last update of this page: 29 July 2005
click here for medline reference of publication preceding this additional discussion
The letter to the editor of this chapter explains that a higher incidence in the screened arm of a trial relative to the control arm may very well be a very temporary effect while on the long run there is hardly any extra incidence due to breast cancer screening. The letter concentrates on how alarming the difference of incidence in the two trial arms is after limited follow up, but does not extensively go into estimating how large extra incidence caused by breast cancer screening is on the longer run. For the model that was used for this letter, we assumed exponential distributions of the dwelling times of the preclinical disease states. This assumption has shown to be in agreement with observations in screening studies that measure detection rates and interval cancer incidence. However the amount of (long term) extra incidence is largely determined by the tail of the dwelling time distribution. The shape of the sojourn time distribution would appear as the difference between the interval cancer incidence by time since last screening and the expected incidence in a situation without screening. It is not possible to measure this difference with any satisfactory precision for long follow up after last screening. Therefore one may wonder if the model estimate for extra incidence has any precision. If one would consider dwelling time distributions to be Weibull with shape parameter 0.5 and the same average duration, then the model expectation for extra incidence caused by the current Dutch screening programme is 5.6% as compared to 5.2% expected extra incidence when assuming exponential distributions. That difference is so small because in order to have the same average duration, increasing the number of very slow growing cancers needs to be compensated by increasing the number of very fast growing cancers and they lead to very little extra incidence.