Lynch Shannon M, Mitra Nandita, Ross Michelle, Newcomb Craig, Dailey Karl, Jackson Tara, Zeigler-Johnson Charnita M, Riethman Harold, Branas Charles C, Rebbeck Timothy R
Fox Chase Cancer Center, Cancer Prevention and Control, Philadelphia, Pennsylvania, United States of America.
University of Pennsylvania, Perelman School of Medicine, Philadelphia, Pennsylvania, United States of America.
PLoS One. 2017 Mar 27;12(3):e0174548. doi: 10.1371/journal.pone.0174548. eCollection 2017.
Cancer results from complex interactions of multiple variables at the biologic, individual, and social levels. Compared to other levels, social effects that occur geospatially in neighborhoods are not as well-studied, and empiric methods to assess these effects are limited. We propose a novel Neighborhood-Wide Association Study(NWAS), analogous to genome-wide association studies(GWAS), that utilizes high-dimensional computing approaches from biology to comprehensively and empirically identify neighborhood factors associated with disease.
Pennsylvania Cancer Registry data were linked to U.S. Census data. In a successively more stringent multiphase approach, we evaluated the association between neighborhood (n = 14,663 census variables) and prostate cancer aggressiveness(PCA) with n = 6,416 aggressive (Stage≥3/Gleason grade≥7 cases) vs. n = 70,670 non-aggressive (Stage<3/Gleason grade<7) cases in White men. Analyses accounted for age, year of diagnosis, spatial correlation, and multiple-testing. We used generalized estimating equations in Phase 1 and Bayesian mixed effects models in Phase 2 to calculate odds ratios(OR) and confidence/credible intervals(CI). In Phase 3, principal components analysis grouped correlated variables.
We identified 17 new neighborhood variables associated with PCA. These variables represented income, housing, employment, immigration, access to care, and social support. The top hits or most significant variables related to transportation (OR = 1.05;CI = 1.001-1.09) and poverty (OR = 1.07;CI = 1.01-1.12).
This study introduces the application of high-dimensional, computational methods to large-scale, publically-available geospatial data. Although NWAS requires further testing, it is hypothesis-generating and addresses gaps in geospatial analysis related to empiric assessment. Further, NWAS could have broad implications for many diseases and future precision medicine studies focused on multilevel risk factors of disease.
癌症是生物、个体和社会层面多个变量复杂相互作用的结果。与其他层面相比，邻里层面在地理空间上产生的社会影响尚未得到充分研究，评估这些影响的实证方法也很有限。我们提出了一种新颖的全邻里关联研究（NWAS），类似于全基因组关联研究（GWAS），它利用生物学中的高维计算方法来全面、实证地识别与疾病相关的邻里因素。
宾夕法尼亚癌症登记数据与美国人口普查数据相链接。在一种逐步更严格的多阶段方法中，我们评估了邻里因素（n = 14,663个人口普查变量）与前列腺癌侵袭性（PCA）之间的关联，其中白人男性中有n = 6,416例侵袭性病例（分期≥3/ Gleason分级≥7）与n = 70,670例非侵袭性病例（分期<3/ Gleason分级<7）。分析考虑了年龄、诊断年份、空间相关性和多重检验。我们在第一阶段使用广义估计方程，在第二阶段使用贝叶斯混合效应模型来计算比值比（OR）和置信区间/可信区间（CI）。在第三阶段，主成分分析对相关变量进行分组。
我们识别出17个与PCA相关的新邻里变量。这些变量代表收入、住房、就业、移民、医疗服务可及性和社会支持。与交通相关的最显著变量（OR = 1.05；CI = 1.001 - 1.09）和贫困（OR = 1.07；CI = 1.01 - 1.12）。
本研究介绍了高维计算方法在大规模公开可用地理空间数据中的应用。尽管NWAS需要进一步测试，但它能够生成假设，并解决与实证评估相关的地理空间分析中的空白。此外，NWAS可能对许多疾病以及未来专注于疾病多层面风险因素的精准医学研究具有广泛影响。
