Discrepancy between results and abstract conclusions in industry – versus non-industry-funded studies comparing topical prostaglandins
Methodological quality of the review: Low confidence
Author: Alasbali T, Smith M, Geffen N, Trope GE, Flanagan JG, Jin Y, Buys YM
Geographical coverage: Not reported
Sector: Efficacy of topical prostaglandins on glaucoma treatment.
Sub-sector: As above
Equity focus: None specified
Review type: Other review
Quantitative synthesis method: Narrative synthesis
Qualitative synthesis method: Not applicable
Background: There has been a dramatic increase in the funding of medical research by drug companies over the last two decades. This may result in industry bias where the source of funding for clinical trials either affects the results in a systematic way, or leads to selective presentation of the results.
It was found that there is an association between industry funding and pro-industry results and publication bias, which can affect the interpretation and presentation of outcomes. This leads to conclusions that overstate results without statistical support. Therefore, it is important to assess the impact of the source of funding on the interpretation of results, and the conclusions of studies funded by different sources.
Objectives: To investigate the relationship between industry versus non-industry-funded publications comparing the efficacy of topical prostaglandin analogues by evaluating the correspondence between the statistical significance of the publication’s main outcome measure and its abstract conclusions.
Main findings: The authors included 39 publications in this review, of which 29 were industry-funded and 10 were non-industry funded. In 62% of the industry-funded studies, the published abstract conclusion was not consistent with the results of the main outcome measure, while it was 100% consistent in the non-industry-funded studies. Also, it was found that 90% of the industry-funded studies had pro-industry abstract conclusions. The authors recommended that both readers and reviewers should scrutinise publications carefully to ensure that data supports the authors’ conclusions.
Methodology: Authors searched MEDLINE from 1966 to the 2nd week of November 2007 using combinations of the keywords latanoprost, travoprost and bimatoprost. English language publications comparing the intra-ocular pressure (IOP)-lowering efficacy of any combination of latanoprost, travoprost or bimatoprost were included in the initial search.
Reference lists from initially identified articles were searched to identify relevant publications. Each publication was reviewed by three independent reviewers using a standardised data collection sheet evaluating source of funding, industry author, study quality, main outcome measure, statistical significance of main outcome measure, abstract conclusion, correspondence between statistical significance of main outcome measure and abstract conclusion, total number of IOP outcomes compared and journal impact factor.
A standard checklist was used to assess the quality of identified papers. The main outcome measure was the correspondence between the statistical significance of the publication’s main outcome measure and its published abstract conclusion. Fisher exact test and the Student T test were used for the statistical analysis of categorical data and continuous data, respectively.
Applicability/external validity: No methods were used either to assess the applicability /external validity of the results or to discuss how generalizable the results are
Geographic focus: Not specified.
Summary of quality assessment: Authors conducted a systematic review of literature to identify relevant articles. The criteria for assessing the quality of selected studies were clear. However, this review had some important limitations. The authors did not avoid publication and language bias in the search, but only one relevant database was searched and only articles written in English were included in the review. Authors did not report methods used to select studies and did not discuss the risk of bias of included studies. Therefore, there is low confidence in the conclusions about the effects of this study.
