Semiparametric estimation of treatment effect in a pretest-posttest study with missing data

Marie Davidian, Anastasios A. Tsiatis, Selene Leon, Hyonggin An, Roderick Little, Babette A. Brumback, Lyndia C. Brumback, Geert Molenberghs, Joseph L. Schafer, Joseph D Y Kang

Research output: Contribution to journalReview article

54 Citations (Scopus)

Abstract

The pretest-posttest study is commonplace in numerous applications. Typically, subjects are randomized to two treatments, and response is measured at baseline, prior to intervention with the randomized treatment (pretest), and at prespecified follow-up time (posttest). Interest focuses on the effect of treatments on the change between mean baseline and follow-up response. Missing posttest response for some subjects is routine, and disregarding missing cases can lead to invalid inference. Despite the popularity of this design, a consensus on an appropriate analysis when no data are missing, let alone for taking into account missing follow-up, does not exist. Under a semiparametric perspective on the pretest-posttest model, in which limited distributional assumptions on pretest or posttest response are made, we show how the theory of Robins, Rotnitzky and Zhao may be used to characterize a class of consistent treatment effect estimators and to identify the efficient estimator in the class. We then describe how the theoretical results translate into practice. The development not only shows how a unified framework for inference in this setting emerges from the Robins, Rotnitzky and Zhao theory, but also provides a review and demonstration of the key aspects of this theory in a familiar context. The results are also relevant to the problem of comparing two treatment means with adjustment for baseline covariates.

Original languageEnglish
Pages (from-to)261-301
Number of pages41
JournalStatistical Science
Volume20
Issue number3
DOIs
Publication statusPublished - 2005 Aug 1
Externally publishedYes

Fingerprint

Semiparametric Estimation
Pre-test
Treatment Effects
Missing Data
Baseline
Efficient Estimator
Covariates
Adjustment
Estimator
Missing data
Treatment effects
Semiparametric estimation
Class
Inference

Keywords

  • Analysis of covariance
  • Covariate adjustment
  • Influence function
  • Inverse probability weighting
  • Missing at random

ASJC Scopus subject areas

  • Statistics and Probability
  • Mathematics(all)
  • Statistics, Probability and Uncertainty

Cite this

Davidian, M., Tsiatis, A. A., Leon, S., An, H., Little, R., Brumback, B. A., ... Kang, J. D. Y. (2005). Semiparametric estimation of treatment effect in a pretest-posttest study with missing data. Statistical Science, 20(3), 261-301. https://doi.org/10.1214/088342305000000151

Semiparametric estimation of treatment effect in a pretest-posttest study with missing data. / Davidian, Marie; Tsiatis, Anastasios A.; Leon, Selene; An, Hyonggin; Little, Roderick; Brumback, Babette A.; Brumback, Lyndia C.; Molenberghs, Geert; Schafer, Joseph L.; Kang, Joseph D Y.

In: Statistical Science, Vol. 20, No. 3, 01.08.2005, p. 261-301.

Research output: Contribution to journalReview article

Davidian, M, Tsiatis, AA, Leon, S, An, H, Little, R, Brumback, BA, Brumback, LC, Molenberghs, G, Schafer, JL & Kang, JDY 2005, 'Semiparametric estimation of treatment effect in a pretest-posttest study with missing data', Statistical Science, vol. 20, no. 3, pp. 261-301. https://doi.org/10.1214/088342305000000151
Davidian, Marie ; Tsiatis, Anastasios A. ; Leon, Selene ; An, Hyonggin ; Little, Roderick ; Brumback, Babette A. ; Brumback, Lyndia C. ; Molenberghs, Geert ; Schafer, Joseph L. ; Kang, Joseph D Y. / Semiparametric estimation of treatment effect in a pretest-posttest study with missing data. In: Statistical Science. 2005 ; Vol. 20, No. 3. pp. 261-301.
@article{4f28c2d170b3498688708d78c38f4fef,
title = "Semiparametric estimation of treatment effect in a pretest-posttest study with missing data",
abstract = "The pretest-posttest study is commonplace in numerous applications. Typically, subjects are randomized to two treatments, and response is measured at baseline, prior to intervention with the randomized treatment (pretest), and at prespecified follow-up time (posttest). Interest focuses on the effect of treatments on the change between mean baseline and follow-up response. Missing posttest response for some subjects is routine, and disregarding missing cases can lead to invalid inference. Despite the popularity of this design, a consensus on an appropriate analysis when no data are missing, let alone for taking into account missing follow-up, does not exist. Under a semiparametric perspective on the pretest-posttest model, in which limited distributional assumptions on pretest or posttest response are made, we show how the theory of Robins, Rotnitzky and Zhao may be used to characterize a class of consistent treatment effect estimators and to identify the efficient estimator in the class. We then describe how the theoretical results translate into practice. The development not only shows how a unified framework for inference in this setting emerges from the Robins, Rotnitzky and Zhao theory, but also provides a review and demonstration of the key aspects of this theory in a familiar context. The results are also relevant to the problem of comparing two treatment means with adjustment for baseline covariates.",
keywords = "Analysis of covariance, Covariate adjustment, Influence function, Inverse probability weighting, Missing at random",
author = "Marie Davidian and Tsiatis, {Anastasios A.} and Selene Leon and Hyonggin An and Roderick Little and Brumback, {Babette A.} and Brumback, {Lyndia C.} and Geert Molenberghs and Schafer, {Joseph L.} and Kang, {Joseph D Y}",
year = "2005",
month = "8",
day = "1",
doi = "10.1214/088342305000000151",
language = "English",
volume = "20",
pages = "261--301",
journal = "Statistical Science",
issn = "0883-4237",
publisher = "Institute of Mathematical Statistics",
number = "3",

