In this study conducted among individuals in charge of biomedical graduate programs in Europe, we found that 47% of programs accepted SRs as research methodology that can partly or fully fulfil the criteria for a PhD thesis. However, most of the participants had negative attitudes about such a model for a PhD thesis, and most had insufficient knowledge about the basic aspects of SR methodology. These negative attitudes and lack of knowledge likely contribute to low acceptance of SRs as an acceptable study design to include in a PhD thesis.
A limitation of this study was that we relied on participants’ responses and not on assessments of formal rules of PhD programs. Due to a lack of familiarity with SRs, it is possible that the respondents gave incorrect answers. We believe that this might be the case since we received answers from different programs in the same university, where one person claimed that SRs were accepted in their program, and the other person claimed that they were not accepted in the other program. We had five such cases, so it is possible that institutions within the same university have different rules related to accepted research methodology in graduate PhD programs. This study may not be generalisable to different PhD programs worldwide that were not surveyed. The study is also not generalisable to Europe, as there are no universal criteria or expectations for PhD theses in Europe. Even in the same country, there may be different models and expectations for a PhD in different higher education institutions.
A recent study indicated a number of opposing views and disadvantages related to SRs as research methodology for graduate theses, including lack of knowledge and understanding by potential supervisors, which may prevent them from being mentors and assisting students to complete such a study [5]. This same manuscript emphasised that there may be constraints if the study is conducted in a resource-limited environment without access to electronic databases, that there may be a very high or very low number of relevant studies that can impact the review process, that methods may not be well developed for certain types of research syntheses and that it may be difficult to publish SRs [5].
Some individuals believe that a SR is not original research. Indeed, it has been suggested that SRs as ‘secondary research’ are different than ‘primary or original research’, implying that they are inferior and lacking in novelty and methodological rigour as compared to studies that are considered primary research. In 1995, Feinstein suggested that such studies are ‘statistical alchemy for the 21st century’ and that a meta-analysis removes or destructs ‘scientific requirements that have been so carefully developed and established during the 19th and 20th centuries’ [7]. There is little research about this methodological issue. Meerpohl et al. surveyed journal editors and asked whether they consider SRs to be original studies. The majority of the editors indicated that they do think that SRs are original scientific contributions (71%) and almost all journals (93%) published SRs. That study also highlighted that the definition of original research may be a grey area [8]. They argued that, in an ideal situation, ‘the research community would accept systematic reviews as a research category of its own, which is defined by methodological criteria, as is the case for other types of research’ [8]. Biondi-Zoccai et al. pointed out that the main criteria to judge a SR should be its novelty and usefulness, and not whether it is original/primary or secondary research [9].
In our study, 80% of the participants reported negative attitudes, and more than half of the respondents agreed with a statement that SRs are ‘not a result of the candidate’s independent work since systematic reviews tend to be conducted by a team’. This opinion is surprising since other types of research are also conducted within a team, and single authorship is very rare in publications that are published within a PhD thesis. On the contrary, the mean number of authors of research manuscripts is continuously increasing [10]. At the very least, the authors of manuscripts within a PhD will include the PhD candidate and a mentor, which is a team in and of itself. Therefore, it is unclear why somebody would consider it a problem that a SR is conducted within a team.
The second most commonly chosen argument against such a thesis was that SRs ‘do not produce enough new knowledge for a dissertation’. The volume of a SR largely depends on the number of included studies and the available data for numerical analyses. Therefore, it is unfair to label a SR as a priori lacking in new knowledge. There are SRs with tens or hundreds of included studies, and some of them not only include meta-analyses, but also network meta-analyses, which are highly sophisticated statistical methods. However, limiting SRs within a thesis only to those with meta-analysis would be unfair because sometimes meta-analysis is not justified due to clinical or statistical heterogeneity [11] and the presence or absence of a meta-analysis is not an indicator of the quality of a SR. Instead, there are relevant checklists for appraising methodological and reporting quality of a SR [12, 13].
The third most commonly chosen argument against SRs within PhD theses was ‘lack of adequate training of candidates in methodology of systematic reviews’. This could refer to either insufficient formal training or insufficient mentoring. The graduate program and the mentor need to ensure that a PhD candidate receives sufficient knowledge to complete the proposed thesis topic. Successful mentoring in academic medicine requires not only commitment and interpersonal skills from both the mentor and mentee, but also a facilitating institutional environment [14]. This finding could be a result of a lack of capacity and knowledge for conducting SRs in the particular institutions where the survey was conducted, and not general opinion related to learning a research method when conducting a PhD study. Formal training in skills related to SRs and research synthesis methods [15, 16], as well as establishing research collaborations with researchers experienced in this methodology, could alleviate this concern.
One third of the participants indicated a ‘lack of appreciation of systematic review methodology among faculty members’ as a reason against such a thesis model. This argument, as well as the prevalent negative attitude towards SRs as PhD theses, perhaps can be traced to a lack of knowledge about SR methodology; however, although the level of knowledge was quite low in our study, there was no statistically significant correlation between knowledge and negative attitudes. Of the nine questions about SR research methodology, only three questions were correctly answered by more than half of the participants. This could be a cause for concern because it has been argued that any health research should begin with a SR of the literature [17]. It has also been argued that the absence of SRs in the context of research training might severely hamper research trainees and may negatively impact the research conducted [18]. Thus, it has been recommended that SRs should be included ‘whenever appropriate, as a mandatory part of any PhD program or candidature’ [18].
It has recently been suggested that the overwhelming majority of investment in research represents an ‘avoidable waste’ [19]. Research that is not necessary harms both the public and patients, because funds are not invested where they are really necessary, and necessary research may not be conducted [17]. This is valid not only for clinical trials, but also for other types of animal and human experiments [20]. SRs can help improve the design of new experiments by relying on current evidence in the field and by helping to clarify which questions still need to be addressed. SRs can be instrumental in improving methodological quality of new experiments, providing evidence-based recommendations for research models, reducing avoidable waste, and enabling evidence-based translational research [20].
Four respondents from three institutions indicated that empty SRs are accepted as a PhD thesis. While it makes sense to include such a SR as a part of the thesis to indicate lack of evidence in a certain field, it is highly unlikely that an entire thesis can be based on an empty SR, without a single included study.
There are many advantages of a SR as a graduate thesis [4, 5], especially as a research methodology suitable for low-resource settings. A PhD candidate can prepare a Cochrane SR as a part of the PhD thesis, yielding a high-impact publication [4]. Non-Cochrane SRs can also be published in high-impact journals. A PhD candidate involved in producing a SR within a PhD thesis goes through the same research process as those conducting primary research, from setting up a hypothesis and a research question, to development of a protocol, data collection, data analysis and appraisal, and formulation of conclusions. Graduate programs can set limits, such as the prevention of empty reviews and the recognition of updated reviews as valid for a PhD thesis, and engage experienced researchers as advisors and within thesis evaluation committees, to ensure that a candidate will conduct a high-quality SR [4]. Conducting a SR should not be mandatory, but candidates and mentors willing to produce such research within a graduate program should be allowed to do so.
Further studies in this field could provide better insight into attitudes related to SRs as graduate theses and explore interventions that can be used to change negative attitudes and improve knowledge of SRs among decision-makers in graduate education.