Article

Effects of Adjusting for Instrumental Variables on Bias and Precision of Effect Estimates

Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA 02120, USA.
American journal of epidemiology (Impact Factor: 4.98). 12/2011; 174(11):1213-22. DOI: 10.1093/aje/kwr364
Source: PubMed

ABSTRACT Recent theoretical studies have shown that conditioning on an instrumental variable (IV), a variable that is associated with exposure but not associated with outcome except through exposure, can increase both bias and variance of exposure effect estimates. Although these findings have obvious implications in cases of known IVs, their meaning remains unclear in the more common scenario where investigators are uncertain whether a measured covariate meets the criteria for an IV or rather a confounder. The authors present results from two simulation studies designed to provide insight into the problem of conditioning on potential IVs in routine epidemiologic practice. The simulations explored the effects of conditioning on IVs, near-IVs (predictors of exposure that are weakly associated with outcome), and confounders on the bias and variance of a binary exposure effect estimate. The results indicate that effect estimates which are conditional on a perfect IV or near-IV may have larger bias and variance than the unconditional estimate. However, in most scenarios considered, the increases in error due to conditioning were small compared with the total estimation error. In these cases, minimizing unmeasured confounding should be the priority when selecting variables for adjustment, even at the risk of conditioning on IVs.

1 Follower
 · 
127 Views
  • 01/2014; 3:187. DOI:10.12688/f1000research.4801.1
  • [Show abstract] [Hide abstract]
    ABSTRACT: To assess the current practice of propensity score (PS) analysis in the medical literature, particularly the assessment and reporting of balance on confounders.Study Design and SettingA PubMed search identified studies using PS methods from December 2011 through May 2012. For each article included in the review, information was extracted on important aspects of the PS such as the type of PS method used, variable selection for PS model, and assessment of balance.ResultsAmong 296 articles that were included in the review, variable selection for PS model was explicitly reported in 102 studies (34.4%). Covariate balance was checked and reported in 177 studies (59.8%). P-values were the most commonly used statistical tools to report balance (125 of 177, 70.6%). The standardized difference and graphical displays were reported in 45 (25.4%) and 11 (6.2%) articles, respectively. Matching on the PS was the most commonly used approach to control for confounding (68.9%), followed by PS adjustment (20.9%), PS stratification (13.9%), and inverse probability of treatment weighting (IPTW, 7.1%). Balance was more often checked in articles using PS matching and IPTW, 70.6% and 71.4%, respectively.Conclusion The execution and reporting of covariate selection and assessment of balance is far from optimal. Recommendations on reporting of PS analysis are provided to allow better appraisal of the validity of PS-based studies.
    Journal of Clinical Epidemiology 11/2014; 68(2). DOI:10.1016/j.jclinepi.2014.08.011 · 5.48 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Background Lithium has been reported in some, but not all, studies to be associated with reduced risk of suicide death or suicidal behavior. The objective of this nonrandomized cohort study was to examine whether lithium was associated with reduced risk of suicide death in comparison to the commonly-used alternative treatment, valproate.MethodsA propensity score-matched cohort study was conducted of Veterans Health Administration patients (n¿=¿21,194/treatment) initiating lithium or valproate from 1999¿2008.ResultsMatching produced lithium and valproate treatment groups that were highly similar in all 934 propensity score covariates, including indicators of recent suicidal behavior, but recent suicidal ideation was not able to be included. In the few individuals with recently diagnosed suicidal ideation, a significant imbalance existed with suicidal ideation more prevalent at baseline among individuals initiating lithium than valproate (odds ratio (OR) 1.30, 95% CI 1.09, 1.54; p¿=¿0.003). No significant differences in suicide death were observed over 0¿365 days in A) the primary intent-to-treat analysis (lithium/valproate conditional odds ratio (cOR) 1.22, 95% CI 0.82, 1.81; p¿=¿0.32); B) during receipt of initial lithium or valproate treatment (cOR 0.86, 0.46, 1.61; p¿=¿0.63); or C) after such treatment had been discontinued/modified (OR 1.51, 95% CI 0.91, 2.50; p¿=¿0.11). Significantly increased risks of suicide death were observed after the discontinuation/modification of lithium, compared to valproate, treatment over the first 180 days (OR 2.72, 95% CI 1.21, 6.11; p¿=¿0.015).Conclusions In this somewhat distinct sample (a predominantly male Veteran sample with a broad range of psychiatric diagnoses), no significant differences in associations with suicide death were observed between lithium and valproate treatment over 365 days. The only significant difference was observed over 0¿180 days: an increased risk of suicide death, among individuals discontinuing or modifying lithium, compared to valproate, treatment. This difference could reflect risks either related to lithium discontinuation or higher baseline risks among lithium recipients (i.e., confounding) that became more evident when treatment stopped. Our findings therefore support educating patients and providers about possible suicide-related risks of discontinuing lithium even shortly after treatment initiation, and the close monitoring of patients after lithium discontinuation, if feasible. If our findings include residual confounding biasing against lithium, however, as suggested by the differences observed in diagnosed suicidal ideation, then the degree of beneficial reduction in suicide death risk associated with lithium treatment would be underestimated. Further research is urgently needed, given the lack of interventions against suicide and the uncertainties concerning the degree to which lithium may reduce suicide risk during active treatment, increase risk upon discontinuation, or both.
    BMC Psychiatry 12/2014; 14(1):357. DOI:10.1186/s12888-014-0357-x · 2.24 Impact Factor

Full-text (2 Sources)

Download
37 Downloads
Available from
May 28, 2014

Similar Publications