}

TY - JOUR

T1 - Semiparametric estimation of treatment effect in a pretest-posttest study with missing data

AU - Davidian, Marie

AU - Tsiatis, Anastasios A.

AU - Leon, Selene

AU - An, Hyonggin

AU - Little, Roderick

AU - Brumback, Babette A.

AU - Brumback, Lyndia C.

AU - Molenberghs, Geert

AU - Schafer, Joseph L.

AU - Kang, Joseph D Y

PY - 2005/8/1

Y1 - 2005/8/1

N2 - The pretest-posttest study is commonplace in numerous applications. Typically, subjects are randomized to two treatments, and response is measured at baseline, prior to intervention with the randomized treatment (pretest), and at prespecified follow-up time (posttest). Interest focuses on the effect of treatments on the change between mean baseline and follow-up response. Missing posttest response for some subjects is routine, and disregarding missing cases can lead to invalid inference. Despite the popularity of this design, a consensus on an appropriate analysis when no data are missing, let alone for taking into account missing follow-up, does not exist. Under a semiparametric perspective on the pretest-posttest model, in which limited distributional assumptions on pretest or posttest response are made, we show how the theory of Robins, Rotnitzky and Zhao may be used to characterize a class of consistent treatment effect estimators and to identify the efficient estimator in the class. We then describe how the theoretical results translate into practice. The development not only shows how a unified framework for inference in this setting emerges from the Robins, Rotnitzky and Zhao theory, but also provides a review and demonstration of the key aspects of this theory in a familiar context. The results are also relevant to the problem of comparing two treatment means with adjustment for baseline covariates.

AB - The pretest-posttest study is commonplace in numerous applications. Typically, subjects are randomized to two treatments, and response is measured at baseline, prior to intervention with the randomized treatment (pretest), and at prespecified follow-up time (posttest). Interest focuses on the effect of treatments on the change between mean baseline and follow-up response. Missing posttest response for some subjects is routine, and disregarding missing cases can lead to invalid inference. Despite the popularity of this design, a consensus on an appropriate analysis when no data are missing, let alone for taking into account missing follow-up, does not exist. Under a semiparametric perspective on the pretest-posttest model, in which limited distributional assumptions on pretest or posttest response are made, we show how the theory of Robins, Rotnitzky and Zhao may be used to characterize a class of consistent treatment effect estimators and to identify the efficient estimator in the class. We then describe how the theoretical results translate into practice. The development not only shows how a unified framework for inference in this setting emerges from the Robins, Rotnitzky and Zhao theory, but also provides a review and demonstration of the key aspects of this theory in a familiar context. The results are also relevant to the problem of comparing two treatment means with adjustment for baseline covariates.

KW - Analysis of covariance

KW - Covariate adjustment

KW - Influence function

KW - Inverse probability weighting

KW - Missing at random

UR - http://www.scopus.com/inward/record.url?scp=26044449172&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=26044449172&partnerID=8YFLogxK

U2 - 10.1214/088342305000000151

DO - 10.1214/088342305000000151

M3 - Review article

AN - SCOPUS:26044449172

VL - 20

SP - 261

EP - 301

JO - Statistical Science

JF - Statistical Science

SN - 0883-4237

IS - 3

ER -