects). Authors conclude that despite the (unweighted) significant findings, there isp-value for the difference between groups insufficient evidence for the <strong>intervention</strong>was providedAppendix 7continued

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

Saved successfully!

Ooh no, something went wrong!