Clinical Scenario and PICO question

The idea for this JC sprung out of the perceived clinical variability in our department in the treatment of COPD exacerbations with antibiotics.  As background, the 2004 ATS guidelines state that antibiotics may be initiated for outpatient or inpatient COPD exacerbations in patients with a change in their sputum characteristics (purulence and/or volume) and should be given if the patient is admitted to the ICU.  The 2010 GOLD (Global Initiative for Chronic Obstructive Lung Disease-joint project of NHLBI and WHO) guidelines recommend antibiotics for COPD exacerbations in patients with increased sputum purulence + increased sputum volume + increased dyspnea, or increased sputum purulence + increased volume or dyspnea, or intubated patients.

What is the evidence behind these expert recommendations?  As discussed below, it’s pretty modest.  Article #1 is the only recent RCT on the topic, Article #2 is a large retrospective cohort study evaluating outcomes in patients treated early with antibiotics, and Article #3 is a Cochrane review of all RCTs on the subject through 2005.

A historical word:  the background article by Anthonisen is still frequently referenced.  This study derived the most widely cited scale for grading the severity of acute COPD exacerbations, and combines pertinent symptoms into three categories:

Type 1 exacerbation:  All of the following symptoms are present: increased dyspnea, sputum volume, and sputum purulence.

Type 2 exacerbation:  Two Type 1 symptoms are present.

Type 3 exacerbation:  At least one Type 1 symptom is present plus upper respiratory tract infection symptoms or fever or increased wheeze or cough or increased resp. rate or heart rate.

In Anthonisen’s study, patients with Type 1 or Type 2 exacerbations had higher rates of clinical success and fewer episodes of deterioration.  Ok, it’s 24 years old, and a lot has changed in COPD management since then (BIPAP anyone?), but it’s good to be familiar with where it all started.

#1.  Daniels JMA, Snijders D, de Graaff CS et al.  Antibiotics in Addition to Systemic Corticosteroids for Acute Exacerbations of COPD. Am J Respir Crit Care Med 2010;181:150-157.

This RCT from the Netherlands compared outcomes in 265 COPD exacerbations (defined as increased dyspnea + increased sputum volume or purulence) treated with either doxy + steroids or placebo + steroids in hospital.   Patients with pneumonia were excluded.  The primary outcome measure was clinical response on Day 30.  Secondary outcomes included clinical success on Day 10, clinical cure on Days 10 and 30, lung function, CRP, symptoms, microbiological response and antibiotic treatment for lack of efficacy.  There was no difference in the primary outcome (61% clinical success in doxy group at Day 30, 53% in placebo group).  The authors stressed the positive results of the secondary outcome, clinical success at Day 10 (80% doxy, 69% placebo).  Clinical cure at Day 10 also was significantly higher in the doxy group.  Lung function results did not differ between the 2 groups, and symptom scores on a non-validated scale favored the doxy group at Day 10 but not at Day 30.  Bacteriological response was higher in the doxy group, but the clinical importance of this result is unknown, as colonization versus infection is difficult to determine, and persistent colonization even after appropriate antibiotic treatment is well described.  Serious adverse events occurred in 11 patients (9%) of the doxy group, including 7 deaths, as compared to 7 patients (5%) and 3 deaths in the placebo group.

Issues/Discussion:  As Christine brought up, Day 30 outcome is a strange choice.  We expect improvement earlier, and as Erin discussed, 10 day improvement can be very important with regards to improved quality of life, and also to help break the cycle of recurrent exacerbations.  Michelle mentioned the large number of exclusion criteria, limiting the external validity of the study (can we apply the results to our patients?).  Erik pointed out that the authors left out the number of eligible patients sampled (top of the fishbone diagram).  This is a recurrent theme, and increases the likelihood of a Type I (false positive) error, and therefore limits the internal validity of the study.  Erik also brought up the doubled mortality rate in the doxy group, which is not discussed further in the paper.

#2.  Rothberg MB, Pekow PS, Lahti M, et al.  Antibiotic Therapy and Treatment Failure in Patients Hospitalized for Acute Exacerbations of COPD. JAMA 2010;303:2035-2042.

