PROSPECTIVE COHORT STUDY OF DIET AND CANCER

该项目继续对癌症发病率进行前瞻性监测 1975年至2015年间接受采访的41,400名夏威夷居民的队列 1980年。受试者构成总体的随机样本，并且 包括夏威夷的所有主要种族群体。食物频率信息， 吸烟史和饮酒史、身高、体重、职业和人口统计 收集了每个人的数据。在队列中的一小部分(4890 人)，获得了更广泛、更定量的饮食史。 因为我们已经表明，这个队列中的外迁率是 极低的高度，完全可以由计算机进行监控 与全州范围的、以人口为基础的夏威夷肿瘤登记中心的联系。由 1996年底(续期的第四年)，我们预计会出现以下情况 感兴趣的癌症病例数目：肺癌544例，结肠癌360例，癌症531例 乳房，410个前列腺，187个胃，116个胰腺，82个口咽部，142个 膀胱和100个子宫内膜。 在续期期间将追求两个具体目标：1) 既往饮食与特定部位癌症发生的关系 将会被调查。这项分析将包括更多的癌症 由于病例数量较少，以前无法进行研究 (口咽、胃、胰腺、膀胱和子宫内膜)。为了更多 常见的癌症部位(肺癌、结肠癌、乳腺癌和前列腺癌)，分析将是 扩展到包括特定种族的比较和按地点进行的比较 出生地。这些分析将利用最近开发的特定于亚组的分析 估计的份量。此外，对质量的进一步评估 调查问卷数据将利用更完整的 关于该队列的特殊子集的信息。2)总死亡率和 相互竞争的死亡风险将被检查与饮食和 人体测量危险因素。 为了检验饮食假说(目标1)，我们将使用成比例的危险 回归分析比较暴露者和非暴露者的发病率 成员，同时针对潜在的混杂因素进行调整。举例 要检查不太常见的网站的关系包括：肉类 摄入与胰腺癌风险；咸鱼和腌制蔬菜 消费与胃癌风险；以及新鲜水果和生蔬菜 消耗和口咽、胃、胰腺、膀胱和 子宫内膜癌的风险。为了评估饮食和饮食的效果 人体测量因素对总死亡率的影响(目标2)，我们将使用 协变量上死亡时间回归的比例风险模型 值得关注的是，在对混淆变量进行调整后。使用这种方法， 我们还将研究不同程度的药物对死亡率的影响 遵守NCI和其他机构推荐的饮食指南 经纪公司。 这一群体的一个主要优势是效率和低成本， 可以进行监测和分析，以产生有用的 关于饮食成分与癌症风险之间关系的信息。

This project continues surveillance for cancer incidence on a prospective cohort of 41,400 Hawaii residents who were interviewed between 1975 and 1980. The subjects constitute a random sample Of the population and include all main ethnic groups in Hawaii. Food frequency information, smoking and alcohol histories, height, weight, occupation, and demographic data were collected on each person. On a small subset of the cohort (4,890 persons), a more extensive, quantitative diet history was obtained. Because we have shown that out-migration rates in this cohort are extremely low, surveillance can be carried out entirely by computer linkage with the statewide, population-based Hawaii Tumor Registry. By the end of 1996 (fourth year of the renewal period), we expect the following numbers of cases of the cancers of interest: 544 lung, 360 colon, 531 breast, 410 prostate, 187 stomach, 116 pancreas, 82 oropharynx, 142 bladder, and 100 endometrium. Two specific aims will be pursued in the renewal period: 1) The relationship of antecedent diet to cancer occurrence at specific sites will be investigated. This analysis will include additional cancers that could not be studied previously because of small numbers of cases (oropharynx, stomach, pancreas, bladder, and endometrium). For the more common cancer sites (lung, colon, breast, and prostate), analyses will be extended to include ethnic-specific comparisons and comparisons by place of birth. These analyses will utilize recently-developed subgroup-specific portion size estimates. In addition, further assessments of the quality of the questionnaire data will be made, utilizing the more complete information on the special subset of the cohort. 2) Total mortality and competing risks of death will be examined in relation to dietary and anthropometric risk factors. To test the dietary hypotheses (aim 1), we will use proportional hazards regression to compare incidence among exposed and non-exposed cohort members, while adjusting for potential confounders. Examples of relationships to be examined for the less common sites include: meat intake and pancreatic cancer risk; salted fish and pickled vegetable consumption and stomach cancer risk; and fresh fruit and raw vegetable consumption and oropharyngeal, stomach, pancreatic, bladder, and endometrial cancer risk. To evaluate the effects of dietary and anthropometric factors on total mortality (aim 2), we will use the proportional hazards model to regress the time-to-death on the covariate of interest, after adjusting for confounding variables. Using this method, we will also examine the effects on mortality of various degrees of adherence to the recommended dietary guidelines from NCI and other agencies. A major strength of this cohort is the efficiency and low cost with which the surveillance and analyses can be carried out to yield useful information on associations between dietary components and cancer risk.