Protocol, registration and definition
This systematic review was registered in the PROSPERO database (PROSPERO 2018 CRD42018084452), conducted according to a prespecified protocol and reported in compliance with best-practice reporting guidelines for systematic reviews [14] and research involving patients [15]. For this study, we employed the PICO model to define the key elements in our research, which included peer-reviewed publications that have investigated the effects of having patients (P – Participants) engaged as authors of or contributors to (I – Intervention) peer-reviewed publications. We quantified the number of these publications (Outcomes) and compared this with the number of publications on patient involvement in the preparation of lay summaries of clinical trial results (Comparator).
To minimize the risk of research waste, we searched (5 June 2017) the PROSPERO database to ensure we were not duplicating a planned or ongoing systematic review. We also registered our review on SYNAPSE, the Patient Focused Medicines Development (PFMD) repository for patient engagement initiatives (https://synapse.pfmd.org/initiatives/patient-involvement-in-preparing-clinical-research-peer-reviewed-publications-or-results-summaries-a-systematic-review).
Within this study, ‘patient’ was defined in broad terms, based on an existing definition [16, 17] and input from our patient partners. For this research, ‘patient’ refers to “people having or at risk of having medical condition(s), whether or not they currently receive medicines or vaccines to prevent or treat a disease” as well as “the family and those voluntarily caring for those with the medical condition(s), patient advocates and patient groups.”
Eligibility criteria
We included peer-reviewed publications that studied the effects of patient involvement for sharing results of health research studies (all phases and all designs) either through a peer-reviewed publication or through a lay summary of clinical trial results (i.e. consistent with the regulatory requirements described in the EU Clinical Trials Regulation 536/2014 [Article 37] [13]). We considered publications, in any language, that were detected in searches of electronic bibliographic databases or secondary sources (see Information sources and search strategies section). We did not include unpublished data or abstract-only articles.
Information sources and search strategies
We used primary and secondary information sources. For our primary source, we searched three electronic bibliographic databases (MEDLINE, EMBASE and Cochrane Database of Systematic Reviews) using the OVID database search interface. We used Medical Subject Headings (MeSH) and words/phrases related to patient involvement in health research peer-reviewed publications and lay summaries of clinical trial results. An experienced search strategist conducted the search of the three databases on 9 October 2017 (limits: 1 January 2015–9 October 2017; see Additional File 2 for full search strategies). References were exported to a reference management software program (EndNote) and saved into a project-specific library within EndNote. Duplicate publications were removed.
We supplemented our primary information source with secondary information sources. These sources comprised the repository of patient involvement in research literature curated by the Patient-Centered Outcomes Research Institute (PCORI), PubMed’s ‘similar articles’ listings and the authors’ personal reference files. A search of the PCORI repository was conducted on 10 June 2018 via PCORI’s Engagement in Health Research Literature Explorer (limits: dissemination, 1 January 2017–10 June 2018); a search of the top 10 PubMed articles most related to eligible articles retrieved from our primary information sources was conducted on 11 June 2018; and authors’ personal references were scanned up until 30 May 2018.
Study selection
We assessed publications for eligibility using a standardized process and a dedicated assessment team (three authors with predefined roles). One author (KW) selected potentially eligible articles based on a title and abstract screen, then two authors (LA, TG) independently selected eligible articles based on a review of the full text of each potential article. Differences in assessment were resolved by consensus among the assessment team.
Data collection process and data items
Based on the protocol, we prepared a spreadsheet template to collect data from all eligible publications. The template was pilot tested and refined before use. Data were collected using a standardized process and a dedicated data collection team (three authors with predefined roles). Two authors (RJ, AS) independently extracted data from each publication and a third author (ACW) verified 100% of the data extracted for the primary outcome measure from every eligible publication, and verified 100% of the data extracted for the secondary outcome measures from a random sample of 25% of the eligible publications. Data (Additional File 3) were extracted for publication title, journal, year, authors, first author’s country, research sponsor, number of patient authors, patient author background, patient author recruitment process, author communication process, extent of patient involvement in preparing the publication or lay summary of clinical trial results, how patient involvement outcomes were assessed, benefits (to patient authors, other authors, the project), risks (to patient authors, other authors, the project) and best practice recommendations.
Risk of bias in individual studies
We anticipated that there would be few randomized controlled trials that directly addressed our research question. As per the protocol, the Cochrane Collaboration tool for assessing bias risk (https://handbook-5-1.cochrane.org/chapter_8/table_8_5_a_the_cochrane_collaborations_tool_for_assessing.htm) was to be used to assess randomized controlled trials. The Newcastle-Ottawa Scale was used to evaluate the risk of bias for other study designs, although we recognize this instrument does have limitations (http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp).
Summary measures
Our primary outcome measure was the number of peer-reviewed publications that investigated the effect of patient involvement on preparing peer-reviewed publications.
Our secondary outcome measures were:
- a.
The number of peer-reviewed publications that investigated the effect of patient involvement on preparing regulatory-standard lay summaries of clinical trial results;
- b.
The quality of the evidence reported in the eligible publications;
- c.
The number and the background (e.g. patient experts, clinical trial participants, patient advocacy group members) of patients contributing to the preparation of the publications or lay summaries of clinical trial results;
- d.
The type of patient involvement (e.g. as authors, as non-author contributors);
- e.
The number and type of patient involvement outcomes assessed (e.g. benefits, risks, best practice recommendations, other).
Synthesis of results
As anticipated in the protocol, we did not synthesize data quantitatively due to heterogeneity in the eligible publications. We conducted a qualitative narrative synthesis of the data to describe patient involvement outcomes.
Additional analyses
To assess the quality of the patient authorship experience in the present study, one author (KW) customized the original PFMD Patient Engagement Quality Guidance [16], which PFMD co-created with over 100 people from 51 organizations, representing patient associations, industry, academics, researchers, and external experts. The resulting draft Patient Authorship Experience (PAE) self-assessment tool was then updated with co-author feedback. The final PAE tool evaluates eight publication-relevant domains using a bipolar, five-point, psychometric, Likert scale (strongly disagree to strongly agree) [18]. The total score is determined by adding up the individual scores for each question and calculating a percentage out of a maximum score of 80. Two versions were created: one for patient authors and another for non-patient authors. All authors in this study completed the PAE tool.
To document the involvement of patients in this project, the GRIPP2 Short Form was completed [15].