Bioanalytical method analytical range - non-linearity
Posted: Thu Jul 14, 2011 9:47 am
We employ a triple quad (API3000) LC-MS system in the course of analysing tk samples from toxicology studies. Sample clean up of the plasma samples is typically SPE or SLE (supported liquid extraction). We tend to set the analytical ranges of the methods at 500-fold as we are not normally able to achieve four orders of magnitude. We have many methods validated over such a range, however, from time to time we observe a compound for which the calibration line curves away at the high end excessively (a result of "no intercept" is returned for high calibrators). In my mind I have a level of signal which we try not to exceed to prevent the excessive curvature (we do employ quadratic regressions sometimes). We have one current compound for which the excessive curvature is happening, and not even a quadratic fit resolves the issue. 20uL of plasma are extracted by SLE and ultimately the extracts are redissolved in 300uL, so I would consider that the samples are relatively pure (10% suppression at the most and 85% recovery) and dilute, and that the signal we calculate for the high calibrator should be achievable by the detector. I wondered if others have experienced this and resolved matters. I should add that mobile phases are A 0.1% formic acid and B methanol with a gradient employed of 10% B to 90% B over two minutes at 0.8mLmin using a 50x2mm, 5uM Luna C18 column. Does replacing the formic acid with a buffer like ammonium formate (i.e. increased ionic strength) in general lead to greater linear ranges ? Am I experiencing a column/ion source loading issue ? I'm confident we are injecting little in the way of drug/matrix components.