Publication bias, meta analysis

In contrast, the lower part of Figure 15.2 shows a funnel plot for trials evaluating streptokinase and the meta-analysis, which combined these trials. The pattern of the individual trial results supports the absence of publication bias and the two large trials shown in the plot, ISIS-2 and GISS-1 show treatment effects entirely in line with the earlier meta-analysis. 

From a clinical practice perspective, a clinician s treatment of a patient may be influenced by the results of a meta-analysis. Therefore, if the result is influenced by the fact that the articles included in the analysis were not truly representative of all evaluations of the treatment, the result is not likely to be representative either. The issue of publication bias therefore is of critical importance in the context of evidence-based medicine. 

In summary, different SNPs and haplotypes in the G72/G30 region showed BD and/or SZ association in different studies. Strong between-study heterogeneities were observed for almost all the five SNPs that were analyzed in the meta-analysis. All these data suggest potential strong allelic heterogeneity at the G72/G30 locus. As usual, the conclusions from meta-analyses must be read with caution because of possible publication bias toward positive reports. 

The third major category of bias, Publication bias, applies to the evaluation of a putative risk factors across studies that are reported in the published literature. Positive findings are probably published more often than negative studies, and thus a meta analysis or review of the literature would be biased toward finding a greater effect than the true effect. 

Another commentary (136) emphasized the potential bias of observational studies and also publication bias in meta-analysis. The contemporary relevance of the findings is further reduced, since many of the studies included patients taking very high doses of thiazides. It is difficult to disentangle a drug-related effect from the association between hypertension and renal cell carcinoma. 

Numerous risk factors have been identified for infection with vancomycin-resistant enterococci. In one study the severity of mucositis in cancer patients was significantly associated with vancomycin-resistant enterococci (100). Previous vancomycin therapy was also believed to be a risk factor. However, a meta-analysis concluded that the reported strong association between vancomycin treatment and hospital-acquired vancomycin-resistant enterococci results from selection bias, confounding by duration of hospitalization, and publication bias (101). 

Chalmers has clearly restated the principles on which meta-analysis is based and discussed some of the empirical evidence that enables the validity of the technique to be considered. In particular, meta-analysis uses objective statistical procedures, specifies how the choice of studies to include was made, and describes the results of the studies on which the analysis was based. Sometimes, studies are designed with the specific intention of being combinable with each other, a strategy that can be referred to as a prospective meta-analysis. This is sometimes the case in drug development. However, when a meta-analysis is retrospective , that is a search has to be carried out to identify studies which address a particular question, the problem of publication bias has to be considered, in that studies with positive results may be more likely to be submitted and accepted for publication. Hence, meta-analyses based solely on published work may be biased in favour of finding an effect. 

The protocol should discuss what information was available prior to designing the meta-analysis and what information motivated the research objectives of the meta-analysis. The reporting should clearly state which trials were conducted and which trial results were known by study investigators prior to the design of the meta-analysis. The protocol should state the trial and patient inclusion, including discussion on any possible publication bias. The protocol should state the sources of the trial and patient data. 

Stewart L, Tierney J, Burdett S (2005) Do systematic reviews based on individual patient data offer a means of circumventing biases associated with trial publications In Rothstein HR, Sutton AJ, Borenstein M (eds) Publication bias in meta-analysis prevention, assessment and adjustments. Wiley, Chichester... 

Systematic review The application of strategies that limit bias in the assembly, critical appraisal and synthesis of all relevant studies on a specific topic. Systematic reviews focus on peer-reviewed publications about a specific health problem and use rigorous, standardized methods to select and assess articles. A systematic review differs from a meta-analysis in not including a quantitative summary of the results (Porta 2008). 

---