Analysis of cervical cancer mortality and incidence data from England and Wales: evidence of a beneficial effect of screening

Incidence and mortality rates for invasive cervical cancer in England and W ales have changed considerably since records began in the 1950s. It is well known that cervical cancer rates show substantial birth cohort effects and these make the interpretation of secular trends difficult. Here incidence data (from 1971) and mortality data (from 1950) are analysed by using gener alized additive models. The goodness of fit is evaluated by using residual plots and scaled changes in deviance. The incidence of squamous cell and ad enocarcinoma of the cervix were analysed separately after making an adjustm ent for those with unknown histology. Although the age and cohort model pro vides an adequate description of the data up to the mid-1980s, additional t erms are required to model the more recent data. Reasonable fits are obtain ed by adding functions of calendar year since 1978 in each of three age ban ds. Estimates of the cohort effects obtained from the incidence of squamous cancer and mortality data are remarkably similar and fit with hypothesized effects of changing sexual norms on cervical cancer rates. Secular trends suggest that cervical screening had little effect on cervical cancer rates before the mid-1980s and that since then the effect on mortality and incide nce of squamous cell carcinoma of the cervix has been substantial whereas t he effect on the incidence of adenocarcinoma of the cervix has been modest. Secular trends are greatest in younger women. They are minimal in women ag ed 70 years and over. If the secular trends are attributed to cervical scre ening, the models estimate that, as a result of screening, approximately 60 00 fewer deaths occurred between 1991 and 1997 and that approximately 4000 fewer squamous cancers were diagnosed between 1989 and 1992.