Cyclicity in variations of incidence rates for breast cancer in different countries.

The aim was to analyze breast cancer (ICD9; Dx:174) in the view of studying and forecasting registration and incidence dynamics. Cycles in incidence rates for breast cancer in Bulgaria (1952-1985, period T = 17.63 and TH = 73.50 years) were restated. Patterns of annual incidence rates for different time stretches over the years 1964-1993 in the North West Region (NWR, UK), Ontario (Canada), USA and Bulgaria were compared. Monthly registration of new cases in NWR (1972-1993) and Bulgaria (1991-1994) was also analysed. Analysis of seasonality, non-linear regression, Fourier spectra, ARIMA and periodogram regression analysis with basic, detrended or decycled trigonometric approximations were applied. Cyclic variations in incidence for NWR (T = 10.25 years, p < 0.05), Ontario (T = 17.75 years, p < 0.05) and USA (TH = 20.50 years, p < 0.05) were revealed and previous findings confirmed. The cycle of 20.50 years for USA was considered a single "first-degree hypercycle" (n = 17 years). A stable "hypercycle" was found only for NWR (TH = 44 years, p < 0.01) describing a long-term cyclic trend. Short-term cyclicity for registrations in NWR (T = 4, 7.5 and 10.5 months, p < 0.05) and Bulgaria (T = 3, 6 and 11.5 months, p < 0.05), with peaks in March and June-July for both countries and a peak in October for NWR (all at p < 0.05), was also revealed. Cyclicity in occurrence might contribute to better understanding of breast cancer epidemiology and promote creation of more accurate predictive models. The lack of stable "hypercycles" in variations of incidence rates for USA and Ontario, for example, might be seen as an indicator of effectiveness of screening and/or influence of risk factors.