Background
National agencies, including the Food and Drug Administration (FDA) in the United States (US) and the European Medicines Agency (EMA) in Europe, regulate the marketing of drugs and medical devices. In the US, a drug may only be marketed legally after it receives FDA approval. Once the drug has at least one FDA-approved use, physicians may prescribe it for other unapproved uses, based on their clinical judgment; this is referred to as off-label use. In general, marketing of drugs for off-label uses is illegal both in the US and Europe, although regulation has changed over the years [
1,
2].
Pharmaceutical companies have used various strategies within their legal rights to market drugs for off-label uses [
3]. For example, current FDA guidance states that pharmaceutical companies may use peer-reviewed articles to disseminate evidence of a drug’s effectiveness for off-label uses, as long as certain conditions are met [
4]. Perhaps the most important of the conditions is that the information disseminated must not be false or misleading [
4]. However, there is no regulation that all research findings must be published, and, therefore, a company may choose to selectively disseminate favorable findings. There is now ample evidence that selective reporting of study results, based on the strength and direction of findings (reporting biases), is widely practiced by the pharmaceutical industry [
5,
6].
Gabapentin (Neurontin®, Pfizer, Inc.) was first approved by the FDA for marketing in the US in 1993 as an adjunctive therapy for epilepsy in adults. Subsequently, the drug was approved for use as an adjunctive therapy for epilepsy in children (2000) and for the management of postherpetic neuralgia (2002). Gabapentin was initially marketed by Warner-Lambert’s Parke-Davis division, which was later acquired by Pfizer (thus, hereafter referred to as Pfizer and Parke-Davis).
Shortly after gabapentin's initial FDA approval for use in epilepsy, Pfizer and Parke-Davis conducted a series of “marketing assessments” for several off-label uses of the drug. The “marketing assessments” for each off-label indication provided a detailed analysis of the importance of the clinical condition, current treatment alternatives, market share commanded by each intervention, and a forecast of financial profits associated with two alternatives that were at the company’s disposal to market the drug for each indication — either obtaining FDA approval for marketing associated with a given indication (an indication strategy) or adopting a publication strategy (these are the terms used in the “marketing assessments”). The goal of the publication strategy was to conduct clinical trials and ‘to disseminate the information as widely as possible through the world’s medical literature’ [
7].
To date, internal company documents related to gabapentin have been made publicly available through two separate legal actions. Documents from a whistleblower-initiated government lawsuit involving violation of off-label marketing regulations, settled in 2004 (hereafter referred to as the 2004 whistleblower litigation), have previously been described [
8]. In addition, a large number of previously undisclosed internal company documents were made available through subsequent litigation, brought by consumers and health insurers that paid for prescriptions of gabapentin, involving allegations of fraud (In re Neurontin Marketing and Sales Practices Litigation, MDL 1629, Civil Action NO. 04-cv-10981-PBS, United States District Court, District of Massachusetts) [
9]. Documents related to this litigation (hereafter referred to as the 2008 consumers and health insurers litigation) first became publicly available in 2008. While internal company documents made available in the 2004 whistleblower litigation and the 2008 consumers and health insurers litigation contain overlapping and unique information, the 2008 documents contain considerably new information. Both sources were used for research presented here and in a previous study [
10]. The documents covered by these two lawsuits provided us with a unique longitudinal view of company deliberations and decisions concerning marketing of one of its drugs. We distilled the literally thousands of pages of new and older information into themes of interest to those who study clinical trial conduct and integrity.
The objective of the current study was to describe the implementation of a publication strategy for off-label marketing of gabapentin, within the context of reporting biases and spin of Pfizer and Parke-Davis’s clinical trial findings, for four off-label uses: migraine prophylaxis, treatment of bipolar disorders, neuropathic pain, and nociceptive pain. We use snapshots of internal company documents to illustrate our findings.
Methods
Source documents for this study were provided to the senior author (KD) for preparation of an expert report for the plaintiffs’ lawyers in 2008 as part of the litigation brought by consumers and health insurers against Pfizer. We included information about all trials sponsored by Pfizer and Parke-Davis and documents relating to the four off-label uses revealed by the company during the discovery process that accompanied the litigation. Through communications with counsel involved in the litigation, Pfizer agreed to waive any confidentiality claims concerning all documents examined as part of KD’s expert report. In addition, we retrieved internal company documents that were made publicly available in the Drug Industry Documents Archive (DIDA;
http://dida.library.ucsf.edu) following the 2004 whistleblower litigation [
8]. In general, full versions of examples of internal company documents shown in the Additional file
1 may be accessed from DIDA using the identification number on the bottom of the relevant page. We redacted the names of individuals in the internal company documents provided in the Additional file
1 and excerpts shown in the manuscript. In one case, however, we did not redact the individual’s name since the company personnel referred to the study using the individual’s (principal investigator’s) name and not the study number.
