Article type
Year
Abstract
Background: The emphasis of the Collaboration has been mainly on reviews of interventions. We suggest that it would be useful to have Cochrane-style reviews of factors which cause, or are alleged to cause, diseases. This would help provide a basis for primary research, and could also be used by clinicians, patients, parents and any others who want to know whether a specific factor is or is not linked to a disease. The methodologies would need to be developed.
Objectives: To explore the problems of doing systematic reviews in epidemiology, using the incidence of type 1 diabetes as the example. There have been reports that the incidence of childhood diabetes is higher in rural areas. Other factors include maternal age, birth by Caesarean section, and nitrates in drinking water. To assess by discussion at the Colloquium, the level of interest in epidemiological reviews. Aims were two-fold - to do some reviews; and to examine methodological problems.
Methods: Four systematic reviews. Quality assessment is three-fold of numerator, denominator and exposure using grades A, B, or C.
Results: A systematic review of urban/rural associations with incidence found 30 studies, but only 8 were scored as having AAA standards. These showed a rural excess in countries with high incidence, but not in countries with low incidence. The review of Caesarean section showed a 'positive-first' bias which we think may be a common phenomenon - until one study has shown a positive association, studies showing no link are either not submitted or not published.
The methods require different search strategies; different QA methods; there are different hierarchies of study design. The literature searches are more complex than for RCTs of interventions, and reflect Bradford Hill's criteria for causation. Studies show heterogeneity of design; of methods of measuring exposure; of different lengths and intensities of exposure.
Conclusions: We think it would be of value if there were Cochrane review in epidemiology, and invite any people interested in this tpoic to discuss it with us at the poster sessions.
Objectives: To explore the problems of doing systematic reviews in epidemiology, using the incidence of type 1 diabetes as the example. There have been reports that the incidence of childhood diabetes is higher in rural areas. Other factors include maternal age, birth by Caesarean section, and nitrates in drinking water. To assess by discussion at the Colloquium, the level of interest in epidemiological reviews. Aims were two-fold - to do some reviews; and to examine methodological problems.
Methods: Four systematic reviews. Quality assessment is three-fold of numerator, denominator and exposure using grades A, B, or C.
Results: A systematic review of urban/rural associations with incidence found 30 studies, but only 8 were scored as having AAA standards. These showed a rural excess in countries with high incidence, but not in countries with low incidence. The review of Caesarean section showed a 'positive-first' bias which we think may be a common phenomenon - until one study has shown a positive association, studies showing no link are either not submitted or not published.
The methods require different search strategies; different QA methods; there are different hierarchies of study design. The literature searches are more complex than for RCTs of interventions, and reflect Bradford Hill's criteria for causation. Studies show heterogeneity of design; of methods of measuring exposure; of different lengths and intensities of exposure.
Conclusions: We think it would be of value if there were Cochrane review in epidemiology, and invite any people interested in this tpoic to discuss it with us at the poster sessions.