We examined whether the Cretan cohort of the Seven Countries Study (SCS) is representative for the entire population in the island using cancer mortality registries. The analysis was carried out on the Cretan cohort of the SCS cancer mortality data and a similar cancer registry for the general population during a 51-year follow-up (1960-2011). Information about the causes of mortality was obtained from official death certificates and classified according to the International Classification of Diseases, 9th Revision (ICD-9). Two time-series models of mortalities from cancer, using data from the Cretan cohort and the Hellenic Statistical Office (EL. STAT), were developed using Matlab software. The existence of long-term memory in the data was tested by rescaled range analysis (Hurst-Mandelbrot). State-space reconstruction was applied to identify the simplest system that was able to re-create the present time-series. In the cohort, cancer mortalities accounted for 18.9% of total mortalities. The EL.STAT time-series analysis generated mean V statistics (95%CI) of 0.69815 (0.398-0.999) and 0.677143 (0.301-0.897) for the general population and the seven countries cohort, respectively. The embedding dimension for the EL.STAT data was equal to 1 for the general population and for the Cretan cohort (m=1). The exponent H values for the two time-series were almost equal. In the two time-series the proposed time delay of cancer mortalities was 2. The Cretan cohort of the SCS and the entire population of the island followed similar patterns of cancer mortality over time.
Keywords: Seven Countries Study; cancer mortality; rescaled range analysis; state-space reconstruction; time-series.