Presentation (CTI)

OBJECTIVES: Synthesizing observational data on the natural history of burden of illness (BOI) for rare and/or clinically heterogeneous diseases can be challenging, as there are often insufficient standardized data to permit formal meta-analysis. Evidence syntheses that lack formal meta-analysis are often described as being narrative; and without clearly-outlined methods, particularly for summarizing quantitative data. This review was conducted to identify synthesis and reporting guidelines, for topics where a formal meta-analysis would be inappropriate.
METHODS: A targeted review was conducted in Google Scholar and Pubmed to identify and summarize published guidelines. A checklist was developed to address current gaps in available guidelines with respect to synthesizing data from natural history and BOI studies.
RESULTS: One identified guideline described approaches for synthesizing treatment data when meta-analysis is not possible. While thorough, the guidance was limited to methods for synthesizing treatment effects from a relatively homogenous group of interventional studies. Key components outlined were 1) grouping studies for synthesis, 2) prioritizing results for synthesis, 3) handling heterogeneity and assessing certainty of findings, and 4) data presentation/reporting. Guidance for characterizing drivers of variability, or methods for synthesizing data from observational studies with heterogeneity in study design and populations was not available. Building from the existing guideline, a checklist was developed outlining an iterative approach for synthesizing natural history and BOI data. It includes: 1) conceptual model development to inform PECOS criteria, 2) initial review and subsequent grouping of data systematically identified, with prioritization for synthesis, 3) a reporting structure for characterizing drivers of variability in tabular and narrative forms, and 4) effective data visualization.
CONCLUSIONS: Synthesis methods that consider the wider evidence base, rather than estimating a central tendency, are valuable to understand drivers of heterogeneity in observational study data. The proposed checklist may be implemented for synthesizing natural history data, and characterizing components of disease burden.