1
|
Shaw PA, Gustafson P, Carroll RJ, Deffner V, Dodd KW, Keogh RH, Kipnis V, Tooze JA, Wallace MP, Küchenhoff H, Freedman LS. STRATOS guidance document on measurement error and misclassification of variables in observational epidemiology: Part 2-More complex methods of adjustment and advanced topics. Stat Med 2020; 39:2232-2263. [PMID: 32246531 PMCID: PMC7272296 DOI: 10.1002/sim.8531] [Citation(s) in RCA: 33] [Impact Index Per Article: 8.3] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 12/09/2018] [Revised: 02/27/2020] [Accepted: 02/28/2020] [Indexed: 12/24/2022]
Abstract
We continue our review of issues related to measurement error and misclassification in epidemiology. We further describe methods of adjusting for biased estimation caused by measurement error in continuous covariates, covering likelihood methods, Bayesian methods, moment reconstruction, moment-adjusted imputation, and multiple imputation. We then describe which methods can also be used with misclassification of categorical covariates. Methods of adjusting estimation of distributions of continuous variables for measurement error are then reviewed. Illustrative examples are provided throughout these sections. We provide lists of available software for implementing these methods and also provide the code for implementing our examples in the Supporting Information. Next, we present several advanced topics, including data subject to both classical and Berkson error, modeling continuous exposures with measurement error, and categorical exposures with misclassification in the same model, variable selection when some of the variables are measured with error, adjusting analyses or design for error in an outcome variable, and categorizing continuous variables measured with error. Finally, we provide some advice for the often met situations where variables are known to be measured with substantial error, but there is only an external reference standard or partial (or no) information about the type or magnitude of the error.
Collapse
Affiliation(s)
- Pamela A Shaw
- Department of Biostatistics, Epidemiology, and Informatics, University of Pennsylvania Perelman School of Medicine, Philadelphia, Pennsylvania, USA
| | - Paul Gustafson
- Department of Statistics, University of British Columbia, Vancouver, British Columbia, Canada
| | - Raymond J Carroll
- Department of Statistics, Texas A&M University, College Station, Texas, USA
- School of Mathematical and Physical Sciences, University of Technology Sydney, Broadway, New South Wales, Australia
| | - Veronika Deffner
- Statistical Consulting Unit StaBLab, Department of Statistics, Ludwig-Maximilians-Universität, Munich, Germany
| | - Kevin W Dodd
- Biometry Research Group, Division of Cancer Prevention, National Cancer Institute, Bethesda, Maryland, USA
| | - Ruth H Keogh
- Department of Medical Statistics, London School of Hygiene and Tropical Medicine, London, UK
| | - Victor Kipnis
- Biometry Research Group, Division of Cancer Prevention, National Cancer Institute, Bethesda, Maryland, USA
| | - Janet A Tooze
- Department of Biostatistics and Data Science, Wake Forest School of Medicine, Winston-Salem, North Carolina, USA
| | - Michael P Wallace
- Department of Statistics and Actuarial Science, University of Waterloo, Waterloo, Ontario, Canada
| | - Helmut Küchenhoff
- Statistical Consulting Unit StaBLab, Department of Statistics, Ludwig-Maximilians-Universität, Munich, Germany
| | - Laurence S Freedman
- Biostatistics and Biomathematics Unit, Gertner Institute for Epidemiology and Health Policy Research, Sheba Medical Center, Tel Hashomer, Israel
- Information Management Services Inc., Rockville, Maryland, USA
| |
Collapse
|
2
|
McNally RJQ, Wakeford R, James PW, Basta NO, Alston RD, Pearce MS, Elliott AT. A geographical study of thyroid cancer incidence in north-west England following the Windscale nuclear reactor fire of 1957. JOURNAL OF RADIOLOGICAL PROTECTION : OFFICIAL JOURNAL OF THE SOCIETY FOR RADIOLOGICAL PROTECTION 2016; 36:934-952. [PMID: 27893453 DOI: 10.1088/0952-4746/36/4/934] [Citation(s) in RCA: 6] [Impact Index Per Article: 0.8] [Reference Citation Analysis] [Abstract] [MESH Headings] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 06/06/2023]
Abstract
