Things corticosteroids have been shown to be effective for since I started my medical training: community-acquired pneumonia, bacterial meningitis, and (just for kicks) sore throat. Seems like a little prednisone can make just about anyone feel better, and while the long-term adverse effects of steroids are well-documented, the collective wisdom of physicians seems to be that a Medrol Dosepak here and there is unlikely to result in significant morbidity. However, Akbar K Waljee and a few other troublemakers at the University of Michigan decided to test this assumption in a registry data-mining operation to examine the risks of adverse effects of corticosteroids over less than 30 days of exposure.
So if our piece last week on the dangers of NSAIDs had you wondering if maybe you should leave off the “N” and just prescribe steroids for, say, acute back pain, wonder no more. It may be an even worse idea.
If You’re Only Going to Read One Paragraph
In this retrospective cohort study with nested self controlled case series, patients with exposure to a brief course of corticosteroids (< 30 days) were significantly more likely to experience deep vein thrombosis, sepsis requiring hospitalization, or fracture than patients who did not receive steroids. These same steroid-exposed patients were also more likely to have these events following the prescription of steroids as compared to the period before. These increased incidence rates peaked in the 30 days immediately following prescription of the steroid course. Absolute rates of these events remained low, in the single digits per thousand person-years (save fracture, at about 20 per thousand person-years). This study should serve as a caution about being too liberal in prescribing even a short course of oral corticosteroids if reasonable alternatives exist.
A Bit of Context
Since glucocorticoids came to market in 1949, physicians have endured patient jokes along the lines of “How come I don’t look like Arnold Schwarzenegger if I keep taking all these steroids?” Which I mean, hilarious the first 50 times, yes, but come on. Steroids are miracle drugs, powerful immunomodulators capable of turning off the inflammation underlying disease processes from rheumatoid arthritis to Crohn’s disease. But their long-term complications became obvious shortly after their introduction. The toxicity of steroids over prolonged exposure is a major underpinning of the treatment of autoimmune disease, and pretty much every professional society that deals with these diseases has published practice guidelines dedicated to minimizing steroid exposures and mitigating their adverse effects. However, when it comes to shorter courses of steroids, such as those used to treat COPD exacerbations or acute flares of gout, we have very little data regarding their adverse effects. We know the cumulative dose can contribute to long-term effects, even if only given in recurrent short-term use, but what about the effects from even one round of steroids?
If we look at the package insert for methylprednisolone, we can see the litany of adverse effects about which the FDA feels consumers should be warned (scroll to page 13). Some of these happen quickly — decreased glucose tolerance, for example, as well as the three outcomes on which the investigators of our current study decided to focus: thromboembolic disease, sepsis, and fracture. There has been no significant prior investigation of increased risk of these adverse effects with only short courses of steroids.
What do you mean it’s self-controlled? A look at study design
The first part of this study is a straightforward retrospective cohort based on a database of patients maintained by a private insurance provider. The investigators used this database to examine de-identified data for patients 18-64 years old (because patients qualify for medicare at age 65, these patients were not represented in this database) between 2011 and 2014. So as not to inadvertently measure cumulative effects of steroids rather than the effects of a single course, they excluded patients who had received steroids in the 12 months prior to the study period. Patients were excluded for malignancy or for any transplant. They were also excluded if they had received more than 30 days of corticosteroid exposure, as, y’know, that was the point of the study. To avoid confounding, patients with a history of the outcome in question (VTE, sepsis, or fracture) were also excluded from the trial. This leaves us with an impressively large 1,548,945 patients to follow, though given that we started with 2,234,931 that’s still a ~30% exclusion rate.
In order to be sure about whether patients met these exclusion criteria, the investigators limited their population to patients who had data in the registry for at least 12 months prior. While this is a totally reasonable strategy, you can imagine that restricting your patient population to people who have not only private insurance but have in fact been on the same plan for more than 12 months leaves out most vulnerable populations. This is not likely to be a group for whom economic instability is a major issue.
Here’s a little wrinkle: because patients who receive steroids are likely to have a reason to need steroids, we might assume that they might be older, have more comorbidities, or otherwise differ from than their steroid-deprived peers. To attempt to adjust for this, the authors employed a technique known as a self controlled case series (SCCS). I think I used a lot of political capital on making you sit through my explanation of patient versus study-level metanalyses, so I’ll be quick here: it’s essentially a way to compare patients to themselves rather than matching them with controls to adjust for confounding factors. In our study, the period from 5-180 days prior to glucocorticoid prescription served as the reference for event rates, with the period 0-90 days after prescription serving as the comparator. In this way, we can observe whether patients with the exact same risk factors (because they are the exact same patients) are more likely to have events after steroid exposure than before. This technique was initially developed to examine vaccine safety, and you can imagine how it might be useful to look at all patients who received MMR to see if they developed, say, aseptic meningitis with increased frequency in the 90 days following vaccination as compared to before.
