Appraising published claims about drug treatment to implement best therapy in clinical practice

Abstract

The validity and applicability of publications about individual clinical studies and systematic overviews regarding interventions with drugs need to be established and perceived in quantitative terms to implement evidence-based, best current therapy. This requires an understanding of study design, various types of bias, intention to treat analysis, clinical versus statistical significance, and other considerations. The quantitative appreciation of drug effects may be facilitated by arranging results from case-control studies, cohort studies, and controlled trials in suitable contingency tables. Relative risks, relative risk reductions, odds ratios, and absolute risk reductions (in a given period of time), as well as corresponding numbers needing treatment (to prevent one event) may then be calculated. Systematic overviews of multiple clinical trials and assessment of their combined quantitative significance (meta-analyses) were developed to enhance statistical power, to enhance the level of confidence about small differences in effect, and to reconcile conflicting claims. The results of a meta-analysis are usually represented by so-called 'forest plots' of point estimates (corresponding to medians) and their respective confidence intervals, as well as a combined point estimate and confidence interval. Heterogeneity (important differences between findings from individual trials is a special problem incurred with this relatively new tool. The meta-analysis are also specially prone to other special sources of bias' a greater likelihood that trials reporting 'favourable' effects are published, covert duplicate inclusion of results from the same patients, and non-blinded meta-analysers