An Optimal Therapy for Community-Acquired Pneumonia?

Andrew F. Shorr, MD, MPH


July 30, 2015

This feature requires the newest version of Flash. You can download it here.

This is Andy Shorr from George Washington University in Washington, DC, with the Pulmonary and Critical Care Literature Update. I want to discuss an article by Postma and colleagues[1] in the April 2 issue of the New England Journal of Medicine. These investigators asked a very important question for those of us who care for hospitalized patients with pneumonia: What is the optimal therapy for pure community-acquired pneumonia, in terms of mortality?

Conflicting Practice

Historically, our three options have been beta-lactam monotherapy, beta-lactam plus a macrolide, and fluoroquinolone monotherapy. Guidelines from the American Thoracic Society and the Infectious Diseases Society of America[2] stress the use of an agent that will cover atypical pathogens in hospitalized patients, especially in non-ICU patients who will have pathogens that may be different from those found in ICU patients with pneumonia.

Thus, for the general hospital patient, the guidelines recommend monotherapy with a fluoroquinolone, or combination therapy with, for example, ceftriaxone and azithromycin. Although US guidelines have never recommended beta-lactam monotherapy, numerous European guidelines[3,4,5,6,7] have recommended the use of beta-lactam monotherapy in patients admitted to hospital with community-acquired pneumonia.

Controversial Literature

The literature on this is controversial. Most randomized controlled trials that have compared macrolide combination therapy vs fluoroquinolone monotherapy were conducted simply to attain regulatory approval of novel agents. These studies included low-risk populations and were never designed to look at anything other than short-term efficacy and safety.

Several large, observational, retrospective studies[8,9] suggest that the addition of a macrolide to any regimen for community-acquired pneumonia in patients admitted to the hospital, particularly critically ill patients, is associated with a mortality benefit. Those data are very limited, however, precisely because they typically come from large, retrospective datasets; they do not include very granular patient information. Thus, these have led to a lot of controversy. Moreover, some recent studies suggest that macrolides may be associated with safety issues, although, again, that remains controversial.

In short, this is a crucial question, given the frequency with which patients with community-acquired pneumonia are admitted to hospitals and about which we do not have strong data. That is why this new study is so thought-provoking.

The CAP-START Clinical Trial

The CAP-START study[1] was a cluster randomized trial. Investigators did not randomly assign individual patients to one of the three treatment strategies. Instead, they randomly assigned an entire cohort of patients who were admitted to a given hospital during a set period of time to receive one of three regimens in 4-month rotations: beta-lactam monotherapy, beta-lactam-macrolide combination therapy, or fluoroquinolone monotherapy. Thus, the level of randomization was essentially the rotation of the protocol at the hospital. The study was designed as a noninferiority study looking at 90-day mortality.

In all, approximately 2200 patients were enrolled. The authors found that 90-day mortality rates were similar across the three groups; overall, 90-day mortality rates were about 10%.

Proceed With Caution

These results should be viewed very cautiously. First, the study was conducted in The Netherlands, where there is notoriously little antibiotic resistance. Pneumococcal resistance to beta-lactam therapy occurs infrequently there, whereas it is much more of a concern in the United States. Whether beta-lactam resistance among pneumococcal infections portends worse outcomes is controversial because pneumococcal resistance can be overcome with continuous infusion of a beta-lactam, high doses of beta-lactam, and what have you. But one must raise a question about the generalizability of these data, given that they come from a region with very little pneumococcal resistance compared with a region such as the United States, where there may be much more beta-lactam resistance.

Second, only about three quarters of the patients in this study had radiographically confirmed infiltrates. If you do not have an infiltrate on a chest x-ray, I am not quite sure how you have pneumonia. We all know that a portable chest x-ray is a poor screening test and may have missed many abnormalities one would see on a lateral view or at the bases. In the absence of a clear radiographic infiltrate, I am not sure that it is appropriate to enroll these patients in a study of community-acquired pneumonia. Although radiographic infiltrates may not be sufficient, they certainly should be a necessary condition to qualify for enrollment. But in this study, as many as 25% of the patients did not have radiographic infiltrates.

In a subgroup analysis, these authors did look at outcomes as a function of whether there was a radiographic infiltrate. This resulted in no difference in their overall conclusions; however, I still believe that it raises concerns, especially given the design of this trial as a cluster randomization rather than an individual patient-level randomization.

Another Caveat: Critically Ill Patients Excluded

This study specifically excluded patients in the critical care unit. The authors are quite straightforward about that and should be commended for this; however, excluding critically ill patients raises more questions about generalizability of results. The median CURB score was 1. (The CURB score assesses confusion, elevated blood urea nitrogen, respiratory rate, and low blood pressure.) Many patients with a CURB score of 1 would never be admitted to a US hospital. Similarly, the Pneumonia Severity Index (PSI) score was about 85 for a median PSI class of 3 with a predicted in-hospital mortality rate of less than 1%. Thus, in addition to not being admitted to a US hospital, these patients do not represent a population in the United States that would have been eligible for the study.

In the same vein, the length of stay, which was one of their secondary endpoints, was similar across the three arms. The median length of stay for these not very sick patients with community-acquired pneumonia was 6 days. A patient with CURB-1 pneumonia who is admitted to the hospital would never stay for 6 days unless some social catastrophe prevented the patient's discharge. It would not make sense. If the median length of stay is 6 days, then half the patients were in the hospital longer than that, and so, again, you have to wonder about who they were studying. Some of these patients probably did not have pneumonia, some had bronchitis, some patients never needed to be in the hospital, or certainly would not be in a US hospital, given the low severity of illness.

More Issues

I have no doubt that the authors executed the study very well. They are a well-respected group of investigators. They took on a challenging and difficult question and conducted this study with limited resources. Despite the shortcomings, they executed it well. They even looked at how frequently patients adhered to the strategy that they were assigned. It is a very thought-provoking study, but when it comes down to issues about generalizability and applicability in the United States, we need to be cautious.

Another issue needs to be pointed out. They could not conclusively determine beta-lactam monotherapy noninferiority to fluoroquinolone therapy. In some cases, at the 95% confidence interval, the fluoroquinolone monotherapy may have been associated with some benefits relative to beta-lactam monotherapy. This is a question about statistics and what the threshold for noninferiority should be, or the best way to adjudicate it or to determine it; however, it does raise concerns that we cannot make as strong enough a conclusion from the data as the authors are trying to do.

This is a very thought-provoking article about a very crucial question. I urge you to read and discuss it in journal club with your fellows if you are at an academic program, or simply read it for your own practices. I am sure it will be discussed at guidelines committees at local hospitals.

This is Andy Shorr from Washington, DC.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: