Author + information
- Robert M. Califf, MD∗ ()
- Duke Forge, Duke University School of Medicine, Durham, North Carolina
- Department of Medicine, Stanford University, Stanford, California
- Verily Life Science, South San Francisco, California
- ↵∗Address for correspondence:
Dr. Robert M. Califf, Duke Forge, Duke University School of Medicine, 40 Medicine Circle, Durham, North Carolina 27710.
Do fluoroquinolone antibiotics increase the systemic risk of medical problems, including aortic and mitral regurgitation, due to effects on collagen and the extracellular matrix? In this issue of the Journal, Etminan et al. (1) produce additional evidence supporting the view that we can add aortic and mitral valve disease to the list of manifestations of fluoroquinolone toxicity. Using a combination of methods, they analyzed data from the U.S. Food and Drug Administration’s (FDA’s) Adverse Event Reporting System database to conduct a disproportionality analysis to identify cases. They then used the PharMetrics Plus database to construct a nested case-control study of current and recent fluoroquinolone use that found an adjusted risk rate ratio of 2.40 compared with amoxicillin, and 1.75 when compared with azithromycin for “current users.” They also noted less impressive but clinically important point estimates when current and recent users were combined in the analysis. One particularly alarming finding is that valvulopathy appears acutely after starting the course of fluoroquinolones, consistent with animal models cited by the authors.
If these estimates are accurate, then given the serious nature of regurgitant valvular lesions, the existing guidance that alternative antibiotic regimens should be used when possible is further reinforced. On the other hand, if the many known complexities and limitations of observational analyses of spontaneous adverse event reports have led to an erroneous estimate of harm when none exists, a highly effective class of antibiotics may have been inappropriately relegated to “last resort” status. The authors present a spirited defense of their methods and only acknowledge that “a number of possible limitations may also be present.” Although they overstep a bit by implying that the large sample size and the depth of clinical detail of the database enables the researchers to overcome bias by allowing them to control for possible confounders, they call for additional studies to confirm or refute their finding. There seems to be a common misperception that a larger database reduces bias.
There is significant evidence supporting a role for fluoroquinolones in other connective tissue problems (2), most notably rupture of the Achilles tendon and retinal detachment, and the authors cite a smattering of peer-reviewed published data finding an effect on connective tissue in a remarkable short interval after exposure to fluoroquinolones. In addition, these drugs are known to prolong the QT interval on the electrocardiogram and to cause the lethal arrhythmia torsades de pointes (3).
What does it say about our system when a class of drugs—one that has been on the market for over 20 years—continues to have new, concerning information emerge in the face of 32.5 million outpatient prescriptions in the United States in 2015 alone (4)? The FDA and European Medicines Agency have both recently updated warnings for prescribers (5,6), but we should all hope for a system that exposes these risks in a quantitative fashion either before application for marketing, or quickly after a drug is marketed for risks with a small effect size; similarly, a quality system should also be in place so that critical decisions can be made by balancing benefits and risks. The low frequency of the risk of valvulopathy points out the danger of complacency when a drug or device is in widespread use and no major risks have been identified. In the case of fluoroquinolone use, when a patient develops mitral or aortic regurgitation, there has been no prior reason for a clinician to suspect that the antibiotic may have been the etiologic factor. Low-frequency risks that could plausibly be due to multiple factors can only be evaluated by a systematic, large-scale study.
Spontaneous adverse events reports are like many important parts of life—it seems that we cannot live with them and we cannot live without them. Their limitations have been described in detail in other publications. Most notably, only a small fraction of adverse events are reported. When they are reported, except in rare, specific situations, causal relationships cannot be inferred from the temporal association of an adverse outcome with taking a drug. When researchers or regulators attempt to sift such data, they are extraordinarily limited by the high risk of bias in reporting and by a high degree of missingness in the data.
Why has so little work been done on the underlying biology of fluoroquinolones and their putative effects on connective tissue? Unfortunately, the greatest concentration of expertise about a given drug often resides within the company that develops and markets it. Some argue that this provides an incentive to investigate clues that adverse events may be seen in the clinic because a company would want to prevent later liability. However, others believe that given the complexity of the biology in play in systemic drug administration, and the many cases in which industries downplay risks of their products, the ability to independently pursue evidence suggestive of potential toxicities is needed. The National Institutes of Health (NIH) rarely funds such investigations, and the FDA has limited capacity to perform this type of research, although its recent study of sunscreen (7), prompted by the failure of industry and the clinical community to explore potential risks (8), received significant interest. In any case, the push for transparency and data sharing in research should be extended into the preclinical arena, taking into account a reasonable need to protect trade secrets and intellectual property. A more robust federal research capacity across the FDA and NIH should also be considered.
