Progression of critically ill patients from Systemic Inflammatory Response Syndrome (SIRS) to Multiple Organ Dysfunction Syndrome (MODS) accounts for more than 75% of deaths in adult surgical intensive care units. Currently, there is no practical clinical technique to predict the progression of SIRS or MODS. In this report, we describe an NMR-based metabonomic method to aid detection of these conditions based on abnormal metabolic signatures. We applied pattern recognition methods to analyze one-dimensional (1)H NMR spectra of SIRS and MODS patient sera. By using Principal Component Analysis (PCA) and Partial Least Squares-Discriminant Analysis (PLS-DA), we could distinguish critically ill patients (n = 52) from healthy controls (n = 26). After noise reduction by Orthogonal Signal Correction (OSC), PLS-DA was also able to clearly discriminate SIRS and MODS patients. The corresponding coefficients indicated that spectra responsible for the discrimination were located in delta3.06-3.86 NMR integral regions from SIRS, mainly composed of sugars, amino acids and glutamine signals, and delta1.18-1.3 and delta4.02-4.1 integral regions of MODS serum samples, principally consisted of various proton signals of fatty acyl chains and glycerol backbone of lipids, along with creatinine and lactate. Our results are consistent with the clinical observations that carbohydrate and amino acid levels changes in the early course of critical illness (SIRS stage) and significant disturbances in fat metabolism and development of organ abnormalities become the characteristics in the late stage (MODS). These data suggest that NMR-based metabonomic approach can be developed to diagnose the disease progress of critically ill patients.