Myth-busting: Azithromycin does not cause torsade de pointes or increase mortality (2024)

.

Introduction

.

In 2012 a NEJM article by Ray et al. reported a correlation between azithromycin and cardiovascular death. This received extensive press and ultimately led the FDA to issue a drug safety communication warning about the risk of QT prolongation and torsade de pointes. Subsequent studies have failed to replicate this result. Nonetheless, suspicion lingers: Does azithromycin increase mortality?

.

Basic science & electrophysiology: Is it plausible that azithromycin would cause torsade de pointes?

.

Antibiotics generally cause QT-prolongation and sudden death by blocking the hERG potassium channel and thus slowing cardiac repolarization. Drugs are more dangerous if they have a higher affinity for the hERG channel and if they are cleared by the CYP enzyme system (rendering them susceptible to more drug interactions). Azithromycin is not cleared by the CYP system, and has a low affinity for the hERG channel (27 times lower, for example, than erythromycin)(Giudicessi 2013). Thus, from a molecular perspective azithromycin would be expected to be fairly safe.

.

Azithromycin appears to cause a small prolongation of QTc, averaging ~10ms. Unfortunately the best study of this was performed by Pfizer and never published (it is briefly described in the package insert). One study failed to detect any change in QT intervals at all (Shin 2014).

.

More importantly, not all QT prolongation is created equal. Some drugs may prolong the QT interval without increasing the risk of arrhythmia (Thomsen 2006). In a rabbit heart model, supratherapeutic azithromycin levels prolong a differentcomponent of the action potential compared to erythromycin (Milberg 2002). Rather than prolonging repolarization (a pattern which tends to cause Torsade de Pointes), azithromycin prolongs the action potential itself. Azithromycin does not predispose to torsade de points, but instead it actually blocks the pro-arrhythmic activity of erythromycin. The inability of azithromycin to cause torsade de pointes has been confirmed in two other studies using a dog model, even with enormous doses of azithromycin (Thomsen 2006, Ohara 2015).

.

Thus, azithromycin seems to effect the heart in a fundamentally different way than erythromycin. Azithromycin prolongs the action potential without signs of proarrhythmia, giving it the properties of “an ideal anti-arrhythmic agent.” (Milberg 2002) This doesprolong the QT interval, which is typically misinterpreted to be a sign of increased risk of arrhythmia.

.

Case Reports

.

Case reports more convincingly relate azithromycin to asymptomatic QT prolongation than actual torsade de pointes. Indeed, every case report of torsade de pointes involved at least two other risk factors (Hancox 2013). Given the millions of courses of azithromycin which have been prescribed, if azithromycin caused torsade de pointes one would expect to see more persuasive reports of this.

.

Ray WA et al NEJM 2012: The paper that started it all

.

This was a retrospective correlational study of Medicaid recipients in Tennessee comparing patients who received various outpatient antibiotics (azithromycin, amoxicillin, ciprofloxacin, or levofloxacin) or patients who were not ill and not receiving any antibiotic. Azithromycin was associated with an increase in mortality on the fourth day after starting therapy, compared either to amoxicillin or no antibiotic:

.

.

Although relegated to the paper's appendix, this same exact pattern of excess death on day #4 was observed with levofloxacin, but not with ciprofloxacin:

.

.

However, compared to patients receiving amoxicillin, patients receiving levofloxacin or azithromycin were more likely to be undergoing treatment for pneumonia or COPD. Such patients tended to have more comorbidities and receive more medications. Thus, use of azithromycin or levofloxacin may have merely correlatedwith a higher risk of death, rather than causingdeath.

.

The fact that both levofloxacin and azithromycin correlate with an identical pattern of increased mortality on day #4 is strange. Levofloxacin reaches steady-state levels well before day #4, so a pharmacologic effect of levofloxacin would be expected to begin earlier and last longer. The fact that two dissimilar drugs correlate with an identical mortality pattern suggests that this mortality pattern is not being caused by either drug.

.

Subsequent observational study fails to validate Ray et al.

.

Svanstrom 2013 performed a similar study using a national database of Danish adults. These authors also used patients receiving a beta-lactam as their control group (penicillin), given that these represented an acutely ill group of patients receiving a safe antibiotic. This study found that compared to patients receiving penicillin, patients receiving azithromycin had lower mortality (rate ratio of 0.93, with a 95% confidence interval of 0.56-1.55).

.

.

The results from Ray et al and Svanstrom et al are summarized in the chart above (1). The traditional interpretation of this data is that Ray et al found a higher mortality among the azithromycin group than the beta-lactam group, demonstrating an increased mortality due to azithromycin. Alternatively, Svaranstrom et al found the same mortality among the azithromycin group and the beta-lactam group, demonstrating no increased mortality due to azithromycin.

