Separate and combined analysis of successive dependent outcomes after breast-conservation surgery: recurrence, metastases, second cancer and death
1 INSERM, U897 (Biostatistique), ISPED, Bordeaux, F-33076, France
2 Université Victor Segalen Bordeaux 2, Bordeaux, F-33076, France
3 Institut Bergonié - Centre Régional de Lutte Contre le Cancer du Sud-Ouest, Bordeaux, F-33076, France
4 INSERM, U897 (Team of Epidemiology for Cancer Prevention), Bordeaux, F-33076, France
BMC Cancer 2010, 10:697 doi:10.1186/1471-2407-10-697Published: 31 December 2010
In the setting of recurrent events, research studies commonly count only the first occurrence of an outcome in a subject. However this approach does not correctly reflect the natural history of the disease. The objective is to jointly identify prognostic factors associated with locoregional recurrences (LRR), contralateral breast cancer, distant metastases (DM), other primary cancer than breast and breast cancer death and to evaluate the correlation between these events.
Patients (n = 919) with a primary invasive breast cancer and treated in a cancer center in South-Western France with breast-conserving surgery from 1990 to 1994 and followed up to January 2006 were included. Several types of non-independent events could be observed for the same patient: a LRR, a contralateral breast cancer, DM, other primary cancer than breast and breast cancer death. Data were analyzed separately and together using a random-effects survival model.
LRR represent the most frequent type of first failure (14.6%). The risk of any event is higher for young women (less than 40 years old) and in the first 10 years of follow-up after the surgery. In the combined analysis histological tumor size, grade, number of positive nodes, progesterone receptor status and treatment combination are prognostic factors of any event. The results show a significant dependence between these events with a successively increasing risk of a new event after the first and second event. The risk of developing a new failure is greatly increased (RR = 4.25; 95%CI: 2.51-7.21) after developing a LRR, but also after developing DM (RR = 3.94; 95%CI: 2.23-6.96) as compared to patients who did not develop a first event.
We illustrated that the random effects survival model is a more satisfactory method to evaluate the natural history of a disease with multiple type of events.