Multimodal hazard rate for relapse in breast cancer: quality of data and calibration of computer simulation

Abstract

Much has occurred since our 2010 report in Cancers. In the past few years we published several extensive reviews of our research so a brief review is all that will be provided here. We proposed in the earlier reports that most relapses in breast cancer occur within 5 years of surgery and seem to be associated with some unspecified manner of surgery-induced metastatic initiation. These events can be identified in relapse data and are correlated with clinical data. In the last few years an unexpected mechanism has become apparent. Retrospective analysis of relapse events by a Brussels anesthesiology group reported that a perioperative NSAID analgesic seems to reduce early relapses five-fold. We then proposed that primary surgery produces a transient period of systemic inflammation. This has now been identified by inflammatory markers in serum post mastectomy. That could explain the early relapses. It is possible that an inexpensive and non-toxic NSAID can reduce breast cancer relapses significantly. We want to take this opportunity to discuss database quality issues and our relapse hazard data in some detail. We also present a demonstration that the computer simulation can be calibrated with Adjuvant-on-line, an often used clinical tool for prognosis in breast cancer.

Figure 1

Recurrence hazard rate for Milan premenopausal patients.

Figure 2

Milan data in disease-free survival format for premenopausal patients. Modified from Bonadonna et al. [3].

Figure 3

Disease-free-survival reported (modified) from Fisher et al. [41].

Figure 4

Saphner et al. data (Saphner et al. ECOG trials, Hazard of Relapse). These data which combine approximately 10 clinical trials for a variety of controls and adjuvant therapies clearly show bimodal hazard. The error bars are small as a result of the large number of patients. Modified from Saphner et al. [39]. ECOG is the Eastern Cooperative Oncology Group.

Figure 5

Simulation of Milan data in blue and actual Milan data for premenopausal patients in red. Format is hazard of relapse. We are comparing the Milan data for premenopausal patients and the simulation. We adjusted the simulation to try and agree with hazard for the Milan data. Time scale is 120 months.

Figure 6

Simulation of Milan data in blue and actual Milan data for premenopausal patients in red. Format is disease free survival.