.

However, closer examination shows that the real difference between these studies is the behavior of the beta-lactam group (the “control” group). Ray et al found a slightly lower mortality rate among the beta-lactam group compared to the no-antibiotic group, which doesn't make sense. Patients receiving beta-lactams were acutely ill (unlike the no-antibiotic group), so the beta-lactam group ought to have a higher mortality. Overall, Svanstrom's results seem more plausible, suggesting that there is no increase in mortality due to azithromycin.

.

Meta-analyses of prospective RCTs

.

Ultimately these observational studies are primarily correlational. As such they can only be used for hypothesis generation. To really determine whether azithromycin causes increased mortality, prospective RCTs are needed.

.

Baker 2007performed a meta-analysis of prospective RCTs investigating the use of azithromycin for secondary prevention of coronary artery disease. Six studies involving 13,778 patients were analyzed. Azithromycin was associated with a nonsignificant trend towards reduced mortality.

.

.

Almalki 2014 performed a meta-analysis of prospective RCTs that compared azithromycin vs. placebo for various conditions (mostly COPD, severe sepsis, and cardiovascular disease). This study involved 12 RCTs with a total of 15,588 patients. Many of these studies involved prolonged administration of azithromycin for up to a year. This increased the power of the meta-analysis, which includes ~1.2 million person-days of azithromycin exposure. There was a trend towards reduced mortality in patients receiving azithromycin.

.

.

.

Conclusions

.

.

Erythromycin may prolong the QT interval and occasionally cause torsade de pointes. Since azithromycin and erythromycin are closely related, it has often been assumed that these drugs would act similarly. For example, a prominent review article recently lumped these two drugs together (Albert et al. 2014).

.

There are two reasons that azithromycin does not share erythromycin's ability to cause torsade de pointes. First, azithromycin's affinity for cardiac potassium channels is 27 times lower than erythromycin's. Second, azithromycin prolongs the QT interval due to prolongation of the action potential itself, unlike erythromycin which delays repolarization. This could actually give azithromycin anti-arrhythmic properties. These factors explain why azithromycin has been shown to cause very small increases in the QT interval, but has notbeen convincingly linked to torsade de pointes.

.

The highest quality of evidence to evaluate azithromycin's safety are meta-analyses of prospective RCTs which compare azithromycin to placebo. Such meta-analyses have shown a trend towards decreased mortality with the use of azithromycin.

.

At this point, the concept that azithromycin causes torsade de pointes and cardiovascular death should be discarded. It is not supported by molecular, electrophysiological, or clinical evidence.

.

Epilogue: The Acontextual Fallacy

.

In retrospect, there was only one study that made us worry about azithromycin: Ray et al. This study was inconsistent with preceding evidence, and has now been shown to be inconsistent with subsequent evidence as well. So why did we care so much about this study?

.

.

This is an example of what might be called the acontextual fallacy. Based on the prevailing use of frequentist statistics and p-values, we approach every hypothesis with a pre-test probability of 50%. As discussed previously, Bayesian statistics might be a better approach to encourage adjustment of the pre-test probability based on prior knowledge. Unfortunately, our current approach to interpreting papers focuses primarily on looking inward at the details of the paper. This naturally leads to the acontextual fallacy, wherein the paper is interpreted within a vacuum.

.

Avoiding the acontextual fallacy requires a thorough understanding of prior basic science and clinical data. Without context, our opinions are easily swayed by whatever the latest study shows. Azithromycin was safe last year, it's dangerous this year, but it will probably be safe again next year. Unfortunately, understanding context is labor-intensive, so this component of interpreting studies is often neglected.

.

• Azithromycin and Erythromycin have different effects on cardiac electrophysiology.
• Azithromycin does prolong the QT interval, but does not cause Torsade de Pointes. It may actually have anti-arrhythmic activity.
• Azithromycin does not increase mortality.
• Azithromycin is a safe drug but should still be prescribed responsibly (it is not intended for anti-viral, anti-pyretic, anti-tussive, or anti-anxiety therapy).

Conflicts of Interest: None.

Notes

(1) For comparison's sake, the rate of death for patients receiving no antibiotic has been set equal to one. Relative rates of cardiovascular mortality are based on the primary analysis in each paper (coincidentally, Table 2 in both publications).

Image credits:

– Opening cartoon from http://www.acphospitalist.org/weekly/archives/2013/5/8/

– Subsequent cartoon fromhttps://adai.files.wordpress.com/2006/12/borgman042797_600x385.jpg