Due to the need of accurate control of their therapeutic doses, warfarin-containing medicines are marketed in preparations with low, but different warfarin contents. A single near infrared reflectance procedure is proposed for warfarin determination in four different doses of a commercial preparation, with PLS as multivariate calibration tool and with an active pharmaceutical ingredient proportion ranging from one to five per cent of mass tablet. A mixed calibration sample set, composed of commercial samples and laboratory compacted tablets was prepared. In order to achieve suitable accuracy and precision, the difference in scattering between laboratory and commercial tablets was reduced by studying the effect of the compaction pressure. Also a spectral pre-treatment was applied, consisting of a second derivative followed by the Standard Normal Variate (SNV) normalization of the derivative spectra. The wavelength range 1480-1560 nm had to be deleted since a shift of an absorption band of lactose was detected after tablet compaction under laboratory conditions. No significant differences were found between the NIR prediction and HPLC reference data of the validation sample set. © 2013 The Royal Society of Chemistry.