A model for the effect of cigarette smoking on lung cancer incidence in Connecticut

Population based data on smoking history derived from NCHS surveys were used to develop a model for lung cancer incidence in Connecticut. Trends in smoking prevalence suggest that, while the prevalence in men increased earlier than women, more male smokers have quit than their female counterparts. These trends in smoking prevalence suggest striking gender differences in a period effect for the smoking prevalence. Estimates of the proportion of current smokers, ex-smokers, and the mean duration of smoking were used in a model for the lung cancer incidence rates. The form for the relationship between smoking history and the incidence rate for these subgroups was based on information from cohort studies. The models represented a mixture of the smoking subgroups where the effect of smoking was considered to be either a multiplicative effect on the underlying age distribution, or a separate effect in which the level of exposure was the sole contribution to risk among smokers. The multiplicative model explained more than 80 per cent of the deviance for the period and cohort effects, while the non-multiplicative model could only account for trends in females. Hence, these results suggest that a sizeable portion of the period and cohort contributions to the lung cancer incidence trends in Connecticut can be attributed to the multiplicative model that utilizes this smoking information, although the lack of more detailed information is a limiting factor in developing the model.

Estimating age, period and cohort effects using the multistage model for cancer

To understand cancer aetiology better, epidemiologists often try to investigate the time trends in disease incidence with year of diagnosis (period) and birth cohort. Unfortunately, one cannot identify these factors uniquely in the usual regression model owing to a linear dependence between age, period and cohort, so that one requires additional information about the underlying biology of the disease. Carcinogenesis models provide one type of information that can result in a unique set of parameters for the effects of age, period and cohort. We use the multistage carcinogenesis model and its extensions to obtain a unique set of parameters for an age-period-cohort model of lung cancer trends of Connecticut males and females from 1935 to 1988. Some of these models do not seem to provide a reasonable set of model parameters, but we found that a model that included second-order terms and a multistage mixture model both gave a good fit to the data and realistic parameter estimates.

Time trends of non-Hodgkin's lymphoma: are they real? What do they mean?

Factors that need to be considered in the analysis of time trends in disease incidence are age, year of diagnosis, and birth cohort. When these are included in a log-linear model, a nonidentifiability problem arises from the linear dependence among these three time factors so that only specified functions of the parameters can be unambiguously determined. One of these invariant functions is the drift or the sum of the period and cohort trend. Non-Hodgkin's lymphoma incidence rates from Connecticut for the period 1935-1989 were analyzed for males and females. In addition to an age effect, both period and cohort significantly improved the fit of the model. The estimated drift shows that there has been a 10.3% increase in risk every 5 years since 1965 for females and 9.2% for males. It is unlikely that a trend of this magnitude can be attributed entirely to data artifact.

Trends in female breast cancer in Connecticut and the United States

Trends for female breast cancer were examined by age, period and cohort for the years 1950-1984 in U.S. mortality. Connecticut mortality and Connecticut incidence. Birth cohort patterns were evident for all three sets of data. The results confirm a continuing increase in invasive breast cancer by providing evidence of a strong birth cohort pattern, over a time series of 90 years of birth cohorts. This trend appears to be real for the most part because of the cohort patterns and because there is minimal underdetection in data obtained by autopsy and blind biopsy. Secondly, when cohort modeling is applied to breast cancer mortality, there is an indication of a modest decline in recent breast cancer mortality (in the face of an apparent long-term increase), which suggests that control of breast cancer mortality may have developed in recent decades, perhaps through earlier detection and improved treatment. Finally, in contrast with a prior report, there is little evidence for a clinically important difference in time trend between pre- and postmenopausal breast cancer.