Search methods
The aim of the qualitative evidence synthesis was to examine people's attitudes towards the taking of agents or supplements that may be used in the primary prevention of colorectal cancer, i.e. NSAIDs (including aspirin), vitamins, minerals, folic acid or folate, selenium, calcium and dietary supplements generally. The synthesis included studies that focused on exploring the views, beliefs or attitudes of people who took any of these agents for any purpose. A systematic search to identify relevant studies was performed by an information specialist following piloting of appropriate search strategies. The search combined terms describing the agents of interest (NSAIDs, aspirin, vitamins, etc.) with a published, validated filter for identifying qualitative studies, together with the medical subject heading "qualitative research"[11]. The full search strategy is available in the Appendix. Databases searched for published and unpublished material included MEDLINE, PreMEDLINE, CINAHL, EMBASE, AMED, ASSIA, IBSS, PsycINFO, Science Citation Index, and Social Science Citation Index, and the HMIC and King's Fund databases. Studies were limited to those in English published from 2003 onwards to capture contemporary views and attitudes. Searches were undertaken in June 2008. Given the problems with identifying social science or qualitative literature through systematic searching of electronic databases alone[12, 13], the reference lists of all included studies were checked for additional literature, and a "berry-picking approach" utilising supplementary, non-systematic searching[14] testing various combinations of terms was also performed by two of the authors (AB, KC). This iterative, pragmatic approach to searching aimed to identify a set of studies providing relevant information on views and attitudes towards the taking of potential chemopreventive agents.
Study selection
To be included in the review, a study had to focus on exploring the attitudes, perceptions and beliefs of adults (any country) surrounding the taking of the agents listed above, through qualitative data from interviews or focus groups, and cross-sectional data from satisfaction surveys, i.e. unstructured and structured, but often textual data describing people's own, personal, subjective experiences, views or attitudes relating to the intervention of interest. Previous reviews have also adopted this inclusive approach to "views" studies, i.e. including qualitative data describing people's attitudes and beliefs from satisfaction surveys as well as more traditional qualitative study designs[6, 15]. The authors each screened a third of the citations for relevance (based on the inclusion criteria) and references for potential inclusion were discussed within the team. Disagreements or uncertain inclusions were resolved by discussion or by retrieval of the full paper to make a definitive judgment. Full papers of all potentially relevant citations were screened using the same process. Data from the included studies were extracted by two of the authors (CC, KC) using a review-specific form developed following piloting on one included paper.
"Best fit" approach to framework synthesis
The authors chose the framework synthesis approach because a published model was identified from the literature that conceptualised attitudes of adult women to the taking of vitamins and minerals[16]. The approach therefore was augmentative and deductive (building on this existing model or framework), rather than grounded or inductive (starting with a completely blank sheet). The model identified did not entirely match the topic under study, but it was a "best-fit" and provided a relevant pre-existing framework and themes against which to map and code the data from the studies identified for this review. A list of themes was derived from this model (see Figure 1) and provided the a priori framework of themes against which to code the data extracted from the included studies.
Data for analysis consisted either of verbatim quotations from study participants or findings reported by authors that were clearly supported by study data, for example, 'four of the five interviewees reported that the views of family and friends affected their decision-making' or '75% of respondents said that they were concerned about side effects of NSAIDs'. These data were extracted from the "Results" sections of included studies only, as it was felt that the Discussion and Conclusion sections would not present any new data, only additional interpretation or contextualisation of a study's 'findings'. Two of the authors (CC, KC) each extracted data from half of the included studies. Where any relevant data from the included studies did not translate into any pre-existing themes, a method was required to capture these data for the analysis. The published descriptions of framework synthesis do not specify a particular method for this, so the authors applied secondary thematic analysis, an interpretive, inductive approach grounded in the data based on methods from primary research, whereby additional themes were created as needed based on the study data[17]. In this way, the existing model acted as the basis for the synthesis and could be built-upon, expanded upon, reduced or added to by these new data. Each reviewer checked and examined critically the extraction and categorisation or coding of data performed by the other. The principal aim of this process was to examine the first reviewer's categorisation of the data, i.e. either to verify the coding or to challenge it by offering an alternative.
The authors then discussed the data and resulting themes, both those from the pre-existing model and those generated by the novel, inductive thematic analysis of the extracted study data. A consensus was reached on which a priori themes were supported by the data, and whether new themes identified by the reviewers did actually map either to a pre-existing theme or to one another (c.p. reciprocal translation[2]). The result was a finalised list of themes. The primary reviewer (CC) then offered an interpretation of the relationships between the themes based in part on the relationships as they were represented in the original model (see Figure 1), and also based on the data itself, which suggested, for example, that "the media" inputted into the central procedural themes of both perceived need and decision-making. The new model was then critically considered by all reviewers. A revised conceptual model was therefore developed building on the earlier, identified model, to describe and explain people's views around the taking of potential chemopreventive agents.
Consideration of study quality
Published descriptions of framework synthesis typically exclude studies of lower quality. However this was not the approach used in this case, representing a further innovative deviation from the published method[2]. All studies that satisfied the relevance criteria were included because there is an increasingly strong case for not excluding qualitative data studies from evidence synthesis based on quality assessment[1, 18, 19]. Studies were assessed using key quality criteria derived from relevant critical appraisal checklists for qualitative studies[20] and other systematic reviews of people's views[1, 2]. These elements also appear in recent guidance from the Cochrane Qualitative Research Methods Group[21]. The assessment consisted of querying whether the following are clearly and adequately described in the publication: the question and study design; how the participants were recruited or selected; and the methods of data collection and analysis used (See Additional file 1). The "better-reported" studies provided details on two or more criteria, whereas the "inadequately-reported" studies clearly described no more than one. The decision only to focus on these four elements, and what was reported or clearly described by the included studies, was taken for two reasons. Firstly, these elements of the study were potentially more easily judged and apprehended than others, as they were either described or not. Secondly, it has been pointed-out previously that any appraisal checklist is only assessing what has been reported in a publication[22]. The focus therefore was on the reporting of basic methods and not potentially subjective judgements regarding studies' validity or reliability[18].
While it is acknowledged that there is always uncertainty concerning how well or poorly a study has been conducted, if authors clearly describe their approach and sampling, and data collection and analysis methods, then this potentially lends greater robustness to the study's findings. This is because any inherent "risk of bias" may be better determined than if this information was absent, regardless of the study's findings. This does not preclude the possibility that an "inadequately-reported" study has actually been well-conducted, but it does form a reasonable basis for making a quality assessment. This relatively small number of easily-defined criteria can also be seen to apply to qualitative studies universally and may be more practical than checklists with much larger numbers of questions, especially as these have been found to generate low inter-rater reliability scores among otherwise experienced qualitative systematic reviewers[18]. This was one of the first practical attempts to utilise assessment criteria based specifically and exclusively on the description or reporting of a study's method and sampling strategies, and methods of data collection and analysis. No study was excluded on the basis of the adequacy of its reported processes, but the assessment aimed to explore quality of reporting as an explanation for differences in the results of otherwise similar studies, and to consider its impact on the internal validity of the review[23]. A sensitivity analysis would be performed in the event of the inclusion of "inadequately-reported" studies.