Background: Acute respiratory distress syndrome (ARDS) is a challenging clinical problem. To date, no standardized diagnostic biomarker has been validated for community-acquired pneumonia (CAP)-induced ARDS. An integrated analysis of changes in the metabolic profile could help detect biomarkers for early prediction of ARDS development and evaluation of treatment efficacy.Methods: A total of 88 patients were enrolled for the final analysis and divided into two groups: the ARDS group (n = 43) and the no-ARDS group (n = 45). We examined differences in serum and urine metabolites and explored dynamic changes with nuclear magnetic resonance (NMR) spectroscopy.Results: A total of 20 serum and 42 urine metabolites were identified using NMR spectroscopy. Serum metabolites, including leucine, 3-hydroxybutyrate, lactate, acetone, citrate, and choline, and urine metabolites, including creatine and creatinine, could distinguish patients with CAP with and without ARDS, with areas under the receiver operating characteristic curve (AUC) values of 0.790 and 0.747, respectively. The treatment efficacy of patients with ARDS was achieved at an AUC of 0.862 with serum metabolites 3-hydroxybutyrate, lactate, acetone, acetoacetate, citrate, and choline, and of 0.691 with urine metabolites taurine and glucose. The treatment efficacy of patients without ARDS was achieved at an AUC of 0.845 with serum metabolites alanine, acetate, acetoacetate, glutamine, creatine, and glucose and 0.891 with urine metabolites choline, tryptamine, and 3-indoxyl sulfate. We also proposed a combined biomarker of associated serum and urine metabolites to predict ARDS in patients with CAP (AUC = 0.865) and evaluate treatment efficacy for patients with and without ARDS (AUC = 0.921 and 0.893, respectively).Conclusions: Serum and urine analyses showed that metabolomics provides potential circulatory markers for early prediction and evaluation of treatment efficacy in patients with CAP with and without ARDS.