We retrospectively analyzed the records of 61 hospitalized patients with community-acquired pneumonia (CAP) caused by Streptococcus pneumoniae or Legionella pneumophila. We found that serum procalcitonin and sodium concentrations were significantly lower, and ferritin levels were significantly higher, in patients infected with L. pneumophila than in those infected with S. pneumoniae. The ratio of C-reactive protein to procalcitonin significantly distinguished between the groups. High procalcitonin levels were associated with an adverse clinical course.Community-acquired pneumonia (CAP) may be caused by either Streptococcus pneumoniae or Legionella pneumophila (5,19). The initiation of adequate empirical antimicrobial therapy can be challenging, since S. pneumoniae and L. pneumophila have partly contrasting antimicrobial susceptibility patterns (11). We thus questioned whether the determination of C-reactive protein (CRP) and procalcitonin (PCT) levels in serum could be helpful for the differential diagnosis of S. pneumoniae or L. pneumophila infection. While determination of CRP levels can be helpful for the differentiation between viral and bacterial pulmonary infections (7), elevated PCT levels have been linked to a poor prognosis in CAP and specifically in Legionella pneumonia (9, 15). However, while PCT was linked to the severity of the underlying cause, CRP was more closely linked to the presence of infections in a prospective study (8).We retrospectively analyzed the records of 61 patients admitted to the University Hospital of Innsbruck (Innsbruck, Austria) between 2005 and 2008 with CAP caused either by S. pneumoniae (n ϭ 37) or by L. pneumophila (n ϭ 24). Pneumonia was diagnosed on the basis of clinical criteria and laboratory evidence of infection and was confirmed by radiographic evidence. For species diagnosis, urinary antigen tests for S. pneumoniae or L. pneumophila, the latter detecting serotype 1 (BinaxNOW; Inverness Medical, ME), were used (1,4,5,10,17).Baseline laboratory parameters were determined by routine automated tests, and data were available from day 1 (admission) and days 5 to 7. PCT levels were determined by a timeresolved amplified cryptate emission technology assay (Brahms AG, Henningsdorf, Germany) (2).All statistical analyses were performed with the SPSS statistical package (version 11.5; SPSS, Chicago, IL). For nonnormally distributed data. nonparametric tests were applied (Mann-Whitney test). The Spearman rank correlation technique was used for analysis of associations. All tests were two-sided, and a P value of Ͻ0.05 indicated statistical significance. Binary logistic regression analysis was used to identify parameters predictive of the patient outcomes. The ratio of CRP concentrations to PCT concentrations was calculated, and the logarithmic values were used to identify cutoffs for discrimination between L. pneumophila and S. pneumoniae infections. The diagnostic value of these cutoffs was evaluated by calculating sensitivity, specificity, and positive and negative predictive ...