Simultaneous analysis of various anticancer drugs by supercritical fluid chromatography-mass spectrometry. Part I: Optimization of chromatographic separation

Abstract

This work presents a generic SFC-MS method for the simultaneous analysis of 22 anticancer drugs (fluorouracil, busulfan, cyclophosphamide, cytarabine, dacarbazine, daunorubicin, docetaxel, doxorubicin, epirubicin, etoposide, gemcitabine, idarubicin, ifosfamide, irinotecan, methotrexate, paclitaxel, pemetrexed, raltitrexed, topotecan, treosulfan, vinblastine, vincristine). The separation conditions were optimized by screening nine stationary phases (2-picolylamine, bare hybrid silica, 2-ethylpyridine, fluoro-phenyl, octadecyl, diethylamine, diol, 1-aminoanthracene, zwitterionic modification), evaluating additives effects (2–5 % water, 20–50 mM ammonium formate, 0–1 mM ammonium fluoride), and adjusting the organic modifier composition (methanol, ethanol, isopropanol, acetonitrile). The optimized SFC-MS method successfully analyzed 22 anticancer drugs, along with 5 additional challenging compounds (azacitidine, mitomycin, cisplatin, oxaliplatin, carboplatin), in 12 min, using a diol column (100 × 3 mm, 1.7 µm) and a gradient of 2–100 % methanol containing 2 % water and 50 mM ammonium formate. To overcome overpressure generated by high organic solvent content, a backpressure gradient (110–150 bar) and a flow rate gradient (0.6–1.5 mL/min) were applied. The diol column was selected as the most promising based on five predefined chromatographic criteria. Additives with 5 % water or ammonium fluoride were excluded due to overpressure and signal loss, respectively. Increasing ammonium formate concentration improved peak symmetry by 29 %. For the organic modifier, pure methanol was chosen since ternary mixtures led to system overpressure without improving separation. Comparison with the LC-MS method using real samples confirmed the potential applicability of the SFC method, as the same trace compounds were detected with comparable concentrations. Sensitivity optimization and method validation will be discussed separately in a later paper.