To the Editor:
Meta-analyses and pooling studies are valuable tools for generating evidence. They typically have enough power to detect associations that are not detectable by individual studies with low sample size and their large size allows for a fine statistical adjustment of the results. Furthermore, they enhance external validity. Pooling studies are becoming more popular, through such initiatives as the Genetic and Susceptibility to Environmental Carcinogens (GSEC) or Human Genome Epidemiology (HuGE) reviews.1–3 While authors have methodologic guidance on performing meta-analysis and systematic reviews (Moose, Quorom),4,5 there are no similar requirements regarding which observational studies should be included in these meta-analyses.
Observational studies may be excluded from meta-analyses because of lack of definition of inclusion and exclusion criteria, lack of information on the adjustment variables and how they were used (eg, continuous or dichotomous), and differences in the selection of the reference category used in calculating the effect measure. Alternately, the meta-analyst can contact the researcher to obtain the necessary information. Exclusion is easier; requests for further information is often frustrated by nonresponse or delay. However, exclusion increases the probability of biased results in the final meta-analysis.
Author’s reports on observational studies should anticipate their inclusion in meta-analyses or pooling studies, and, provide the information necessary for this task. Such reports could be published with a label indicating that they fulfill the requirements for inclusion in a meta-analysis or in a pooling study. These requirements would include the availability of authors to perform new analysis of the data if requested for a meta-analysis and to share their original data in the context of a pooling study. Papers meeting this condition would be labeled on the first page as “good for meta-analysis or pooling.”
As a starting point, we propose these minimum criteria for observational studies to be included in meta-analyses. 1) Present the number of subjects included in the final statistical analysis, categorized by exposure status for cohort studies and disease status for case-control studies. 2) Provide details on how the independent and dependent variables have been included in the analysis, indicating the reference categories for both. 3) Explain what variables were used for statistical adjustment and on which scale. 4) If subgroup analysis is performed, the same descriptive requirements should apply as for the general analysis. 5) State that the authors are willing to share their data or perform additional analysis if requested for a meta-analysis.
This proposal extends the recommendations of the Quorom, Moose, and STROBE6 statements by highlighting with utility of a published observational study for further analysis. This proposal does not rule out the use of quality criteria to be determined by the meta-analyst. A final list of criteria should be developed by consensus among authors, editors, and other professionals involved in scientific publishing.
Juan M. Barros-Dios
Department of Preventive Medicine and Public Health
University of Santiago de Compostela
Santiago de Compostela, Spain
1. Taioli E. International collaborative study on genetic susceptibility to environmental carcinogens. Cancer Epidemiol Biomarker Prev
2. Khoury MJ, Little J. Human genome epidemiologic reviews: the beginning of something HuGE. Am J Epidemiol
3. Raimondi S, Paracchini V, Autrup H, et al. Meta- and pooled analysis of GSTT1 and lung cancer: a HuGE-GSEC review. Am J Epidemiol
4. Stroup DF, Berlin JA, Morton SC, et al, for the Meta-analysis of observational studies in epidemiology (MOOSE) Group. Meta-analysis of observational studies in epidemiology: a proposal for reporting. JAMA
5. Moher D, Cook DJ, Eastwood S, et al. for the QUOROM Group. Improving the quality of reports of meta-analyses of randomised controlled trials. Lancet
6. von Elm E, Altman DG, Egger M, et al, for the STROBE initiative. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Epidemiology