Multiple linear regression describing evolution of the Northern Hemisphere middle latitude total ozone during February-May (1979-2001) has been applied to estimate contributions of changes in the atmospheric circulation to observed long-term trends in total ozone. We use 4 predictors, namely: the 100 mb zonal mean heat flux averaged over the 40-85N region, the Arctic Oscillation index, quasi-biennial oscillation (-1 year lag) and solar cycle. The best fitting has been found using the heat flux and the AO index for the January-February period. The regression accounts for 70% of the observed total ozone variance in May, 79% in April, 75% in March, and 76% in February. There are no significant trends in the model residuals from March to May. The February trend is reduced by more than two times. Including in the model a term describing stratospheric aerosol loading does not significantly increase the regression sum of squares. Obtained results show that changes in the atmospheric circulation may be one of the important factors determining the long-term trends in the midlatitude ozone.