Sharing interim trial results by the Data Safety Monitoring Board with those responsible for the trial’s conduct and progress: a narrative review

A narrative review was considered the most appropriate method to use because it allowed us to explore the literature and inductively evaluate qualitative and quantitative information. We anticipated that most of the literature we would find would be opinions, policies or guidelines, and a narrative review would require an inductive approach for theme analysis and categorisation. To find literature discussing the issue of DSMBs sharing interim results, a broad and comprehensive systematic search of the literature was done in December 2015 within the databases of PubMed (includes all MEDLINE citations), Web of Science, EMBASE and CINAHL from the inception of all four databases using a detailed search strategy for each of them, as outlined in Additional file 1. Key phrases related to ‘Data Safety Monitoring Boards’ were utilised in each of the four databases as well as a filter for articles in the English language.

The title and abstract for each citation that came up within the search of each of the four databases were reviewed. Citations were eligible and included for full-text review if the title or abstract associated with a particular citation met the following inclusion criteria: [(1) related to DSMB issues OR (2) related to the management, operation, conduct, use, experience, or discussion of DSMBs] AND (3) the article associated with the citation was published in English. Subsequently, citations from the full-text review were eligible and included for full-text information extraction if there was some focused discussion or statement, within the full-text, about sharing interim trial results with parties outside of the DSMB. Reference lists from included articles were searched for other unique articles discussing the issue of DSMBs sharing interim trial results using the same inclusion criteria. Additionally, two major textbooks that solely focused on the operation and management of DSMBs were also reviewed and consulted [1, 2]. These were found upon discussion with a professor who is a health methodologist with expertise in clinical trials.

We also searched regulatory, governmental and guideline groups from the USA, Canada, UK, European Union and Australia, and two international groups, the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) [8] and the World Health Organisation (WHO) [9]. For a list and details of the organisations that were searched directly for relevant literature or information from their respective websites, see Additional file 1. Two strategies were used in combination to find relevant literature on the issue of interim result sharing by the DSMB within major governmental/regulatory/funding bodies and guideline groups (Additional file 1). Documents and webpages on an organisation’s website that had information about clinical trial research and DSMBs, as indicated from a webpage’s or document’s title or within the full text, were eligible and included for full-text information extraction. A full-text review was done immediately because most webpages or documents found on these organisation websites are not structured like most journal articles with an abstract. Documents and pages from the full-text review were eligible and included for full-text information extraction if there was some focused discussion or statement within the full text about sharing interim trial results with parties outside of the DSMB. Please note that within the Results and Discussion sections, not all the citations included from our systematic search of the literature for this review are cited in this paper. A full citation list of articles found to support this review in its entirety is given in Additional file 2. All screening and full-text extraction was done independently by one of the authors (VBD) for this narrative review and then checked by co-reviewers (LM and LT). Information extracted from the included full text pertained to the sharing of interim trial results with parties outside of the DSMB. Any disputes or concerns were resolved by consulting co-reviewers (LM and LT). We worked inductively and thus retrospectively with the included literature to perform a categorisation and thematic literature analysis. Analyst triangulation was used among co-reviewers (LM and LT) with VBD to ensure that categorisation and thematic analysis of literature were sound. No changes were made to the review process that was initially planned. Our review is principled in pragmatism review because we were not initially sure what categories or themes would emerge from the literature. Please see the RAMESES checklist [10] attached as Additional file 3.

A total of 162 articles, documents, policies, guidelines books or webpages were included for this review. See Fig. 1 for a flow diagram showing the inclusion process and the number of articles included.

There are mixed views and opinions on the issue of the DSMBs sharing interim results during the conduct of a clinical trial with non-DSMB members. Out of the 162 included articles, 129 (79.6%) were based on author’s views, 27 (16.7%) on DSMB or trial research guidelines and 6 (3.7%) on surveys of trialists. The literature falls into three categories of opinions: Category 1 - literature that expresses the opinion or argument against sharing interim results, stating that they should remain confidential with the DSMB (71/162, 43.8%); Category 2 - literature that expresses the opinion or argument against sharing interim results with exceptions (73/162, 45.0%); and Category 3 - literature that expresses the opinion or argument in favour of sharing interim results by the DSMB with a non-DSMB group (18/162, 16.8%). No one group held the outright majority. However, the largest two literature groups that were very close percentage wise were Category 1 (43.8%) and Category 2 (45.0%).

