Mid-infrared (MIR) spectroscopy is a valuable analytical method for patient monitoring within point-of-care diagnostics. For implementation, quantum cascade lasers (QCL) appear to be most suited regarding miniaturization, complexity and eventually also costs. External cavity (EC) - QCLs offer broad tuning ranges and recently, ultra-broadly tunable systems covering spectral ranges around the mid-infrared fingerprint region became commercially available. Using such a system, transmission spectra from the wavenumber interval of 780 to 1920 cm-1, using a thermoelectrically cooled MCT-detector, were recorded while switching the aqueous glucose concentrations between 0, 50 and 100 mg/dL. In order to optimize the system performance, a multi-parameter study was carried out, varying laser pulse width, duty cycle, sweep speed and the optical sample pathlength for scoring the absorbance noise. Exploratory factor analysis with pattern recognition tools (PCA, LDA) was used for the raw data, providing more than 10 significantly contributing factors. With the glucose signal causing 20 % of the total variance, further factors include short-term drift possibly related to thermal effects, long-term drift due to varying atmospheric water vapour in the lab, as well as wavenumber shifts and drifts of the single tuners. For performance testing, the noise equivalent concentration was estimated based on cross-validated Partial-Least Squares (PLS) predictions and the a-posteriori obtained scores of the factor analysis. Based on the optimized parameters, a noise equivalent glucose concentration of 1.5 mg/dL was achieved.