How many days of antibiotics do I need to treat this infection? The academic answer to this question has historically been based on the numbers of fingers on one’s hand, or, for more serious infections, on the number of days in a week. In practice, durations recommended by expert guidelines are regularly extended, sometimes over concern for patient safety but perhaps just as often for the ever-important purpose of improving physician comfort. Trials of shorter courses for antibiotics are rarer than we might like, in no small part because nobody is lining up to fund studies about using FEWER expensive pharmaceuticals. But in a much-needed and well-designed trial, Uranga and colleagues seek to validate existing guidelines and prove that 5 days of antibiotics is plenty for most cases of community-acquired pneumonia (CAP).
If You’re Only Going to Read One Paragraph
In inpatients with community-acquired pneumonia, 5 days of antibiotics with extension only for patients with persistent fever or predefined signs of instability was noninferior to antibiotic course at the discretion of the treating clinician. This is a validation of current IDSA/ATS guidelines. As outpatients are unlikely to exhibit signs of instability (given the selection bias inherent in choosing not to hospitalize them), it is doubtful that any outpatient should receive more than 5 days of antibiotics for CAP unless they deteriorate to the point of requiring hospitalization.
Patients, Intervention, Comparator, and Outcomes
In this multi-center randomized controlled trial, 312 patients with CAP (defined as a new infiltrate and at least one of fever, dyspnea, cough, or chest pain) were randomized at day 5 of hospitalization to one of two groups. In the intervention group, antibiotics were discontinued after the 5th day provided the patients had been afebrile for 48 hours (Tmax < 37.8 C / 100.4 F) and had no more than 1 “CAP-associated sign of clinical instability” (more on this in a second). The control group received usual care, with antibiotic duration at the discretion of the treating clinician. The choice of antimicrobial therapy was left at the discretion of the treating physician in both groups.
Patients were excluded for immunosuppression including any HIV infection, having been an inpatient in any setting in the past 14 days (i.e. not having CAP), or having taken any antibiotics in the prior 30 days. Culture confirmed staph aureus or pseudomonas were also criteria for exclusion, as were confirmed extrapulmonary sites of infection (i.e. meningitis or endocarditis) — reasonable, as these could change the recommended duration of antibiotic therapy. Patients requiring ICU transfer either at admission or before randomization on day 5 were excluded, a sensible but important limitation of this trial.
The main outcomes were clinical success at day 10 and day 30 (defined as improvement or resolution in symptoms without additional antibiotic therapy), as well as scores on a validated questionnaire regarding CAP-related symptoms. Due to low event rates, mortality was not a primary outcome. Remember: these patients were randomized at day 5, and ICU admission was criteria for exclusion, so this was not a study well-suited to evaluate for a mortality effect. A slew of secondary outcomes included duration of antibiotic therapy and time until clinical improvement.
A Bit of Context
The benefits of using fewer antibiotics, if safe for the patient, are self-evident. It’s cheaper, generates fewer adverse effects, and can decrease rates of antimicrobial resistance. 2007 consensus guidelines from the Infectious Disease Society of America and the American Thoracic Society recommend a minimum of 5 days of antibiotic therapy for CAP. They recommend prolonging antibiotic therapy in patients who have had fever within the preceding 48 hours or who exhibit at least two of the following criteria: systolic BP < 90, HR > 100, RR > 24, SpO2 < 90% or PaO2 < 60%. These are collectively termed CAP-associated signs of clinical instability (see, I told you we’d get back to it).
So is it unreasonable that some patients get more than 5 days of antibiotics? Not at all. These expert recommendations assert that 5 days of therapy is likely enough for most people, but allow that there are many circumstances that warrant extended therapy. After all, not all patients are created equal, and chronic disease in the underlying substrate may delay clinical convalescence. However, it is reasonable to wonder whether longer courses are always determined after careful reflection over the risks and benefits, or might rather be reflexive decisions by nervous doctors. Accordingly, the trial we’re reviewing today is NOT examining a strict 5-day duration; it’s examining a default course of 5 days, overruled only by pre-specified criteria. The authors state that their purpose is to validate the current IDSA/ATS guidelines, and their intervention group is a good representation of strict adherence to these.
