Meta-Analysis Concludes Angiotensin Receptor Blocker Use Increases the Risk of Developing Cancer: Concerns About the Science and the Message
Address for correspondence: Michael J. Bloch, MD, Risk Reduction Center, Saint Mary’s Regional Medical Center, 645 North Arlington Street, Suite 460, Reno, NV 89503 E-mail:email@example.com
On June 14, 2010, a new meta-analysis by Sipahi and colleagues1 that examined the potential relationship between angiotensin receptor blocker (ARB) use and risk of cancer was published online by Lancet Oncology. The accompanying editorial commentary by Nissen entitled “Angiotensin-Receptor Blockers and Cancer: Urgent Regulatory Review Needed” concluded that, pending further regulatory review, “we should use ARBs, particularly telmisartan, with greater caution. These drugs are often overprescribed as a result of aggressive marketing and in the absence of evidence that they are better than angiotensin-converting enzyme (ACE) inhibitors.”2
Within hours of online publication, this story was picked up by a number of major news organizations, including ABC news, CBS news, The Wall Street Journal, and USA Today.3–6 While the televised national news stories were often balanced and did highlight the potential weaknesses of meta-analyses, the printed headlines were anything but; for example, The Wall Street Journal titled their online story: “Some Blood Pressure Drugs Tied to Cancer Risk,” and other headlines were similarly declarative. Many of us can remember past headlines, all of which turned out to be untrue, suggesting reserpine, calcium channel blocker, and diuretic therapy caused cancer.
Citing concerns, based on this reporting, that patients may suddenly stop taking their antihypertensive medications, many leaders of the national and international hypertension specialist community quickly went on record denouncing the methodology of the meta-analysis and, in some cases, questioning the integrity of the authors, the editorial commentary, and the Lancet Oncology editorial staff.7
Clearly, this was not the most effective way to communicate new medical information to the public, nor did it represent the best means of evaluating the strength of the evidence. To determine what went wrong and the best way forward, we need to critically examine the strengths and weaknesses of the published meta-analysis; determine what, if any, further steps are needed to clarify whether there is a relationship between ARB use and cancer incidence; and try to find lessons in communicating study results to media outlets.
For their meta-analysis, Sipahi and coworkers included all publicly available randomized controlled studies for all seven clinically approved ARBs, published before November 2009, that had a mean follow-up of more than 12 months, involved at least 100 patients, had a control group, and reported cancer data. They searched both online clinical trial databases and obtained information from the public Web site of the US Food and Drug Administration (FDA), which posts data not necessarily included in the originally published manuscripts from each study.
The primary objective was to determine the effect of ARB use on new cancer occurrence, with data available on 61,590 patients from five trials (the Losartan Intervention for Endpoint Reduction in Hypertension [LIFE] trial, the Trial for Preventing Hypertension [TROPHY], the Telmisartan Randomized Assessment Study in ACE Intolerant Subjects With Cardiovascular Disease [TRANSCEND], the Ongoing Telmisartan Alone and in Combination With Ramipril Global Endpoint Trial [ONTARGET], and the Prevention Regimen for Effectively Avoiding Second Strokes [PROFESS] trial). There were two separate secondary end points. One examined the effect of ARB use on cancer-related deaths, for which data were available on 93,515 patients from eight trials (LIFE, the Candesartan in Heart Failure: Assessment of Reduction in Mortality and Morbidity—Overall [CHARM-Overall] trial, TRANSCEND, ONTARGET, PROFESS, the Optimal Trial in Myocardial Infarction With the Angiotensin II Antagonist Losartan [OPTIMAAL], the Valsartan in Acute Myocardial Infarction [VALIANT] trial, and the Valsartan Heart Failure Trial [Val-HEFT]). The other secondary end point examined the incidence of specific types of solid organ tumors (lung, prostate, and breast), for which data were available on 68,402 patients from five trials (LIFE. CHARM-Overall, TRANSCEND, ONTARGET, AND PROFESS).
Telmisartan was the study drug in 86% of the studies evaluated. The primary finding was that the use of ARBs, compared with active control (other antihypertensive agents) or placebo, was associated with a modest, but statistically significant, increase in the risk of incident cancer (7.2% for patients receiving ARBs compared with 6.0% for those not receiving ARBs) (risk ratio [RR]=1.08; 95% confidence interval [CI], 1.01–1.15; P=.016). The strongest data, however, come from three of the five clinical trials that examined new cancer incidence as a prespecified end point, LIFE, ONTARGET, and TRANSCEND (evaluating only telmisartan and losartan). In this analysis, there was an 11% increase in lung cancer risk with ARB use compared with control (RR, 1.11; CI, 1.04–1.18; P=.001). Importantly, there was not an increased risk of cancer deaths with ARBs.
