Instrumental Variables: Application and Limitations

To correct for confounding, the method of instrumental variables (IV) has been proposed. Its use in medical literature is still rather limited because of unfamiliarity or inapplicability. By introducing the method in a nontechnical way, we show that IV in a linear model is quite easy to understand and easy to apply once an appropriate instrumental variable has been identified. We also point out some limitations of the IV estimator when the instrumental variable is only weakly correlated with the exposure. The IV estimator will be imprecise (large standard error), biased when sample size is small, and biased in large samples when one of the assumptions is only slightly violated. For these reasons, it is advised to use an IV that is strongly correlated with exposure. However, we further show that under the assumptions required for the validity of the method, this correlation between IV and exposure is limited. Its maximum is low when confounding is strong, such as in case of confounding by indication. Finally, we show that in a study in which strong confounding is to be expected and an IV has been used that is moderately or strongly related to exposure, it is likely that the assumptions of IV are violated, resulting in a biased effect estimate. We conclude that instrumental variables can be useful in case of moderate confounding but are less useful when strong confounding exists, because strong instruments cannot be found and assumptions will be easily violated.

[1]  J. Concato,et al.  Randomized, controlled trials, observational studies, and the hierarchy of research designs. , 2000, The New England journal of medicine.

[2]  T. Gyorkos,et al.  Does aggressive care following acute myocardial infarction reduce mortality? Analysis with instrumental variables to compare effectiveness in Canadian and United States patient populations. , 2003, Health services research.

[3]  Richard Startz,et al.  Some Further Results on the Exact Small Sample Properties of the Instrumental Variable Estimator , 1988 .

[4]  S. Ebrahim,et al.  Mendelian randomization: prospects, potentials, and limitations. , 2004, International journal of epidemiology.

[5]  R. F.,et al.  Mathematical Statistics , 1944, Nature.

[6]  J. Paul Leigh,et al.  Instrumental variables technique: cigarette price provided better estimate of effects of smoking on SF-12. , 2004, Journal of clinical epidemiology.

[7]  Yi-Ting Hwang,et al.  An exploratory instrumental variable analysis of the outcomes of localized breast cancer treatments in a medicare population. , 2003, Health economics.

[8]  A. McMahon,et al.  Approaches to combat with confounding by indication in observational studies of intended drug effects , 2003, Pharmacoepidemiology and drug safety.

[9]  David V Conti,et al.  Commentary: the concept of 'Mendelian Randomization'. , 2004, International journal of epidemiology.

[10]  D. Rubin,et al.  The central role of the propensity score in observational studies for causal effects , 1983 .

[11]  Takeshi Amemiya,et al.  The nonlinear two-stage least-squares estimator , 1974 .

[12]  M. M. Langenhuysen,et al.  VIRUS INFECTION OF THE URINARY TRACT , 1974 .

[13]  J. Angrist,et al.  Split Sample Instrumental Variables , 1995 .

[14]  Joshua D. Angrist,et al.  Identification of Causal Effects Using Instrumental Variables , 1993 .

[15]  David R. Cox.  The analysis of binary data , 1970 .

[16]  D A Savitz,et al.  Econometric approaches to epidemiologic data: relating endogeneity and unobserved heterogeneity to confounding. , 1997, Annals of epidemiology.

[17]  N. Wermuth,et al.  A Comment on the Coefficient of Determination for Binary Responses , 1992 .

[18]  M. McIntosh.  Instrumental variables when evaluating screening trials: estimating the benefit of detecting cancer by screening. , 1999, Statistics in medicine.

[19]  Joshua D. Angrist,et al.  Split-Sample Instrumental Variables Estimates of the Return to Schooling , 1995 .

[20]  Sean D Sullivan,et al.  Methods to assess intended effects of drug treatment in observational studies are reviewed. , 2004, Journal of clinical epidemiology.

[21]  Keith R Abrams,et al.  An integrated approach to the meta-analysis of genetic association studies using Mendelian randomization. , 2004, American journal of epidemiology.

[22]  M. Weinstein,et al.  Effectiveness of chemotherapy for advanced lung cancer in the elderly: instrumental variable and propensity analysis. , 2001, Journal of clinical oncology : official journal of the American Society of Clinical Oncology.

[23]  M. B. Katan,et al.  Apolipoprotein E isoforms, serum cholesterol, and cancer , 2004 .

[24]  Takamitsu Sawa,et al.  The Exact Sampling Distribution of Ordinary Least Squares and Two-Stage Least Squares Estimators , 1969 .

[25]  T. Permutt,et al.  Simultaneous-equation estimation in a clinical trial of the effect of smoking on birth weight. , 1989, Biometrics.

[26]  David A. Jaeger,et al.  Problems with Instrumental Variables Estimation when the Correlation between the Instruments and the Endogenous Explanatory Variable is Weak , 1995 .

[27]  J. Angrist,et al.  Two-Stage Least Squares Estimation of Average Causal Effects in Models with Variable Treatment Intensity , 1995 .

[28]  John M Brooks,et al.  Was breast conserving surgery underutilized for early stage breast cancer? Instrumental variables evidence for stage II patients from Iowa. , 2003, Health services research.

[29]  B J McNeil,et al.  Does more intensive treatment of acute myocardial infarction in the elderly reduce mortality? Analysis using instrumental variables. , 1994, JAMA.

[30]  J. Stock,et al.  Instrumental Variables Regression with Weak Instruments , 1994 .

[31]  Roger J. Bowden,et al.  A Comparative Study of Instrumental Variables Estimators for Nonlinear Simultaneous Models , 1981 .

[32]  William E. Griffiths,et al.  Principles of Econometrics , 2008 .