Partly conditional survival models for longitudinal data

University of Washington Seattle, Seattle, Washington, United States
Biometrics (Impact Factor: 1.52). 07/2005; 61(2):379-91. DOI: 10.1111/j.1541-0420.2005.00323.x
Source: PubMed

ABSTRACT It is common in longitudinal studies to collect information on the time until a key clinical event, such as death, and to measure markers of patient health at multiple follow-up times. One approach to the joint analysis of survival and repeated measures data adopts a time-varying covariate regression model for the event time hazard. Using this standard approach, the instantaneous risk of death at time t is specified as a possibly semi-parametric function of covariate information that has accrued through time t. In this manuscript, we decouple the time scale for modeling the hazard from the time scale for accrual of available longitudinal covariate information. Specifically, we propose a class of models that condition on the covariate information through time s and then specifies the conditional hazard for times t, where t > s. Our approach parallels the "partly conditional" models proposed by Pepe and Couper (1997, Journal of the American Statistical Association 92, 991-998) for pure repeated measures applications. Estimation is based on the use of estimating equations applied to clusters of data formed through the creation of derived survival times that measure the time from measurement of covariates to the end of follow-up. Patient follow-up may be terminated either by the occurrence of the event or by censoring. The proposed methods allow a flexible characterization of the association between a longitudinal covariate process and a survival time, and facilitate the direct prediction of survival probabilities in the time-varying covariate setting.

  • [Show abstract] [Hide abstract]
    ABSTRACT: Endogenous hormones and growth factors that increase mammographic breast density could increase ovarian cancer risk. We examined whether high breast density is associated with ovarian cancer risk. We conducted a cohort study of 724603 women aged 40 to 79 years with 2506732 mammograms participating in the Breast Cancer Surveillance Consortium from 1995 to 2009. Incident epithelial ovarian cancer was diagnosed in 1373 women. We used partly conditional Cox regression to estimate the association between breast density and 5-year risk of incident epithelial ovarian cancer overall and stratified by 10-year age group. All statistical tests were two-sided. Compared with women with scattered fibroglandular densities, women with heterogeneously dense and extremely dense breast tissue had 20% and 18% increased 5-year risk of incident epithelial ovarian cancer (hazard ratio [HR] = 1.20, 95% confidence interval [CI] = 1.06 to 1.36; HR = 1.18, 95% CI = 0.93 to 1.50, respectively; P trend = .01). Among women aged 50 to 59 years, we observed a trend in elevated risk associated with increased breast density (P trend = .02); women with heterogeneously and extremely dense breast tissue had 30% (HR = 1.30; 95% CI = 1.03 to 1.64) and 65% (HR = 1.65; 95% CI = 1.12 to 2.44) increased risk, respectively, compared with women with scattered fibroglandular densities. The pattern was similar but not statistically significant at age 40 to 49 years. There were no consistent patterns of breast density and ovarian cancer risk at age 60 to 79 years. Dense breast tissue was associated with a modest increase in 5-year ovarian cancer risk in women aged 50 to 59 years but was not associated with ovarian cancer at ages 40 to 49 or 60 to 79 years.
    CancerSpectrum Knowledge Environment 12/2013; DOI:10.1093/jnci/djt341 · 15.16 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Epoetin is commonly used to treat anemia in chronic kidney disease and End Stage Renal Disease subjects undergoing dialysis, however, there is considerable uncertainty about what level of hemoglobin or hematocrit should be targeted in these subjects. In order to address this question, we treat epoetin dosing guidelines as a type of dynamic treatment regimen. Specifically, we present a methodology for comparing the effects of alternative treatment regimens on survival using observational data. In randomized trials patients can be assigned to follow a specific management guideline, but in observational studies subjects can have treatment paths that appear to be adherent to multiple regimens at the same time. We present a cloning strategy in which each subject contributes follow-up data to each treatment regimen to which they are continuously adherent and artificially censored at first nonadherence. We detail an inverse probability weighted log-rank test with a valid asymptotic variance estimate that can be used to test survival distributions under two regimens. To compare multiple regimens, we propose several marginal structural Cox proportional hazards models with robust variance estimation to account for the creation of clones. The methods are illustrated through simulations and applied to an analysis comparing epoetin dosing regimens in a cohort of 33,873 adult hemodialysis patients from the United States Renal Data System.
    The Annals of Applied Statistics 02/2015; 8(4). DOI:10.1214/14-AOAS774 · 1.69 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: OBJECTIVE As use of standard depression questionnaires in clinical practice increases, clinicians will frequently encounter patients reporting thoughts of death or suicide. This study examined whether responses to the Patient Health Questionnaire for depression (PHQ-9) predict subsequent suicide attempt or suicide death. METHODS Electronic records from a large integrated health system were used to link PHQ-9 responses from outpatient visits to subsequent suicide attempts and suicide deaths. A total of 84,418 outpatients age ≥13 completed 207,265 questionnaires between 2007 and 2011. Electronic medical records, insurance claims, and death certificate data documented 709 subsequent suicide attempts and 46 suicide deaths in this sample. RESULTS Cumulative risk of suicide attempt over one year increased from .4% among outpatients reporting thoughts of death or self-harm "not at all" to 4% among those reporting thoughts of death or self-harm "nearly every day." After adjustment for age, sex, treatment history, and overall depression severity, responses to item 9 of the PHQ-9 remained a strong predictor of suicide attempt. Cumulative risk of suicide death over one year increased from .03% among those reporting thoughts of death or self-harm ideation "not at all" to .3% among those reporting such thoughts "nearly every day." Response to item 9 remained a moderate predictor of subsequent suicide death after the same factor adjustments. CONCLUSIONS Response to item 9 of the PHQ-9 for depression identified outpatients at increased risk of suicide attempt or death. This excess risk emerged over several days and continued to grow for several months, indicating that suicidal ideation was an enduring vulnerability rather than a short-term crisis.
    Psychiatric services (Washington, D.C.) 09/2013; 64(12). DOI:10.1176/ · 1.99 Impact Factor