| 摘要: | 惡性腫瘤(癌症)威脅人類生命日劇，許多國家數十年前已開始建立癌症發生與癌症死亡登記檔其能此重要疾病能有更深的瞭解。對此檔案的研究大都個別針對癌症發生或癌症死亡進行探討，較少同時去瞭解其間的關係。本研究即是以台灣地區為研究區域深入探討國人癌症發生與死亡之間於空間、時間分佈之相關性。癌症資料包括1979至1999年資料，研究癌症有全癌症、肺癌、肝癌、直腸癌、胃癌、口腔癌、血癌、子宮頸癌及乳癌。癌症指標包括五歲年齡組粗發生率及死亡率、每年年齡標準化發生率及死亡率、五年年齡標準化發生率及死亡率、五分法危險等級(以20,40,60,80百分位為切點)、二分法危險危險等級(以60百分位為切點)、發生與死亡勝算比及死亡/發生比值。空間特質因子有都市化程度指標、居住地形指標、人口老化程度、醫療資源分佈及人口密度指標。所用統計方法有空間相關係數(Moran’s I)、時間序列相關係數(自相關、交叉相關)、卡方趨勢檢定、等比勝算迴歸模式、對數線性迴歸、慨括性線性模式等，此外並結合地理資訊系統繪製地圖。 空間分析結果顯示癌症發生及死亡在空間上有顯著的正相關。由癌症地圖的比較可看出不同癌症有不同的分佈情形。聚集分析顯示癌症發生的聚集高於癌症死亡。癌症發生與都市化程度、醫療分佈程度、居住地形(離島、山地、平地)與人口密度呈正相關，但與人口老化程度呈負相關。時間分析結果顯示癌症死亡與發生隨年代漸增(胃癌死亡率例外呈漸減)，相較之下，癌症死亡曲線較平緩而癌症發生曲線則增加較快。由研究中我捫發現癌症死亡與相關( )比值指標可以與空間及時間因子結合用以解釋癌症發生與死亡之關係。預測模式結果顯示年代效應、存活效應與 ( )比值呈負相關同時年齡效應與( )比值呈正相關，此外我們亦可將空間特質間接納入我們的模式。透過預測模式不同癌症之間年代(Period)效應、年齡(Age)效應、世代(Cohort)效應及空間特質影響對癌症發生曲線與癌症死亡曲線的相對位置，均可透過此模式做有系統的解釋。此研究結果除其能幫助我們對癌症發生與死亡之趨勢相關有更深入瞭解外亦期能提供政府相關單位施政參考之依據。; Throughout the world cancer is one of the leading causes of death. Many countries have collected data with regard to cancer mortality and cancer incidence for the past several decades in order to better understand this important disease. In recent years many studies have used these data to access cancer mortality and cancer incidence separately, but there is little literature about their relationship. The purpose of this study was to access the relationship between cancer mortality and cancer incidence with regard to temporal and spatial factors, and build a model to explain their relationship. The cancer data included in our study were from 1979-1999 and as follows: total cancer, lung cancer, liver cancer, colon/rectal cancer, stomach cancer, oral cancer, leukemia, cervical cancer and breast cancer. The indices for cancer mortality and incidence included rates for each 5-year age group, age standardized rate, 5-year cumulative age standardized rate, 5-scaled risk (at percentiles 20th, 40th, 60th and 80th), 2-scaled risk (at the 60th percentile), odds ratio and death/incidence (D/I) ratio. The spatial factors were urbanization, topographic area (outlying islands, mountainous regions and plains), proportion of elderly (aged 65 years and over), medical resources and population density. The temporal factors were time period, age effect, cohort effect and survival rates. The statistical methods used were as follows: spatial correlation index (Moran’s I), time-series correlation index, chi-squared test for trends, proportional hazards model, log-linear model and generalized linear model, Geographical Information System (GIS) was used to create the cancer map. The spatial analyses showed that there were positive correlations between cancer mortalities and cancer incidences. The cancer map indicated that the distribution of cancers varied among different cancers. There was a greater clustering effect for cancer incidence than for cancer mortality. There was a positive correlation of total cancer incidence with urbanization, topographic area (outlying islands, mountainous regions and plains), proportion of elderly, medical resources and population density, but proportion of elderly negatively. The temporal analyses showed that cancer mortality rates have been generally stable compared to cancer incidence rates which have risen at a faster rate. The ratio of mortality to incidence (D/I) was found to be a good index to show the relationship of cancer mortalities and cancer incidences with temporal factors such as time periods, age and survival rates and spatial factors such as urbanization, topographical areas, and population. Our proposal model shows that there were negative correlations of D/I ratio with time period effect and 5-year survival rate. There was a positive correlation of D/I ratio with age effect. Our model did not indicate a direct relationship between D/I ratio and spatial factors. However, this relationship may be determined indirectly. It is possible to systematically predict trends for different cancers according to their |
