Ensuring Integrity in Industry-Sponsored Research: Title and subTitle BreakPrimum Non Nocere, Revisited

The most fundamental principle of medicine, primum non nocere, holds for every physician, whether functioning as a clinician providing direct patient care; as a researcher, reviewer, or editor involved in medical publishing; or as an administrator overseeing an academic institution, health care organization, or pharmaceutical company research program. In all situations affecting patients, physicians must do no harm.

The Commentary in this issue of JAMA by Nissen¹ describes a disturbing example of inappropriate conduct surrounding an industry-sponsored clinical trial of rosiglitazone and reveals a situation in which concerns about preserving market share apparently trumped concerns about the potential for causing patient harm.² Analyzing this situation and others involving misleading reporting and possible misrepresentation of industry-sponsored research^{3 - 7} has become only too common, and it appears that physicians functioning at several levels failed to put the well-being of patients first.

Several important questions include why would a peer reviewer for a medical journal violate ethical standards and provide a confidential manuscript to a for-profit company; why would physician managers in a company misuse that confidential information in attempts to counter research findings suggesting that a company's product was harmful; why would any physician in a for-profit company participate in, tolerate, or condone deceptive or unethical conduct or lack sufficient ability to stop such conduct; why would any physician-researcher agree to report findings from an unscheduled interim analysis of a clinical trial, potentially undermining the statistical validity and credibility of the overall results; why would any physician serving on a steering committee and realizing that patients were at risk remain silent when a drug is shown to be potentially harmful; and why would studies be published without complete assurance that the data management was appropriate and the statistical analysis was accurate and ethically conducted? Perhaps some of these concerns can be explained with legitimate reasons; however, based on the report prepared by the staff of the US Senate Committee on Finance² involving the investigation of rosiglitazone (Avandia) and its manufacturer, GlaxoSmithKline (GSK), it appears that reassuring explanations are unlikely.

According to the article by Nissen,¹ the report of the Senate investigation,² and published media accounts,^{8 - 9} the manufacturer of rosiglitazone exerted inappropriate influence during the conduct of a pivotal safety study of this drug, the RECORD (Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycaemia in Diabetes) clinical trial,^{10 - 11} which included undertaking nonprespecified unblinding of study data; attempting to undermine the authority and responsibilities of the study steering committee; expediting publication of an unscheduled interim analysis,¹⁰ specifically to counter² the publication of a meta-analysis by Nissen and Wolski¹² that suggested increased cardiovascular risk associated with rosiglitazone; having employees extensively involved in statistical analysis,¹¹ and preparation of the manuscript¹⁰ reporting the results of the trial; and reportedly² failing to fully acknowledge the significant cardiovascular risk associated with this drug.

According to the staff report of the US Senate Committee on Finance, which was based on a 2-year investigation involving review of more than 250 000 pages of documents from GSK, the US Food and Drug Administration (FDA), universities, and others, along with numerous interviews, “[T]he totality of evidence suggests that GSK was aware of the possible cardiac risks associated with Avandia years before such evidence became public. Several years prior to Nissen's study, it can be argued that GSK was on notice that Avandia may have problems. Based on this knowledge, GSK had a duty to sufficiently warn patients and the FDA of its concerns in a timely manner. Instead, GSK executives intimidated independent physicians, focused on strategies to minimize findings that Avandia may increase cardiovascular risk, and sought ways to downplay findings that the rival drug Actos (pioglitazone) might reduce cardiovascular risk.”²

The circumstances surrounding the RECORD trial and the investigation involving rosiglitazone and its manufacturer² represent another serious example of corporate motives apparently outweighing scientific standards and perhaps ethical principles, thereby exposing patients to potential harm.^{1 ,4 ,8 - 9} Journals have been concerned about these issues, and some have been increasingly reporting detailed descriptions about the level of involvement of sponsors in clinical trials. For instance, JAMA requires detailed reporting of all financial and material support for the research and the work in the article and also includes in each article the specific role of the funding organization or sponsor in each of the following: “design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.”¹³

Similarly, the article reporting the RECORD interim analysis¹⁰ includes information about end point assessment: “Members of an independent committee evaluating clinical end points (five cardiologists, a neurologist, and a diabetologist) were unaware of study-group assignments and used prespecified criteria to adjudicate all potential outcomes reported by investigators. Evaluators in the trial's contract organization (Quintiles) were unaware of study-group assignments in screening all serious adverse events for potential end points.” The article also includes information about study oversight: “An independent data and safety monitoring board meets twice annually to review unblinded safety data for the ongoing study. . . . Members of the steering committee (seven academic investigators and one representative of the sponsor) developed the study design, had full access to the interim data, were responsible for the decision to publish the results, and wrote the manuscript. The committee members vouch for the accuracy and completeness of the data reported.” In addition, the methods section of the article indicates that “This interim report follows a prespecified plan for statistical analysis,” and the acknowledgment section of the article thanks an individual for “conducting confirmatory statistical analyses.” However, it would have been helpful to provide more detail about who conducted the primary statistical analysis, and more specific information about what the confirmatory analyses involved.

