Shi Min, Umbach David M, Weinberg Clarice R
Biostatistics and Computational Biology Branch, National Institute of Environmental Health Sciences, National Institutes of Health, Department of Health and Human Services, Research Triangle Park, North Carolina, USA.
Environ Health Perspect. 2024 Dec;132(12):127004. doi: 10.1289/EHP14476. Epub 2024 Dec 24.
Large prospective cohort studies have been fruitful for identifying exposure-disease associations. In a cohort where biospecimens (e.g., blood, urine) were collected at enrollment, analysts can exploit a case-cohort approach: Biospecimens from a random sample of cohort participants, called the "subcohort," plus a sample of incident cases that were not part of the subcohort are assayed. Reusing subcohort data for multiple disease outcomes can reduce costs and conserve specimen archives. Pooling biospecimen samples before assay could both save money and reduce depletion of the archive but has not been studied for cohort studies.
We develop and evaluate a biospecimen pooling strategy for case-cohort analyses that relate an exposure to risk of a rare disease.
Our approach involves constructing pooling sets for cases not in the subcohort after grouping them according to time of diagnosis (e.g., age). In contrast, members of the subcohort are grouped by age at entry before constructing pooling sets. The analyst then fits a logistic regression model that jointly stratifies by age at risk and pooling set size and adjusts for confounders. We used simulations (288 sampling scenarios with 1,000 simulated datasets each) to evaluate the performance of this approach for several sizes of pooling sets and illustrated its application to environmental epidemiologic studies by reanalyzing Sister Study data.
Parameter estimates were nearly unbiased, and 95% confidence intervals constructed using a bootstrap estimate of the standard error performed well. In statistical tests also based on the bootstrap standard error, pooling up to 8 specimens per pool caused only modest loss of power. Assigning more cohort members to the subcohort and commensurately increasing the number of specimens per pool improved power and precision substantially while reducing the number of assays.
When using case-cohort analysis to study disease outcomes in relation to exposures assessed using biospecimens in a cohort study, epidemiologists should consider biospecimen pooling as a way to improve statistical power, conserve irreplaceable archives, and save money. https://doi.org/10.1289/EHP14476.
大型前瞻性队列研究在识别暴露与疾病的关联方面成果丰硕。在一个入组时收集了生物样本（如血液、尿液）的队列中，分析人员可以采用病例-队列方法：对队列参与者的随机样本（称为“亚队列”）的生物样本，加上不属于亚队列的新发病例样本进行检测。将亚队列数据用于多种疾病结局的分析可以降低成本并保存样本档案。在检测前合并生物样本可以节省资金并减少档案的消耗，但尚未在队列研究中进行过研究。
我们开发并评估一种用于病例-队列分析的生物样本合并策略，该分析涉及一种暴露与罕见病风险的关联。
我们的方法包括在根据诊断时间（如年龄）对亚队列之外的病例进行分组后，为这些病例构建合并组。相比之下，亚队列成员在构建合并组之前按入组时的年龄进行分组。然后，分析人员拟合一个逻辑回归模型，该模型按风险年龄和合并组大小进行联合分层，并对混杂因素进行调整。我们使用模拟（288种抽样场景，每种场景有1000个模拟数据集）来评估该方法在几种合并组大小下的性能，并通过重新分析姐妹研究数据来说明其在环境流行病学研究中的应用。
参数估计几乎无偏，使用标准误差的自助估计构建的95%置信区间表现良好。在同样基于自助标准误差的统计检验中，每个合并组合并多达8个样本只会导致适度的效能损失。将更多队列成员分配到亚队列并相应增加每个合并组的样本数量，在减少检测数量的同时，显著提高了效能和精度。
在队列研究中使用病例-队列分析来研究与使用生物样本评估的暴露相关的疾病结局时，流行病学家应考虑将生物样本合并作为提高统计效能、保存不可替代的档案和节省资金的一种方法。https://doi.org/10.1289/EHP14476 。
