Systematic Error and Confounding: Meta-Analyses of Alcohol and Disease

描述（由申请人提供）：当在多个流行病学研究中重复出现系统偏差和混淆时，它们可能导致错误的结论。本应用程序解决了酒精使用与一系列前哨疾病之间关联的队列和病例对照研究中的偏差和混淆，并具有纠正这些疾病的潜力。如果不加以纠正，这些错误模糊了对酒精与疾病关系的科学理解，破坏了公共卫生政策的质量和准确性，并导致误导性信息的传播。2009年，我们的研究团队获得了NIH挑战基金（CG）。成功的CG应用只是整个研究计划的一小部分。当前的应用程序填充了整个计划，并为其添加了新的重要元素。首先，我们检验了流行病学研究中关于酒精作为7种前哨疾病和全因死亡率相关的发病率和死亡率的危险因素可能存在的系统性错误的假设：(i) 2型糖尿病，（ii）胆石症，（iii）女性乳腺癌，（iv）胃癌，(v)肾癌，（vi）肺癌，（vii）冠心病（用新研究更新我们以前的研究结果），（viii）全因死亡率（也更新我们以前的研究结果）。在过去的评论中，这些疾病中的每一种都被确定为与饮酒有以下一种关系：(i) j型曲线（对“轻度/中度”饮酒者有“保护”作用）；（ii）统计上显著的正线性曲线（对中度和重度饮酒者有害）；（三）混合交叉研究效应（无效应）。除了我们在CG中测试的4种疾病结果外，还检查了这些情况。其次，我们采用了一种新方法，至少在一定程度上解释了不同饮酒类别中混杂因素的不均匀分布，纠正了我们目前所知道的结果。第三，我们使用来自大型队列研究的数据，测试了关于“不平等”混杂因素（从CG中出现）的新假设。第四，我们利用我们的发现在人口水平上重新估计不同类别饮酒的潜在危害和益处。荟萃分析将用于确定研究设计中是否存在系统误差，潜在混杂因素测量的质量和全面性，或研究中这些因素的组合是否会对结果产生不适当的偏倚：(i)对某些疾病产生明显的酒精保护；（ii）掩盖潜在的统计上显著的关联；（iii）导致无法明确解决的“混合”结果；（iv）影响已建立的酒精疾病关联的大小和显著性。这项拟议的工作旨在促进对酒精对一系列慢性疾病的贡献的理解，并改进对酒精导致的死亡率和发病率的估计。它应阐明临床实践建议的证据基础，并有助于理解减少酒精饮料危害的国家和国际战略。

DESCRIPTION (provided by applicant): When systematic errors of bias and confounding are repeated across multiple epidemiological studies, they may lead to false conclusions. This application addresses bias and confounding in cohort and case-control studies of associations between alcohol use and a range of sentinel diseases with the potential of correcting for them. Uncorrected, these errors obscure scientific understanding of the alcohol-disease relation, undermine the quality and accuracy of public health policy and lead to the dissemination of misleading information. In 2009, our research team was awarded an NIH Challenge Grant (CG). The successful CG application represented a small portion of a comprehensive research plan. The present application fills out the total plan, and adds new and important elements to it. First, we test hypotheses regarding possible systematic errors in epidemiological studies of alcohol as a risk factor for morbidity and mortality associated with 7 sentinel diseases as well as all-cause mortality: (i) Type 2 diabetes, (ii) cholelithiasis, (iii) female breast cancer, (iv) stomach cancer, (v) renal cancer, (vi) lung cancer, (vii) coronary heart disease (updating our previous findings with new studies), and (viii) all-cause mortality (also updating findings from our previous studies). Each of these diseases has been identified in past reviews as having one of the following relations to alcohol consumption: (i) a J-shape curve (a 'protective' effect for 'light/moderate' drinkers); (ii) a statistically significant positive linear curve (a detrimental effect for moderate and heavier drinkers); and (iii) a mixed cross-study effect (no effect). These conditions are examined in addition to the 4 disease outcomes we are testing in the CG. Second, we apply a new method that may account, at least in part, for the unequal distribution of confounders within distinct drinking categories, correcting results as we currently know them. Third, we test newly developed hypotheses regarding 'unequal' confounders (which has since emerged from the CG) using data from a large cohort study. Fourth, we utilize our findings to re-estimate potential harms and benefits of various categories of drinking at the population level. Meta-analyses will be used to determine whether the presence of systematic error in study design, the quality and comprehensiveness of potential confounder measurement, or combinations of these in studies can unduly bias results towards: (i) creating apparent protection from alcohol use against some diseases; (ii) masking potentially statistically significant associations; (iii) contributing to 'mixed' outcomes that defy a clear resolution; (iv) influencing the size and significance for established alcohol-disease associations. This proposed work is intended to advance the understanding of alcohol's contribution to a range of chronic diseases and improve estimates of alcohol-attributable mortality and morbidity. It should clarify the evidence basis for advice from clinical practice and contribute understanding to national and international strategies to reduce the harm from alcoholic beverages.