Confirmation of double-peaked time distribution of mortality among Asian breast cancer patients in a population-based study.

Division of Clinical Trials and Epidemiological Sciences, National Cancer Centre, Singapore.
Breast cancer research: BCR (Impact Factor: 5.87). 02/2007; 9(2):R21. DOI: 10.1186/bcr1658
Source: PubMed

ABSTRACT Double-peaked time distributions of the mortality hazard function have been reported for breast cancer patients from Western populations treated with mastectomy alone. These are thought to reflect accelerated tumour growth at micrometastatic sites mediated by angiogenesis after primary tumour removal as well as tumor dormancy. Similar data are not available for Asian populations. We sought to investigate whether differences exist in the pattern of mortality hazard function between Western breast cancer patients and their Asian counterparts in Singapore, which may suggest underlying differences in tumor biology between the two populations.
We performed a retrospective cohort study of female unilateral breast cancer patients diagnosed in Singapore between October 1994 and June 1999. Data regarding patient demographics, tumour characteristics and death were available. Overall survival curves were calculated using the Kaplan-Meier method. The hazard rate was calculated as the conditional probability of dying in a time interval, given that the patient was alive at the beginning of the interval. The life table method was used to calculate the yearly hazard rates.
In the 2,105 women identified, 956 patients (45.4%) had mastectomy alone. Demographic characteristics were as follows: 86.5% were Chinese, 45.2% were postmenopausal, 38.9% were hormone receptor positive, 54.6% were node negative and 44.1% had high histological grade. We observed a double-peaked mortality hazard pattern, with a first peak in mortality achieving its maximum between years 2 and 4 after mastectomy, and a second large peak in mortality during year 9. Analyses by subgroups revealed a similar pattern regardless of T stage, or node or menopausal status. This pattern was also noted in high-grade tumors but not in those that were well to moderately differentiated. The double-peaked pattern observed in Singaporean women was quantitatively and qualitatively similar to those reported in Western series.
Our study confirms the existence of a double-peaked process in Asian patients, and it gives further support to the tumour dormancy hypothesis after mastectomy.

  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: To explain bimodal relapse patterns observed in breast cancer data, we have proposed that metastatic breast cancer growth commonly includes periods of temporary dormancy at both the single cell phase and the avascular micrometastasis phase. The half-lives of these states are 1 and 2 years respectively. We also suggested that surgery to remove the primary tumor often terminates dormancy resulting in accelerated relapses. These iatrogenic events are very common in that over half of all metastatic relapses progress in that manner. Assuming this is true, there should be ample and clear evidence in clinical data. We review here the breast cancer paradigm from early detection, through treatment and follow-up, and consider how dormancy and surgery-driven escape from dormancy would be observed. We examine mammography data, effectiveness of adjuvant chemotherapy, heterogeneity and aggressiveness, timing of surgery within the menstrual cycle and racial differences in outcome. Dormancy can be identified in these diverse data but most conspicuous is the sudden escape from dormancy following primary surgery. These quantitative findings provide linkage between experimental studies of tumor dormancy and clinical efforts to improve patient outcome.
    Apmis 01/2008; 116(7-8):730-41. · 2.07 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: An improved "weighted information" method of spectral estimation has recently been introduced [1,2]. The method treats the problem of estimating autoregressive (AR) process parameters from discrete time observations corrupted by additive independent noise, with known power spectral density. The estimation procedure is cast as the problem of minimizing a weighted information measure and has a theoretical foundation relating it to asymptotic maximum likelihood and minimum information divergence procedures. This paper develops several topological properties of the weighted information measure. Computational procedures appropriate to the estimation problem are discussed elsewhere [1,2]; procedures appropriate to the detection ("vector quantization") problem are presented here. Simulation results and a real speech example demonstrate the advantages of the method.
    Acoustics, Speech, and Signal Processing, IEEE International Conference on ICASSP '84.; 04/1984
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The presence of non-conforming components instead of, or in addition to, the usual assembly errors results in N- or W-shaped hazard rate (HR) functions rather than the usual bathtub (i.e. U-shaped) ones. Although there have been numerous models for bathtub-shaped HR functions, N- and W-shaped HR functions are usually modelled using mixtures of two or more distributions. While this approach does sometimes lead to tidy interpretation, there can be a degree of overparameterization, with consequent problems in stability and fitting. For this reason, the present paper revisits the natural approach of modelling N- and W-shaped HR functions using polynomial functions of degree three or four. Although the non-negativity of the hazard rate function becomes non-trivial, this ensures a minimal number of parameters. The polynomial approach also allows the use of a parametric model without imposing a particular shape of hazard rate function on the data, which usually requires a non-parametric approach. The possible hazard rate shapes obtainable are characterized, and detailed formulae for local minima and maxima of the functions provided. The performance of the models is compared to that of several generalizations of the Weibull distribution, with promising results.
    Proceedings of the Institution of Mechanical Engineers Part O Journal of Risk and Reliability 02/2009; 223:59-69. · 0.62 Impact Factor

Full-text (2 Sources)

Available from