The limitations of meta-analyses are widely known. As opposed to prospective randomized clinical trials where investigators pre-specify their methodology, meta-analyses are retrospectively conducted and are prone to “cherry picking” inclusion criteria and end point definitions that often bias the results and lead to misleading and unreliable conclusions relating to questions that the original studies were never designed to answer. In addition, they often become fishing expeditions, based simply on the laws of probability, where one out of several variables examined may show a statistically significant relationship where none truly exists. Meta-analyses are best when the investigators make every effort to obtain patient-level data and include all relevant studies. We agree with the authors that the lack of patient-specific data, including the age, sex, and especially the smoking status of individual enrollees, greatly limits this meta-analysis.
To put things in perspective, the risk of lung cancer from common activities such as smoking is much greater than the potential risk from any antihypertensive agent and this information was not available. In addition, the lack of data on the time of occurrence of the cancers does not allow a time-to-event analysis on cancer occurrence. At most, the present data suggest that ARBs may unmask an already-developing cancer rather than causing de novo cancer, as the duration of treatment and follow-up is relatively short (mean follow-up, 1.7–4.8 years). There certainly is not adequate time for cancers to develop de novo, which from the time of initial mitotic occurrence to their clinical presentation often takes between eight and 16 years in the case of lung cancer, for example.
In addition, it does not appear that every effort was made to include all relevant data that could have been made available to the authors. In particular, the authors have been criticized for not including data from the Valsartan Antihypertensive Long-Term Use Evaluation (VALUE) study, which may have changed their results.7 While the authors may not have had access to unpublished information from additional trials, personal phone calls to the published investigators of those ARB trials who have access to specific patient-level data can often be shared and should have occurred. While all information should ideally be available and published on the Web, this does not always occur, especially when a question now being asked was not part of the original study hypothesis.
Given these shortcomings, should this analysis have been published? Contrary to the opinions expressed by some in the hypertension community, we agree with the publication of these results but object to the choice of editorial commentator and subsequent media attention that the publication and editorial garnished. As early as 2003, the CHARM study reported an increase in cancer death with the use of candesartan,8 which, at the time, was felt to be a statistical anomaly.
While experimental studies implicate the angiotensin II type 1 and type 2 receptors as important in cellular proliferation, angiogenesis, and tumor progression (suggesting potential mechanistic plausibility), preclinical studies with ARBs in both rats and mice have not been associated with carcinogenicity. The authors were clearly aware of this information. Yet, they performed a hypothesis-generating meta-analysis, which some might argue is a reasonable way to address this question. But the paper and editorial conclusions were too strong, suggesting that ARBs be stopped as antihypertensive agents, based on the information published.
We feel that it is critically important that all information gathered from industry-sponsored trials be published or made available to the scientific community. For example, we find it worrisome that the ONTARGET publications did not include cancer data that was apparently made available to the FDA. While there has been a modest imbalance in malignancies in the recently completed outcome studies with telmisartan, as the present authors note, it was primarily in the telmisartan/ramipril combined arm of ONTARGET, as there was no difference when comparing the telmisartan vs ramipril monotherapy arms. Accordingly, we feel that the accompanying commentary should have been more cautionary and discriminating in its conclusions.
Finally, the choice of author for this commentary was potentially problematic since it has been reported that Sipahi was a former student of Nissen, the author of the editorial commentary. The editors of major journals need to exercise greater caution when selecting the author of an editorial, providing appropriate and balanced commentary and, when necessary, presenting even more than one viewpoint.
In conclusion, we think that this situation highlights the need for us to better educate the mass media about levels of evidence. The fact that the publication of this finding from a study with a low weight of evidence was given such widespread national attention is troubling. According to well-established rules of evidence-based medicine, studies such as this should never dictate clinical care and should remain hypothesis-generating. Perhaps the American Society of Hypertension (ASH) should develop a panel of experts that could be available to help the media appropriately evaluate the strength and significance of publications related to hypertension.
We feel that based on the evidence presented to date, there is no need for physicians to change their prescribing of ARBs or for any regulatory change to occur at this time. We do recommend the following steps be taken to clarify the issue of a potential link between ARBs and incident cancer: (1) a better-designed patient level meta-analysis of all relevant ARB studies, including VALUE, be performed; (2) administrative claims databases such as those available from the Veterans Administration (VA) should be examined to determine whether there is an association between ARB use and incident cancer; (3) any clinical investigators with cancer-related data and ARB use should contact the authors of the present study so they can perform an updated meta-analysis (submitted as a Letter to the Editor of Lancet Oncology); and (4) future trials using ARBs should collect cancer data as a pre-specified end point of interest. Societies such as ASH need to work with the media to properly position and explain the results.
Disclosure: Dr Bloch wishes to disclose that he has received research support, consulting fees, and/or honoraria from Novartis, Takeda, and AstraZeneca, each of which investigate/market an ARB. Dr Basile has received research support from Novartis and consulting fees and/or honoraria from AstraZeneca, Boehringer Ingelheim, Daiichi Sankyo, Novartis, and Takeda.