Dimitrov B D
Biostatistics Unit, Clinical Research Center for Rare, Diseases "Aldo e Cele Dacco", Vila Camozzi, Ranica, Bergamo, Italy.
Croat Med J. 2000 Jun;41(2):197-202.
To analyze temporal dynamics and model trends and variations of the annual incidence rates of stomach cancer in the North-Western Region of England.
The data consisted of 23,465 new cases of stomach cancer as provided by the population-based registry of the Centre for Cancer Epidemiology (Manchester, England, UK). The parameter studied was the annual incidence rate of stomach cancer per 100,000 persons as age-adjusted to the world standard population and presented as time-series over the interval from 1971 to 1990. The hypotheses to be tested, regarding the annual incidence rates, were: 1) existence of specific temporal characteristics; 2) appearance of cyclic patterns of variability; and 3) usefulness of cyclicity in predictive modeling.
The decreasing tendency of annual incidence rates of stomach cancer for both men and women was best fitted by a quadratic trend (y=a+b?t2) out of 13 available linear/nonlinear regression models. This trend explained only about 49% of variability. Cyclic patterns of variability in incidence rates were established (short-term cycle of 8.9 years; long-term hypercycle of about 22-23 years). The best fit to the real incidence rates was achieved by bi-cyclic regression model. The summary cosine-sine model contained both cycles of 8.875 and 22-23 years; it showed the least variance of regression (men - Sy2=1.09; women - Sy2=0.65) and best prognostic index (PI=1.47 for men, and PI=1.29 for women). This trigonometric model explained about 83-86% of variability of incidence rates of stomach cancer.
Cyclic patterns of variability of the annual incidence rates of stomach cancer have been established. Such cyclicity might not only find likely implications in predictive modeling and forecasting of incidence rates, but it could also be considered useful in research on risk/prevention factors for stomach cancer.
分析英格兰西北地区胃癌年发病率的时间动态变化，并对其趋势和变异进行建模。
数据来源于癌症流行病学中心（英国曼彻斯特）基于人群的登记处提供的23465例胃癌新发病例。研究参数为按世界标准人口年龄调整后的每10万人胃癌年发病率，并呈现为1971年至1990年期间的时间序列。关于年发病率要检验的假设为：1）存在特定的时间特征；2）出现变异的周期性模式；3）周期性在预测建模中的有用性。
在13种可用的线性/非线性回归模型中，二次趋势（y = a + b×t²）最能拟合男性和女性胃癌年发病率的下降趋势。该趋势仅解释了约49%的变异性。确定了发病率变异的周期性模式（短期周期为8.9年；长期超周期约为22 - 23年）。双周期回归模型对实际发病率的拟合效果最佳。汇总余弦 - 正弦模型包含8.875年和22 - 23年两个周期；它显示出最小的回归方差（男性 - Sy² = 1.09；女性 - Sy² = 0.65）和最佳的预后指数（男性PI = 1.47，女性PI = 1.29）。该三角模型解释了胃癌发病率约83 - 86%的变异性。
已确定胃癌年发病率变异的周期性模式。这种周期性不仅可能在发病率的预测建模和预测中有潜在意义，还可能在胃癌风险/预防因素研究中有用。
