The Balanced Survivor Average Causal Effect
-
Tom Greene
,
Marshall Joffe
Abstract: Statistical analysis of longitudinal outcomes is often complicated by the absence of observable values in patients who die prior to their scheduled measurement. In such cases, the longitudinal data are said to be “truncated by death” to emphasize that the longitudinal measurements are not simply missing, but are undefined after death. Recently, the truncation by death problem has been investigated using the framework of principal stratification to define the target estimand as the survivor average causal effect (SACE), which in the context of a two-group randomized clinical trial is the mean difference in the longitudinal outcome between the treatment and control groups for the principal stratum of always-survivors. The SACE is not identified without untestable assumptions. These assumptions have often been formulated in terms of a monotonicity constraint requiring that the treatment does not reduce survival in any patient, in conjunction with assumed values for mean differences in the longitudinal outcome between certain principal strata. In this paper, we introduce an alternative estimand, the balanced-SACE, which is defined as the average causal effect on the longitudinal outcome in a particular subset of the always-survivors that is balanced with respect to the potential survival times under the treatment and control. We propose a simple estimator of the balanced-SACE that compares the longitudinal outcomes between equivalent fractions of the longest surviving patients between the treatment and control groups and does not require a monotonicity assumption. We provide expressions for the large sample bias of the estimator, along with sensitivity analyses and strategies to minimize this bias. We consider statistical inference under a bootstrap resampling procedure.
References
[1] ZhangJL, RubinDB.Estimation of causal effects via principal stratification when some outcomes are truncated by “death.”J Educ Behav Stat2003; 28:353–68.10.3102/10769986028004353Search in Google Scholar
[2] RubinD.Causal inference through potential outcomes and principal stratification: application to studies with “censoring” due to death. Stat Sci2006; 21:299–309.10.1214/088342306000000114Search in Google Scholar
[3] McConnellS, StuartE, DevaneyB.The truncation by death problem. What to do in an experimental evaluation when the outcome is not always measured. Eval Rev2008; 32:157–86.10.1177/0193841X07309115Search in Google Scholar
[4] DlehrP, PatrickD.Trajectories of health for older adults over time: accounting fully for death. Ann Intern Med2003; 139:416–20.10.7326/0003-4819-139-5_Part_2-200309021-00007Search in Google Scholar
[5] LachinJ.Worst-rank score analysis with informatively missing observations in clinical trials.Control Clin Trials1999; 20:408–22.10.1016/S0197-2456(99)00022-7Search in Google Scholar
[6] KurlandB, JohnsonL, EglestonB, DiehrP.Longitudinal data with follow-up truncated by death: finding a match between analysis method and research aims. Stat Sci2009; 24:211–22.10.1214/09-STS293Search in Google Scholar
[7] KalbfleischJ, PrenticeR.The statistical analysis of failure time data, 2nd ed. Hoboken, NJ: Wiley-Interscience, 2002.Search in Google Scholar
[8] RobinsJ.A new approach to causal inference in mortality studies with a sustained exposure period – application to control of the healthy worker survivor effect. Math Model1986; 7:1393–512.10.1016/0270-0255(86)90088-6Search in Google Scholar
[9] FrangakisCE, and RubinDB.Principal stratification in causal inference. Biometrics2002; 58:20–9.10.1111/j.0006-341X.2002.00021.xSearch in Google Scholar
[10] SjolanderA, HumphreysK, VansteelandtS, BelloccoR, PalmgrenJ.Sensitivity analysis for principal stratum direct effects, with an application to a study of physical activity and coronary heart disease. Biometrics2009; 65:514–20.10.1111/j.1541-0420.2008.01108.xSearch in Google Scholar PubMed
[11] EglestonB, ScharfsteinD, MacKenzieE.On estimation of the survivor average causal effect in observational studies when important confounders are missing due to death, Biometrics2009; 65(2):497–504.10.1111/j.1541-0420.2008.01111.xSearch in Google Scholar PubMed PubMed Central
[12] RobinsJM.An analytic method for randomized trials with informative censoring: Part I. Lifetime Data Anal1995; 1:241–54.10.1007/BF00985759Search in Google Scholar
