Eric Winer is a consultant for GlaxoSmithKline, Bristol-Meyers Squibb, Sanofi-Aventis, Boerringer-Ingleheim, Berlex, Genomic Health, Genentech, Wyeth, Pfizer, and Astra Zeneca and receives research funding for clinical trials from Genentech, GlaxoSmithKline, Astra Zeneca, and Pfizer.
Ann Partridge MD, MPH
Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, Massachusetts
Presented as Abstract 6057 at the Annual Meeting of the American Society of Clinical Oncology, May 13-17, 2005, Orlando, Florida.
Since 1992, pharmaceutical industry support has surpassed National Institutes of Health funding for clinical research in the United States. In this study, the authors sought to evaluate the impact of this shift in funding from public to private sponsors on the nature of published breast cancer clinical research.
All published breast cancer clinical trials from 2003, 1998, and 1993 were reviewed from 10 select English-language medical journals to evaluate pharmaceutical involvement over time and the association between sponsorship, trial design, and results. Clinical studies that reported disease-specific outcomes from medical therapy for breast cancer were eligible for analysis. Pharmaceutical involvement was defined as reported pharmaceutical industry funding, provision of drug, and/or authorship for each publication.
In total, 140 eligible studies were identified, including 45 studies that were published in 1993, 39 studies that were published in 1998, and 56 studies that were published in 2003. Among those, 67 publications (48%) reported pharmaceutical industry involvement, 36 publications (26%) had at least ≥1 pharmaceutical industry author, And 100 publications (71%) were considered positive. Pharmaceutical involvement was identified in 44% of the studies published in 1993, in 38% of the studies published in 1998, and in 58% of the studies published in 2003. Pharmaceutical authorship was reported in 22% of the 1993 studies, in 21% of the 1998 studies, and in 34% of the 2003 studies. For studies that were published in 2003, those that reported pharmaceutical involvement were more likely to be positive (84% vs 54%; P = .02; Fisher exact test), to be single-arm studies (66% vs 33%; P = .03), and to evaluate metastatic disease (72% vs 46%; P = .06).
In 1992, research and development spending by pharmaceutical companies exceeded the research and development spending by the National Institutes of Health (NIH) for the first time.1 Subsequently, pharmaceutical spending for research and development has continued to increase and outpace NIH spending.2 A major focus of pharmaceutical industry research is the development of new cancer therapies: Half of all pharmaceutical company-sponsored drugs under development are designed to fight cancer.2 This strong interest in cancer therapy, combined with the trends in research funding, suggests that clinical cancer research and the resulting changes in clinical care increasingly may be influenced by the pharmaceutical industry-sponsored research.
In recent years, increased investment in oncology research, including collaborations between academic researchers and the pharmaceutical industry, has yielded new, more effective therapies to fight cancer.3, 4 However, increasing investment and influence of the pharmaceutical industry has raised concerns regarding conflicts of interest and the integrity of the design, conduct, and reporting of clinical trials.5–7 These concerns have led to greater oversight and scrutiny of the clinical trials process and reporting of results.8 Medical journals have established more stringent requirements for disclosure of financial incentives,9 and a national clinical trial registry has been established, in part to promote awareness of unpublished studies and to allow greater scrutiny of trial conduct and outcomes.10
The growing importance of pharmaceutical industry involvement in clinical research has led to efforts to better understand its impact. Previous studies evaluating nononcologic research have revealed that 1) pharmaceutical sponsorship of clinical research is increasing over time11, 12 and 2) pharmaceutical-sponsored studies are more likely to report positive results than studies without clear industry connections.13–15 In oncology, there has been only 1 published study evaluating the impact of pharmaceutical involvement on clinical research. This research demonstrated an association between pharmaceutical sponsorship and positive results in clinical trials for multiple myeloma.16 Several additional studies evaluating the economics of cancer therapy also have suggested that the source of funding may affect research design and outcome.17–19
Breast cancer is the most common cancer in American women and affects >200,000 women in the United States20 and as many as 1.1 million women globally each year.21 Thus, the factors that affect the nature and quality of breast cancer research may have a large impact on public health. To our knowledge, there has been no published research to date regarding the correlation between pharmaceutical sponsorship and breast cancer clinical research.
