In a meta-ethnography, a popular way of analysing data is the translation of the concepts or metaphors of one study into another, while preserving the structure of relationships between concepts within any given study . We will thus show how we sampled different studies and how this influenced the translation exercise based on an example of three example concepts from three articles included in our review. Note that the decision to work with three concepts only was taken to increase the clarity of the procedures we describe in this paper, not to describe all the actual results and complete line-of-argument.
First step: Arriving at a “jumping off point” through intensity sampling
We will illustrate these decisions of intensity sampling by describing the inclusion of 3 articles [34–36] which - according to our parameters described above - have a great degree of overlap with the research goal, a high methodological quality and strong conceptual clarity.
On the articles that were included through intensity sampling, we performed a reciprocal translation of the concepts, which is the translation of one study’s findings into another, using metaphors and overarching concepts.  In what follows, we give a worked example of how we did this reciprocal translation for 3 concepts identified in the initial set of studies considered for the synthesis, as this is a necessary step towards the illustration of the subsequent sampling methodology. In order to be explicit about how the concepts compared to one another, we created a table into which we placed and compared the concepts of each paper (See Table 3). Each row of the table represents a key concept. In the left collumn, we labelled the rows with concepts that encompassed all the relevant concepts from each paper.
The first concept we retrieved through intensity sampling is “sexual struggling”, encompassing the different ways of struggling with the sexual changes due to cancer. In Walker’s study (2011) it is formulated as having a sense of loss . In the study of Gilbert (2013), this is formulated as patients having an altered body image . In Juraskova’s study (2013) it is formulated as “reduced vaginal lubrification” .
Another overarching concept that we retrieved was “exacerbation of struggling”, encompassing strategies, situations, characteristics that were leading to an increasing struggling with the sexual changes. In Gilbert’s study (2013), this is formulated as “sticking to the coital imperative”, which means that intercourse is the most normal and natural form of heterosexuality, and condemns those who cannot perform as dysfunctional. In Walker’s study (2012), this is formulated as avoidance of communication about the sexual changes. In Juraskova (2003), exacerbation of struggling is the case when the patients are “ Receiving radiotherapy combined with external radiation and brachytherapy”.
A third overarching concept we found was the “sexual adjustment” to changes due to having cancer, encompassing the different ways of adaptation to sexual changes. Gilbert’s study (2010) describes that there is “a renegociation of the practices of sexual intimacy”, which means that the couple included sexual practices that had previously been marginalized in relation to sexual intercourse. Walker (2011) formulates this adjustment as “accepting the decision to stop sexuality”. Juraskove (2003) formulates it as “sexual adjustment and quality of life”.
The articles were sampled by the main author, but all articles included by intensity sampling were read and analysed by two authors (CB and MS). After a certain point which we call the “jumping off point”, we began to discover certain key dimensions of variation between the studies, which we explored further through maximum variation sampling. In the worked example that we explain here was the discovery that the studies varied on the scientific approach they took on, resulting in a different interpretation of the overarching concepts. To illustrate this: Gilbert (2010) used a social-constructionist lens to investigate sexual adjustment, Walker (2011) used a more psychological approach to investigate the subject, and Juraskova (2003) underscores more the biological aspects of sexual changes after cancer. Through the maximum variation sampling, we thus want to further explore how these different approaches lead to different interpretations of the phenomenon.
Second step: Apply a maximum variation sampling strategy to construct a preliminary line of argument
To explore the consequence of variation on the key dimension, we used maximum variation sampling to include studies that varied on the above cited dimension (i.e. scientific approach, socio-, psycho, or biological perspective). In this worked example, we show through the inclusion of three more papers [37–39] how we arrived – through comparison of the papers- at a preliminary line of argument.
The sampling was also done by one researcher, but the articles were read and analysed by 2 researchers. As a result of this maximum variation sampling and constant comparison between the papers, could develop relationships between the different concepts and constructing a preliminary line of argument (see Table 4).
First, with regard to the concept of struggling, we found that articles who work with a psychological approach, describe the concept of struggling on an emotional level, analog with the stages of grief (anger, depression,..) while the sociological articles describe it more on a level of identity, analog with the theory of biographical disruption. Articles who have a more biological approach reduce the struggling on a level of sexual dysfunction.
Second, with regard to the concept of exacerbation of struggling, articles who work with a psychological approach again describe a stage of the grief theory, which is denial. Sociological oriented articles work with the adherence to hegemonic discourses, and biological oriented articles use certain characteristics of the cancer treatment as barriers towards adjustment.
Third, with regard to the concept of sexual adjustment, articles who are psychological oriented again use a stage of the grief theory to encompass this adjustment, which is acceptance. Sociological oriented article worked with a “rediscovery” of what sexuality is. The changes are thus not merely accepted, rather they are incorporated in a new definition of the self and sexuality. Biological oriented articles worked with “sexual recovery”, which –in contrast to the sociological oriented articles- means that there is no difference in what sexuality means , but a reuptake of sexual activity , similar to what it was before the cancer.
Our preliminary line of argument consisted of three different pathways the articles worked with. First, there are articles following the grief theory to describe the adjustment process In this case, sexual changes are depicted in terms of losses, and the adjustment occurs through the process of grief and mourning.
