Skip to main content

Languages

An approach to deal with the absence of cohort studies in prognosis systematic reviews: a meta-analysis of cross-sectional studies

Date and Location

Session: 

P4.019

Date

Monday 23 September 2013 - 10:30 - 12:00

Location

Presenting author and contact person

Presenting author

Regina El Dib

Contact person

Regina El Dib
Abstract text
Background: Systematic reviews aim to provide an exhaustive summary of literature relevant to a certain topic through the organization of similar primary studies and, when possible, summarize them in a meta-analysis. This means that systematic reviews should offer clinicians, health professionals and policy makers with level 1 of evidence for health care decision making. However, it does not often happen as we frequently note an absence of high-quality primary studies in the health field. There are a large number of studies with lower level of evidence in the literature provided by non-randomized studies and/or single case studies that could be used to provide an alternative interpretation of the available evidence in the literature while waiting for the results from high level of evidence. Objectives: To describe a statistical analysis of evaluating cross-sectional studies in prognosis health care reviews when there is absence of cohort studies. Methods: We provided illustrations from recent experience related to the clinical question “Is the immunoexpression of p53 a surrogate marker for laryngeal cancer lesions?” and, discussed the impact of the level of evidence in the clinical practice. Studies were included if the participants of interesting have received a diagnosis of cancer laryngeal lesion at baseline. The control group was patients who have received a diagnosis of benign laryngeal lesions (i.e., polyps, nodules, cysts) or normal mucosae. This review example planned to include cohort studies as they have an observational and prospective design features with the aim to verify the immunoexpression of p53 as a surrogate marker in recidivism laryngeal cancer lesions compared to the p53 positive rates in laryngeal benign lesions. We also considered to include all available cross-sectional studies as an anticipation of not finding many cohort studies in this area. For the cross-sectional studies the occurrence of positive p53 immunohistochemical expression rates were treated as dichotomous variable with their respectively 95% confidential intervals (CIs) and, the statistically significant difference between both interventions studied was defined if their combined 95% CIs did not overlap. The software used to plot the cross-sectional studies into a meta-analysis was StatsDirect. Results: 35 studies met the methodological requirements proposed in this review: only two cohort studies with insufficient data to plot in a conventional meta-analysis and, 33 cross-sectional studies. As only two studies from level I of evidence for prognosis purposes met the inclusion criteria of this systematic review, we decided to provide a different overview through a proportional meta-analysis of cross-sectional studies. Figures 1 and 2 show a proportional meta-analysis of cross-sectional studies regarding the occurrence of positive p53 immunohistochemical expression in benign and cancer laryngeal lesions. There was a statistically significant difference favoring cancer laryngeal lesions compared to benign laryngeal lesions on positive p53 immunohistochemical expression rates (31,9% vs. 62%) as their CIs did not overlap. Conclusions: We described a statistical analysis to evaluate cross-sectional studies in health care prognosis reviews. In our example, there is currently some evidence suggesting a correlation between positive p53 immunohistochemical expression and the occurrence of cancer laryngeal lesions higher than in benign laryngeal lesions. This method is extended to be used in the absence of cohort studies. The use of this method leads to substantial gains in the scientific community as it provides health professionals with the available information for their clinical practice until higher-quality primary studies are conducted, although we cannot ruled out the possibility of clinical heterogeneity and methodological failures due to the lower-quality level of evidence from these studies.