Likewise, the published report of the RECORD trial¹¹ includes information about data management: “The study was monitored by a clinical trials organisation (Quintiles, Bracknell, UK), which also coordinated data collection. . . . The study was overseen by a steering committee. An independent data safety and monitoring board reviewed conduct of the study and unblinded data at about 6-month intervals. Interim analyses were not made available to the steering committee, except for those required for the interim publication. Adverse-event data were made available to the pharmacovigilance department of the sponsor, but not to staff involved with the study.” The article also includes information about the statistical analysis: “The statistical analysis was done by the sponsor's statisticians according to a detailed predefined statistical analysis plan agreed by the steering committee. . . . Confirmatory analyses were done by an independent investigator (London School of Hygiene and Tropical Medicine, UK).” A summary of the role of the sponsor is also included: “Sponsor statisticians were involved in the design, reporting plan, and data analysis. The steering committee had responsibility for study conduct, data collected, data analysis, the writing of this report, and the decision to publish.” However, it would have been informative to have a detailed explanation of the specific approach used for the “confirmatory analyses” and to indicate whether the results of the independent analysis were those reported in the article.

While these efforts at reporting details regarding data management, statistical analysis, and sponsor involvement in clinical trials represent important progress in achieving more transparent reporting of industry-sponsored studies, it is virtually inevitable that other instances of incomplete or inappropriate data analysis, selective reporting, and misrepresentation of reports involving industry-sponsored research may occur. In an effort to counteract such problems, nearly 10 years ago JAMA began to require that for all studies reporting original data, regardless of funding source, at least 1 author must indicate that she or he “had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.”¹⁴ For industry-sponsored studies, this statement must be provided by an academic investigator or government researcher (preferably the principal investigator) who is not employed by any commercial funding source. Moreover, in industry-sponsored studies, data collection and data management should be conducted primarily or solely by the academic investigators independently of the study sponsor or other for-profit research organization, and with additional monitoring and oversight, such as under the auspices of an academic independent data and safety monitoring committee.^{3 - 4}

In addition, industry-sponsored studies in which the data analysis has been conducted only or primarily by statisticians employed by the company sponsoring the research are not accepted for publication in JAMA.^{3 - 4} This does not mean that the names of industry-employed statisticians, epidemiologists, or others involved with the data management or analyses should be removed from the manuscript reporting these studies; the roles of these individuals as authors and the contributions of those who do not meet authorship criteria should be clearly identified.

For industry-sponsored studies, an independent analysis of the data must be conducted by an independent statistician at an academic institution, rather than by statisticians employed by the sponsor or by a commercial research organization. The independent biostatistician must be a faculty member at a medical school or academic medical center or an employee of a government research institute, such that the academic organization has oversight over the person conducting the analysis and is independent of the commercial sponsor.

Ideally, the independent academic statistical analysis should be planned prospectively and specified in the trial design. For analyses of ongoing or recently completed trials, the independent academic biostatistician should receive the entire raw data set (not simply the summarized data), along with the study protocol and the prespecified plan for data analysis. The independent biostatistician should verify the appropriateness of the analytic plan and conduct an independent analysis of the raw data. The independent statistician should clearly describe his or her involvement in conducting the analyses and provide written confirmation of the data analysis. Details of this independent statistical analysis, the name and institutional affiliation of the independent statistician, and whether compensation or funding was received for conducting the analyses are reported in the manuscript. Perhaps most important, the results of the independent statistical analysis must be the results that are reported in the manuscript.

We recognize that this requirement for an independent statistical analysis of industry-sponsored studies entails additional effort, time, and cost. However, given concerns about the integrity and reporting of some industry-sponsored research, this independent assessment of the data and statistical analysis is essential and also provides an additional layer of institutional oversight for the integrity of the data analysis and reporting.^{3 - 4} For instance, if concerns arise about data integrity or completeness of reporting, or about data manipulation or misrepresentation, a mechanism for investigation would be in place, such as by an investigative committee appointed by the dean of the academic medical center at which the academic investigators and independent statisticians are faculty members. These policies are not intended to question the ability or integrity of company-employed statisticians; rather, the requirement provides an additional layer of institutional oversight intended to help reduce the ability of the sponsoring company to interfere.

If other journals also would adopt formal policies that require all industry-sponsored studies to have an academic investigator attest to having access to all of the data and to the integrity of the data and analysis, and that require an independent statistical analysis as a condition for review and publication, these requirements could serve as an effective deterrent to inappropriate or selective data analysis or reporting by industry, could encourage prospective planning for the required independent statistical analysis and data management to be the responsibilities of academic investigators, could provide leverage for academic investigators in their negotiations with companies to conduct trials to ensure complete data access, and may help decrease episodes of misrepresentation or incomplete reporting in industry-sponsored clinical trial results.

Similar to the experience with clinical trial registration,^{15 - 16} involvement of numerous journals in adopting a unified approach to requirements for complete data access and independent statistical analysis in industry-sponsored studies will be necessary to ensure data integrity and complete reporting of industry-sponsored research. While we recognize that these approaches are not infallible, we maintain that widespread adoption of a unified approach will be essential to maintain public confidence and trust, and to ensure scientific integrity of biomedical research.

Given the circumstances surrounding rosiglitazone^{1 - 2} and the RECORD trial,^{10 - 11} and perhaps other industry-sponsored trials, it is now time for all editors to require that academic researchers have full access to all trial data and that all industry-sponsored trials include independent statistical analysis and assurance. This approach would add powerful support to the fundamental principle that physicians must first do no harm.