We examined marketing assessments prepared by or for the company for each of the four off-label uses of gabapentin (migraine prophylaxis, treatment of bipolar disorders, neuropathic pain, and nociceptive pain) to identify discussion of a publication strategy as a possible marketing option. We examined over 20,000 pages of internal company communications, memos, and other documents to identify the definition of a publication strategy and details related to its implementation. We also examined the documents for evidence of reporting biases (dissemination of research findings influenced by the nature and direction of results) within our sample of clinical trials.
To identify reporting biases, we compared internal company protocols and research reports with the main publication, if one existed, for each study included in our analysis. The internal company research reports were prepared for or by Pfizer and Parke-Davis at the end of each study and provide documentation of the methods for the trial and results of analyses. Typically, the internal company research reports also included as appendices the study protocol and its amendments, the analysis plan, and output from statistical analyses. We selected one main publication for each study in the following order of priority: full-length publication, letter to the editor, non-systematic review showing a pooled analysis of selected trials, and a conference abstract [
10]. We assessed the documents for appearance of the following reporting biases: (a) publication bias (that is, publishing a study report based on strength and direction of findings), (b) outcome reporting bias (that is, selective reporting of outcomes based on the strength and direction of findings), (c) location bias (that is, in this study, reporting in journals with lower or higher circulation, based on findings), and (d) time lag bias (that is, timing of publication based on the strength and direction of findings) [
6,
11]. Publication bias and outcome reporting bias in the context of the documents we examined for this study have been explicitly explored in a previous article and are not discussed here [
10].
In this study, we also examined control of what was published, assessed the documents for evidence of ghost authorship (that is, failure to report as an author individuals who have made a substantial contribution to the study or the article) [
12], and assessed spin of the study findings. We considered spin to exist when we observed either an explicit description of spinning study findings in the internal company documents or a description in the main publication that appeared to re-frame the study results in order to explain away unfavorable findings or to emphasize favorable findings. Our definition of spin was derived from terminology used in internal company documents and other research related to spin [
13].
We obtained data on circulation of journals publishing the main publication for studies from internal company documents (dated November 28, 2001) for 10/12 trials (see Additional file
1: Figure S9 for an example). For the remaining two trials, we obtained the 2001 journal circulation from other sources: in one case from the 2001 edition of the
Ulrich’s Periodicals Directory[
14], and in the other through communication with the publisher. To check for validity of the data retrieved from
Ulrich’s, we compared the circulations in the internal company documents with those in
Ulrich’s. In general, the journal circulation data were comparable, though not identical.
We examined internal company research reports and published reports for each clinical trial and considered the direction of trial findings to be statistically significant if the primary outcome(s) was reported to be statistically significant or if the P value was less than 0.05. If more than one primary outcome was described in the reports or if there was no distinction between primary and secondary outcomes, we selected the smallest P value favoring gabapentin to determine statistical significance of findings in the trial. To determine the time to publication for each trial, we calculated the time in months between the date issued written on the internal company research report and the date of the published report. For estimating the time to publication, we did not include trials where an internal company research report was not available (n = 2) or where the date of issue was not available (n = 1).
In preparation for litigation and the jury trial, all authors reviewed all documents utilized for this paper. Assessments of reporting biases, spin, and journal circulation were made by SV and KD and discrepancies in the assessments between the two authors were resolved through discussion and reference to the original documents. All authors discussed the findings and agreed upon their interpretation.
We present narrative description of events, counts, and snapshots to depict reporting biases and we illustrate our findings from relevant internal company documents.