For the current investigation, we sought to identify the extent of documented involvement of pharmaceutical companies in published clinical breast cancer research and to explore the correlation between pharmaceutical company involvement and study design and outcome. We also evaluated changes in documented pharmaceutical company involvement in published breast cancer clinical trials over time.
MATERIALS AND METHODS
We conducted a systematic audit of breast cancer clinical trials that were published in the years 1993, 1998, and 2003 in 10 select English-language medical journals to evaluate the correlations between pharmaceutical company involvement, study design, and study outcome and to explore changes in these areas over time.
Eligible journals were selected based on the frequency of breast cancer publications in selected years, as determined by a search of the MEDLINE database using the index terms breast cancer and clinical trial, and expert opinion regarding top journals for breast cancer clinical research (see Table 1). All published articles from the selected journals that reported original clinical research involving medical therapy for the treatment of breast cancer and that reported results for ≥1 disease-specific outcome were included in this analysis. Phase I trials, trials of supportive care agents, and trials that tested surgical or radiotherapy interventions were excluded. Studies were identified by a hand search of all volumes of the chosen journals in the selected years; this was supplemented by a MEDLINE search using the year of publication, journal name, and index search terms to identify potential articles that were missed during the hand search. Studies were screened for eligibility by reviewing the title and abstract.
Table 1. Trial Characteristics
Year of publication: No. of articles (%)
Total, N = 140
1993, N = 45 (32%)
1998, N = 39 (28%)
2003, N = 56 (40%)
J Clin Oncol
Breast Cancer Res Treat
Eur J Cancer
Br J Cancer
J Natl Cancer Inst
Lancet, JAMA, N Engl J Med
Randomized, Phase II
Randomized, Phase III
A study-specific abstraction tool was developed and piloted using breast cancer clinical trials from nonindex years. A medical oncologist (J.P.) reviewed and abstracted relevant data from each eligible publication. We also randomly selected 10% of eligible articles to undergo repeat audit and abstraction by another oncologist (A.P.) to determine the potential for miscoding or disagreement over abstraction items. For 6 select items (pharmaceutical involvement, pharmaceutical authors, study outcome, cancer stage, study design, and size of trial) among 14 articles (≈10%) there was agreement for 82 of 84 items for a concordance rate of 97.6%. It was determined on review that the 2 differences represented miscoding.
Definitions of Major Outcomes
Pharmaceutical versus nonpharmaceutical studies
Any indication of pharmaceutical industry involvement in an article, including the provision of drugs, funding of the study or the investigator, or ≥1 pharmaceutical industry author(s), meant that it was categorized as a pharmaceutical study (PS). Lack of such documented indication of pharmaceutical support meant that the publication was classified as a nonpharmaceutical study (NPS).
Pharmaceutical industry authorship
Reported author affiliations were reviewed. Studies that reported ≥1 author(s) who were affiliated with a biotechnology company or pharmaceutical company were classified as having a pharmaceutical industry author. Disclosures of potential conflicts of interest through grant support, consultant relationships, or other connections with the pharmaceutical industry were not classified as pharmaceutical authorship.
Studies were classified as positive or negative based on the authors' conclusions regarding the safety and efficacy of the intervention compared with standard therapy. For randomized controlled trials, we classified trials as positive or negative and recorded the strength of the authors' conclusions by using the 6-point scale developed by Kjaergard and Als-Nielsen.22 This scale allows for the abstraction not only of the direction of the authors' conclusion (positive or negative) but also of the strength of statements regarding the merits of the experimental intervention versus the control. For a random subset of 20 randomized controlled trials from our sample, study outcomes were abstracted by 2 independent reviewers (A.P. and J.P.) using the 6-point scale. There was agreement between reviewers on the exact score for 9 of 20 trials (45%). Using scores of 4 through 6 as evidence of a positive trial and scores of 1 through 3 as evidence of a negative trial, there was 100% agreement between the reviewers regarding whether the trial was positive or negative. Thus, we used the 6-point scale for abstraction purposes but dichotomized the results from the scale as positive or negative only for our analyses. For uncontrolled and for randomized Phase II studies, we abstracted the authors' conclusions regarding the results in terms of the safety and efficacy of the experimental agent as it was used in the trial. Because of the varied ways in which sponsorship may be indicated in published articles, we did not believe that blinding to the funding source was feasible.