You could be excused for concern at the fact 227 of the 539 patients assessed were found to be ineligible for the study. However, these patients were assessed on day one and not randomized until day 5; so of the 227 patients excluded, 39 died before day 5 and 57 went to the ICU before that time. 75 were excluded for antibiotics within the prior 30 days (which is too bad, because whether “failure” of outpatient treatment affects antibiotic duration is an interesting question, albeit one for another study). So this high exclusion rate really reflects just patients meeting the predefined exclusion criteria in between assessment and randomization.
As the study is out of Spain, the population is homogeneously white. Just shy of 2/3 of patients were male. ~15% of patients were diabetic. Mean pneumonia severity index was 83.7 in the control group and 81.8 in the intervention group. All baseline characteristics were similar between groups, as were vital signs on hospital day 2.
And the results? Five is fine.
In the intention-to-treat analysis, the primary outcome of clinical success at day 10 was 48.6% in the control group and 56.3% in the intervention group (P = 0.18), with a similar non-significant difference in the per protocol analysis. At day 30, it was 88.6% in the control group vs 91.9% in the intervention group (P = 0.33), again similar in the per protocol analysis. Symptoms as assessed by questionnaire were also similar at both periods in both analyses. I could keep writing out the ways in which there was no difference between groups, but take my word for it – I looked. No difference in time to symptomatic improvement or return to normal activity. No difference in length of stay. Seriously: there was no difference.
Oh, save this one — the intervention group got a median of 5 days of antibiotics, versus a whopping 10 days in the physician discretion group. This, needless to say, was significant.
Though there is this oddity: readmission by day 30 was more common in the control group than the intervention group (6.6% vs 1.4%, P = 0.02). This is a bit weird; combine this with the non-significant trend towards better clinical success in the intervention group and you might start to wonder if patients in the intervention arm were less sick to begin with. However, their baseline Pneumonia Severity Index (PSI) scores were similar (83.7 for control vs 81.8 for intervention), and their vitals at days 2 and 5 were similar as well. The intervention group was even moderately more symptomatic (P = 0.049) at time of randomization. So it’s reasonable to wonder if the intervention group could have been sicker at the start, but there’s no evidence that this was the case.
In a deviation from common clinical practice in the United States, >80% of patients in both groups were treated with fluoroquinolones, whereas <10% received beta lactam therapy plus a macrolide. Should this matter for generalizability? The short answer is no. The long answer is noooooooooooooooooo. Beta lactams have better coverage overall than quinolones for common CAP organisms, so it’s unlikely that there would be more treatment failure on this regimen, especially when you consider that most patients received 500mg rather than 750mg daily of levofloxacin — lower doses than are commonly used in CAP.
Indeed, it’s hard to take issue with the study. Sure, it doesn’t apply to ICU patients, but it did include very sick patients (severe CAP with PSI > 130). You could argue that the open label design altered the prescribing habits of physicians with patients in the control group, but they certainly didn’t cut their courses short. The biggest limitation is probably patient population and geography; four hospitals in Basque Spain doubtless have a higher concentration of certain genotypes and phenotypes than the average urban medical center in the United States, but it’s difficult to predict exactly how this could have affected the results of this study. A validation on a more heterogenous population would be welcome, but there’s no clear impetus for this to be a priority.
At the end of the day, this study tells me that the IDSA/ATS guidelines got it right. Five days is enough for patients who have become stable within that time. There are certainly good reasons to extend the course, but fever or CAP-associated instability appears to be a sufficiently complete list of those reasons. In the absence of one of these: CAP gets five days, no excuses.
Read the article here.