We wish to thank Dr. Dimopoulos for his interest in our study.1 Dr. Dimopoulus raises three issues. First, he asks if only respiratory fluoroquinolones were included in the fluoroquinolone group. We suspect that Dr. Dimopoulus is concerned that the poor activity of ciprofloxacin against Gram-positive organisms such as Streptococcus pneumonaie could explain why patients receiving fluoroquinolones had poorer adjusted outcomes than patients receiving macrolides. While we included patients receiving ciprofloxacin in the fluoroquinolone group, only 12 received fluoroquinolone monotherapy (1.3% of the fluoroquinolone patients), and only 1 of these patients died (8.3%); thus, it is unlikely that poorer outcomes in this small group explained our results. Dr. Dimopoulus correctly notes that various dosing regimens for levofloxacin are used around the world, and suggests that dosages > 500 mg/d have theoretical benefit based on pharmacokinetic and pharmacodynamic parameters. Although we did not abstract antibiotic dosages, we suspect that most patients received 500 mg/d of levofloxacin because other dosing regimens were not commonly used for community-acquired pneumonia prior to 2001 in the United States. However, we are not aware of any convincing data to suggest that higher doses of levofloxacin result in improved outcomes in patients with community-acquired pneumonia. Finally, Dr. Dimopoulus notes that in contrast to our results, randomized, controlled trials have demonstrated equivalent or superior outcomes with fluoroquinolones when compared to regimens including macrolides. As mentioned in our article,,1others have noted disparate results of randomized trials (almost all antibiotic registration studies) vs retrospective studies2–3 of patients with community-acquired pneumonia. This may be because patients enrolled in antibiotic registration trials are not representative of most pneumonia patients,2 or because retrospective studies are subject to bias from potentially important patient-related factors not being included in the multivariable analyses. Only a large randomized trial can answer this important question.