Discussion
Parke Davis and Pfizer’s decision to use a publication strategy to market off-label uses of gabapentin was recognized by the courts in the 2004 whistleblower case, and this decision was documented in the biomedical literature in 2006 by an expert witness and his colleagues [
8]. What is new in our study and in the 2008 consumers and insurers litigation is the discovery that various forms of reporting biases and spin were used in conjunction with the publication strategy. We previously reported failure to publish trials with statistically non-significant findings and outcome reporting bias [
10]. In this study we also observed that publication occurred in journals with higher or lower circulations related to statistical significance of findings; delay in publishing statistically non-significant findings; tailoring of publication content to reflect key marketing messages; adding spin to scientific publications such that conclusions favoring gabapentin were emphasized and conclusions that did not favor gabapentin were explained away; and indicators of ghost authorship. Each form of bias, and spin, on its own, could be seen as a relatively minor issue. The value of our findings is in the overall picture that emerges from what appears to be the simultaneous use of many forms of reporting bias and spin, all within the context of a pharmaceutical company’s publication strategy, implemented for marketing purposes.
Our manuscript builds on previous research [
8] by examining longitudinally the initial company decision to adopt a publication strategy for marketing gabapentin and increasing sales, and the consequences of that decision in terms of publication. Our work also provides a framework for distinguishing a publication plan from a publication strategy. A publication plan is typically employed by industry to manage publication of clinical trial findings [
43,
44]. In contrast, a publication strategy involves conducting clinical trials not for the purpose of obtaining regulatory approval to market for a condition but to publish the trial findings in order to market the drug for off-label uses.
Our definition for spin extends work previously done by others on this topic, which used only published literature and focused on trials with statistically non-significant findings [
13]. Our findings provide a more detailed elucidation of strategies for spin, for example, emphasizing favorable findings from outcomes not pre-specified in the trial protocol or from statistically significant secondary outcomes, reaching conclusions that are not consistent with trial design or findings, using extensive rationale to explain away unfavorable findings, and in the case of one trial, explicitly describing attempts to spin trial findings (see Figure
4 and Additional file
1: Figure S7 and Additional file
1: Table S1).
Our study has limitations. Our findings in this study related to reporting biases are influenced by what we identified in the internal company documents provided through the legal discovery process. We include as authors of this article all persons involved in this aspect of the legal case, for transparency. We adopted a case study approach, whereby we described the findings from our observations instead of conducting a detailed, qualitative content analysis of the internal documents. We included in our analysis all available internal company documents. However, there may be other documents that are unavailable to us that could provide additional information regarding issues discussed in this article. We did not examine pharmaceutical trials on other topics or supported by other companies, nor did we examine trials sponsored by not-for-profit entities. Thus, we are unable to say that our findings are uniquely or generally related to industry funding. Indeed, reporting biases have been documented in trials sponsored by both industry and not-for-profit entities [
45,
46]. We did not conduct an exhaustive search for all conference abstracts related to each included trial. Our choice of the main publication for each trial may have affected our observations on reporting biases. However, our criteria for selecting the main publication were based on the published report that provided the most information among all available published reports of the trial. We cannot be certain that the reporting biases and spin we observed are a direct consequence of the discussion and actions reported in the internal company documents we examined. We did not examine study data for the number of prescriptions for off-label uses. Analyses presented as testimony in the 2008 consumers and health insurers litigation indicate that there was a general rise in the number of prescriptions for gabapentin over the time period covered by the publications examined in our study [
47], although it is not possible to establish to what extent publications made an independent contribution to this increase.
On its own, a goal to disseminate research findings via the published literature is broadly applicable, whether the investigators are in academia, industry, government, or elsewhere. In the case of gabapentin, the internal company documents illustrate that as part of implementation of the publication strategy, the company’s marketing personnel appear to have controlled many aspects of publication and presentation of study results, including content (see Figure
7). Although trial investigators outside the company were frequently keen to publish study findings (including situations when tests of gabapentin’s effectiveness were not statistically significant), internal company emails indicate that decisions about the content and timing of journal publication may have been strongly influenced by the company’s marketing personnel.
Pharmaceutical companies have a strong financial incentive to pursue off-label marketing because such marketing can influence physicians’ prescribing practices and, consequently, the companies’ profits. Indeed, published reports of research studies play an important role in marketing by pharmaceutical companies. Documents released during litigation against Pfizer and Parke-Davis describe ideas discussed during training sessions for medical liaisons within Parke-Davis [
48]:
"‘
Notice all the studies we talk about, nothing gets a doc more interested in a drug than a study[
48]
.’"