[13] ChibaY, VanderWeeleT.A simple method for principal strata effects when the outcome has been truncated due to death. Am J Epidemiol2011; 173(7):745–51.10.1093/aje/kwq418Search in Google Scholar
[14] LeeK, DanielsM.Marginalized models for longitudinal ordinal data with application to quality of life studies. Stat Med2008; 27:4359–80.10.1002/sim.3352Search in Google Scholar
[15] ShalitH, YitzhakiS.Marginal conditional stochastic dominance. Manag Sci1994; 40:670–84.10.1287/mnsc.40.5.670Search in Google Scholar
[16] SchechtmanE, ShelefA, YitzhakiS, ZitikrisR.Testing hypotheses about absolute concentration curves and marginal conditional stochastic dominance. Economet Theory2008; 24:1044–62.10.1017/S0266466608080407Search in Google Scholar
[17] GreeneT, BeckG, GassmanJ, GotchFA, KusekJW, LeveyAS, LevinNW, SchulmanG, EknoyanG. for the HEMO Study Group. Design and statistical issues of the hemodialysis (HEMO) study. Control Clin Trials2000; 21(5):502–25.10.1016/S0197-2456(00)00062-3Search in Google Scholar
[18] EknoyanG, BeckG, CheungA, DaugirdasJ, GreeneT, KusekJ, et al.. Effect of dialysis dose and membrane flux in maintenance hemodialysis. New Engl J Medi2002; 347:2010–19.Search in Google Scholar
[19] MiskulinDC, AthienitesNV, YanG,MartinA, OrntD, KusekJ, MeyerK, LeveyA, for the Hemodialysis (HEMO) Study Group. Comorbidity assessment using the Index of Coexistent Diseases in a multicenter clinical trial. Kidney Int2001; 60:1498–510.10.1046/j.1523-1755.2001.00954.xSearch in Google Scholar PubMed
[20] RobinsJ.Causal models for estimating the effects of weight gain on mortality. Int J Obes32, S15–S41, 200810.1038/ijo.2008.83Search in Google Scholar PubMed
[21] JoffeM, SmallD, HsuC.Defining and estimating intervention effects for groups that will develop an auxiliary outcome. Stati Sci2007; 22:74–97.10.1214/088342306000000655Search in Google Scholar
[22] PearlJ.Principal stratification – A goal or a tool? Int J Biostat.2011: 7, Article 20, DOI: 10.2202/1557–4679.1322.Search in Google Scholar
[23] SjolanderA.Reaction to Pearl’s critique of principal stratification. Int J Biostat2011; 7(1), Article 36. DOI: 10.2202/1557–4679.1341.Search in Google Scholar
[24] JofferM.Principal stratification and attribution prohibition: Good ideas taken too far. Int J Biostat2011; 7(1), Article 35. DOI: 10.2202/1557–4679.1367.Search in Google Scholar
©2013 by Walter de Gruyter Berlin / Boston
Articles in the same Issue
- Masthead
- Masthead
- Research Articles
- Sensitivity Analysis for Causal Inference under Unmeasured Confounding and Measurement Error Problems
- Assessing the Causal Effect of Policies: An Example Using Stochastic Interventions
- Novel Point Estimation from a Semiparametric Ratio Estimator (SPRE): Long-Term Health Outcomes from Short-Term Linear Data, with Application to Weight Loss in Obesity
- Exact Nonparametric Confidence Bands for the Survivor Function
- Semiparametric Regression Analysis of Clustered Interval-Censored Failure Time Data with Informative Cluster Size
- A Weighting Analogue to Pair Matching in Propensity Score Analysis
- Alternative Monotonicity Assumptions for Improving Bounds on Natural Direct Effects
- Estimation of Risk Ratios in Cohort Studies with a Common Outcome: A Simple and Efficient Two-stage Approach
- Distance-Based Mapping of Disease Risk
- The Balanced Survivor Average Causal Effect
- Commentary
- Principal Stratification: A Broader Vision
Articles in the same Issue
- Masthead
- Masthead
- Research Articles
- Sensitivity Analysis for Causal Inference under Unmeasured Confounding and Measurement Error Problems
- Assessing the Causal Effect of Policies: An Example Using Stochastic Interventions
- Novel Point Estimation from a Semiparametric Ratio Estimator (SPRE): Long-Term Health Outcomes from Short-Term Linear Data, with Application to Weight Loss in Obesity
- Exact Nonparametric Confidence Bands for the Survivor Function
- Semiparametric Regression Analysis of Clustered Interval-Censored Failure Time Data with Informative Cluster Size
- A Weighting Analogue to Pair Matching in Propensity Score Analysis
- Alternative Monotonicity Assumptions for Improving Bounds on Natural Direct Effects
- Estimation of Risk Ratios in Cohort Studies with a Common Outcome: A Simple and Efficient Two-stage Approach
- Distance-Based Mapping of Disease Risk
- The Balanced Survivor Average Causal Effect
- Commentary
- Principal Stratification: A Broader Vision