The Windscale nuclear reactor fire at Sellafield, United Kingdom, in October 1957 led to an uncontrolled release of iodine-131 (radioactive half-life, 8 d) into the atmosphere. Contamination from the accident was most pronounced in the counties of Cumbria and Lancashire, north-west England. Radioiodine concentrates in the thyroid gland producing an excess risk of thyroid cancer, notably among those exposed as children, which persists into later life. For an initial investigation of thyroid cancer incidence in north-west England, data were obtained on cases of thyroid cancer among people born during 1929-1973 and diagnosed during 1974-2012 while resident in England, together with corresponding populations. Incidence rate ratios (IRRs), with Poisson 95% confidence intervals (CIs), compared thyroid cancer incidence rates in Cumbria and in Lancashire with those in the rest of England. For those aged <20 years in 1958, a statistically significantly increased IRR was found for those diagnosed during 1974-2012 while living in Cumbria (IRR = 1.29; 95% CI 1.09-1.52), but the equivalent IRR for Lancashire was marginally non-significantly decreased (IRR = 0.91; 95% CI 0.80-1.04). This pattern of IRRs was also apparent for earlier births, and the significantly increased IRR in Cumbria extended to individuals born in 1959-1963, who would not have been exposed to iodine-131 from the Windscale accident. Moreover, significant overdispersion was present in the temporal distributions of the IRRs, so that Poisson CIs substantially underestimate statistical uncertainties. Consequently, although further investigations are required to properly understand the unusual patterns of thyroid cancer IRRs in Cumbria and Lancashire, the results of this preliminary study are not consistent with an effect of exposure to iodine-131 from the Windscale accident.
Collapse
Affiliation(s)
- Richard J Q McNally
- Institute of Health and Society, Newcastle University, Sir James Spence Institute, Royal Victoria Infirmary, Queen Victoria Road, Newcastle upon Tyne, NE1 4LP, UK
| | | | | | | | | | | | | |
Collapse
|
3
|
Kwon D, Hoffman FO, Moroz BE, Simon SL. Bayesian dose-response analysis for epidemiological studies with complex uncertainty in dose estimation. Stat Med 2015; 35:399-423. [DOI: 10.1002/sim.6635] [Citation(s) in RCA: 24] [Impact Index Per Article: 2.7] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/29/2014] [Revised: 07/31/2015] [Accepted: 08/10/2015] [Indexed: 11/09/2022]
Affiliation(s)
- Deukwoo Kwon
- Sylvester Comprehensive Cancer Center; University of Miami; Miami FL U.S.A
| | | | - Brian E. Moroz
- Division of Cancer Epidemiology and Genetics; National Cancer Institute, National Institutes of Health; Bethesda MD U.S.A
| | - Steven L. Simon
- Division of Cancer Epidemiology and Genetics; National Cancer Institute, National Institutes of Health; Bethesda MD U.S.A
| |
Collapse
|
4
|
Schlumberger M, Chevillard S, Ory K, Dupuy C, Le Guen B, de Vathaire F. Cancer de la thyroïde après exposition aux rayonnements ionisants. Cancer Radiother 2011; 15:394-9. [DOI: 10.1016/j.canrad.2011.05.002] [Citation(s) in RCA: 9] [Impact Index Per Article: 0.7] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 04/29/2011] [Accepted: 05/04/2011] [Indexed: 10/18/2022]
|
5
|
Wing S, Richardson DB, Hoffmann W. Cancer risks near nuclear facilities: the importance of research design and explicit study hypotheses. ENVIRONMENTAL HEALTH PERSPECTIVES 2011; 119:417-421. [PMID: 21147606 PMCID: PMC3080920 DOI: 10.1289/ehp.1002853] [Citation(s) in RCA: 11] [Impact Index Per Article: 0.8] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Subscribe] [Scholar Register] [Received: 08/11/2010] [Accepted: 12/06/2010] [Indexed: 05/30/2023]
Abstract
BACKGROUND In April 2010, the U.S. Nuclear Regulatory Commission asked the National Academy of Sciences to update a 1990 study of cancer risks near nuclear facilities. Prior research on this topic has suffered from problems in hypothesis formulation and research design. OBJECTIVES We review epidemiologic principles used in studies of generic exposure-response associations and in studies of specific sources of exposure. We then describe logical problems with assumptions, formation of testable hypotheses, and interpretation of evidence in previous research on cancer risks near nuclear facilities. DISCUSSION Advancement of knowledge about cancer risks near nuclear facilities depends on testing specific hypotheses grounded in physical and biological mechanisms of exposure and susceptibility while considering sample size and ability to adequately quantify exposure, ascertain cancer cases, and evaluate plausible confounders. CONCLUSIONS Next steps in advancing knowledge about cancer risks near nuclear facilities require studies of childhood cancer incidence, focus on in utero and early childhood exposures, use of specific geographic information, and consideration of pathways for transport and uptake of radionuclides. Studies of cancer mortality among adults, cancers with long latencies, large geographic zones, and populations that reside at large distances from nuclear facilities are better suited for public relations than for scientific purposes.