Six surveys were also found out of the 162 included articles that looked globally at trial monitoring practice as described in detail in Table 1. These surveys did not specifically focus on the issue of DSMBs sharing interim results during the conduct of a clinical trial with non-DSMB members. However, all of them asked at least one question that was related to the issue and surveyed those who were somehow involved in trials. Two of the six surveys report results that are quantitatively unclear, as their results are qualitatively described. The remaining four of the six surveys report results quantitatively. The target populations for these surveys varied, ranging from directors of the statistical centres from 12 cancer cooperative groups sponsored by the National Cancer Institute (NCI) in the USA from 1993 [11], trialists of past and ongoing trials [5, 12], PIs and biostatisticians on DSMBs and Institutional Review Board (IRB) community representatives [13], major funders of trials, regulatory agencies and other relevant organisations related to trial research [5]. Methods to obtain the sample also varied. Sampling frames included statistical centres from 12 cancer cooperative groups sponsored by the NCI in the US in 1993 [11], the US National Institutes of Health (NIH) [12], ClinicalTrials.gov, a MEDLINE search of articles pertaining to randomised controlled trials (RCTs) from Biometrics or Statistics in Medicine, the Office of Human Research Protection website [13], a database of Health Technology Assessment (HTA) programme and Medical Research Council (MRC) trials [5] and a list of 25 handpicked organisations that are major funders of trials, regulatory agencies and other relevant organisations related to trial research [5]. Response rates ranged from 40% to 100% [5, 11‐13]. The number of responses to these surveys varied from 12 to 309 [5, 11‐13]. Respondents were not chosen at random in two cases [11, 12]. The other four surveys had some method to select respondents at random [5, 13]. Survey collection methods included telephone interviews [5], email surveys [5] and mail/paper surveys [5, 11‐13].

Three of the six surveys [5, 11, 13] (one qualitatively and two quantitatively reported) support the view against sharing interim results, stating that they should remain confidential with the DSMB (Category 1 view). For one of the two surveys reported quantitatively [11], all respondents (n = 9) indicated that NCI groups, at the time the survey was administered, did not provide unmasked outcome reports to the participants. For a second question in the same survey asking about which non-DSMB members had access to interim data reports [11], it was reported that 70% of the respondents (n = 10) indicated that non-DSMB members do not access interim data reports. For the remaining three respondents who answered this question, who alternatively stated that non-DSMB members do have access to interim data reports, there is no mention as to who specifically gets this information and why. For the second survey [13], as can be seen in Table 1, the majority of respondents, all of whom were biostatisticians, PIs or IRB community members, indicate that the sponsor should be masked to interim data or results (percentages can be viewed in Table 1). Masking of other non-DSMB members besides the sponsor was not mentioned. For the minority of respondents who said “No” to this question asking if the sponsor should be masked to interim data or results (0% biostatisticians, 21.7% PIs and 37.0% IRB community members), there was no mention as to why this information should be shared with the sponsor. For the one survey that qualitatively reported their results indicating “Views on sharing the interim information with other DMCs were consistent; the investigators were not enthusiastic about the DMC consulting others” [5], there was no mention of whether there were respondents with another view.

One survey [5] (quantitatively reported) supported the Category 2 view: against sharing interim results but with exceptions. The largest group of respondents from this survey, for the question related to interim data or results sharing (50%, 10/20), agreed with DSMBs sharing interim results with non-DSMB members if it was necessary, particularly for concerns related to participant safety [5]. Specifically with whom this information would be shared was not clear. For the remaining respondents to this question, 30% (6/20) of respondents had no provision for, or idea about, the DSMB sharing interim data or results with non-DSMB members, and 20% (4/20) of respondents disagreed with the DSMB interim data or results sharing with non-DSMB members.

