Sensitivity analysis for causal inference using inverse probability weighting

Division of Biostatistics, Department of Medicine, Indiana University School of Medicine, Indianapolis, IN 46202, USA.
Biometrical Journal (Impact Factor: 0.95). 09/2011; 53(5):822-37. DOI: 10.1002/bimj.201100042
Source: PubMed


Evaluation of impact of potential uncontrolled confounding is an important component for causal inference based on observational studies. In this article, we introduce a general framework of sensitivity analysis that is based on inverse probability weighting. We propose a general methodology that allows both non-parametric and parametric analyses, which are driven by two parameters that govern the magnitude of the variation of the multiplicative errors of the propensity score and their correlations with the potential outcomes. We also introduce a specific parametric model that offers a mechanistic view on how the uncontrolled confounding may bias the inference through these parameters. Our method can be readily applied to both binary and continuous outcomes and depends on the covariates only through the propensity score that can be estimated by any parametric or non-parametric method. We illustrate our method with two medical data sets.

19 Reads
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Understanding the role of dietary quality in the progression of adiposity in populations already affected by overweight or obesity is crucial for the guidance of secondary prevention strategies. To examine the association of diet quality, as reflected by the French Nutrition and Health Programme (Programme National Nutrition Santé, PNNS) - Guideline Score (GS), with 6-year-changes in weight and waist circumference. Subjects were 1029 male and 450 female participants of the SUplémentation en VItamines et Minéraux AntioXydants (SU.VI.MAX) cohort (1994-2002) with anthropometric variables at baseline and follow-up and available data for estimating diet quality at baseline. We employed analysis of variance and covariance models to investigate anthropometric changes (% of the initial value) by categories of the PNNS-GS, which contains both dietary components and a physical activity component, and of a modified score (mPNNS-GS) containing dietary components only. In men, a low (<6 points) PNNS-GS was associated with greater 6-year weight gain (adjusted mean: 3.63% [95% confidence interval: 2.87%; 4.39%]) as compared to a high (≥9 points) PNNS-GS (2.10% [1.39%; 2.81%]); p = 0.01. Results for the mPNNS-GS were very similar. In women, no associations between diet scores and weight change were observed. No significant relation between dietary quality and change in waist circumference was present among either men or women. These results support a beneficial role of high dietary quality - as characterized by good adherence to official French nutritional guidelines - in secondary obesity prevention, among men.
    PLoS ONE 02/2014; 9(2):e87083. DOI:10.1371/journal.pone.0087083 · 3.23 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: We implemented six confounding adjustment methods: (1) covariate-adjusted regression, (2) propensity score (PS) regression, (3) PS stratification, (4) PS matching with two calipers, (5) inverse probability weighting and (6) doubly robust estimation to examine the associations between the body mass index (BMI) z-score at 3 years and two separate dichotomous exposure measures: exclusive breastfeeding v. formula only (n=437) and cesarean section v. vaginal delivery (n=1236). Data were drawn from a prospective pre-birth cohort study, Project Viva. The goal is to demonstrate the necessity and usefulness, and approaches for multiple confounding adjustment methods to analyze observational data. Unadjusted (univariate) and covariate-adjusted linear regression associations of breastfeeding with BMI z-score were -0.33 (95% CI -0.53, -0.13) and -0.24 (-0.46, -0.02), respectively. The other approaches resulted in smaller n (204-276) because of poor overlap of covariates, but CIs were of similar width except for inverse probability weighting (75% wider) and PS matching with a wider caliper (76% wider). Point estimates ranged widely, however, from -0.01 to -0.38. For cesarean section, because of better covariate overlap, the covariate-adjusted regression estimate (0.20) was remarkably robust to all adjustment methods, and the widths of the 95% CIs differed less than in the breastfeeding example. Choice of covariate adjustment method can matter. Lack of overlap in covariate structure between exposed and unexposed participants in observational studies can lead to erroneous covariate-adjusted estimates and confidence intervals. We recommend inspecting covariate overlap and using multiple confounding adjustment methods. Similar results bring reassurance. Contradictory results suggest issues with either the data or the analytic method.
    Journal of Developmental Origins of Health and Disease 08/2014; 5(6):1-13. DOI:10.1017/S2040174414000415 · 0.75 Impact Factor


19 Reads
Available from