Active surveillance cohort in CEASAR

Aims:
  • Describe reasons patients come off AS
  • Describe which patients come off AS.

Cohort definition/Inclusion and Exclusion Criteria:

  • No active treatment within one year (n = 505) OR {indication of AS on
medical chart or survey (bl, 6m, 12m) AND no active treatment within 6 months (n = 33)} --> n = 538 For example, these patients would be excluded and considered to be in an "active treatment cohort"
  • anyone receiving treatment within the first 6 months. (even if their
surveys and chart say AS)
  • anyone receiving treatment between 6 and 12 months after dx who have
no indication of AS on chart or surveys

Analysis:

Aim 1:Describe reasons patients come off AS (question 8 of 3-year survey).
  • Frequency table and proportion estimate with CI.
  • Have counts with individual reasons as well as composite reasons (medical vs. non medical)
  • Only include those who we've classified as having come off AS.
  • May need the concept of a "tie breaker." If someone selected both med and nonmed reasons, consider them having a medical reason. Or rather, classify everyone as medical vs nonmedical only. Also show how many selected both.
Aim 2: Predictors of coming off AS.
  • Should probably do this separately for medical and for non-medical reasons. The question/purpose is different, and the predictors will be different Since there were only 8 people who indicated only a medical reason, we can't model them separately. We will just model those who came off for medical reasons.
  • Survival or not survival?
  • Survival model with time to active treatment as response.
  • Dan has some hypothesized predictors as well as some quantities that would be confounders.
  • All the predictors should be baseline predictors.
  • Will make spreadsheet of variables. We will use age, comorbidity, d’amico risk, race, and a SES indicator or two. Mark wants to include the PDM and a RAND quality indicator.
  • Will be limited in the number of observations and thus in df for model
  • Need to reduce dimensionality. Will use principal components regression
  • Vars we really want estimates of: max-pc, education, ss, pdm.

Meeting notes

2016 February 8

  • Want estimates of max pc, ss, pdm, and education
  • Dan: age can be considered a measure of overall health in this context.
  • Can bundle epic function measures
  • PS worry and max-pc will be colinear
  • It would be nice to also make inference about race and marital status
  • Can exnclude erection quality and quality of care

2016 January 18

  • Showed Mark and Matt some tabulations of treatment combos.
  • Mark says that most of the combos we have make sense, but in the ones that do not, I don't need to do any checking.
  • I found a variable in the 12 month survey asking about fathers and brothers with pca. Mark says to put it in table 1 and in the PCA.
  • I should bin patients into tx groups using the same algorithm as we used last time. (Surgery if any surgery except if radiation, hormones if hormones were the only treatment received, etc.) done
  • I showed them some questions in the medical chart database on the follow up form that I thought would be useful at defining recurrence and salvage treatments, but Mark said they don't provide us with useful information.
  • Based on the numbers of responses indicating a medical reason and indicating nonmedical reasons, we will probably just make one model for coming off for a medical reason.

2016 January 6

  • Sketched out tables and figures for manuscript
  • Discussed problems with comparing patients who came off AS with those who didn't in a univariate table in the presence of censoring
  • Discussed framework for a survival analysis and possible strategies for an unadjusted analysis
    • Could show nelson aalen plot. Discussed a little about whether to plot prob of survival or to plot prob of failure. _I did some reading on this. There are both Nelson-Aalen and Kaplan-Meier estimates of both survival and cumulative hazard. H = -log S. The cum hazard is the "accumulation" of the hazard over time. (It is the integral of the hazard rate.) I think we should decide separately which quantity we want to plot (survival, cum hazard, or 1 - survival) and then decide whether we want to use km or nelson aalen. _
  • Table 3 can be the results of the survival analysis or whatever multivariable regression addressing aim 2.
  • I need to remind myself that salvage therapy is not the same as coming off AS(?)
  • Where should race go? We are probably interested in finding this out.

2015 November 23

  • Refined aims to the above.
  • One major task is defining the AS cohort for this sub study.
  • When patients go on AS, they typically have a PSA test every 3 months and a repeat biopsy within 12 months. If the PSA is elevated, they might have a repeat biopsy sooner. We would expect very few patients to have a repeat biopsy by 6 months. This info is useful for defining the study cohort. We want to exclude patients who got an active treatment by some time point after dx, and we need to define that cutoff. The earliest it would be is probably 6 months, and the latest it would be is 12 months.
  • We want to incorporate info from the chart about AS.
  • Need to include checks for people who indicated on a survey or chart that they are undergoing AS. Decided to integrate the treatment info with chart or patient report of AS.
  • Proposed defn: no active treatment within one year OR (indication on chart or bl, 6m, or 12m survey AND no active treatment within 6 months).

-- JoAnnAlvarez - 04 Jan 2016
Topic revision: r9 - 09 Feb 2016, JoAnnAlvarez
 

This site is powered by FoswikiCopyright © 2013-2017 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Vanderbilt Biostatistics Wiki? Send feedback