An improved LC-MS/MS method for the determination of tobacco specific nitrosamines has been developed, utilizing four internal deuterated standards instead of two. This was found to be necessary due to an analyte dependent sample matrix effect, which can lead to both a decrease in accuracy and in precision. In addition, two ion transition pairs (instead of one) were utilized, for confirmation and quantification. The linear range for the method was from 0.5 to 200 ng/mL with an analyte dependent limit of detection of from 0.027 to 0.049 ng/mL. This method and Health Canada methods T111 (mainstream smoke) and T309 (cigarette filler) were applied to 6 brands of Canadian cigarettes (ISO tar 4-15 mg) collected in 2003, 2004 and 2005. Smoke yields of tar, nicotine, CO, benzo[a]pyrene, and selected volatiles were also determined as a basis for comparison (Health Canada methods T115, T103 and T116 respectively). Yields of NNN, NNK, and NAT (expressed as ng/mg tar) all decreased by about 60% during this time period with smoke yields of NAB falling below detection limits. During the same period, yields of other smoke constituents showed little change. By 2005, levels of NNN, NNK and NAB in cigarette filler fell below detection limits. Consequently, with respect to Canadian cigarettes, significant progress has been made during the past three years in reducing the TSNA content of cigarette filler and yields in mainstream smoke.