Department of Epidemiology and Biostatistics and MRC-HPA Centre for Environment and Health, Imperial College London, London, UK.
Int J Health Geogr. 2011 Feb 28;10:16. doi: 10.1186/1476-072X-10-16.
The authors analyse the spatio-temporal variations of the incidence of bladder cancer between 1973 and 2004 in Utah at the census tract level (496 areas) to highlight areas of high and low relative risks that remained so throughout the 32 year period. Using these identified areas, a novel strategy is used to carry out a geographical case-control study of association between the risk of bladder cancer and presence of Toxic Release Inventory sites, where areas with stable high RRs are 'case areas' and all remaining areas with stable non increased risks are 'control areas'.
The time trend of bladder cancer risk fluctuated over the study period: A steady decrease was observed, followed by an abrupt increase from 1992 to 2004. Using a Bayesian space-time model, 93 census tracts were classified as having an excess relative risk and 81 a lower relative risk, sustained over the 32 years. We showed that these high relative risk areas for bladder cancer were associated with the presence of Toxic Release Inventory sites, after adjusting for the proportion of Latter-Day Saint Church members as an area level proxy for smoking habits.
Our study has demonstrated that the modeling of data in time and space has additional benefits over a purely spatial analysis. In addition to highlighting the areas with high and low relative risks, this model also allows the simultaneous study of persistency of spatial patterns over time and detection of 'unusual' time trends that may warrant further investigation.
作者分析了 1973 年至 2004 年犹他州按普查区（496 个地区）划分的膀胱癌发病率的时空变化，以突出高和低相对风险的地区，这些地区在 32 年期间保持相对稳定。利用这些确定的地区，采用一种新策略，对膀胱癌风险与有毒物质释放清单（Toxic Release Inventory，TRI）站点之间的相关性进行地理病例对照研究，其中相对风险稳定较高的地区为“病例地区”，所有剩余风险稳定无增加的地区为“对照地区”。
膀胱癌风险的时间趋势在研究期间波动：观察到稳定下降后，从 1992 年到 2004 年突然增加。使用贝叶斯时空模型，将 93 个普查区归类为具有超额相对风险，81 个具有较低的相对风险，持续了 32 年。我们表明，这些膀胱癌高相对风险地区与 Toxic Release Inventory 站点的存在有关，调整了后期圣徒教会成员比例作为吸烟习惯的区域水平代理变量后。
我们的研究表明，数据在时间和空间上的建模比纯空间分析具有额外的优势。除了突出高和低相对风险的地区外，该模型还允许同时研究时空模式的持久性，并检测可能需要进一步调查的“异常”时间趋势。