Collapse
Affiliation(s)
- Steve Wing
- Department of Epidemiology, University of North Carolina-Chapel Hill, Chapel Hill, North Carolina 27599-7435, USA.
| | | | | |
Collapse
|
6
|
Meliker JR, Goovaerts P, Jacquez GM, Nriagu JO. Incorporating individual-level distributions of exposure error in epidemiologic analyses: an example using arsenic in drinking water and bladder cancer. Ann Epidemiol 2010; 20:750-8. [PMID: 20816314 DOI: 10.1016/j.annepidem.2010.06.012] [Citation(s) in RCA: 7] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 03/01/2010] [Revised: 06/09/2010] [Accepted: 06/20/2010] [Indexed: 10/19/2022]
Abstract
PURPOSE Epidemiologic analyses traditionally rely on point estimates of exposure for assessing risk despite exposure error. We present a strategy that produces a range of risk estimates reflecting distributions of individual-level exposure. METHODS Quantitative estimates of exposure and its associated error are used to create for each individual a normal distribution of exposure estimates which is then sampled using Monte Carlo simulation. After the exposure estimate is sampled, the relationship between exposure and disease is evaluated; this process is repeated 99 times generating a distribution of risk estimates and confidence intervals. This is demonstrated in a bladder cancer case-control study using individual-level distributions of exposure to arsenic in drinking water. RESULTS Sensitivity analyses indicate similar performance for categorical or continuous exposure estimates, and that increases in exposure error translate into a wider range of risk estimates. Bladder cancer analyses yield a wide range of possible risk estimates, allowing quantification of exposure error in the association between arsenic and bladder cancer, typically ignored in conventional analyses. CONCLUSIONS Incorporating distributions of individual-level exposure error results in a more nuanced depiction of epidemiologic findings. This approach can be readily adopted by epidemiologists assuming distributions of individual-level exposure.
Collapse
Affiliation(s)
- Jaymie R Meliker
- Graduate Program in Public Health, Department of Preventive Medicine, Stony Brook University, Stony Brook, NY 11794-8338, USA.
| | | | | | | |
Collapse
|
7
|
Hofer E. How to account for uncertainty due to measurement errors in an uncertainty analysis using Monte Carlo simulation. HEALTH PHYSICS 2008; 95:277-290. [PMID: 18695409 DOI: 10.1097/01.hp.0000314761.98655.dd] [Citation(s) in RCA: 7] [Impact Index Per Article: 0.4] [Reference Citation Analysis] [Abstract] [MESH Headings] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 05/26/2023]
Abstract
Two kinds of error are considered, namely Berkson and classical measurement error. The true values of the measurands will never be known. Possibly true sets of values are generated by the Monte Carlo simulation of the uncertainty analysis. This is straightforward for Berkson errors but requires the modeling of statistical dependence between measured values and errors in the classical case. A method is presented that enables this dependence modeling as part of the uncertainty analysis. Practical examples demonstrate the applicability of the method. Two "quick fixes" are also discussed together with their shortcomings. The uncertainty analysis of the application of a small computer model from the area of dose reconstruction illustrates, by example, the effect both kinds of error can have on model results like individual dose values and mean value and standard deviation of the population dose distribution.
Collapse
|