RATIONALE. The risk for selection of antibiotic resistance (ABR) is increased by unnecessary use of antibiotics (ABs). Unnecessary AB use occurs when patients:
i) with a suspected bacterial infection are treated with ABs empirically while the infection is in fact viral;
ii) are treated with a broad-spectrum AB when a narrow-spectrum AB could be used, but cannot be chosen because the underlying bacterial pathogen is unknown;
iii) are treated with ABs but do not respond adequately, while this lack of response is not immediately observed;
iv) are treated with ABs while the infection is already eradicated and AB treatment can be stopped.
To address these situations, diagnostic- and treatment response biomarkers are of critical importance. Diagnostic biomarkers should discriminate bacterial- from viral infections to allow early stopping of AB treatment, or should identify bacterial infections for which a narrow spectrum AB can be used. Treatment response biomarkers identify when the infection is eradicated and treatment can be stopped, or when a switch should be made to another AB in case of treatment failure. Assays for such biomarkers should be fast, highly specific and sensitive. However, current standard diagnostic assays still rely on bacterial culturing and have turnaround times of >48h, while treatment response biomarkers such as C-reactive protein lack sensitivity and specificity. As such, a need exists to improve the performance of such biomarkers in order to reduce unnecessary AB use. Metabolomics is emerging as a key bioanalytical method for biomarker discovery. Its relevance has been shown for bacterial infection including tuberculosis and bacteremia [1–7]. Community-acquired pneumonia (CAP) represents a major infection type where improved biomarkers may significantly reduce ABR. CAP is highly prevalent, and patients receive extensive and precautious AB treatment, while for up to 50% of CAP cases no microbial organism is found.
OBJECTIVES. This project develops and validates predictive metabolomic fingerprint biomarkers (MFBs) for diagnosis and treatment response monitoring of patients with suspected bacterial infections. We will develop MFBs based on patients with CAP. The MFBs are generated as follows: state-of-the-art metabolomic profiling is applied to patient plasma samples, after which the metabolomic profiles are processed using computational algorithms that generate the MFBs. The MFBs can be considered as clinically and biologically weighted fingerprints of a selected set of molecules associated with the host response to infections. The goal of this project is to deliver an optimized rapid (<1 h) metabolomic assay, in conjunction with a computational algorithms to generate the diagnostic and treatment response MFBs. The MFBs will be validated using additional plasma samples from patients with CAP and patients with other types of suspected bacterial infections. Finally, we will make preparations to guide further clinical implementation.
APPROACH. This project will use plasma samples from study cohorts associated with 3 previously conducted randomized controlled trials (RCTs) Cohort 1 is used for MFB development, while cohorts 2 and 3 are used for MFB validation. We will develop the diagnostic and treatment response MFBs based on plasma samples from cohort 1 (CAP patients). Samples taken at the start of treatment will be used for the diagnostic MFBs, while samples taken during treatment will be used for the treatment response MFBs. To this aim, samples from cohort 1 will be analyzed using global metabolomic profiling, entailing 6 bioanalytical platforms covering a large number of biologically relevant molecules. Subsequently, we will develop an optimized rapid (<1h) metabolomic assay for selected molecules. The resulting metabolomic profiles will be used to develop the actual MFBs through computational analysis, in combination with information about the cause of infection and clinical treatment response metrics. Computational analysis will include state-of-the-art modeling approaches, and will include patient characteristics, biological knowledge, and AB dosing, in order to fully optimize their predictive value. Finally, the performance to predict diagnosis and treatment response by the MFBs will be validated using RCT study cohorts 2 and 3.
SIGNIFICANCE AND INNOVATION. This project is based on an large number of high-quality clinical samples. The use of metabolomics is highly innovative and may overcome some of the intrinsic limitations of alternative technologies. The use of the proposed state-of-the-art computational modeling methods has proven to be clinically relevant in other areas of therapeutic treatment optimization. Combining these approaches with an experienced and inter-disciplinary team with clinical, metabolomic and computational expertise, this project will make a significant step towards reduction of AB use and ABR development.