Given my above speculation on these patients being better off than my mostly-Medicaid primary care panel, let’s start by talking about demographics. It’s tough to tease apart socioeconomic status based on the data reported, but they do report educational attainment, and if you think I’m not going to shoehorn that into a surrogate for poverty, think again! Compare to US census data for educational attainment and you’ll see that the patients we’re currently examining are indeed better off than average — just 0.4% of patients in the study population had failed to obtain at least a high school diploma, compared to 11.7% nationally. Similarly, the 53.9% of patients who had at least some amount of college education dramatically outstrips the 19.1% of national respondents who report this outcome. Given the well-established correlation between educational attainment and income, well, shoehorning complete. Which I guess just means the shoe is on? Point is, patients are less poor than a representative US sample.
Back to the study: 327,452 (21.1%) of the included patients received a qualifying steroid dose. Just over 9% of patients were black, with just over 10% identifying as Hispanic. Steroid users were statistically older (mean age 45.5 versus 44.1) than non-users, and were also more likely to be female, to be white, and to have more comorbidities.
The most commonly prescribed dose of steroids was a great testament to marketing, the Medrol Dosepak. This 6 day tapering dose of methylprednisolone accounted for 46.9% of all steroid prescriptions, and you will be thus unsurprised to hear that the median steroid course for those who received a prescription was 6 days. Most patients (70.5%) received only one course of treatment. There was significant regional variability, with statistically lower rates of prescriptions in the Pacific region compared to the eastern and western south central regions. Most prescriptions were written by general internists or family physicians.
The punchline: every adverse event examined was more common in those who received steroids than those who did not. Rates of hospital admission for sepsis were 0.05% versus 0.02%, rates of VTE 0.14% versus 0.09%, and rates of fracture 0.51% versus 0.39%. All of these are significant.
Now, an easy objection here would be that we have already established that patients receiving steroids had more comorbidities and tended to be older than patients who did not, which could explain the fracture and sepsis risks. This is where our self controlled case series comes in! In the 30 days following a steroid prescription, hospitalization for sepsis occurred more frequently, with an incidence rate ratio of 5.30, as did VTE (IRR 3.33) and fracture (IRR 1.87). These rates of incidence remained elevated during days 31-90, but the increased risk was attenuated somewhat during this period. This is a pretty elegant demonstration that it likely is the steroids themselves that are increasing this risk, not simply the baseline characteristics of patients likely to be prescribed steroids.
It’s worth noting here that not all increased risks are created equal. While these do represent a significant relative risk increase, the absolute rates are still extremely low, with a number needed to harm >3300 for sepsis, 2000 for VTE, and 833 for fracture. On a population level, these can become significant. But if you have a patient who has a good reason to receive a short course of steroids, be it an acute COPD flare or Bell’s palsy, these risks may be in the back of your mind but are necessarily going to outweigh the benefits of treatment.
But the thing is, the indications for which steroids were prescribed just weren’t these things. Look at the list of common indications in our current study: upper respiratory tract infections, spinal conditions and intervertebral disc disorders, allergies, bronchitis, and (non-bronchitic) lower respiratory tract disorders. Not exactly situations in which the benefit to enteral steroid therapy is clear.
It’s far easier to prescribe a drug than not to do so, and the temptation to prescribe a short course of steroids for symptomatic relief alone is understandable. But if we start doing this on a broad scale, we have every reason to think we will start to see more fractures, more infections, and more blood clots. But for back pain? I’m not sold. Heck, even opioids may have a better short-term safety profile.
So for me, I will continue to use steroids when I think they offer significant benefit, as while they do increase risks of complications even with short-term use, these complications continue to happen with very low absolute rates. However, I’m in no hurry to start prescribing them for symptomatic relief alone, as even short courses clearly come with complications. I’m in no hurry to start giving prednisone out like it’s candy — after all, that’s what statins are for.
Read the original study here.
Next Time on IM HEAT
Thanks for your patience with the lack of update this Monday. We’re back on track now — I promised you a piece about cystic fibrosis, and you’re going to get it! Tune back in on Monday for an interesting look into what insurance coverage means to those with pre-existing conditions.