A joint industry‐sponsored data monitoring committee model for observational, retrospective drug safety studies in the real‐world setting

Abstract Purpose To share better practice in establishing data monitoring committees (DMCs) for observational, retrospective safety studies with joint‐industry sponsorship. Methods A DMC model was created to monitor data from an observational, retrospective, post‐authorization safety study investigating risk of medullary thyroid cancer in patients treated with long‐acting glucagon‐like peptide‐1 receptor agonists (LA GLP‐1RAs) (NCT01511393). Sponsors reviewed regulatory guidelines, best practice and sponsors' standard operation procedures on DMCs. Discussions were held within the four‐member consortium, assessing applicability to observational, retrospective, real‐world studies. A DMC charter was drafted based on a sponsor‐proposed, adapted DMC model. Thereafter, a kick‐off meeting between sponsors and DMC members was held to receive DMC input and finalize the charter. Results Due to this study's observational, retrospective nature, assuring participant safety – central for traditional explanatory clinical trial models – was not applicable to our DMC model. The overall strategy and key indication for our real‐world model included preserving study integrity and credibility. Therefore, DMC member independence and their contribution of expert knowledge were essential. To ensure between‐sponsor data confidentiality, all study committees/corporations and sponsors, besides the DMC, received blinded data only (adapted to refer to data blinding that revealed the specific marketed LA GLP‐1RA/sponsor). Communication and blinding/unblinding of these data were facilitated by the contract research organization, which also provided crucial operational oversight. Conclusions To our knowledge, we have established the first DMC model for joint industry‐sponsored, observational, retrospective safety studies. This model could serve as a precedent for others performing similar post‐marketing, joint industry‐sponsored pharmacovigilance activities.

blinding/unblinding of these data were facilitated by the contract research organization, which also provided crucial operational oversight.
Conclusions: To our knowledge, we have established the first DMC model for joint industry-sponsored, observational, retrospective safety studies. This model could serve as a precedent for others performing similar post-marketing, joint industrysponsored pharmacovigilance activities.
K E Y W O R D S data monitoring committee, joint sponsorship, observational, real-world data, retrospective

| INTRODUCTION
The importance of real-world evidence to support clinical trial data is increasingly recognized and requested by stakeholders, for example, health regulatory agencies, health policy decision-makers, patient groups, public funders and academia. Clinical trial limitations may include short duration, small populations and selective inclusion criteria.
Moreover, results from highly selected and controlled environments provide high internal validity but are not necessarily generalizable to real-world settings (i.e. failure to provide external validity). [1][2][3][4][5] Consequently, the potential existence of unforeseen safety signals/safety concerns (unknowns) may persist, despite evaluation in well-performed clinical trials and following marketing authorization.
Industry sponsors with marketing authorizations for same-class drugs are encouraged, by regulatory authorities, to conduct joint collaborative real-world studies. 6 Hence, inconvenience to patients, physicians and registries is minimized while data exposure is maximized.
Increasing recognition of the value of drug safety studied in such settings has resulted in comprehensive databases that may warrant enhanced monitoring.
Data monitoring committees (DMCs) were first described in 1967. 7 Despite increased and evolved use over the last 50 years, DMCs remain within the clinical trial realm. We believe joint multidisciplinary and multi-institutional efforts are needed to establish robust infrastructures supporting real-world studies, including DMC empowerment beyond clinical trials, lending study integrity and credibility to real-world studies. The DMC model we share monitors data from an observational, retrospective, real-world study. This model accommodated additional challenges resulting from one DMC serving four industry sponsors within one study.

KEY POINTS
• Observational, retrospective studies are increasingly performed to assess real-world drug safety. They are important tools for detecting rare events (e.g. medullary thyroid cancer [MTC]) that are difficult to measure in drug development programs.
• Stakeholders (regulatory health authorities, payers, physicians) request real-world data to support clinical trial data (internal validity) with external validity.
• Data monitoring committees (DMCs) may play a central role in validation and ongoing interpretation of the increasingly extensive data emerging from real-world studies.
• Our DMC model was shown to be effective for an observational, retrospective, joint industry-sponsored study.
• This model may be of interest to others involved in this evolving, real-world pharmacovigilance discipline, particularly in challenging/rare diseases such as MTC.

| RESULTS
The following results are set out according to (i) findings following review of the DMC guidelines/best practice, as detailed in the Methods (Section 2.2); and (ii) applicability of these "recommendations" to our DMC model (also summarized in Appendix S1).

| Precedence according to DMC guidelines and best practice
Health regulatory and government funding agencies created DMC guidelines (henceforth, referred to as "guidelines") 9-13 based on the Greenberg DMC model. 7 Like the Greenberg model, guidelines focused on DMCs for explanatory clinical trials. Lilienfeld et al. recognized necessary DMC role evolution from pre-clinical to postmarketing trial settings, including observational studies. 14 However, no solid experience with their proposed DMC model was shared.
Except for a DMC acting in a prospective, open-label, observational study, 15 authors have described and discussed consensus and controversies of the DMC set-up for traditional explanatory clinical trial settings. [16][17][18][19][20][21][22][23][24] Hereafter, these are referred to as "best practice." Only explanatory clinical trials were in scope for each sponsor's DMC SOPs.
Although there is precedence for a single DMC reviewing multiple trials, 22,25 these involved only one sponsor. Hence, no best practice for a joint industry-sponsored DMC model was identified.

