In the USA, macrolides are the dominant choice for initial outpatient treatment for community-acquired pneumonia (CAP). However, there is no consensus on second-line therapy. Cephalosporins, quinolones, beta-lactams and other macrolides are all commonly prescribed. OBJECTIVE: To compare outcomes for different second-line treatment strategies in outpatient treatment of CAP. METHODS: We employed a previously developed and validated probabilistic model to calculate treatment failure rates for adult CAP with no major comorbidities. We considered four choices for second-line therapy following initial treatment with the macrolide azithromycin (AZI): cefuroxime (CEF), levofloxacin (LEV), amoxycillin-clavulanate (AMC) and clarithromycin (CLA). Distribution of causative pathogens was based on literature. We used surveillance data for the USA from the 1999-2000 PROTEKT study, and applied pharmacokinetic/pharmacodynamic breakpoints to determine resistance levels for streptococcus pneumoniae (SP) and haemophilus influenzae (HI). Intracellular organisms were assumed 100% susceptible to CLA, LEV but resistant to CEF, AMC. Other bacteria were assumed 50% susceptible to all four drugs. We used published estimates of pathogen-specific rates of spontaneous resolution. We assumed 20% of initial treatment failures and 100% of second-line treatment failures were hospitalized. We estimated that 90% of susceptible organisms would respond to treatment. RESULTS: In SP, the most common cause of CAP, there was substantial cross-resistance between AZI and the second-line drugs, except for LEV. For example, SP was 86.6% susceptible to CEF when it was susceptible to AZI, but only 16.3% susceptible to CEF when resistant to AZI. The model estimated that in 59.5% of cases the initial treatment with AZI would be successful and 5.9% would be hospitalized following initial treatment failure. An additional 5.2% would be hospitalized if treated with LEV, 9.1% with AMC, 13.3% with CEF and 19.3% with CLA. CONCLUSION: Selection of appropriate second-line therapy should be based on coverage and local cross-resistance patterns.