Interpretation of disease time trends: is cancer on the increase? A simple cohort technique and its relationship to more advanced models.

Analysis of trends in mortality from respiratory cancer among women shows that, contrary to previous interpretation, there is no suggestion of any increase over time which might be due to recent increases in exposures to carcinogens. Although there are upward trends in the number of deaths, the crude death rate, and the age standardised rate, these are shown to be related to the aging of earlier cohorts of women who have experienced high mortality rates. More recent cohorts, born since the middle 1920s, show a decline. A simple technique to identify trends in different cohorts is described, and it is shown to be linked to the age-period-cohort modelling approach to investigating time trends. Dangers inherent in ignoring either period or cohort effects when describing one of these factors are discussed.