Something that I feel differentiates experienced from novice providers in the emergency department is a true recognition of the evidence and rationale behind why we do what we do. It is certainly important to recognize the right treatment for a given illness. Perhaps equally important it is necessary to understand the evidence behind the reason that treatment is given in the first place. This enables us to practice not only the most up-to-date evidence-based care but also provide the best education to our patients, families and our colleagues. It is with that intro that I would like to introduce a new feature on the blog called “why we do what we do.”
I’m going to tackle some commonly seen clinical problems and scenarios and explain the evidence behind our shared practice. I hope to provide some education not only on the treatment and or diagnostics but also on the evidence-based medicine concepts behind them. A great place to start is with the common winter illness acute laryngotracheobronchitis AKA croup.
Many of you instinctively give dexamethasone for patients with varying degrees of severity in croup. But why? Well, physiologically patients with croup experience edema and swelling within the lumen of their tracheae. Steroids direct their action to reduce the production of inflammatory mediators. This effect is not instantaneous and can take 4 to 6 hours to affect a relevant clinical impact. So to start, when we give steroids to patients with croup we should explain that this is not an acute therapy but is being given to reduce the inflammation in what is an ongoing infectious process without an antimicrobial cure. Simply stated, it’s not fixing the immediate problem. Well then, why do we seem to give it out like candy during croup season?
This is where I think the concept of number needed to treat is valuable. The NNT is the average number of patients who need to be treated with a particular therapy in order to prevent one additional bad outcome. It is the inverse of the absolute risk reduction. The number needed to treat is a function of four elements: The condition, the intervention, the event being prevented and the duration of follow-up. In the context of croup we are looking for the number of patients we need to treat with dexamethasone in order to prevent one additional bad outcome. Well, what is that additional bad outcome? Recall that dexamethasone does nothing to impact the immediate disease process – so it’s not going to prevent intubation or immediate stridor. However, I think that we can truly say that the emergency department is the only part of the hospital were we can truly and legitimately say that we do not want to see our patients again. Therefore it would make sense given the time it takes for dexamethasone to work to look at reduction of return ED visits in terms of the number needed to treat for croup.
Fortunately, there are studies that do just that.
96 patients were randomized to dexamethasone or not. The results showed a decreased risk of return visits for those that received steroids. You need to do the math here, but the authors give us the numbers.
|
Return visit |
No return visit |
|
|
Dexamethasone |
0 |
48 |
|
Placebo |
9 |
39 |
To calculate the NNT you must use math. So, I’ll show my work.
Control event rate = 9/(9+39) = 0.1875
Experimental event rate = 0/(0+48) = 0
Absolute risk reduction = 0.1875 – 0 = 0.1875
Number needed to treat = 1/0.1875 = 5.3
So, rounding to the nearest whole number we see that you must treat five patients with croup with dexamethasone to prevent one additional return visit.
Luria, 2001 – JAMA Pediatrics (Archives of Pediatrics & Internal Medicine)
The authors compared oral dexamethasone, nebulized dexamethasone, and placebo and noted that in 264 patients with mild croup randomized to one of the three therapies those treated with oral dexamethasone sought additional care less frequently. The NNT was 10. It is important to note that enrollment criteria included any child whom the physician felt might need dex and/or racemic epi. The post HOC power calculation was 58%. So there was a type II error in effect somewhat limiting the study’s availability to see the observed effect.
A double-blind placebo RCT of 720 children with croup found that there was a reduction in the risk of return visits to the ED within 7 days of initial presentation. 54 (15.3% of the placebo group returned for care as opposed to 7.3% of dexamethasone (p<0.001; 95 % CI 3.3% – 12.5%). The adjusted odds ratio was 2.4 (95 % CI, 1.4 to 3.9), after adjustment for study center, age, type of croup, baseline Westley croup score, and duration of barking cough before presentation. The number needed to treat in order to prevent one return visit is 13 (95% CI, 8 to 31).
Okay – so we see through studies of varying size over the course of the last 20 years the dexamethasone reduces the risk of a return visit to the ED. The number needed to treat is relatively low, only hitting 13 on the largest study. Given that the risk of giving a patient an oral dose of dexamethasone is very low, and that you will likely see at least 13 patients with croup during a rotation in the ED it makes sense to give any and all patients that you have diagnosed with croup a dose of oral steroid. In terms of how to explain this to the family, I think the approach varies based on how sophisticated their level of understanding is. When asked why are we giving this medicine you could simply state that it reduces your child’s risk of coming back to the emergency department. For many parents that’s enough. You could also explain that the odds of their child needing to come back are approximately 2 ½ times decreased if we give the steroid versus not. I think the concept of number needed to treat makes more sense when teaching your colleagues. Especially given that the volume of patients with an individual condition that you will see is substantial. So in essence the way that you explain the evidence really depends on which lens you are viewing it through.