Second, there are articles following the “restructuring theory” during illness. Unlike the case of grief theory, where the patient and partner are working through some emotional stages, in the restructuring pathway patient and partner are more cognitively dealing with sexuality after cancer through the development of a new sexual paradigm. Flexibility is the central aspect of this adjustment.
Thirdly, there are articles following the pathway of sexual rehabilitation. This pathway is embedded in a more positivistic paradigm where the adaptation does not emphasize psychological changes or cognitive restructuring, but sexual changes as a bodily dysfunction that needs treatment and behavioural strategies.
Refining the preliminary line of argument by means of disconfirming case sample.
To test, refine, and deepening our preliminary line-of-argument, , we included 3 articles out of the pool of 58 articles that consist of a theory and concepts opposing the preliminary line-of argument. We will give an example with including 1 article (see Table 5).
In this phase of sampling, we worked together with a researcher who was not involved in the analysis process before (JB). This is because we wanted to have a fresh and “unambiguous view” of our line of argument. This researcher, together with the first researcher, read the articles and tested them against the line of argument.
In our preliminary line-of-argument, we assumed that the three pathways of adjustment all followed a linear pattern from the struggling towards the adjustment. However, Ramirez (2009) counter argues this linear approach by stating that patients could refine their definition of sexuality, but could also return to it at a certain moment . These disconfirming findings led us to re-analyse the included articles, where we came eventually to the conclusion that the sexual adjustment as a cognitive restructuring process does not have a linear pattern with an endpoint, but rather makes on oscillating movement between following hegemonic definitions of sexuality, and challenging them.
Challenges and opportunities
In the process of conducting a qualitative evidence synthesis through purposeful sampling, we encountered several challenges. But this process also created a few opportunities that would not have occurred if we had used an exhaustive sampling and analysis strategy. In what follows, we discuss how we have bridged obstacles and maximized benefits in terms of the opportunities arising.
First, it proved to be difficult to define what exactly to look for, since the concept of e.g. an intensity sample on the meta-level could not readily been borrowed from the logic applied in basic research projects. In an original research project, as opposed to a qualitative evidence synthesis project, purposeful sampling can often easily be conducted, for example by using a brief questionnaire as a screening tool to search for participants with specific characteristics . However, with research reports, this is more difficult in practice. We chose to search for literature by means of electronic databases with the use of search strings. Finding a specific search string to detect a specific information-rich research report which meets the sampling criteria would be difficult, because the search terms are usually based on population and setting characteristics as well as the topic of interest, rather than on conceptual or theoretically interesting leads.
Therefore we decided to conduct a scoping of the literature prior to applying a purposeful sampling technique. The scoping review was intended to create a pool/or archive of primary research reports that are easily accessible and can be used later as material for purposeful sampling. In fact, our purposeful sampling strategy did not start at the level of data-collection. It was initiated at the level of data extraction and analysis. The consequence of this decision was that the sampling procedure was rather labour-intensive as we had to perform a scoping review before the actual mixed purposeful sampling could start.
We illustrated through our worked example that using purposeful sampling techniques also has several advantages.
First of all, although some researchers argued that reducing the number of included articles by means of purposeful sampling could result in neglecting important data [18, 42], we showed throughout this worked example that the opposite can be true. With the use of this combination of three purposeful sampling techniques – intensity sampling, maximum variation sampling and confirming/disconfirming case sampling - we arrived at a line-of-argument.
Because of this emphasis on conceptual robustness instead of generalization of the data, we were more sensitive to “deviant data”, i.e. data that may not have been picked up when synthesizing information from an exhaustive sample of the literature, because review authors are generally more focused on detecting commonalities between articles. When using an exhaustive sampling technique, researchers will arrive at results that describe the “greatest common devisor” of all included papers.
Furthermore, deviant data that has been derived through maximum variation sampling and confirming/disconfirming case sampling may add new perspectives or a new space of understanding to the line-of-argument, while sampling randomly may run the risk of preventing enhanced insight and knowledge.
Moreover, the combination of sampling techniques – instead of a random sample or just one method of purposeful sampling- could enhance the quality and diversity of the papers being included, and could make the results more conceptually aligned with the synthesis purpose. This would further enhance the possible impact a qualitative evidence synthesis could have on informing healthcare practice .
Such an approach, however, demands a considerable amount of flexibility from review authors, mainly because inclusion criteria may change progressively during the process. This fact, together with the experience described above of doing a labour-intensive scope of the literature, goes against the argument of many authors  that using purposeful sampling provides a pragmatic solution or a short cut for reviewers who have limited time for searching and screening. However, we felt we did gain some time in the analytical process, since the number of articles from which data were extracted was modest in number. This strategy is therefore recommended for authors who are left with a high number of relevant articles after screening for inclusion.
However, the choice of using this particular combination of sampling techniques should also be motivated from a theoretical perspective. Authors who want to build a theoretical model out of the qualitative evidence synthesis could use this scheme of sampling methods, as it aligns well with the different stages of analysis, and is parallel to what Corbin and Strauss suggested for primary research .