- Defining Clinical Trial Transparency
- Clinical Trial Spin and Suppression
- Benefits and Costs of Transparency
- Legislative Strategies to Encourage Transparency
- Bottom-Up Strategies for Clinical Trial Transparency
- Conclusions: Towards Greater Transparency
- Clinical Trial Transparency References
Why worry about transparency? To justify new measures to enforce the reporting of clinical trials, it must first be shown that a problem exists with the status quo of results reporting. We will consider both the global, statistical evidence that trial results are not adequately reported in the literature, and will then review a few specific instances of information suppression and biopharmaceutical malfeasance that might have been prevented by stronger incentives and rules to report trial results.
Several analyses have been done to assess the possibility of reporting and publication bias in clinical trials. Dwan et al.  recently reported a comprehensive statistical evaluation of bias in controlled clinical trials. Bias may take several forms. Studies with positive results are much more likely to be published. Dwan reports that even when negative results are published, they are susceptible to “time lag” or “pipeline” bias, which means that studies with negative results take longer to be published than positive studies. The tendency to publish negative results may not even be malicious; negative results are generally seen as less interesting to the scientific community, so researchers may be less enthused about working to publish negative results, or feel that the negative results are not adequately interesting to justify publishing. It is also very likely that trials with negative results are more difficult to get accepted for publication in top-tier peer-reviewed journals, so even if a clinical researcher makes their best effort to publish a negative result, doing so may be difficult or impossible. In addition to bias in reporting the outcomes of trials as originally specified, another form of publication bias noted by Dwan is “within-study selective reporting bias,” which occurs when a subset of the original variables are chosen for reporting in a publication, such as reporting of different patient subgroups or time-points.
Dwan et al. performed a meta-analysis in which they reviewed the results of 16 different smaller studies that each evaluated a cohort of clinical trials to assess publication bias. Eleven of these studies looked for publication bias, while five examined outcome reporting bias, by comparing final publication to the pre-specified trial protocols. Approximately half of the trials evaluated were found to have at least one major outcome that was modified or omitted, suggesting that outcome reporting bias is prevalent and should be of concern. Looking at publishing bias, across the sixteen cohorts, trials with positive results were consistently more likely to be published than those with negative results, and the total frequency of publication was less than 50%. From the limited data available, it also appeared that this bias was due more to negative trials not ever being submitted for publication, rather than being more likely to be rejected after submission to a journal.
Turner et al. assessed evidence for selective publication of antidepressant trials . Turner accessed trial results for drug approval studies directly from the FDA, and compared these results to outcomes reported (or not reported) by publication. The fact that Turner was able to access FDA-reported results is itself unusual; it is difficult for researchers to even study whether or not publication bias exists when the existence of and data from unpublished studies is not available. Turner found that 31% of FDA-registered studies were not published at all. 37 studies with FDA-positive results were published; only a single study seen as positive was not published. Most studies with negative or questionable results were either not published (22) or published in a way that conveyed a positive outcome (11 studies). When primary outcome failed to have a positive significant result, a secondary result was often reported as the headline and the major failure was subordinated or omitted. So according to the literature, it appeared as if 94% of the trials were positive, whereas FDA data analysis showed that only 51% were positive.
This statistical evidence shows that there are systematic problems in the reporting and publication of clinical trial results. Several specific cases discussed below illustrate the potential harm that may arise when clinical trials are not transparent and their results not subject to adequate scrutiny. Medicine generally attempts to be evidence-based, in which clinical methods are developed based on the empirical results reported in the literature; if the data in the published literature is incomplete or biased, evidence-based medicine cannot function effectively. I will describe five cases that have occurred over the last decade which illustrate the risks and consequences of opaqueness in clinical trials.
Actonel, manufactured by Procter and Gamble, is a drug intended to help strengthen the bones of post-menopausal women, and competes with a similar drug called Fosamax made by Merck & Co. In 2002, P&G contracted with a British physician, Dr. Aubrey Bluhmson, to conduct and analyze a study comparing the effectiveness of Actonel with Fosamax. After the study concluded, P&G refused to grant Bluhmson access to the randomization codes required to analyze the double-blinded data, and provided Bluhmson with only a limited subset of the results that the researcher claimed biased the results to favor Actonel.  P&G then published ghost-written publications about the Actonel studies under Bluhmson’s name, despite his disagreement with the study conclusions and lack of access to the dataset. Even though there may be legitimate disagreements in the interpretation of such trials, it is impossible for third parties to assess the controversy in the absence of full, primary data set. If the investigator’s allegations are true, patients taking Actonel may be at unnecessary risk and missing out on the benefits of a better drug. This case illustrates the risks of clinical trial investigator agreements in which researchers agree to give up rights to access data, and to control authorship of publications, and reveals the difficulties that may derive from the conflicting interests of trial sponsors, individual investigators, their institutions, and the journals that publish results from clinical trials. It also shows that even when the summarized results of a trial are published, it is difficult to evaluate the veracity of the conclusions without full access to the primary research data. If the primary data had been deposited in a publicly accessible database, disagreements about the interpretation of results could be assessed by independent third-parties.
Glaxo Smith Kline has been involved in major controversies over two of its drugs. Paxil (Paroxetine; Seroxat) is an anti-depressant medication in the family of selective-serotonin reuptake inhibitors. Paxil was approved in 1992 for treatment of depression in adults, and grew to become a blockbuster that generated more than $1B in sales for GSK by 1997, and the set of indications for which Paxil was marketed included panic disorder, social anxiety disorder, and premenstrual dysphoric disorder. By 2006, evidence had come to light of significant problems with Paxil, including ineffectiveness and a significant increased risk of suicide when administered to children.
