Accommodating measurements below a limit of detection: A novel application of cox regression

Gregg E. Dinse, Todd A. Jusko, Lindsey A. Ho, Kaushik Annam, Barry I. Graubard, Irva Hertz-Picciotto, Frederick W. Miller, Brenda W. Gillespie, Clarice R. Weinberg

Research output: Contribution to journalArticle

16 Citations (Scopus)

Abstract

In environmental epidemiology, measurements of exposure biomarkers often fall below the assay's limit of detection. Existing methods for handling this problem, including deletion, substitution, parametric regression, and multiple imputation, can perform poorly if the proportion of "nondetects" is high or parametric models are misspecified. We propose an approach that treats the measured analyte as the modeled outcome, implying a role reversal when the analyte is a putative cause of a health outcome. Following a scale reversal as well, our approach uses Cox regression to model the analyte, with confounder adjustment. The method makes full use of quantifiable analyte measures, while appropriately treating nondetects as censored. Under the proportional hazards assumption, the hazard ratio for a binary health outcome is interpretable as an adjusted odds ratio: the odds for the outcome at any particular analyte concentration divided by the odds given a lower concentration. Our approach is broadly applicable to cohort studies, case-control studies (frequency matched or not), and cross-sectional studies conducted to identify determinants of exposure. We illustrate the method with cross-sectional survey data to assess sex as a determinant of 2,3,7,8-tetrachlorodibenzo-p-dioxin concentration and with prospective cohort data to assess the association between 2,4,4′-trichlorobiphenyl exposure and psychomotor development.

Original languageEnglish (US)
Pages (from-to)1018-1024
Number of pages7
JournalAmerican Journal of Epidemiology
Volume179
Issue number8
DOIs
StatePublished - Apr 15 2014

Fingerprint

Limit of Detection
Cross-Sectional Studies
Health
Case-Control Studies
Epidemiology
Cohort Studies
Biomarkers
Odds Ratio
Polychlorinated Dibenzodioxins

Keywords

  • 2,3,7,8-tetrachlorodibenzo-p-dioxin
  • 2,4,4′-trichlorobiphenyl
  • hazard identification
  • limit of detection
  • National Health and Nutrition Examination Survey
  • nondetects
  • proportional hazards

ASJC Scopus subject areas

  • Epidemiology

Cite this

Accommodating measurements below a limit of detection : A novel application of cox regression. / Dinse, Gregg E.; Jusko, Todd A.; Ho, Lindsey A.; Annam, Kaushik; Graubard, Barry I.; Hertz-Picciotto, Irva; Miller, Frederick W.; Gillespie, Brenda W.; Weinberg, Clarice R.

In: American Journal of Epidemiology, Vol. 179, No. 8, 15.04.2014, p. 1018-1024.

Research output: Contribution to journalArticle

Dinse, GE, Jusko, TA, Ho, LA, Annam, K, Graubard, BI, Hertz-Picciotto, I, Miller, FW, Gillespie, BW & Weinberg, CR 2014, 'Accommodating measurements below a limit of detection: A novel application of cox regression', American Journal of Epidemiology, vol. 179, no. 8, pp. 1018-1024. https://doi.org/10.1093/aje/kwu017
Dinse, Gregg E. ; Jusko, Todd A. ; Ho, Lindsey A. ; Annam, Kaushik ; Graubard, Barry I. ; Hertz-Picciotto, Irva ; Miller, Frederick W. ; Gillespie, Brenda W. ; Weinberg, Clarice R. / Accommodating measurements below a limit of detection : A novel application of cox regression. In: American Journal of Epidemiology. 2014 ; Vol. 179, No. 8. pp. 1018-1024.
@article{115a3b344483453597d0941e48f1fb76,
title = "Accommodating measurements below a limit of detection: A novel application of cox regression",
abstract = "In environmental epidemiology, measurements of exposure biomarkers often fall below the assay's limit of detection. Existing methods for handling this problem, including deletion, substitution, parametric regression, and multiple imputation, can perform poorly if the proportion of {"}nondetects{"} is high or parametric models are misspecified. We propose an approach that treats the measured analyte as the modeled outcome, implying a role reversal when the analyte is a putative cause of a health outcome. Following a scale reversal as well, our approach uses Cox regression to model the analyte, with confounder adjustment. The method makes full use of quantifiable analyte measures, while appropriately treating nondetects as censored. Under the proportional hazards assumption, the hazard ratio for a binary health outcome is interpretable as an adjusted odds ratio: the odds for the outcome at any particular analyte concentration divided by the odds given a lower concentration. Our approach is broadly applicable to cohort studies, case-control studies (frequency matched or not), and cross-sectional studies conducted to identify determinants of exposure. We illustrate the method with cross-sectional survey data to assess sex as a determinant of 2,3,7,8-tetrachlorodibenzo-p-dioxin concentration and with prospective cohort data to assess the association between 2,4,4′-trichlorobiphenyl exposure and psychomotor development.",
keywords = "2,3,7,8-tetrachlorodibenzo-p-dioxin, 2,4,4′-trichlorobiphenyl, hazard identification, limit of detection, National Health and Nutrition Examination Survey, nondetects, proportional hazards",
author = "Dinse, {Gregg E.} and Jusko, {Todd A.} and Ho, {Lindsey A.} and Kaushik Annam and Graubard, {Barry I.} and Irva Hertz-Picciotto and Miller, {Frederick W.} and Gillespie, {Brenda W.} and Weinberg, {Clarice R.}",
year = "2014",
month = "4",
day = "15",
doi = "10.1093/aje/kwu017",
language = "English (US)",
volume = "179",
pages = "1018--1024",
journal = "American Journal of Epidemiology",
issn = "0002-9262",
publisher = "Oxford University Press",
number = "8",

