The validity of methods for reconstructing historical rates of smoking prevalence has not been assessed before. Our objective was to assess their validity. We reconstructed smoking prevalence rates for each calendar year from 1940 to 2007 for men and women in Spain, using data on ages of smoking initiation and cessation available in the Spanish National Health Surveys of 2003-2004 and 2006-2007. To assess the validity of the reconstruction, we computed the differences between the reconstructed smoking prevalence and the contemporary observed smoking prevalence measured in the Spanish National Health Surveys of 1987, 1993, 1995, 1997, and 2001. We also compared reconstructed smoking prevalence trends with 35-year lagged lung cancer mortality rates in Spain as a proxy for the real prevalence trends. Reconstructed smoking prevalence rates compared with contemporary measured rates showed small differences in men (between -2.1% and 2.1%) and an overestimation in women (between 2.0% and 5.7%). Reconstructed smoking prevalence trends were significantly correlated with lagged lung cancer mortality trends (P = 0.004 for men, P < 0.0001 for women). The reconstruction of smoking prevalence rates through this methodology offers a feasible tool with which countries lacking previous smoking surveys can understand historical trends in their tobacco epidemic, which aids in designing and implementing adequate tobacco control interventions.
Keywords: developing countries; prevention and control; smoking; statistical models; validation studies.