Statistical Methods for Assessing Patterns of Change in Cancer-Control Behavior

描述(由申请人提供)：有规律的体力活动已被证明与降低癌症风险有关(Lee等人，2006和McTiernan，2008)。尽管意识到了这一点，但大多数美国成年人目前并不符合体力活动指南(国家健康统计中心，2007)。我们的研究团队已经进行了许多行为干预，旨在鼓励以前久坐不动的成年人采用体育活动(Marcus等人，2007；Marcus，Lewis，Williams，Dunsiger等人，2007；Pekmezi，Neighters，Lee，Gans，Bock，Morrow，Marquez，Dunsiger&Marcus，2009)。分析这些纵向试验数据的标准统计方法可能不足以描述体力活动采用、维持和复发的模式。我们提出了一种更复杂的方法，专门用于在六种体力活动干预中发现行为变化的模式。能够在不同研究之间比较由此产生的行为变化的子类型(这些研究本身是重叠的，因为它们至少包含一个共同的治疗臂)将使我们能够确定在各个研究中是否存在一致的模式，以及这些子类型的分布是否因治疗臂而异。此外，我们将通过修改我们建议的计算程序来确定行为模式变化的潜在预测者和调节者，以允许其他基线变量对子类型的分布做出贡献。最后，我们将使用这些亚型来确定治疗阶段的关键时间，在此之后，久坐行为复发的可能性更大。这些目标中的每一个都对未来体力活动干预措施的设计和分析都有具体的影响。使用目前的方法，我们将无法进行同样的分析，因此可能会错过参与者如何改变他们的癌症风险的关键信息。

DESCRIPTION (provided by applicant): Regular physical activity has been shown to be associated with reduced cancer risk (Lee et al, 2006 & McTiernan, 2008). Despite being aware of this, the majority of American adults do not currently meet physical activity guidelines (National Center for Health Statistics, 2007). Our research team has conducted numerous behavioral interventions aimed at encouraging physical activity adoption amongst previously sedentary adults (Marcus et al, 2007; Marcus, Lewis, Williams, Dunsiger et al, 2007; Pekmezi, Neighbors, Lee, Gans, Bock, Morrow, Marquez, Dunsiger & Marcus, 2009). Standard statistical methodology for analyzing data from these longitudinal trials may not suffice to describe patterns of physical activity adoption, maintenance and relapse. We propose a more sophisticated methodology aimed specifically at finding patterns of behavior change in six physical activity interventions. The ability to compare resulting sub-types of behavior change across studies (which themselves overlap in that they contain at least one treatment arm in common) will allow us to determine whether there are consistent patterns across studies and whether the distribution of these sub-types differs by treatment arm. In addition, we will identify potential predictors and moderators of pattern of behavior change by modifying our proposed computing routine to allow for other baseline variables to contribute to the distribution of sub-type. Finally, we will use these sub-types to identify key times in the treatment phase after which relapse to sedentary behavior is more likely. Each of these aims has specific implications for both design and analysis of future physical activity interventions. Using current methods we would be unable to carry out these same analyses and thus might risk missing key information about how participants change their cancer risk.