Incorporating data from atypical experimental study designs in systematic reviews: experiences from Cochrane Infectious Diseases

Session Type
Oral presentation
Non-randomised studies for interventions
Chaplin M1, Lissenden N1, Choi L2, Pryce J3, Gleave K1, Fox T1, Malone D4, Villanueva G5
1Liverpool School of Tropical Medicine, UK
2Cochrane, UK
3Bradford Teaching Hospitals NHS Foundation Trust, UK
4Bill & Melinda Gates Foundation, UK
5Cochrane Response, UK

Background: For ethical, logistical and cost-related reasons, studies examining interventions to prevent occurrence and transmission of infectious diseases often use atypical experimental designs. Analysis methods for these study designs may be established, but inclusion of data from these studies in systematic reviews often raises unique challenges, which methodological experts at Cochrane Infectious Diseases must overcome by determining appropriate analysis approaches.
Objectives: To present examples of atypical experimental study designs and outline how Cochrane Infectious Diseases authors have included data from these study designs in systematic reviews.
Methods: We qualitatively reviewed examples of the inclusion of atypical experimental study designs in Cochrane Infectious Diseases reviews in recent years. We extracted information on the key features of these study designs and the analysis approach used by review authors to synthesise data from these studies. A descriptive account of the main issues that were encountered, and lessons learnt are presented.
Results: We identified three key examples of inclusion of data from atypical experimental study designs in Cochrane Infectious Diseases reviews; namely, the inclusion of: 1) interrupted time series studies in a review of space spraying for preventing malaria transmission; 2) experimental hut studies in a review of piperonyl butoxide combined with pyrethroids in insecticideÔÇÉtreated nets to prevent malaria; 3) a cluster-randomised, test-negative study in a review of Wolbachia-carrying Aedes mosquitoes for preventing dengue infection. Review authors used various approaches to synthesise data from these studies.
Conclusions: Even when studies with atypical experimental designs have been carefully planned, conducted and analysed, it is not always the case that study authors report effect estimates and standard errors that can be entered directly into RevMan. Review authors must occasionally calculate effect estimates that take special features of the data (for example, clustering or seasonality) into consideration. Where effect estimates and standard errors are reported by the study authors, literature presenting the development of the atypical experimental study design and accompanying analysis methods should be reviewed carefully. Collaboration between Cochrane statisticians is key to ensuring robustness of methods and results reported from reviews that include studies with atypical experimental designs.
Patient, public and/or healthcare consumer involvement: None