Novel criteria to exclude the surrogate paradox and their optimalities

Yunjian Yin, Lan Liu, Zhi Geng, Peng Luo

Research output: Contribution to journalArticle

Abstract

When the primary outcome is hard to collect, a surrogate endpoint is typically used as a substitute. However, even when a treatment has a positive average causal effect (ACE) on a surrogate endpoint, which also has a positive ACE on the primary outcome, it is still possible that the treatment has a negative ACE on the primary outcome. Such a phenomenon is called the surrogate paradox and greatly challenges the use of surrogates. In this paper, we provide criteria to exclude the surrogate paradox. Our criteria are optimal in the sense that they are sufficient and “almost necessary” to exclude the paradox: If the conditions are satisfied, the surrogate paradox is guaranteed to be absent, whereas if the conditions fail, there exists a data-generating process with surrogate paradox that can generate the same observed data. That is, our criteria capture all the observed information to exclude the surrogate paradox.

Original languageEnglish (US)
JournalScandinavian Journal of Statistics
DOIs
StatePublished - Jan 1 2019

Fingerprint

Paradox
Optimality
Causal Effect
Surrogate Endpoint
Observed Information
Substitute
Sufficient
Necessary
Causal effect

Keywords

  • average causal effect
  • optimality
  • surrogate paradox

Cite this

Novel criteria to exclude the surrogate paradox and their optimalities. / Yin, Yunjian; Liu, Lan; Geng, Zhi; Luo, Peng.

In: Scandinavian Journal of Statistics, 01.01.2019.

Research output: Contribution to journalArticle

@article{03f160c64bb946e39fdd48a633e24020,
title = "Novel criteria to exclude the surrogate paradox and their optimalities",
abstract = "When the primary outcome is hard to collect, a surrogate endpoint is typically used as a substitute. However, even when a treatment has a positive average causal effect (ACE) on a surrogate endpoint, which also has a positive ACE on the primary outcome, it is still possible that the treatment has a negative ACE on the primary outcome. Such a phenomenon is called the surrogate paradox and greatly challenges the use of surrogates. In this paper, we provide criteria to exclude the surrogate paradox. Our criteria are optimal in the sense that they are sufficient and “almost necessary” to exclude the paradox: If the conditions are satisfied, the surrogate paradox is guaranteed to be absent, whereas if the conditions fail, there exists a data-generating process with surrogate paradox that can generate the same observed data. That is, our criteria capture all the observed information to exclude the surrogate paradox.",
keywords = "average causal effect, optimality, surrogate paradox",
author = "Yunjian Yin and Lan Liu and Zhi Geng and Peng Luo",
year = "2019",
month = "1",
day = "1",
doi = "10.1111/sjos.12398",
language = "English (US)",
journal = "Scandinavian Journal of Statistics",
issn = "0303-6898",
publisher = "Wiley-Blackwell",

}

TY - JOUR

T1 - Novel criteria to exclude the surrogate paradox and their optimalities

AU - Yin, Yunjian

AU - Liu, Lan

AU - Geng, Zhi

AU - Luo, Peng

PY - 2019/1/1

Y1 - 2019/1/1

N2 - When the primary outcome is hard to collect, a surrogate endpoint is typically used as a substitute. However, even when a treatment has a positive average causal effect (ACE) on a surrogate endpoint, which also has a positive ACE on the primary outcome, it is still possible that the treatment has a negative ACE on the primary outcome. Such a phenomenon is called the surrogate paradox and greatly challenges the use of surrogates. In this paper, we provide criteria to exclude the surrogate paradox. Our criteria are optimal in the sense that they are sufficient and “almost necessary” to exclude the paradox: If the conditions are satisfied, the surrogate paradox is guaranteed to be absent, whereas if the conditions fail, there exists a data-generating process with surrogate paradox that can generate the same observed data. That is, our criteria capture all the observed information to exclude the surrogate paradox.

AB - When the primary outcome is hard to collect, a surrogate endpoint is typically used as a substitute. However, even when a treatment has a positive average causal effect (ACE) on a surrogate endpoint, which also has a positive ACE on the primary outcome, it is still possible that the treatment has a negative ACE on the primary outcome. Such a phenomenon is called the surrogate paradox and greatly challenges the use of surrogates. In this paper, we provide criteria to exclude the surrogate paradox. Our criteria are optimal in the sense that they are sufficient and “almost necessary” to exclude the paradox: If the conditions are satisfied, the surrogate paradox is guaranteed to be absent, whereas if the conditions fail, there exists a data-generating process with surrogate paradox that can generate the same observed data. That is, our criteria capture all the observed information to exclude the surrogate paradox.

KW - average causal effect

KW - optimality

KW - surrogate paradox

UR - http://www.scopus.com/inward/record.url?scp=85067354480&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85067354480&partnerID=8YFLogxK

U2 - 10.1111/sjos.12398

DO - 10.1111/sjos.12398

M3 - Article

AN - SCOPUS:85067354480

JO - Scandinavian Journal of Statistics

JF - Scandinavian Journal of Statistics

SN - 0303-6898

ER -