Overview (Part 1 of 4)
When did your doctor last prescribe you a quinolone antibiotic? My last was in 2017, which was before the FDA warning about aortic aneurysm, but after warnings about severe nerve damage. Now I’m curious: Next time I get a urinary tract infection, what will my doctor prescribe…?
When did you last get a quinolone?
Starting in 2008, and over the subsequent years, the US Food and Drug Administration (FDA) has insisted that quinolone antibiotic manufacturers include warnings about some newly-discovered serious adverse effects in their product labels.1 These antibiotics, some of which you might know as Cipro or Levaquin, are used to treat common ailments like urinary tract and respiratory infections. Initially, the safety warnings were about tendonitis, which may have seemed relatively benign, and so widespread use of the antibiotics continued.
Since then, the list of known adverse effects has grown progressively serious, and has included things like tendon rupture, serious nerve damage, and as of late last year, aortic aneurysm and rupture.2 It is not yet known whether prescribing rates have gone down with these recent discoveries, but as of 2015, quinolones were still being prescribed in the US at a rate of more than 30 million prescriptions annually — or about one prescription per 10 Americans per year.3 That estimate is from before the aortic aneurysm warning, but after the one for serious nerve damage. Cipro accounted for about two-thirds of that.
Aortic aneurysms are thought to be rare, especially among young people, but the risk increases as you get older. It is estimated that 2-13% of Americans over 65 has one, although up to two-thirds of them doesn’t even know it.4 But if there are 30 million Americans taking these antibiotics every year, “rare events” not withstanding, quinolones could be causing considerable damage.
Late last year the British Medical Journal published a large cohort study that examined the association between fluoroquinolone use and risk of aortic aneurysm.5 The study informed the FDA’s decision to increase the warnings about the drugs. The authors found that patients who received fluoroquinolones were 66% more likely to have aortic aneurysms compared to amoxicillin, a finding that was “statistically significant.” Because it’s a rare disease at baseline, this would only translate into a very small increase in the absolute risk: it would cause an estimated additional 0.8 events per 1000 people who took quinolones every year. However, with 30 million prescriptions annually, that means we are likely seeing an extra 2,400 aortic aneurysms per year in the US alone!
I sure hope doctors are decreasing use of these agents whenever there are other options.
We recently discussed this article in our journal club class. Overall, we thought it was a reasonably well-done cohort study, and taken in the context of other evidence,6,7 we believe that their conclusion about increased risk is robust. That said, the weaknesses listed below lead us to suspect that there may be some wiggle room in the estimated magnitude of effect. In addition to the strengths and weaknesses listed below, there were also a couple of interesting features of this study that are worth noting. Click the links here if you want to read about them.
- In Part 2, read how this study illustrates that absolute and relative risks can tell different stories.
- In Part 3, read how this study illustrates a curious phenomenon about the trade-off between variance and effect size depending on background risk.
- In Part 4, read how this is a curious case of residual confounding despite good “balance.”
- The study was conducted in Sweden, where the national healthcare system makes it easy to collect data on drug exposures, disease outcomes, and confounding characteristics in a large population. Their sample size was large (N=360,088), which is important when you are looking at rare adverse events.
- They were able to control for a lot of key confounders, showing good covariate balance with matching on propensity scores.
- They used an active comparator (amoxicillin), which is great for minimizing some sources of confounding by indication and detection bias. Several prior studies that had not used an active comparator may have suffered from these biases.
- The study’s authors also conducted some key sensitivity analyses that supported their primary findings, increasing the credibility of their conclusions.
- They were unable to control for smoking history, which is a key risk factor for aneurysm. It’s hard to know exactly how this would have impacted their findings, but it probably hinges on the question of whether clinicians knew about the increased risk of aortic aneurysm. If they did, then there could have been a type of confounding by indication that would have biased the study toward finding a smaller difference in risk, because hopefully more doctors would have avoided quinolones in smokers.
- Clinicians were surely aware that smoking could cause aortic aneurysm, and so smokers would have been preferentially screened for aortic aneurysm. This would have meant that smokers would be more likely than nonsmokers to have subclinical aortic aneurysms diagnosed. The importance of this information bias also depends on the question of whether doctors avoided quinolones in smokers. Either way, this would have biased the study toward finding a smaller difference in risk.
- The use of propensity scores to control for confounding over such a long study period (11 years) was not ideal. Propensity scores are less useful if the drug’s indications and uses are evolving over the study period. They likely were in this case, because new data has been emerging since 2008. (The study period was 2006-2013.) This problem might have been mitigated by sensitivity analyses that divided up the follow-up period into smaller intervals, but the outcome may have been too rare for such an analysis to succeed.
- Another limitation of propensity score matching is that they require the researchers to restrict the analysis to the subset of patients in whom the reasons that doctors would prescribe one of these antibiotics over the other were less compelling, or at least unknown to researchers. This is because it would have eliminated observations without “common support.” For one thing, this could limit the generalizability of their findings. For another, the residual confounding associated with unknown indications probably means that this study can’t provide a precise estimate of the magnitude of effect.
- Food and Drug Administration. FDA updates Warnings for fluoroquinolone antibiotics. July 26, 2016. Available online at http://www.fda.gov/newsevents/newsroom/pressannouncements/ucm513183.htm. Accessed March 20, 2019.
- Food and Drug Administration. FDA warns about the increased risk of ruptures or tears in the aorta blood vessel with fluoroquinolone antibiotics in certain patients. December 20, 2018. Available online at http://www.fda.gov/Drugs/DrugSafety/ucm628753.htm. Accessed March 20, 2019.
- Centers for Disease Control and Prevention. Outpatient antibiotic prescriptions — United States, 2015. September 12, 2017. Available online at https://www.cdc.gov/antibiotic-use/community/programs-measurement/state-local-activities/outpatient-antibiotic-prescriptions-US-2015.html. Accessed March 20, 2019.
- Dalman RL, Mell M. Eidt JF, Mills JL, Collins KA, eds. Patient education: Abdominal aortic aneurysm (Beyond the Basics). UpToDate. January 31, 2018. Available online at https://www.uptodate.com/contents/abdominal-aortic-aneurysm-beyond-the-basics. Accessed March 20, 2019.
- Pasternak B, Inghammer M, Svanström H. Fluoroquinolone use and risk of aortic aneurysm and dissection: nationwide cohort study. British Medical Journal. 2018;360:k678.
- Daneman N, Lu H, Redelmeier DA. Fluroquinolones and collagen associated severe adverse events: A longitudinal cohort study. BMJ Open. 2015;5(11):e010077.
- Lee CC, Lee MT, Chen YS, et al. Risk of aortic dissection and aortic aneurysm in patients taking oral fluoroquinolones. JAMA Internal Medicine. 2015;175(11):1839-47.