Declinig Hazard rate in Breast Cancer

Presented at the conference The Challenge of Breast Cancer
Organized by "The Lancet" in Brugge, Belgium, 1994.
(The Lancet vol. 343, 1085, 1994)

Survival curves and hazard rates of 118,329 females with breast cancer were computed from data of the recoded SEER public access record version June, 1988 (Cancer Surveillance, Epidemiology, and End Results Program). The cases were diagnosed in 1973-85. Cancers were staged into: in situ, localised, regional and distant Patients alive at cutoff date were censored otherwise no censoring was applied. A patient was included even if cause of death was unknown.

I assumed that the observed hazard rate µo consists of two components: aging hazard µa which stands for death due to causes other than cancer, and cancer hazard µc, and that µo = µa+ µc. The change of the two components with time after diagnosis (µc(t), µa(t)), was investigated. Since patients with in situ cancer do not die from their disease, their observed hazard was taken as estimate of the aging hazard (µc=0, µa=µo).

In in-situ cancer the observed hazard increases with age and with time. In localized and regional breast cancer, up to the third year the hazard increases after which it declines. The hazard of distant cancer declines from the time of clinical detection. Obviously the same aging hazard operates in all stages, and in all it increases with time. The cancer hazard µo was the sole contribution since in clinical cancer from the third year and onward the observed hazard declines with time.

The declining hazard indicates that despite the fact that the patient eventually dies, throughout the course of the disease she continually improves her condition. When her resistance is exhausted she is overcome by the disease and dies. The longer the patient lives the better her chances of survival. The nature of this resistance is still unknown.

Click here for the detailed statistical analysis