Background: In mammographic screening for breast cancer in women aged 40-49, previous studies have found very low estimates of sensitivity and predictive value. Methods of estimation have, however, used primitive models with relatively strong assumptions and less than full use of the data.
Methods: Here we estimate the underlying preclinical incidence, mean sojourn time, sensitivity and positive predictive value by a variety of methods in a randomized trial of mammographic screening (The Swedish Two-County Trial) and a service screening programme (The Florence Programme, 1975-1986) in women aged 40-49 years.
Results: In the Two-County study, sensitivity estimates ranged from 72-83% and predictive value from 39-89%. In the Florence programme, sensitivity estimates ranged from 69-85% and predictive value was uniformly estimated as 100%.
Conclusions: The methods involving more explicit modelling of the disease process and fewer assumptions tended to find higher estimates of predictive value in the Two-County study. The results suggest that previously poor sensitivity and predictive value estimates may have been overpessimistic.