If we are to prevent future failures similar to those that have left us ill-equipped to define the benefit-risk profile of fluoroquinolones, we must accelerate ongoing work to create a different information environment for biomedicine and medical practice in general, and specifically for the development and evaluation of drugs and devices. No amount of exhortation of clinicians will lead to high levels of reporting of adverse events, and no amount of statistical manipulation will provide the risk estimates needed to properly balance risks and benefits when decisions are made about the use of a drug or device in populations or individuals. Instead of a system dependent on arbitrary decisions about what to report and when, with only rough approximations of numerators and denominators, we need a system in which rare and idiosyncratic catastrophic toxicities are reported, and one that also enables measurement of common risks and benefits in the context of practice with solid measurement of numerators and denominators. The context can be provided by a compilation of claims and electronic health record (EHR) data complemented by enhanced self-reporting, so that proper epidemiological and statistical methods can be used to deal with confounding, and randomization can be applied to reduce uncertainty when it is needed.
Such a comprehensive system is now in sight. The FDA started years ago with the funding of the Sentinel Network (9) to enable massive curation of claims data (including drug-dispensing data for millions of Americans). The latest iteration of this system will incorporate data from EHRs and other sources and will be “democratized” to ensure that those data are widely available. In the arena of medical devices, the National Evaluation System for Healthcare Technology is organizing a prospective system that uses claims and EHR data to track device use and clinical outcomes (10). The Patient-Centered Outcomes Research Institute has invested in the National Patient-Centered Clinical Research Network (11), which combines EHR and claims data from tens of millions of people to measure outcomes over time. The NIH has also invested in networks of academic centers to conduct similar research (12,13). A system in which prescriptions and device use are recorded and matched with patient outcomes can provide numerators and denominators both acutely and over time, together with the rich data needed to understand clinical context for epidemiological and statistical methods to adjust for risks, examine alternative causal possibilities, and quantify the degree of uncertainty in causal inference.
An exciting new aspect of this ecosystem is the potential presented by the advanced use of patient self-reporting available through social media or direct measurement using personal devices. However, important issues related to confidentiality and information security must be addressed to take advantage of this vast source of data, and appropriate analytical methods must also be developed.
Finally, challenges presented by the use of fluoroquinolone drugs must be considered in the context of the growing problem of antibiotic resistance. Despite policy efforts aimed at reducing unnecessary antibiotic use and creating incentives and pathways for new antibiotic development, it is clear that the current pipeline is not adequate to deal with the constantly evolving pattern of resistance. The use of accelerated development and approval pathways means that new antibiotics will be marketed with incomplete knowledge of the full range of possible adverse effects, such as the types of toxicity manifested by fluoroquinolones. A meaningful response to this problem must include a combination of facilitated pathways, financial support or incentives, and a robust post-market evaluation system.
In the specific case of fluoroquinolones, what should be done now? First, the FDA should use Sentinel to examine the relationship between fluoroquinolones and valvulopathy in detail. Until such work is finished, there should be a substantial reduction in use of fluoroquinolones, which should be prescribed only as a last resort. In the longer run, the FDA and the larger clinical ecosystem should accelerate the creation of the superior information infrastructure described in the previous text. We are facing a critical need for an arsenal of effective, well-characterized antibiotics to deal with the evolving global problem of antibiotic resistance (14), and only an efficient, responsive, high-quality system of evidence generation will be able to supply it in a manner that supports a rational balance of safety and effectiveness.
↵∗ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
Dr. Califf sits on the corporate board for Cytokinetics and is Board Chair for the People-Centered Research Foundation; has received personal fees for consulting from Merck, Amgen, Biogen, Genentech, Eli Lilly, and Boehringer Ingelheim; and has been employed as an advisor for Verily Life Sciences (Alphabet).
- 2019 American College of Cardiology Foundation
- Etminan M.,
- Sodhi M.,
- Ganjizadeh-Zavareh S.,
- Carleton B.,
- Kezouh A.,
- Brophy J.M.
- Alves C.,
- Mendes D.,
- Marques F.B.
- Gorelik E.,
- Masarwa R.,
- Perlman A.,
- et al.
- Centers for Disease Control and Prevention
- US Food and Drug Administration
- European Medicines Agency
- Matta M.K.,
- Zusterzeel R.,
- Pilli N.R.,
- et al.
- Califf R.M.,
- Shinkai K.
- Platt R.,
- Brown J.S.,
- Robb M.,
- McClellan M.,
- Ball R.,
- Nguyen M.D.,
- Sherman R.E.
- US Food and Drug Administration
- ↵The National Patient-Centered Clinical Research Network (PCORnet). The network. Available at:. https://pcornet.org/clinical-research-network/. 2272019.
- Rethinking Clinical Trials
- NIH-DoD-VA Pain Management Collaboratory
- World Health Organization