An expert panel has proposed a checklist of 25 items that researchers should adhere to where reporting data from genetic risk studies.
A workshop of researchers, epidemiologists, geneticists, statisticians, and journal editors sponsored by the Human Genome Epidemiology Network developed a checklist of items to strengthen the reporting of Genetic Risk Prediction Studies, or GRIPS. The recommendations build on previously issued guidelines, and are not meant to support any one risk assessment method or study design.
"The 25 items of the GRIPS statement are intended to maximize the transparency, quality, and completeness of reporting on research methodology and findings in a particular study," said Cecile Janssens of Erasmus University and colleagues in the statement published in 10 journals last week. "It is important to emphasize that these recommendations are guidelines only for how to report research and do not prescribe how to perform genetic risk prediction studies."
Although there are already guidelines that address how to handle tumor markers in prognostic studies, how to report observational studies, and how to assess gene variants, none of these past guidelines are "fully suited to genetic risk prediction studies," the authors note. The panel characterized GRIPS as "an emerging field of investigation with specific methodological issues that need to be addressed, such as the handling of large numbers of genetic variants (from 10s to 10,000s) and flexibility in handling such large numbers in analyses."
The expert panel that authored the GRIPS statement met in December 2009 in Atlanta, Ga., to discuss and draft their recommendations. Over time, they settled on around two dozen items that genetic risk predictions studies — the type of analysis increasingly being factored into diagnostic, preventative, and therapeutic decisions — should include when reported in a journal in order to ensure that the findings from these studies are accurately interpreted, synthesized into other studies, and translated into medical care.
"There is ample evidence that prediction research often suffers from poor design and bias, and these may also have an impact on the results of the studies and on models of disease outcomes based on these studies," Janssens et al. wrote in the published statement. "Although most prognostic studies published to date claim significant results, very few translate to clinically useful applications."
The GRIPS statement comes during a time when health regulators are increasingly paying attention to how genetic risk data is being presented to the public. The need for more transparency and standardization in how direct-to-consumer genomics firms report genetic risk analysis was recently debated by an US Food and Drug Administration advisory committee (PGx Reporter 03/16/11).
Janssens told PGx Reporter that the panel's review of published genetic risk prediction papers showed that these reports often didn't include "details that are crucial to understand what exactly was investigated."
For example, it was common for studies to omit details about how the genetic risk score was developed, which makes it difficult for others to replicate these studies in independent populations, she noted.
"Another shortcoming of most genetic risk prediction studies is that they are investigated in populations that are not of clinical or public health relevance," Janssens said. "For example, we have investigated genetic prediction of type 2 diabetes in a well-characterized prospective study with almost 20 years of follow-up in which the average age of the participants was 69 years. An excellent study population, but not suitable for investigating prevention of diabetes, which you would like to prevent in much younger generations."
The GRIPS statement recommends that authors include some basic items in their publications, such as the words "genetic" and "risk" in the title, and that they state how the study was funded. Other recommendations include specifying the methods used to derive the genetic risk model and to validate it; detailing the features of the study population; identifying the limitations of the study; reporting when risk modeling has been adjusted or unadjusted to consider various factors (ie. genetic, environmental, familial, etc.); and discussing the healthcare relevance of the study results.
Like the multidisciplinary panel that developed these recommendations, the audience the panel is addressing is also broad. According to the authors, the 25-item checklist can be useful for anyone undertaking and reporting genetic risk prediction studies, as well as journal editors and reviewers who have to evaluate these studies for publication.
"To improve the transparency and completeness of reporting, we developed these guidelines," Janssens said. "We hope that indirectly they also make researchers aware of the important issues in genetic risk prediction research. For the latter, we hope and expect that they will have a broader impact than single risk prediction studies only. For example, the same items are relevant for GWAS studies that include the assessment of a genetic risk score."