}

TY - JOUR

T1 - Accommodating measurements below a limit of detection

T2 - A novel application of cox regression

AU - Dinse, Gregg E.

AU - Jusko, Todd A.

AU - Ho, Lindsey A.

AU - Annam, Kaushik

AU - Graubard, Barry I.

AU - Hertz-Picciotto, Irva

AU - Miller, Frederick W.

AU - Gillespie, Brenda W.

AU - Weinberg, Clarice R.

PY - 2014/4/15

Y1 - 2014/4/15

N2 - In environmental epidemiology, measurements of exposure biomarkers often fall below the assay's limit of detection. Existing methods for handling this problem, including deletion, substitution, parametric regression, and multiple imputation, can perform poorly if the proportion of "nondetects" is high or parametric models are misspecified. We propose an approach that treats the measured analyte as the modeled outcome, implying a role reversal when the analyte is a putative cause of a health outcome. Following a scale reversal as well, our approach uses Cox regression to model the analyte, with confounder adjustment. The method makes full use of quantifiable analyte measures, while appropriately treating nondetects as censored. Under the proportional hazards assumption, the hazard ratio for a binary health outcome is interpretable as an adjusted odds ratio: the odds for the outcome at any particular analyte concentration divided by the odds given a lower concentration. Our approach is broadly applicable to cohort studies, case-control studies (frequency matched or not), and cross-sectional studies conducted to identify determinants of exposure. We illustrate the method with cross-sectional survey data to assess sex as a determinant of 2,3,7,8-tetrachlorodibenzo-p-dioxin concentration and with prospective cohort data to assess the association between 2,4,4′-trichlorobiphenyl exposure and psychomotor development.

AB - In environmental epidemiology, measurements of exposure biomarkers often fall below the assay's limit of detection. Existing methods for handling this problem, including deletion, substitution, parametric regression, and multiple imputation, can perform poorly if the proportion of "nondetects" is high or parametric models are misspecified. We propose an approach that treats the measured analyte as the modeled outcome, implying a role reversal when the analyte is a putative cause of a health outcome. Following a scale reversal as well, our approach uses Cox regression to model the analyte, with confounder adjustment. The method makes full use of quantifiable analyte measures, while appropriately treating nondetects as censored. Under the proportional hazards assumption, the hazard ratio for a binary health outcome is interpretable as an adjusted odds ratio: the odds for the outcome at any particular analyte concentration divided by the odds given a lower concentration. Our approach is broadly applicable to cohort studies, case-control studies (frequency matched or not), and cross-sectional studies conducted to identify determinants of exposure. We illustrate the method with cross-sectional survey data to assess sex as a determinant of 2,3,7,8-tetrachlorodibenzo-p-dioxin concentration and with prospective cohort data to assess the association between 2,4,4′-trichlorobiphenyl exposure and psychomotor development.

KW - 2,3,7,8-tetrachlorodibenzo-p-dioxin

KW - 2,4,4′-trichlorobiphenyl

KW - hazard identification

KW - limit of detection

KW - National Health and Nutrition Examination Survey

KW - nondetects

KW - proportional hazards

UR - http://www.scopus.com/inward/record.url?scp=84897398140&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84897398140&partnerID=8YFLogxK

U2 - 10.1093/aje/kwu017

DO - 10.1093/aje/kwu017

M3 - Article

C2 - 24671072

AN - SCOPUS:84897398140

VL - 179

SP - 1018

EP - 1024

JO - American Journal of Epidemiology

JF - American Journal of Epidemiology

SN - 0002-9262

IS - 8

ER -