Most clinical trials testing the effectiveness of drugs are sponsored by two types of entities: the pharmaceutical industry and not-for-profit organizations [
49‐
51], such as the National Institutes of Health in the U.S. The primary motivation for conducting the trials is different for the two entities. The for-profit pharmaceutical industry has a legal fiduciary responsibility to its shareholders, with a corresponding goal of financial profit, achieved either through marketing approval by the FDA or through other marketing strategies. Not-for-profit organizations and their research partners in academia, on the other hand, generally conduct clinical trials in the context of scientific inquiry. The reporting biases we observed represent deliberate dissemination of promotional messages to market gabapentin [
52]. This is not a situation likely unique to gabapentin and Pfizer and Parke-Davis [
53‐
55]; rather, the opportunity to examine internal company documents has provided us with rare insight into industry communication and practices regarding trials included in our study.
The public relies on a regulatory approval system designed to protect them from the marketing of ineffective and potentially harmful drugs. The public is assured that for drugs approved by the FDA, an unbiased examination of safety and effectiveness for the approved indication has been conducted. There is no such system to ensure the safety and effectiveness of a drug used off-label. Because the FDA and public can typically access only the published reports of trials sponsored by a company for off-label uses of a drug, they have to rely on the published literature as a truthful and complete repository of the trial findings. When trials are conducted solely within a publication strategy framework, and the strategy includes reporting findings in a biased manner to further a company’s marketing goals, there is the potential for the scientific record to be distorted.
In addition to standard criminal and civil penalties, several other government policies are in place to hold companies accountable for unethical, off-label marketing practices and prevent future occurrences. Companies that repeatedly conduct clinical trials as part of a publication strategy and intentionally engage in unethical and biased reporting of trial findings could be suspended or even debarred from doing business with the federal government so as ‘to protect the public interest’ and ensure ‘the integrity of Federal programs’ [
56]. Suspension or debarment should be a strong deterrent, because pharmaceutical companies rely for large portions of their revenue from these programs (for example, Medicare, Medicaid) and the sums involved are often far larger than the revenue earned by improperly marketing a single drug. The government may also suspend or debar individual corporate employees [
57], making them unemployable in the industry. This would potentially have a deterrent effect on corporate insiders who are often unaffected by the imposition of large corporate fines. Additionally, under the Park Doctrine, the FDA has the ability to bring misdemeanor charges against high-level corporate officers who were in a position to have corrected or prevented violation of the Federal Food, Drug, and Cosmetic Act (that is, ‘responsible corporate officer’) [
58,
59]. Vigorous application of the Park Doctrine would target high-level officers who are normally harder to prosecute and typically escape liability. Finally, where the improper marketing poses a serious threat to public health, the FDA could require companies to conduct a corrective marketing campaign to fix the misimpression the improper marketing created and formally disavow the erroneous marketing messages the company previously delivered [
60]. Corrective actions could be costly and could erode the revenues reaped by fraudulent marketing.
Competing interests
KD served as an expert witness for the plaintiffs’ lawyers in litigation against Pfizer that provided several source documents used for this study. Funds that would have been paid to KD for serving as an expert witness were donated by the plaintiffs’ attorneys to Johns Hopkins for use in scholarly endeavors related to reporting biases. SV was paid by the plaintiffs’ attorneys in litigation against Pfizer for assistance provided to KD for her work as an expert witness. SV’s work on this manuscript and other ongoing research is being supported using the fund established at Johns Hopkins. IR, PG, and TG are affiliated with Greene LLP, a law firm that represents and has represented various clients in litigation against Pfizer, including the cases in which KD served as an expert witness and other cases referred to in this manuscript. The firm has earned legal fees from a case against Pfizer in the past, and the firm is currently involved in other litigation that may result in additional fees.
Authors’ contributions
Attorneys for the plaintiffs (PG, IR, and TG) provided internal company documents to KD for preparing her report related to reporting biases. Over the course of preparation of the report, and as part of the subsequent research we are conducting, KD and SV requested, and received, further information from plaintiffs’ attorneys related to study conduct, dissemination of findings, and internal discussions about related clinical trials of gabapentin. All authors reviewed all documents used for this study. KD and SV extracted data and assessed the documents for reporting biases, “spin,” and journal circulation. PG, IR, and TG provided insights that facilitated examination of the scientific issues in the context of marketing goals. All findings and interpretations were discussed and agreed among all authors. SV wrote the first draft of the manuscript and KD, PG, IR, and TG critically revised and edited subsequent versions of the manuscript. All authors read and approved the final manuscript.