Beta-lactam alone compared to b-lactam-macrolide combination therapy for community-acquired pneumonia: prospective, observational study using a propensity score

Abstract number: 1733_356

Paul M., Nielsen A., Gafter-Gvili A., Tacconelli E., Andreassen S., Almanasreh N., Goldberg E., Cauda R., Frank U., Leibovici for the TREAT Study Group L.

Objectives: Observational studies have shown improved outcomes for patients treated with b-lactam (BL)-macrolide combinations compared to single BL treatment for CAP. However patients treated for an atypical pathogen are probably a-priori different from patients treated with a BL drug alone. Physicians are likely to reflect in their choice of treatment common wisdom as to the presentation of ``atypical'' pathogens. We used a propensity score to adjust for these differences and compared treatment groups.

Methods: Patients were enrolled as part a multicentre trial assessing a decision support system for antibiotic treatment in Israel, Germany and Italy. We compared all 30-day mortality for patients with radiologically proven CAP treated empirically with BL alone vs. combination therapy. Baseline characteristic of the two treatment groups were used to develop a propensity score for combination therapy and comparisons were matched by patients' propensity score. In addition, we used the propensity score as a covariate in a logistic model for mortality.

Results: Patients treated with BLs alone (N = 169) were older (mean age 70.6±17.3 vs. 65.0±19.6 years), more often residents of nursing homes (9% vs 4%), bed-ridden (60% vs. 40%), had a higher chronic diseases score and a different clinical presentation compared to patients given combination therapy (N = 282). Accordingly, the propensity scores differed markedly, 0.179±0.139 vs. 0.074±0.103, p < 0.001, respectively. Unadjusted fatality was significantly higher with BL monotherapy 22% vs. 7% (p < 0.001) and remained significantly higher on multivariate analysis without adjusting for the propensity score. Only 27 patients in the BL group could be matched to (27) patients in the BL-macrolide group using the propensity score with a precision of 3 figures after the decimal point. Among these patients, mortality was identical in the two study groups, 3 demises each (11%, p = 1.0, OR = 1.0, 95% CI 0.2–5.5). A multivariable analysis for mortality, adjusting for the propensity score showed no significant difference between the study groups (OR = 0.84, 95% CI 0.27–2.51).

Conclusions: Patients given single BL treatment for CAP are markedly different from patients given combination therapy. Only a small percentage of patients could be matched using a propensity score and when matched mortality is identical. Classical multivariable techniques may not adjust correctly for these differences.

Session Details

Date: 31/03/2007
Time: 00:00-00:00
Session name: European Society of Clinical Microbiology and Infectious Diseases
Location: ICC, Munich, Germany
Presentation type:
Back to top