Impact of mis‐specification of the treatment model on estimates from a marginal structural model

Inverse probability of treatment weighted (IPTW) estimation for marginal structural models (MSMs) requires the specification of a nuisance model describing the conditional relationship between treatment allocation and confounders. However, there is still limited information on the best strategy for building these treatment models in practice. We developed a series of simulations to systematically determine the effect of including different types of candidate variables in such models. We explored the performance of IPTW estimators across several scenarios of increasing complexity, including one designed to mimic the complexity typically seen in large pharmacoepidemiologic studies.Our results show that including pure predictors of treatment (i.e. not confounders) in treatment models can lead to estimators that are biased and highly variable, particularly in the context of small samples. The bias and mean-squared error of the MSM-based IPTW estimator increase as the complexity of the problem increases. The performance of the estimator is improved by either increasing the sample size or using only variables related to the outcome to develop the treatment model. Estimates of treatment effect based on the true model for the probability of treatment are asymptotically unbiased.We recommend including only pure risk factors and confounders in the treatment model when developing an IPTW-based MSM.

[1]  G. Schwarz.  Estimating the Dimension of a Model , 1978 .

[2]  Mark J. van der Laan,et al.  A semiparametric model selection criterion with applications to the marginal structural model , 2006, Comput. Stat. Data Anal..

[3]  Samy Suissa,et al.  Warfarin use and the risk of motor vehicle crash in older drivers. , 2006, British journal of clinical pharmacology.

[4]  J. Robins,et al.  Estimating causal effects from epidemiological data , 2006, Journal of Epidemiology and Community Health.

[5]  Patrick Royston,et al.  The design of simulation studies in medical statistics , 2006, Statistics in medicine.

[6]  S. Suissa,et al.  Confounding by indication and channeling over time: the risks of beta 2-agonists. , 1996, American journal of epidemiology.

[7]  J. Robins,et al.  Effect of highly active antiretroviral therapy on time to acquired immunodeficiency syndrome or death using marginal structural models. , 2003, American journal of epidemiology.

[8]  P. Lavori,et al.  Using inverse weighting and predictive inference to estimate the effects of time‐varying treatments on the discrete‐time hazard , 2002, Statistics in medicine.

[9]  Richard F MacLehose,et al.  Improved estimation of controlled direct effects in the presence of unmeasured confounding of intermediate variables , 2005, Statistics in medicine.

[10]  J. Robins,et al.  Marginal Structural Models and Causal Inference in Epidemiology , 2000, Epidemiology.

[11]  J. Robins,et al.  Marginal Structural Models to Estimate the Joint Causal Effect of Nonrandomized Treatments , 2001 .

[12]  J. Robins,et al.  Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. , 2000, Epidemiology.

[13]  J. Neuhaus.  Estimation efficiency with omitted covariates in generalized linear models , 1998 .

[14]  P. Cryer,et al.  Hypoglycemia in diabetes. , 2003, Diabetes care.

[15]  M Alan Brookhart,et al.  Analytic strategies to adjust confounding using exposure propensity scores and disease risk scores: nonsteroidal antiinflammatory drugs and short-term mortality in the elderly. , 2005, American journal of epidemiology.

[16]  S. Suissa,et al.  Confounding by Indication and Channeling over Time: The Risks of β2-Agonists , 1996 .

[17]  Peter C Austin,et al.  A comparison of the ability of different propensity score models to balance measured variables between treated and untreated subjects: a Monte Carlo study , 2007, Statistics in medicine.

[18]  S. Bull,et al.  Confidence intervals for multinomial logistic regression in sparse data , 2007, Statistics in medicine.

[19]  M. Hernán,et al.  Causal knowledge as a prerequisite for confounding evaluation: an application to birth defects epidemiology. , 2002, American journal of epidemiology.

[20]  Erica E M Moodie,et al.  Demystifying Optimal Dynamic Treatment Regimes , 2007, Biometrics.

[21]  J. Robins.  The control of confounding by intermediate variables. , 1989, Statistics in medicine.

[22]  H. Akaike,et al.  Information Theory and an Extension of the Maximum Likelihood Principle , 1973 .

[23]  Joseph Kang,et al.  Demystifying Double Robustness: A Comparison of Alternative Strategies for Estimating a Population Mean from Incomplete Data , 2007, 0804.2958.

[24]  J. Avorn,et al.  Variable selection for propensity score models. , 2006, American journal of epidemiology.

[25]  Paul Zador,et al.  Variable selection and raking in propensity scoring. , 2007, Statistics in medicine.

[26]  Romain Neugebauer,et al.  An application of model-fitting procedures for marginal structural models. , 2005, American journal of epidemiology.

[27]  N. Jewell,et al.  Some surprising results about covariate adjustment in logistic regression models , 1991 .

[28]  S. Weiss,et al.  Antidiabetic therapy and the risk of heart failure in type 2 diabetic patients: an independent effect or confounding by indication , 2005, Pharmacoepidemiology and drug safety.

[29]  J. Robins,et al.  Sensitivity Analyses for Unmeasured Confounding Assuming a Marginal Structural Model for Repeated Measures , 2022 .

[30]  James M Robins,et al.  Marginal structural models for estimating the effect of highly active antiretroviral therapy initiation on CD4 cell count. , 2005, American journal of epidemiology.