Among all of the studies in our sample, descriptive statistics were used to characterize the differences in studies with and without pharmaceutical involvement. The Fisher exact test was used to test the association between pharmaceutical involvement and study characteristics. The Cochran-Armitage test for trend was used to determine whether there was a trend in the proportion of trials that reported pharmaceutical involvement across the 3 years we studied (1993, 1998, and 2003). Because we believed that the reporting of pharmaceutical sponsorship was most stringent in later years, we focused our formal statistical hypothesis tests on the data from trials in 2003.
We identified 140 eligible articles from 10 journals. The journals and characteristics of the articles are presented in Table 1. Forty-five studies were identified from 1993, 39 studies were identified from 1998, and 56 studies were identified from 2003. Based on our definitions of major outcomes, 67 studies (48%) were categorized as PS, and 37 studies (26%) reported pharmaceutical industry authorship. One hundred studies (71%) reported primarily positive results, and 40 studies (29%) reported negative results. The majority of studies were single-arm trials (55%) and/or concerned the treatment of advanced disease (69%).
Differences in Studies According to Sponsorship
A combined analysis of all studies from 1993, 1998, and 2003 demonstrated no significant difference in the percentage of PS reports that were positive compared with NPS reports (78% [52 of 67 studies] vs 66% [48 of 73 studies]; P = .14). Seventy-nine percent of PS reports involved the treatment of advanced disease versus 60% of NPS reports (P = .02). In addition, 55% of PS reports were from single-arm studies (vs randomized Phase II or III studies) compared with 41% of NPS reports (P = .13). Among all of the articles with documented pharmaceutical authorship, 29 of 37 studies (78%) reported positive results compared with 71 of 103 studies (69%) without pharmaceutical authorship (P = .30).
Because the requirements for documentation of pharmaceutical involvement in published clinical research changed during the period evaluated, we conducted additional analyses that focused on the most recent year (2003) (see Table 2). In 2003, 57% of all articles (32 of 56 studies) were PS, and 84% of those studies reported positive results compared with 54% of NPS (P = .02). PS were more likely to report a single-arm, nonrandomized study design (66% vs 33%; P = .03), and PS trials trended toward being more likely to involve treatment of advanced-stage rather than early-stage disease (72% vs 46%; P = .06). In addition, PS appeared more likely to involve experimental hormone or biologic therapies (50% vs 33%; P = .28), but this difference did not reach statistical significance. Eighty-four percent of the 2003 studies with pharmaceutical authorship were positive compared with 62% of studies without pharmaceutical authorship (P = .18).
Table 2. Characteristics of Breast Cancer Clinical Trials Published in 2003 by Sponsor
No. of trials (%)
NPS, N = 24
PS, N = 32
NPS indicates nonpharmaceutical study; PS. pharmaceutical study.
Stage IV disease
Hormone or biologic therapy
Size of trial <100 patients
Changes Over Time
The percentage of PS, including those with pharmaceutical authorship, appeared to increase over the years studied, although the trends were not statistically significant (see Fig. 1). The published studies that reported positive results in all 3 years are presented in Figure 2 according to pharmaceutical involvement and pharmaceutical authorship. In all periods, the majority of both PS and NPS were positive. We also noted several general temporal trends among breast cancer trials. The percentage of randomized trials increased from 36% in 1993 to 48% in 2003 (P = .23). Hormone therapy trials increased from 24% of all trials in 1993 to 33% of all trials in 2003 (P = .3); and, by 2003, 8% of trials involved biologic therapy compared with 0% in 1993 (P = .06). The percentage of studies that enrolled >200 patients also increased over time. In 1993, only 20% of trials involved ≥200 patients; and, by 2003, 57% of trials involved ≥200 patients (P = .0002). The percentage of trials that involved adjuvant or neoadjuvant therapy increased from 16% in 1993 to 34% in 2003 (P = .04).