Two surveys [5, 12] (one qualitatively and one quantitatively reported) supported the view in favour of sharing interim results by the DSMB with a non-DSMB group (Category 3 view). For the survey that was reported quantitatively [5], all of the respondents to the question related to interim data or results sharing (n = 17) indicated that someone outside of the DSMB had access to interim data or results during their trial. With whom interim results were shared was indicated as follows: 41% of respondents indicated everyone except the participants, 35% of respondents indicated key institute staff (one respondent from this group said there was also provision for interim data to be seen occasionally and confidentially by the DSMB of another trial), 18% of respondents indicated only the trial statistician and the DSMB and 6% stated that the data centre personnel had access to interim data or results. Why information was shared with these non-DSMB members was not discussed. For the other survey [12] that reported results qualitatively, it was indicated that respondents (n = 12) from the US NIH support that DSMB reports are to be confidential and privy only to DSMB members. However, access is also granted to selected US NIH staff, these being non-DSMB members. Why information was shared with selected US NIH staff was not discussed.

For the other 158 documents, which were not describing surveys, we assessed for a time trend to see where the views and policies lie for the last ten years, back to 2006, in regard to the three categories we identified. For the literature in Category 1, dating back to 1981, 54% (37/68) of the literature comes from the past ten years alone. For Category 2, the literature dates back to 1998. About 24% (4/17) of the literature comes from the last decade. For the literature in Category 3, dating back to 1991, 52% (38/73) of the literature comes from the past ten years. The most recent literature, count and percentage wise, has predominately supported the DSMB not sharing or not sharing but with some exceptions. We also found that regulation, policy or guideline documents predominately support Category 3 (55%, i.e. 15/27 of the 27 regulation, policy or guideline documents included in our review).

In regard to our second objective, there is a subset of the literature within Category 2 or 3 literature that discusses what interim results the DSMB should share, with whom and the circumstance (why). Eleven circumstances that may warrant the DSMB sharing interim results with non-DSMB members are explained in Table 2, generally categorised under four themes: (1) current usual practice by DSMBs, (2) trial completion is threatened, (3) concern about patient safety and (4) regulatory approval). There is also a category for other special circumstances that includes three unique situations for DSMB sharing of interim results that did not fit into a theme. Six of these eleven circumstances are supported in the literature with real-life examples. What is shared by the DSMB with non-DSMB members varies depending on the particular circumstance. For many of the cases where sharing may be warranted, a risk or counter argument to sharing is indicated where applicable. Most of the risks with sharing in these circumstances are predominately associated with introducing bias in the trial that will affect the final trial results. It is indicated in the literature that there is always the potential that sharing interim results with non-DSMB members may do harm to a trial by disturbing equipoise [14, 15], as people may make inaccurate impressions about what is happening between treatment groups [3, 16‐18]. It is explained that when equipoise is disturbed with knowledge of interim results by those operating and managing the trial and those participating in the trial, there may be actions people can take, either consciously or subconsciously [19], that can bias the trial’s results [1, 16, 19]. The introduction of bias can reduce the credibility and integrity [16, 20, 21] of the trial, rendering the results questionable [16, 20, 22].

Interim result sharing is an important issue because it affects the validity of the results from confirmatory trials on which we base regulatory and practise decisions, impacting the health and lives of many. From this review, two categories of the literature dominate (Category 1 and Category 2), but not in majority as distinct groups. Category 1 is against sharing interim results, stating that they should remain confidential with the DSMB, and Category 2 shares the same sentiment as Category 1 but additionally acknowledges exceptions, that there are circumstances that may warrant the DSMB to share interim results with certain non-DSMB groups/members. What is shared with these non-DSMB members depends on what the situation calls for and should be assessed by the DSMB using their expertise to balance risk(s) with the potential benefit(s) regarding participant safety and trial validity and integrity. Because of the limitations of the evidence found, collecting more empirical evidence through a survey of the general clinical trials community focused on the issue of DSMB sharing interim results (what, if any, interim result(s) to share, with whom and under what circumstance) is needed to better understand and guide DSMB interim information sharing practices.