The New York State Attorney General sued GSK in June 2004, alleging that GSK withheld information from results of 5 internal studies showing that Paxil was ineffective and dangerous in children, despite having this information as early as 1999. At the same time, Paxil was heavily promoted in for use in children. GSK also found itself the target of a number of class-action suits related to Paxil safety issues. The suit brought to light internal GSK memos that showed the company deliberately attempted to “manage the dissemination of data in order to minimize any potential negative commercial impact.” 
The suit against GSK was settled in August 2004 before going to trial . Glaxo agreed to pay $14M and to modify their reporting behaviour. GSK also agreed to create and update a public, internet-accessible database containing summaries of all sponsored clinical trials completed after Dec 2000. Finally, GSK also consented to permit monitoring by the NY Office of the Attorney General, and to submit full clinical trial reports and raw data for inspection up request, subject to a reasonable confidentiality agreement. The format and contents of the clinical trial summaries was specified in the agreement, and was based upon guidelines from recommendations of the International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use, a joint effort between the USA FDA and European regulatory authorities. This agreement provides a good framework for what should be industry-wide standards in reporting clinical trial data: a public, standardized database with full auditability.
Avandia (rosiglitazone) is another drug manufactured by GSK, used in the treatment of type II diabetes. In 2007, Avandia was found to lead to a significant increase in risk of myocardial infarction, and some in the medical community claimed  that the FDA moved too slowly to warn doctors and patients about the risks from Avandia, despite having data about the risks as early as 2002. Once again, while the scientific details about the true risks are debatable, it is impossible for physicians to objectively assess the data when it is not available. In the case of Avandia, it is likely that the evasive behaviour of GSK actually did more damage than the results themselves; rather than engaging in a scientific debate about the risks and benefits of Avandia, GSK appeared to have something to hide, leading many doctors to switch patients to different drugs, and Avandia sales have been declining since 2006.
Zetia is a cholesterol-reducing drug marketed by Merck and Schering-Plough. In 2007, the manufacturers came under criticism for failing to disclose evidence that Zetia posed serious risks of liver damage. Although some of the data from the studies showing these risks had been reported to the FDA, neither the FDA nor the companies took any action. Under pressure from the medical community, some of the data were disclosed via press releases. Concerns were also raised that registration of the trials were delayed, and that the primary outcome measures of the trial were modified after the trial was underway. In this case, the data showing risks did not come from a clinical-outcome trial, showing that to protect patients, it is not sufficient to limit regulation to trials with a clinical outcome, as the FDAAA law currently does.
One of the most notorious instances in which patients were put at risk by a drug was Merck’s Vioxx. A series of large clinical trials that showed a significant cardiovascular risk for Vioxx were performed, but the results of the trials were only made available to the FDA and not published for years after they had been completed. Therefore independent researchers were not able to assess cardiovascular risk, and doctors continued to prescribe Vioxx widely. The Vioxx case demonstrates the puzzling fact that the threat of tort liability is not a sufficient disincentive to pharmaceutical companies to prioritize patient safety and to publicize risks as soon as possible. In the Vioxx case, the NEJM accused Merck of withholding certain data from publication that would have demonstrated a significant cardiovascular risk, while publishing the data that supported Vioxx efficacy. If trial data was automatically deposited in public databases, it would relieve companies of these kinds of accusations.
Issues of drug safety risks and clinical trial non-disclosure are not limited to large pharmaceutical companies. For small companies, whose businesses often depend on the development of only one or two products, the pressures to avoid exposing potential risks and failure may be even than greater than with larger companies that are much more diversified. One case in which a smaller biopharmaceutical firm faced difficult disclosure issues was Telik, which was developing a cancer treatment called Telcyta. After announcing at a conference that three of its trials had failed to show positive results and had shown patient deaths, Telik came under criticism from the FDA and investors for failing to disclose the findings earlier .
Finally, TGN1412 illustrates how poor communication of a trial protocol, even for a Phase I study, can place subjects at risk of dire harm. TGN1412 was an immunomodulatory drug being developed by TeGenero Immuno Therapeutics for treatment of chronic lymphocytic leukemia and rheumatoid arthritis. In phase I trials, conducted by a US contract research organization, PAREXEL at a site in England, healthy human volunteers were given a low dose, thought to be much lower than the dose found to be safe in animals. In fact, the low dose caused a disastrous response that made all six volunteers gravely ill and nearly killed them. After the incident, observers argued that the drug should not have been tested in multiple subjects simultaneously, and that the responses might have been predictable. Had TeGenero made their study protocols available for outside review before the trial, these views might have been considered.
The cases described have illustrated the range of problems that are possible when clinical trial data is not accessible. Patients can be placed at risk, because the economic incentives of companies are not aligned with maximizing patient safety. Clinical trial results may be completely suppressed by sponsors, or they may be reported to the FDA but never published. These examples show that solely relying on the FDA to scrutinize plans and results is insufficient. The failures of companies and the FDA to promptly alert consumers of potential risks are not necessarily due to malfeasance; rather, the larger problem is simply that the FDA lacks adequate resources to adequately monitor and respond to all of the trial data to which it has access. Even if there is no deliberate obfuscation or suppression of results, the limited resources at both FDA and at companies make it likely that data will not be sufficiently scrutinized and safety risks might not be promptly discovered.