Pharmaceutical industry investment in research exceeds the entire operating budget of the NIH: It is rising, and it is focused on the development of cancer therapies. Increased pharmaceutical sponsorship of clinical research also has been noted among published stroke trials12 and among randomly selected trials from 5 medical journals.11 Consistent with these trends, we observed that the majority of the most recently published breast cancer trials that were included in our study were sponsored or supported by the pharmaceutical industry. Whether the increased prevalence of reported pharmaceutical involvement over time reflects increased involvement in breast cancer clinical research, more stringent guidelines for disclosure, or both remains unclear. In any event, the finding that the majority of published studies now involve some degree of collaboration with the pharmaceutical industry makes it important to understand the influence that such industry involvement may have on the nature and direction of breast cancer research.
To date, the full consequences of pharmaceutical industry involvement in clinical research are not well understood. We observed that clinical studies involving the pharmaceutical industry may differ significantly from those that do not involve the pharmaceutical industry. Despite our small sample size, in the subset of 2003 trials, we detected differences both in study design and in study outcome that may have implications for patients, clinicians, and policy makers. Documented pharmaceutical involvement was associated with smaller trials, single-arms studies, and studies that focused on advanced disease. We also observed that trials published in 2003 with pharmaceutical involvement were significantly more likely to report positive results than trials without pharmaceutical involvement.
The finding that the pharmaceutical industry may be more inclined to support or design smaller, single-arm studies for patients with advanced breast cancer likely reflects their important role in the development of novel therapeutics. However, there are many important questions regarding breast cancer therapy that go beyond identifying active new drugs. For many proven therapies, the optimal timing, sequencing, and duration of therapy are unclear. Similarly, identifying subgroups of patients who are most likely to benefit or develop toxicity from available therapies and understanding the long-term outcomes from current therapies are critically important for clinical decision making. Pursuing these questions may not always be in the financial interests of companies that are selling approved therapies. If these types of questions are supported by other funding agencies, then the care of women with breast cancer will continue to improve. However, if increased investment by industry leads to curtailed investment from other funding sources and if pharmaceutical involvement increasingly shapes the research agenda, then critical questions may be neglected. Many of these questions have great practical importance and may have important financial implications for the health care system (eg, the optimal duration of treatment for an expensive new drug). It will be important to continue to monitor trends in breast cancer research over time and to be sure that important clinical questions are not overlooked as pharmaceutical involvement in clinical research continues to increase.
Our finding that, for 2003 studies, there was a correlation between pharmaceutical industry involvement in breast cancer research and the publication of positive trial results is consistent with prior research on the impact of sponsorship on clinical research results. The association was noted first by Davidson et al in 1986, who demonstrated that industry-sponsored, randomized controlled trials were significantly more likely to report positive results than nonindustry-sponsored trials.23 More recently, Bekelman et al conducted a systematic review of studies that investigated the impact of financial conflicts of interest on biomedical research.14 Pooled analyses of 8 studies specifically addressing the impact of sponsorship on research outcomes demonstrated that industry sponsorship was correlated with positive study outcomes or proindustry conclusions with an odds ratio of 3.60 (95% confidence interval, 2.63–4.91). In oncology, Djulbegovic et al. reported that randomized controlled trials for multiple myeloma that were funded by for-profit organizations were more likely to favor the experimental treatments over control treatments compared with randomized controlled trials that were sponsored by nonprofit organizations (74% vs 47%).16 Three additional studies that evaluated the economic impact of trial sponsorship in oncology documented a similar association between industry sponsorship and positive results.17–19 It is important to note that we failed to detect a significant difference between positive outcomes among PS versus NPS when we compared trials in all 3 time periods in our analysis (78% positive among all PS vs 66% positive among all NPS; P = .14). Further research may be needed to clarify whether this discrepancy between our combined analysis and both our 2003 data and prior publications was caused by misclassification bias during earlier periods for our sample or by a change in the association between pharmaceutical involvement and outcomes in breast cancer trials over time.
We evaluated the association between PS, single-arm versus multiarm study design, and study outcomes for all years and for the 2003 subsample (Fig. 3). The significant difference in positive outcomes among multiarm trials between PS and NPS in 2003 suggests that increased prevalence of single-arm studies among PS is not the sole explanation for the association between PS and positive study outcomes during this period.