This retrospective cohort study compared outcomes of hospitalized patients with COPD exacerbations but without pneumonia who either received antibiotics during the first two days of hospitalization or received late or no antibiotics.  The primary outcome was a composite measure of treatment failure that included mechanical ventilation, mortality, or readmission for COPD within 30 days.  Secondary outcomes included hospital cost and LOS and antibiotic associated adverse reactions.  Over the study period, 84,621 admissions from 413 acute care centers were included.  The risk of treatment failure was lower in the antibiotic treated patients (odds ratio 0.87; 95% CI, 0.82-0.92).  Each component of the composite primary endpoint was statistically lower in the group treated with antibiotics.  Patients treated with antibiotics had a higher rate of readmission for Clostridium difficile than those not treated with antibiotics (0.19%; 95% CI, 0.187%-0.193% versus 0.09%; 95% CI, 0.086%-0.094%).  

Issues/Discussion:  With retrospective studies it’s impossible to determine why certain patients did/did not receive treatment (antibiotics).  Although the authors attempted to correct for this using multivariable adjustments, the study remains vulnerable to selection bias.  As Matt discussed, the very large sample size allows the authors to look at rare outcomes (e.g. mortality) but with very large numbers, p values may be “significant” without there also being clinical significance (statistical versus clinical significance).   The readmission rate for C. difficile was low in both groups, but as Ben pointed out, the increasing rates of quinolone use lead to more C. difficile in the future.  Also, the patients treated with antibiotics were healthier overall, and this favors the antibiotic group.  Ben also calculated some numbers needed to treat.  Remember, the NNT = 1/ARR, and for the composite outcome, the NNT is 50.  For each of the components, the NTT is >100, so although the outcomes favor antibiotics, the treatment effect is modest.  Finally, Chintan made the important point that the study design will miss many patients with complications.  To be enrolled, patients needed a principal diagnosis of COPD or respiratory failure with COPD.  If a patient then had an MI, MI would likely be the principal ICD-9 code, and the patient would have fallen out of this study.

#3.  Ram FSF, Rodriguez-Roisin R, Granados-Navarrete A et al.  Antibiotics for exacerbations of COPD.  Cochrane Database of Systematic Reviews 2006, Issue 2. Art No.:  CD004403.

This Cochrane review includes all RCTs comparing antibiotics to placebo in patients with acute COPD exacerbations since 2005.  Eleven trials with 917 patients were included.  Their conclusion:  in COPD exacerbations with increased cough and sputum purulence, antibiotics significantly reduce the risk of short-term mortality (NNT 8; 95% CI 6-17) and treatment failure (NNT 3; 95% CI 3-5) with an increase in the risk of diarrhea (NNH 20; 95% CI 10-100) compared to placebo.  The authors admit that the results should be interpreted with caution due to differences in patient selection, antibiotic choice, the small number of trials, and the lack of controls for other interventions (steroids, BIPAP).  They acknowledge the growing resistance rates to common antibiotics and recommend limiting antibiotics to patients with COPD exacerbations and infectious symptoms who are moderately or severely ill.

Issues/Discussion:  As Jess discussed, there are huge variations in the definitions of COPD, the patient populations included, and there is no standard definition of “usual care”.  Gromis commented that the trials are in general old (only one published in the last 10 years), limiting the ability to generalize to current care of patients with COPD.  In addition, reviews like this suffer from publication bias:  studies demonstrating a positive treatment effect are more likely to be published than negative studies.

So, what’s our take home?  The data aren’t great, but modestly support antibiotics in patients with COPD exacerbations and infectious symptoms (in the absence of pneumonia).  Consensus of the faculty members in the room is to give antibiotics to patients with infectious symptoms.  Some give them to all COPD patients being admitted, and everyone agreed that antibiotics are indicated if the gestalt is that the patient looks “sick”.  The decision may be more difficult if the patient was just recently admitted or just recently finished a course of antibiotics.  Erik and Dan struck the cautionary note that given the doubled mortality in the first study for the antibiotic group, we may not be that far from clinical equipoise, and although they both prescribe a lot of antibiotics for these patients, we may also be doing harm (C. diff, etc).  As Dan pointed out, it’s ironic that in our last JC, we pushed hard against prescribing unnecessary antibiotics for AOM, and here we have a much more liberal prescribing policy.  Both conditions have reasonably high placebo success rates, but there are higher clinical consequences of treatment failure with COPD.  Final thought; if you’re giving antibiotics, think macrolides.  They appear to have anti-inflammatory effects at least partly independent of their anti-microbial effects, and are excellent agents against common respiratory pathogens.

Background Article