Comparing Inverse Probability of Treatment Weighting and Instrumental Variable Methods for the Evaluation of Adenosine Diphosphate Receptor Inhibitors After Percutaneous Coronary Intervention

Federspiel, Jerome J.; Anstrom, Kevin J.; Xian, Ying; McCoy, Lisa A.; Effron, Mark B.; Faries, Douglas E.; Zettler, Marjorie E.; Mauri, Laura; Yeh, Robert W.; Peterson, Eric D.; Wang, Tracy Y

doi:10.1001/jamacardio.2016.1783

Cited by 21 publications

(9 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several other studies have compared statistical techniques and its effect on treatment effect estimates (Tsuchiya et al 2016;Federspiel et al 2016;Stukel et al 2007). They all showed approximately similar effect estimates using all methods and conclude that results have to be interpreted with care.…”

Section: Discussionmentioning

confidence: 99%

Different statistical techniques dealing with confounding in observational research: measuring the effect of breast-conserving therapy and mastectomy on survival

Maaren

Cessie

Strobbe

et al. 2019

J Cancer Res Clin Oncol

View full text Add to dashboard Cite

Purpose Propensity trimming, hierarchical modelling and instrumental variable (IV) analysis are statistical techniques dealing with confounding, cluster-related variation or confounding by severity. This study aimed to explain (dis)advantages of these techniques in estimating the effect of breast-conserving therapy (BCT) and mastectomy on 10-year distant metastasisfree survival (DMFS). Methods All women diagnosed in 2005 with primary T1-2N0-1 breast cancer treated with BCT or mastectomy were selected from the Netherlands Cancer Registry. We used multivariable Cox regression to correct for confounding. Propensity trimming was used to create a more homogeneous population for which the treatment choice was not self-evident. Hospital of surgery was used as hierarchical level to handle hospital-related variation, and as IV to deal with unmeasured confounding. Results Multivariable Cox regression showed higher 10-year DMFS for BCT than mastectomy ]. Propensity trimming on the 10-90th and the 20-80th percentile of the propensity score distribution and hierarchical modelling showed similar HRs. IV analysis showed no significant difference between BCT and mastectomy. Conclusion Unmeasured confounding is very difficult to eliminate in observational research. We cannot conclude that BCT or mastectomy has a causal relationship with 10-year DMFS. It is crucial to critically evaluate all model's assumptions, and to be careful in drawing firm conclusions.

show abstract

Section: Discussionmentioning

confidence: 99%

Different statistical techniques dealing with confounding in observational research: measuring the effect of breast-conserving therapy and mastectomy on survival

Maaren

Cessie

Strobbe

et al. 2019

J Cancer Res Clin Oncol

View full text Add to dashboard Cite

show abstract

“…We are now introducing the list of analytical methods through the stepwise selection strategy: Pseudo treatment : If multiple (≥2) control groups that are systematically different exist, and they are not expected to differ in influencing the outcome, then arbitrarily treat one as a “pseudo treatment” and others as the “clean controls.” A significant difference in the outcome comparison between “pseudo treatment” and “clean controls” with proper adjustment would increase concerns regarding the presence of unmeasured confounding ; Negative control : Pre‐specify outcomes that are highly unlikely to be causally related to the treatment (“negative control outcome”) or interventions that are known to have minimum impact on the original outcome of interest (“negative control exposure”). A significant association between treatment and negative control outcomes or between negative control exposures and original outcomes while adjusting for measured confounders would suggest the existence of unmeasured confounders; Manski's partial identification : Under Rubin's “potential outcome” framework, the average treatment effect (ATE) can be described as:

italicATE = normalΕ ([], Y_{i} ((), 1)) - normalΕ ([], Y_{i} ((), 0))

where 1 = treatment, 0 = control, Y i ( ) represents the outcome of ith subject given interventions, and Ε stands for the expectation. Without assuming “no unmeasured confounding,” the above formula can be split as:

italicATE = normalΕ ([]|, Y_{i} ((), 1), W_{i} = 1) \times P ((), W_{i} = 1) + normalΕ ([]|, Y_{i} ((), 1), W_{i} = 0) \times P ((), W_{i} = 0) - normalΕ ([]|, Y_{i} ((), 0), W_{i} = 0) \times P ((), W_{i} = 0) - normalΕ ([]|, Y_{i} ((), 0), W_{i} = 1) \times P ((), W_{i} = 1)

where W i is the exposure.…”

Section: The Toolbox: Search Strategy and A Summary Of Analytical Metmentioning

confidence: 99%

“…A significant difference in the outcome comparison between "pseudo treatment" and "clean controls" with proper adjustment would increase concerns regarding the presence of unmeasured confounding 21,22; • Negative control: Pre-specify outcomes that are highly unlikely to be causally related to the treatment ("negative control outcome") or interventions that are known to have minimum impact on the original outcome of interest ("negative control exposure"). A significant association between treatment and negative control outcomes or between negative control exposures and original outcomes while adjusting for measured confounders would suggest the existence of unmeasured confounders; [23][24][25][26] • Manski's partial identification: Under Rubin's "potential outcome" framework, the average treatment effect (ATE) can be described as:…”

Section: Key Pointsmentioning

confidence: 99%

Addressing unmeasured confounding in comparative observational research

Zhang

Faries

et al. 2018

Pharmacoepidemiology and Drug

Self Cite

View full text Add to dashboard Cite

show abstract

“…Careful (and critical) selection of analytical techniques should also inform studies. 16 Appropriate measures to curate the quality of the data will be necessary: quality in, quality out. It is important as well for researchers, clinicians, and health service executives to carefully define the questions whose answers may best serve their individual healthcare contexts and to assess whether or not they have appropriate data to ask these questions.…”

mentioning

confidence: 99%