Estimating the odds ratio when exposure has a limit of detection

Stephen R. Cole, Haitao Chu, Lei Nie, Enrique F. Schisterman

Research output: Contribution to journalArticle

Abstract

Background: In epidemiologic research, little emphasis has been placed on methods to account for left-hand censoring of 'exposures' due to a limit of detection (LOD). Methods: We calculate the odds of anti-HIV therapy naiveté in 45 HIV-infected men as a function of measured log10 plasma HIV RNA viral load using five approaches including ad hoc methods as well as a maximum likelihood estimate (MLE). We also generated simulations of a binary outcome with 10% incidence and a 1.5-fold increased odds per log increase in a log-normally distributed exposure with 25, 50 and 75% of exposure data below LOD. Simulated data were analysed using the same five methods, as well as the full data. Results: In the example, the estimated odds ratio (OR) varied by 1.22-fold across methods, from 1.45 to 1.77 per log10 copies of viral load and the standard error for the log OR varied by 1.52-fold across methods, from 0.31 to 0.47. In the simulations, use of full data or the MLE was unbiased with appropriate confidence interval (CI) coverage. However, as the proportion of exposure below LOD increased, substituting LOD, LOD/√2 or LOD/2 was increasingly biased with increasingly inappropriate CI coverage. Finally, exclusion of values below LOD was unbiased but imprecise. Conclusions: In this example and the settings explored by simulation, and among methods readily available to investigators (i.e. sans full data), the MLE provided an unbiased and appropriately precise estimate of the exposure-outcome OR. Published by Oxford University Press on behalf of the International Epidemiological Association.

Original languageEnglish (US)
Article numberdyp269
Pages (from-to)1674-1680
Number of pages7
JournalInternational Journal of Epidemiology
Volume38
Issue number6
DOIs
StatePublished - Aug 10 2009
Externally publishedYes

Fingerprint

Limit of Detection
Odds Ratio
Likelihood Functions
HIV
Viral Load
Confidence Intervals
Hand
Research Personnel
RNA
Incidence
Research

Keywords

  • Biomarkers
  • Epidemiologic methods
  • Limit of detection
  • Statistical method

ASJC Scopus subject areas

  • Epidemiology

Cite this

Cole, S. R., Chu, H., Nie, L., & Schisterman, E. F. (2009). Estimating the odds ratio when exposure has a limit of detection. International Journal of Epidemiology, 38(6), 1674-1680. [dyp269]. https://doi.org/10.1093/ije/dyp269

Estimating the odds ratio when exposure has a limit of detection. / Cole, Stephen R.; Chu, Haitao; Nie, Lei; Schisterman, Enrique F.

In: International Journal of Epidemiology, Vol. 38, No. 6, dyp269, 10.08.2009, p. 1674-1680.

Research output: Contribution to journalArticle

Cole, SR, Chu, H, Nie, L & Schisterman, EF 2009, 'Estimating the odds ratio when exposure has a limit of detection', International Journal of Epidemiology, vol. 38, no. 6, dyp269, pp. 1674-1680. https://doi.org/10.1093/ije/dyp269
Cole, Stephen R. ; Chu, Haitao ; Nie, Lei ; Schisterman, Enrique F. / Estimating the odds ratio when exposure has a limit of detection. In: International Journal of Epidemiology. 2009 ; Vol. 38, No. 6. pp. 1674-1680.
@article{09b77d31b12a4f38bb04049068842904,
title = "Estimating the odds ratio when exposure has a limit of detection",
abstract = "Background: In epidemiologic research, little emphasis has been placed on methods to account for left-hand censoring of 'exposures' due to a limit of detection (LOD). Methods: We calculate the odds of anti-HIV therapy naivet{\'e} in 45 HIV-infected men as a function of measured log10 plasma HIV RNA viral load using five approaches including ad hoc methods as well as a maximum likelihood estimate (MLE). We also generated simulations of a binary outcome with 10{\%} incidence and a 1.5-fold increased odds per log increase in a log-normally distributed exposure with 25, 50 and 75{\%} of exposure data below LOD. Simulated data were analysed using the same five methods, as well as the full data. Results: In the example, the estimated odds ratio (OR) varied by 1.22-fold across methods, from 1.45 to 1.77 per log10 copies of viral load and the standard error for the log OR varied by 1.52-fold across methods, from 0.31 to 0.47. In the simulations, use of full data or the MLE was unbiased with appropriate confidence interval (CI) coverage. However, as the proportion of exposure below LOD increased, substituting LOD, LOD/√2 or LOD/2 was increasingly biased with increasingly inappropriate CI coverage. Finally, exclusion of values below LOD was unbiased but imprecise. Conclusions: In this example and the settings explored by simulation, and among methods readily available to investigators (i.e. sans full data), the MLE provided an unbiased and appropriately precise estimate of the exposure-outcome OR. Published by Oxford University Press on behalf of the International Epidemiological Association.",
keywords = "Biomarkers, Epidemiologic methods, Limit of detection, Statistical method",
author = "Cole, {Stephen R.} and Haitao Chu and Lei Nie and Schisterman, {Enrique F.}",
year = "2009",
month = "8",
day = "10",
doi = "10.1093/ije/dyp269",
language = "English (US)",
volume = "38",
pages = "1674--1680",
journal = "International Journal of Epidemiology",
issn = "0300-5771",
publisher = "Oxford University Press",
number = "6",