| Applicability
No guidance/sponsor SOPs for DMCs serving observational, retrospective studies exist, and no solid experience regarding such DMCs has been shared; thus, there was no best practice precedence.

| Indication for DMC according to DMC guidelines and best practice
Safeguarding the interests and safety of trial participants was a central indication for creating a DMC, particularly relevant for trials where the studied population was fragile, vulnerable or at elevated risk of serious outcomes/death, the studied disease was life-threatening, or the studied treatment was considered invasive/toxic. Other indications for DMCs include preserving trial integrity (scientific validity) and credibility, particularly for large, multicenter, long-duration trials, where interim analyses might ethically require trial termination before planned completion (e.g. due to futility), or trials with complex designs requiring potential modifications, depending on unblinded interim data.

| Applicability
Assuring trial participants' safety was not applicable to our DMC model because the "event of interest" had already occurred when drug exposure data (prior to the studied outcome) were captured in the retrospective studies (i.e. a previous diagnosis of MTC was an inclusion criterion).
Preserving study integrity and credibility was deemed the main indication for our model. The increasing complexity of real-world data from observational, retrospective studies requires expertise for appropriate analysis and interpretation. Furthermore, changes in understanding of the disease, the affected population and standard realworld treatment may happen over the 15+ years of our study, warranting independent DMC oversight.

| Composition
The DMC should include medically qualified clinical trial experts within the areas relevant to the study's population and outcomes.
Additional members should include statisticians, ethicists and patient advocates from patient populations using the product. Previous DMC experience is also desirable. 20 The DMC model was adapted to include two epidemiologists and a statistician with expertise in observational study designs ( Figure 1). Including patient advocates and data-programming experts could also have been relevant for this DMC model reviewing real-world data.
All DMC members received travel reimbursement and honoraria from the sponsors via the CRO.

| DMC roles and responsibilities according to DMC guidelines and best practice
Trial participant safety was considered the primordial DMC role and responsibility. Other proposed DMC goals were related to study conduct and progress, including patient recruitment, protocol compliance and data quality. 16

| Applicability
Blinding of treatment arms was not applicable to our DMC model; there was no treatment randomization and neither physicians nor patients were blinded to treatment prior to MTC diagnosis. We adopted and adapted the terms "blinded" and "unblinded" data. which also included an executive committee session. 11,12,25 No consensus was reached, across the guidelines and best practice, as to whether only DMC members should be allowed to attend the closed sessions. The meeting venue should be in a neutral location.

| Applicability
The open and closed sessions were adapted to our DMC model to align with the model-defined concept of "blinded" data ( Figure 2

| DMC recommendations according to DMC guidelines and best practice
To ensure trial participant safety, the DMC should make formal recommendations to clinical trial sponsors after review of interim data, such as: (i) continue without modifications; (ii) stop wholly or partly; (iii) continue with modifications. Stopping a trial early could be due to futility or a positive treatment effect. Study design modification could be due to an unblinded interim data review.

| Applicability
The recommendation of stopping a study to ensure patient safety was not applicable to our DMC model. Indeed, a significant finding might warrant continuing the study until adequate knowledge is obtained, as a DMC responsibility to societies where the drugs already have been marketed.
Recommendations in our DMC model reflected the study's ongoing validity (futility) and overall credibility of results. Based on the above considerations, and to reflect the retrospective nature of the collected data emerging from the real-world setting, the recommendations shown in Box 1 were incorporated into our model and signed by the DMC chairman after each closed session.
3.8 | DMC communication flow to sponsors following data review meetings as recommended by DMC guidelines and best practice DMC recommendations could be communicated directly to the trial sponsor or a study steering group/third party. The communication should preferably be in writing, maintaining blinded data confidentiality.

| Applicability
Due to the added complexity inherent in a DMC reviewing safety data related to multiple sponsors' study drugs, the above was not directly applicable to our DMC model. DMC communication to sponsors could not reveal other study sponsors' confidential (blinded) data. However, if a safety concern was raised with a study drug(s)/sponsor(s), the remaining study sponsors had to be informed promptly to perform timely pharmacovigilance activities. confidentiality, the agreed wording to be used by the DMC chair informing study sponsors of a safety concern was "[Safety concern XXX] was identified for one or more LA GLP-1 RAs." Each consortium sponsor would facilitate the appropriate actions according to each sponsor's internal SOPs to support their pharmacovigilance obligations. Also, each sponsor responsible for one or more of the drugs triggering the safety concern would be directly informed by the chairman, facilitated by the CRO. As a minimum, if safety concerns/signals (and/or recommendations to continue the study with modifications, or to report to the FDA earlier than the next planned annual report submission) were raised by the DMC, a dialog between the DMC chair, sponsor representatives (according to each company's internal procedure) and the CRO would take place before further actions and decisions were made. The written communication on DMC recommendations/raised safety concerns would be conveyed to the Steering Committee chair by the CRO at the end of the third working day after the closed session. In our model, the sponsors had final responsibility for acting on the DMC's recommendations.
3.9 | DMC kick-off meeting -DMC charter and SAP according to DMC guidelines and best practice The DMC set-up should be described in a document (e.g. DMC charter) before the DMC initiates trial data monitoring. DMC charter templates have been published. 17 according to trial nature. 20 An early DMC meeting was recommended, allowing members to meet and provide input to the DMC charter prior to its finalization.

| Applicability
We introduced a "sponsor pre-DMC kick-off meeting" to our model.