We looked separately at the age cohorts and the geographic regions.1 For the 15 age cohorts, the correlation coefficients between rates of any adjacent years were greater than 0.90 for both incidence and mortality. Random variations can therefore not explain the lack of association observed. For the 15 geographic regions, by using the mean rates from 4 consecutive years (1986– 1989 for incidence and 1992–1995 for mortality) in the calculation of baseline rates, we accounted for the random variation that could have blurred the data at the baseline end of the rate difference equations. Moreover, relying on a single year (1993) for incidence increase, at the other end of the incidence rate difference equation, should have been fairly reliable. The correlation coefficients between the incidence rates of any adjacent years during the period where we measured incidence increase (1990–1995) ranged from 0.55 to 0.68 and were all statistically significant.
However, since the correlation coefficients between mortality rates of any adjacent years of the period where we measured mortality reduction (1995–1999) were weaker (–0.09 to 0.54), relying on a single year (1999) for mortality reduction in the mortality rate differences equation was a limitation. As we mentioned in our Discussion, the change in incidence should be reflected downstream on mortality rates over several years.
Nevertheless, while awaiting a more definitive answer that might come from randomized control trials, we believe that our study produced useful information. When the data become available, the study could be replicated using rates from several years at baseline and after the presumed lead-time period. This will measure the mortality rate differences with greater precision.
Linda Perron Lynne Moore Isabelle Bairati François Meyer Centre de recherche en cancérologie de l'Université Laval Faculté de médecine Université Laval Quebec City, Que. Paul-Marie Bernard Département de médecine sociale et préventive Faculté de médecine Université Laval Quebec City, Que.
Reference
- 1.↵