There are multiple potential explanations for the association between pharmaceutical industry involvement and positive study outcomes. It is possible that pharmaceutical trials, whether by development and selection of better drugs or greater focus on established therapeutic targets, may be more likely on average to yield positive results than the experimental interventions tested in nonpharmaceutical trials. These possibilities have been overlooked in many discussions of the industry-positive study connection.12, 14, 16 Alternatively, the pharmaceutical industry may be more likely to support study designs that are deemed more likely to yield positive results. In randomized controlled trials, this could involve the selection of an inferior control arm or setting up a straw horse to make the experimental therapy look better, as suggested by the number of placebo-controlled PS trials versus publicly funded trials in the study by Djulbegovic et al.16 We reviewed all randomized controlled trials from 2003 in our sample and observed that the control arms of all trials reflected appropriate standards of care. Only 1 study compared an experimental therapy with a placebo control; and, in that setting, the standard treatment was no active therapy, which meant that a placebo was entirely reasonable.24 However, because the biggest difference between PS reports and NPS reports occurs among randomized controlled trials, as detailed in Figure 3, it remains plausible that PS randomized controlled trials are conducted when there is a high level of certainty from some aspect of the trial design or from the selection of control arm therapy that the trial will be positive. An evaluation of the differences between PS randomized controlled trials and NPS randomized controlled trials is indicated for future research. It should be noted that many PS randomized controlled trials may be registration trials, and study design and selection of control arms for these studies will be influenced by U.S. Food and Drug Administration requirements for drug development.
It is also possible that the correlation between positive results and industry funding is an artifact of differential publication. It is well established that, in general, positive studies are more likely to be published than negative studies (publication bias).25 PS with negative results may be even less likely to be completed, submitted, or published than negative NPS because of early termination of trials or sponsor control over publication.7 Future research should seek to identify the reasons for this association and the relative contributions of different factors.
The recently initiated national trial registry may permit an evaluation of the number of trials that are stopped early or that are completed but never published because of negative results and the degree to which this number varies based on sponsorship. It will be important to make the information contained in these registries available to researchers who wish to untangle the correlations between sponsorship, research outcomes, and publication.
There were several limitations to the current study. We relied on documentation of pharmaceutical involvement to classify a study as PS versus NPS. Particularly in the earlier periods sampled, when disclosure requirements were less stringent, it is possible that this led to misclassification bias. Because we classified any report without documentation of pharmaceutical industry involvement as NPS, such misclassification likely would result in an under estimation of the percentage of trials with pharmaceutical involvement. A further limitation was the relatively small number of studies identified in each of the 3 years. Identification of all published breast cancer trials in a given year or expansion of time periods to 2-year blocks would have increased our sample size. Future research should focus on a specific subgroup, such as Phase II studies or adjuvant trials, and on longer time periods and/or articles from all journals. There is no established method for evaluating trends over time in published studies. Our selection of 3 time periods was arbitrary, and other methods could yield different results. Our objective was to make this a hypothesis generating analysis, and the trends suggested should be evaluated in larger studies. We selected the most recent year with complete data at the time of study conception in early 2004 and wanted to evaluate changes over a decade at 5-year intervals, which led to the selection of 1993, 1998, and 2003 as the years of analysis for the study. The major strength of this study is that, despite the relatively small sample size, we observed important differences between published clinical trials that involved the pharmaceutical industry compared with studies that did not report pharmaceutical involvement. These results are supported by other studies in this area. To our knowledge, this is the second study to explore the question of industry impact on clinical research in oncology and the first in the area of breast cancer research.
In conclusion, the impact of the growing pharmaceutical industry involvement in breast cancer clinical research appears similar to the impact of industry sponsorship documented in other fields of medical research. The majority of clinical research currently is associated with the pharmaceutical industry, and this association appears to have an impact on research outcomes and may shape study design. What this association means for the future of breast cancer research is unclear. It may yield better therapies for treatment of breast cancer but, at the same time, may focus research on some clinical problems while neglecting others. Given the importance of breast cancer as a public health problem and the current necessity of the pharmaceutical industry as a supporter of clinical breast cancer research, further study is needed to evaluate and understand the impact of industry involvement on research design and outcomes and, ultimately, to determine the potential impact on patient care.