Lima Isabela da Silva, Silva Sóstenes Jerônimo da, Brighenti Carla Regina Guimarães, Nakamura Luiz Ricardo, Oliveira Tiago Almeida de, Oliveira Milena Edite Casé de, Ramires Thiago Gentil
Programa de Pós-graduação em Estatística e Experimentação Agropecuária, Universidade Federal de Lavras, Lavras, Brasil.
Programa de Pós-graduação em Biometria e Estatística Aplicada, Universidade Federal Rural de Pernambuco, Recife, Brasil.
Cad Saude Publica. 2025 Aug 22;41(8):e00073324. doi: 10.1590/0102-311XEN073324. eCollection 2025.
Cancer is a global public health concern due to its high mortality rates. In Brazil, breast cancer is one of the leading causes of disease and death among women in all regions of the country, with higher mortality rates in less developed regions. Hence, this study analyzes variables associated with survival time in breast cancer patients in Campina Grande, Paraíba State, Brazil. Distributional regression models, also known as generalized additive models for location, scale, and shape (GAMLSS), were used due to their flexibility in explaining complex behaviors of a given response (for example, survival time) based on other variables. Tumor site, age, number of hormone therapy, radiotherapy and chemotherapy sessions, and molecular markers such as estrogen receptor, progesterone receptor, Ki-67 protein, p53, HER2 mutation and molecular subtype were examined. Two different GAMLSS were fitted considering Weibull and log-normal distributions, the former of which is more appropriate per the Akaike information criterion. Using a variable selection procedure specific to GAMLSS, we identified four covariates that directly affect average survival time: number of hormone therapy and chemotherapy sessions, p53 status, and estrogen receptor status. Excepting estrogen receptor status, the other covariates selected to explain average survival time were also used to explicitly explain the variability of these times.
由于癌症的高死亡率，它已成为一个全球公共卫生问题。在巴西，乳腺癌是该国所有地区女性疾病和死亡的主要原因之一，在欠发达地区死亡率更高。因此，本研究分析了巴西帕拉伊巴州大坎皮纳市乳腺癌患者生存时间的相关变量。由于分布回归模型（也称为位置、尺度和形状的广义相加模型，即GAMLSS）在基于其他变量解释给定响应（例如生存时间）的复杂行为方面具有灵活性，所以使用了该模型。研究考察了肿瘤部位、年龄、激素治疗、放疗和化疗疗程的次数，以及雌激素受体、孕激素受体、Ki-67蛋白、p53、HER2突变和分子亚型等分子标志物。考虑到威布尔分布和对数正态分布，拟合了两种不同的GAMLSS，根据赤池信息准则，前者更为合适。使用特定于GAMLSS的变量选择程序，我们确定了四个直接影响平均生存时间的协变量：激素治疗和化疗疗程的次数、p53状态和雌激素受体状态。除雌激素受体状态外，其他用于解释平均生存时间的协变量也被用来明确解释这些时间的变异性。
