Journal Club: Association Between Corticosteroids and 28 Day All-Cause Mortality Among Critically Il
J. Falatko D.O.
The latest issue of JAMA is an interesting one. The entire original investigation section is devoted to steroids for the treatment against COVID-19 in severely ill patients. The researchers are not talking about the steroids Mark McGwire used to hit home-runs. They are investigating corticosteroids. These are medicines used to impair the immune system, mitigate inflammation and support poor adrenal function. The first 3 papers were kind of duds. The final study is a meta-analysis, and it is a beautiful example of the benefit these studies bring to medical research.
The first 3 articles really highlight the difficulty in treating patients severely ill with Acute Respiratory Distress Syndrome (ARDS). Historically, mortality is high at around 60%. Ventilator days are many. Organ complications are many. ICU stays are long. Patients are lucky to punch their ticket out.
The sample sizes were quite small. One trial was stopped early due to difficulty with enrollment. Another trial only showed a reduction in ventilator free days but no significance in other outcomes. The third trial only showed an increase in organ dysfunction free days (you may have to read that outcome twice), without any change in ICU days, ventilator days, or mortality.
All of these studies had trends in the right direction for outcomes like mortality, icu days, ventilator free days, but lacked the size to be able to show a statistical significance. They did not have statistical power. However, when you grouped the results together much of these trends, which weren’t statistically significant, became significant. This helped reveal the potential benefit of steroids in the treatment of severe/ life-threatening COVID, which has been elusive.
Let’s take a look at the study and what they found.
There were 16 total trials identified in their search. These were all randomized trials. Directors of each trial were contacted. Some of the trials were excluded due to ongoing enrollment, or refusal to collaborate with the authors of the meta-analysis. They were left with 7 studies combined in the analysis. All of the trials investigated patients with severe COVID or life-threatening disease. There were 678 patients randomized to steroids and 1025 to usual care. Alternative therapies like hydroxychloroquine and azithromycin varied significantly between the trials. Only 13 patients were lost to follow-up, which is low.
The underlying risk of bias in 6 of the 7 trials was rated as low.
There were 222 deaths in the 678 patients treated with steroids and 425 deaths in the 1025 patients treated with usual care. The odds ratio (OR) had a confidence interval of (0.53-0.82). For the most part Odd's ratios and risk reduction are interchangeable in medical research. Particularly when used in prospective randomized controlled trials and when event rates are rare. The Odd's ratio describes the change in the base rate based on exposure to a variable. For example, in the base death rate is 41%, the best case odds ratio of of 0.53 would represent a change in event rate to 21%, a worst case odds ratio of .82 would be a change in event rate to 33% after exposure to the variable, in this case steroids.
With these numbers the absolute risk reduction from the treatment is 9%. Based on this the number needed to treat (NNT) can be calculated. The NNT tells us the number of patients that need to receive the treatment to prevent the outcome once. The NNT is 11.1. So, for every 11 patients that receive steroids 1 life is saved. In medicine this would be considered a robust result. Usually we are dealing with NNT’s in the 50’s to 100’s.
The heterogeneity assessment was 15.6%. In a meta-analysis you want all the studies to look the same. They should be homogenous. The closer to zero this number is the better.
They found similar results in subgroups for patients > 60 and those requiring mechanical ventilation.
Serious adverse events were not measured consistently in the trial, so they could not comment much on the potential downside. But I’m sure there were. There are very few free lunches in this world. Steroids are known to significantly impair the immune system, increase blood sugar, increase blood pressure, cause psychosis, reduce muscle strength, and thin the stomach lining just to name a few. The longer a patient is on them, the more complicated their care becomes.
This meta-analysis showed that steroids are a helpful therapy for severe and life-threatening COVID. Trials investigating other therapies like Remdesivir, Hydroxychloroquine, and Convalescent plasma have showed to be ineffective in treating this population. The low number needed to treat suggest that the treatment response is robust and can really impact the death rate associated with severe COVID.
This study also showed the beauty of meta-analyses. None of the included studies showed statistical significance. Mainly due to being underpowered. The original studies did not have the resources, patients, and time to get to truly test to hypothesis. Much like the example set by the cartoon “Captain Planet,” When their collective data combined, they formed a super study that solved the hypothesis problem.
Steroids work by mitigated the immune response to the virus. In Acute Respiratory Distress Syndrome (ARDS) the real threat is the attack the immune system initiates to eliminate the infecting agent. Steroids impair white blood cells in almost all of their actions. Their movement is impaired, the release of inflammatory proteins is impaired, and their communication between each other is impaired. Often times in ARDS the infecting agent has already been suppressed, but the immune cascade is so amped up that it takes too long to turn off. Unfortunately, in many cases the patient’s lungs are destroyed before the immune cascade is winded down. So, the science behind this makes sense.
However, this runs counter to much of the previous research investigating steroids for the treatment of ARDS. Recently, steroids have been shown to benefit patients with severe ARDS from pneumonia, but in other causes they are less likely to benefit. They may be associated with worse outcomes in patients with influenza. Current guidelines recommend their use in only refractory ARDS. So why do they work so well with COVID?
There are two possible explanations that I can think of for this. First, the result the researchers found could have been random chance, or the underlying studies could be biased favoring steroids. In that case there is no true effect. However, their event rate was high enough to tell a difference. They had statistical power. The studies they included were similar and of low risk of bias, so I think this is less likely. It’s possible some of the studies that were excluded due to lack of investigator participation were left out because no benefit was seen during interim analysis. This could have significantly impacted the results. Second, studies involving steroids in the treatment for other causes of ARDS are difficult to execute. So many variables effect the treatment of ARDS. Timing is critical. Oxygenation is critical. Underlying causes are many. Complications rates are high. Death rates are high. It is very difficult to find an effective treatment in these scenarios. This creates a lot of variability with this disease. If any of these variables are off, the study could be doomed.
Although there is no magic therapy for COVID, in this high-quality study steroids show promise as the first effective treatment in severe and life-threatening COVID.