}

TY - JOUR

T1 - Estimating the odds ratio when exposure has a limit of detection

AU - Cole, Stephen R.

AU - Chu, Haitao

AU - Nie, Lei

AU - Schisterman, Enrique F.

PY - 2009/8/10

Y1 - 2009/8/10

N2 - Background: In epidemiologic research, little emphasis has been placed on methods to account for left-hand censoring of 'exposures' due to a limit of detection (LOD). Methods: We calculate the odds of anti-HIV therapy naiveté in 45 HIV-infected men as a function of measured log10 plasma HIV RNA viral load using five approaches including ad hoc methods as well as a maximum likelihood estimate (MLE). We also generated simulations of a binary outcome with 10% incidence and a 1.5-fold increased odds per log increase in a log-normally distributed exposure with 25, 50 and 75% of exposure data below LOD. Simulated data were analysed using the same five methods, as well as the full data. Results: In the example, the estimated odds ratio (OR) varied by 1.22-fold across methods, from 1.45 to 1.77 per log10 copies of viral load and the standard error for the log OR varied by 1.52-fold across methods, from 0.31 to 0.47. In the simulations, use of full data or the MLE was unbiased with appropriate confidence interval (CI) coverage. However, as the proportion of exposure below LOD increased, substituting LOD, LOD/√2 or LOD/2 was increasingly biased with increasingly inappropriate CI coverage. Finally, exclusion of values below LOD was unbiased but imprecise. Conclusions: In this example and the settings explored by simulation, and among methods readily available to investigators (i.e. sans full data), the MLE provided an unbiased and appropriately precise estimate of the exposure-outcome OR. Published by Oxford University Press on behalf of the International Epidemiological Association.

AB - Background: In epidemiologic research, little emphasis has been placed on methods to account for left-hand censoring of 'exposures' due to a limit of detection (LOD). Methods: We calculate the odds of anti-HIV therapy naiveté in 45 HIV-infected men as a function of measured log10 plasma HIV RNA viral load using five approaches including ad hoc methods as well as a maximum likelihood estimate (MLE). We also generated simulations of a binary outcome with 10% incidence and a 1.5-fold increased odds per log increase in a log-normally distributed exposure with 25, 50 and 75% of exposure data below LOD. Simulated data were analysed using the same five methods, as well as the full data. Results: In the example, the estimated odds ratio (OR) varied by 1.22-fold across methods, from 1.45 to 1.77 per log10 copies of viral load and the standard error for the log OR varied by 1.52-fold across methods, from 0.31 to 0.47. In the simulations, use of full data or the MLE was unbiased with appropriate confidence interval (CI) coverage. However, as the proportion of exposure below LOD increased, substituting LOD, LOD/√2 or LOD/2 was increasingly biased with increasingly inappropriate CI coverage. Finally, exclusion of values below LOD was unbiased but imprecise. Conclusions: In this example and the settings explored by simulation, and among methods readily available to investigators (i.e. sans full data), the MLE provided an unbiased and appropriately precise estimate of the exposure-outcome OR. Published by Oxford University Press on behalf of the International Epidemiological Association.

KW - Biomarkers

KW - Epidemiologic methods

KW - Limit of detection

KW - Statistical method

UR - http://www.scopus.com/inward/record.url?scp=75349102030&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=75349102030&partnerID=8YFLogxK

U2 - 10.1093/ije/dyp269

DO - 10.1093/ije/dyp269

M3 - Article

C2 - 19667054

AN - SCOPUS:75349102030

VL - 38

SP - 1674

EP - 1680

JO - International Journal of Epidemiology

JF - International Journal of Epidemiology

SN - 0300-5771

IS - 6

M1 - dyp269

ER -