The Impact of Prior Distributions for Uncontrolled Confounding and Response Bias

This article examines the potential for misleading inferences from conventional analyses and sensitivity analyses of observational data, and describes some proposed solutions based on specifying prior distributions for uncontrolled sources of bias. The issues are illustrated in a sensitivity analysis of confounding in a study of residential wire code and childhood leukemia and in a pooled analysis of 12 studies of magnetic-field measurements and childhood leukemia. Both analyses have been interpreted as evidence in favor of a causal effect of magnetic fields on leukemia risk. This interpretation is contrasted with results from analyses based on prior distributions for the unidentified bias parameters used in the original sensitivity-analysis model. These analyses indicate that accounting for uncontrolled confounding and response bias under a reasonable prior can substantially alter inferences about the existence of a magnetic-field effect. More generally, analyses with informative priors for unidentified bias parameters can help avoid misinterpretation of conventional results and ordinary sensitivity analyses.

[1]  J Schüz,et al.  Combined risk estimates for two German population-based case-control studies on residential magnetic fields and childhood acute leukemia. , 1998, Epidemiology.

[2]  Robert Kavet,et al.  Contact voltage measured in residences: Implications to the association between magnetic fields and childhood leukemia , 2002, Bioelectromagnetics.

[3]  George E. P. Box,et al.  Sampling and Bayes' inference in scientific modelling and robustness , 1980 .

[4]  Willis L. Owen,et al.  Modelling Frequency and Count Data , 1996 .

[5]  Carl V Phillips,et al.  Quantifying and Reporting Uncertainty from Systematic Errors , 2003, Epidemiology.

[6]  S. Greenland.  When Should Epidemiologic Regressions Use Random Coefficients? , 2000, Biometrics.

[7]  T Tynes,et al.  Electromagnetic fields and cancer in children residing near Norwegian high-voltage power lines. , 1997, American journal of epidemiology.

[8]  J. Robins,et al.  Sensitivity Analysis for Selection bias and unmeasured Confounding in missing Data and Causal inference models , 2000 .

[9]  B Langholz,et al.  Factors that explain the power line configuration wiring code-childhood leukemia association: what would they look like? , 2001, Bioelectromagnetics.

[10]  J. Olsen,et al.  Residence near high voltage facilities and risk of cancer in children. , 1993, BMJ.

[11]  S Greenland,et al.  The interpretation of multiplicative-model parameters as standardized parameters. , 1994, Statistics in medicine.

[12]  J J Spinelli,et al.  Power-frequency electric and magnetic fields and risk of childhood leukemia in Canada. , 1999, American journal of epidemiology.

[13]  J. Pearl,et al.  Confounding and Collapsibility in Causal Inference , 1999 .

[14]  Edward E. Leamer,et al.  False Models and Post-Data Model Construction , 1974 .

[15]  T. Lash,et al.  A Sensitivity Analysis to Separate Bias Due to Confounding from Bias Due to Predicting Misclassification by a Variable That Does Both , 2000, Epidemiology.

[16]  L Tomenius,et al.  50-Hz electromagnetic environment and the incidence of childhood tumors in Stockholm County. , 1986, Bioelectromagnetics.

[17]  Pekka J Jarvinen,et al.  Public health implications , 2006 .

[18]  R. Christensen,et al.  A New Perspective on Priors for Generalized Linear Models , 1996 .

[19]  H. Jeffreys.  A Treatise on Probability , 1922, Nature.

[20]  S Greenland,et al.  Sensitivity Analysis, Monte Carlo Risk Analysis, and Bayesian Uncertainty Assessment , 2001, Risk analysis : an official publication of the Society for Risk Analysis.

[21]  H. D. De Kanter.  [The philosophy of statistics]. , 1972, Ginecología y Obstetricia de México.

[22]  Takashi Yanagawa,et al.  Case-control studies: Assessing the effect of a confounding factor , 1984 .

[23]  John D. Dockerty,et al.  Electromagnetic Field Exposures and Childhood Cancers in New Zealand , 1998, Cancer Causes & Control.

[24]  Sander Greenland,et al.  How a Court Accepted a Possible Explanation: A Comment on Gastwirth, Krieger, and Rosenbaum , 1997 .

[25]  N. Breslow,et al.  Statistical methods in cancer research: volume 1- The analysis of case-control studies , 1980 .

[26]  S Greenland,et al.  Putting Background Information About Relative Risks into Conjugate Prior Distributions , 2001, Biometrics.

[27]  W. Flanders,et al.  Indirect Assessment of Confounding: Graphic Description and Limits on Effect of Adjusting for Covariates , 1990, Epidemiology.

[28]  Sander Greenland,et al.  Generalized Conjugate Priors for Bayesian Analysis of Risk and Survival Regressions , 2003, Biometrics.

[29]  D C Thomas,et al.  Exposure to residential electric and magnetic fields and risk of childhood leukemia. , 1991, American journal of epidemiology.

[30]  R W Coghill,et al.  Extra low frequency electric and magnetic fields in the bedplace of children diagnosed with leukaemia: a case‐control study , 1996, European journal of cancer prevention : the official journal of the European Cancer Prevention Organisation.

[31]  E. C. Hammond,et al.  Smoking and lung cancer: recent evidence and a discussion of some questions. , 1959, Journal of the National Cancer Institute.

[32]  K. Jaffa,et al.  Pooled analysis of magnetic fields, wire codes, and childhood leukemia. , 2001, Epidemiology.

[33]  W T Kaune,et al.  Residential exposure to magnetic fields and acute lymphoblastic leukemia in children. , 1997, The New England journal of medicine.

[34]  H. Wachtel,et al.  Case-control study of childhood cancer and exposure to 60-Hz magnetic fields. , 1988, American journal of epidemiology.

[35]  P Gustafson,et al.  Case–Control Analysis with Partial Knowledge of Exposure Misclassification Probabilities , 2001, Biometrics.

[36]  N. E. Breslow.  Statistical Methods in Cancer Research , 1986 .

[37]  W T Kaune,et al.  Do confounding or selection factors of residential wiring codes and magnetic fields distort findings of electromagnetic fields studies? , 2000, Epidemiology.

[38]  Edmund A. C. Crouch,et al.  Health risk assessments prepared per the risk assessment reforms under consideration in the U.S. congress , 1997 .

[39]  N. Breslow,et al.  The analysis of case-control studies , 1980 .

[40]  Timothy L Lash,et al.  Semi-Automated Sensitivity Analysis to Assess Systematic Errors in Observational Data , 2003, Epidemiology.

[41]  A. Whittemore.  Collapsibility of Multidimensional Contingency Tables , 1978 .

[42]  B. deFinetti,et al.  Theory of Probability , 1981 .

[43]  S Greenland,et al.  Basic methods for sensitivity analysis of biases. , 1996, International journal of epidemiology.

[44]  M Feychting,et al.  Magnetic fields and cancer in children residing near Swedish high-voltage power lines. , 1993, American journal of epidemiology.