OBJECTIVES: To analyze the evolution of the pharmaceutical public expenditure and to evaluate the impact of supply measures of restraint over its growth. METHODS: Time series analysis through linear autoregressive multivariate models (January 1999 - June 2005). Source: monthly pharmaceutical invoices. Dependent variables: total public expenditure (E), number of prescriptions adjusted by days (P/d), and average public expenditure per prescription (E/P). Independent variables analyzed: trend, seasonal variation, cycles, number of days and lags when necessary. Impacts analyzed: introduction of maximum margins for wholesalers and chemists' and discounts over chemists' turnovers (D), and introduction of two reference pricing groups (RPG1 & RPG2 respectively). Lags are also introduced when residual autocorrelation occurs. Models are validated through normal contrast and no-correlation of residual. RESULTS: Adjusted R2 are 0.818, 0.684 and 0.814 in E, P/d and E/P when trend is considered as unique variable. There are no seasonal variation at any time series (p>0.1) when this variable is related to four seasons as a categorical variable. However, August has been entered in E and P/d models as a dichotomic variable (p<0.0001 & p<0.01). Number of days also entered in E model (p<0.0001). Annual cycle is observed in P/d and E/P model (p<0.05) and half-yearly and cuatrimestral cycles also entered in P/d model (p<0.05). Two supply measures present impacts at the short term. D has entered in the three models: E (p<0.001; Beta:-8.733.966), P/d (Beta: -13.100) and E/P (Beta:-0.18) (p<0.05 both). RPG2 has been also considered in E (Beta: -4.339.412) and E/P (Beta: -0.16) (p<0.05 both). Final adjusted R2: E: 0.937; P/d: 0.918; E/P: 0.906. CONCLUSIONS: trend is the most significant variable in the three models and when an impact is statistically significant, it seems not to present long-term sustainability because these supply measures have a short-term impact