作者bmka (和平 奋斗 写paper)
看板Statistics
标题Re: [问题] 有关存活分析的问题
时间Fri Jul 11 12:15:02 2008
※ 引述《hospice (平安快乐)》之铭言:
: 您好:
: 再次感谢您的回覆!
: 不过在下资质驽钝,
: 有些地方还是不明白。
: Is the prognostic factor an event time? If not, why do you even
: consider survival analysis.
: 您这里的意思是说预後因子是否为 event time吗?
: 对不起我是新手不太清楚您的意思,
: 我看的paper都是将年龄分成两组(>60 or <60),
: 肿瘤大小分为两组(<3cm or <3cm)等等,
: 来看Disease-specific survial(DSS)或Recurrent-free survival(RFS),
: 用的是Kapaln-Meiner Method
: 因此我们才想如法炮制,
: 将tumor recurrence与否分成两组(yes or no),
: 来看Disease-specific survial(DSS)或Recurrent-free survival(RFS)
Okay. Then the endpoint is time to a specific event, subject to censoring.
You should use methods for survival data.
: Are you planing to treat tumor recurrence as a binary variable?
: If yes, you better define a time frame. Otherwise, you may consider
: treating it as a time-dependent covariate.
: 没错,我们正是想把tumor recurrence当成binary variable,
: 只是不知道这样做合不合理?因为没有paper有这样做过,
: 您说的time frame是怎麽一回事?有没有相关的参考资料呢?
Based on the description, it is better to treat "tumor recurrence" as
a time-dependent covariate. Keep in mind that the status of tumor
recurrence for those who loss to follow-up may be unknown.
: Do you mean to estimate the K-M curve for those with and without cancer
: recurrence separately?
: 是的,如同上面所说的,
: 我们是想这样做!不知道行不行的通?
Again, how would you define the value of the "tumor recurence" variable for those
who loss to follow-up and do not have tumor recurrence before they drop out?
: 如果用t test 来看两组存活天数的话,
: 可以得到无 recurrence的mean±SD,以及有recurrence的mean±SD,以及p值,
: 不过这好像不是正确的作法,而且存活天数有极端值,有的只追踪一两周就loss follow up,
: 有的可以追踪到好几年!
The t test is not appropriate for survival data.
※ 编辑: bmka 来自: 71.255.237.189 (07/11 12:17)