• 2022-09
  • 2022-08
  • 2022-07
  • 2022-06
  • 2022-05
  • 2022-04
  • 2021-03
  • 2020-08
  • 2020-07
  • 2020-03
  • 2019-11
  • 2019-10
  • 2019-09
  • 2019-08
  • 2019-07
  • GSK1838705A br RESULTS br Demographics br Our study population


    Our study GSK1838705A consisted of 47,220 female patients with stage I adenocarcinoma of the breast. The mean and median age was 74.6 6.7 years and 74 (range 65 to 114) years, respectively. The majority of patients were classified as white (87%) and most of the tumors were ERþ (84%). At last follow-up, the majority of the patients were alive (86%), and 3% had died of breast cancer, 2% of other cancers, and 10% from other causes.
    Cumulative incidence of mortality
    The overall cumulative incidence of mortality was 4.9% for DOD, 3.7% for DOC, and 21.3% for NCD. With a median follow-up of 50 months, the cumulative inci-dence graph depicting probability of death from breast cancer, other cancers, and non-cancer causes is shown in Figure 2. Although the probability of mortality from any 1 of these 3 causes increases with time from diagnosis, that from non-cancer causes does so precipitously.
    Probability of death table
    The 5- and 8-year probability of breast cancer death (DOD), other cancer death (DOC), and non-cancer death (NCD) is depicted in Table 1. For all patients, the 5- and 8-year prob-ability of DOD was 3% and 4.7%, for DOC 1.9% and 3.5%, and for NCD 9.8% and 18.9%, respectively. As ex-pected, with increasing age the probability of DOC and NCD rises. However, an increase in the probability of DOD with age is also seen with patients older than 81 years having a 7.4% chance of DOD at 8 years compared with 3.1% in patients 65 to 70 years old. Patients classified as black had a higher probability of DOD compared with white or
    Figure 2. Cumulative incidence of mortality graph. DOC, death due to other cancers; DOD, death due to breast cancer; NCD, death due to non-cancer causes.
    other patients. The presence of any major comorbidity (eg cardiovascular or neurologic disorders) significantly increased the probability of an NCD. Estrogen-receptor sta-tus was the strongest predictor of DOD for cancer related variables (5- and 8-year mortality 2.4% vs 6.8% and 4.0% vs 9.2%, respectively, for ERþ vs ER ).
    Cause-specific Cox regression analyses
    For each mortality category the association between age, race, comorbidity, and ER status and the hazard of mor-tality is shown in Table 2. For DOD increasing age, black race, and psychiatric comorbidity were associated with increased risk. Other and Asian race, as well as ERþ status were associated with decreased risk. The most significant variable associated with an increasing hazard of DOC was increasing age. The NCD risk was higher with increasing age and psychiatric, neurologic, as well as car-diovascular morbidity, and lower with ERþ status.
    Fine-Gray competing risk regression
    The sub-distribution hazards of the risk for DOD after competing risk analysis are shown in Table 3. Age con-tinues to have a strong association with increased risk of DOD, as does black race and psychiatric comorbidity. Asian and other race, as well as positive ER status, are associated with decreased risk of DOD.
    Risk calculator
    A risk calculator incorporating probabilities of mortality given patient age, comorbidity, and tumor characteristics generated from the cumulative incidence graph gives an es-timate of competing risks of death from DOD, DOC, and NCD. For example, a 70-year-old woman with no comor-bidity, Other race designation and an ERþ stage I breast can-cer has an 8-year probability of 2.0% for DOD, 2.4% for 
    DOC, and 7.5% of NCD. For a 70-year-old woman with cardiac and neurologic comorbidity, White race and an ER- tumor these probabilities are 9.3% for DOD, 4.2% for DOC, and 15.3% for NCD (risk calculator available in Appendix 1). Calibration plots showed that the risk models were fairly accurate, with estimates generally falling on the dashed diagonal lines, indicating that the predicted scores are generally reflective of the actual risk (eFig. 1).