Evaluating hybrid controls methodology in early-phase oncology trials: A simulation study based on the MORPHEUS-UC trial.

Phase Ib/II oncology trials, despite their small sample sizes, aim to provide information for optimal internal company decision-making concerning novel drug development. Hybrid controls (a combination of the current control arm and controls from one or more sources of historical trial data [HTD]) can be used to increase statistical precision. Here we assess combining two sources of Roche HTD to construct a hybrid control in targeted therapy for decision-making via an extensive simulation study. Our simulations are based on the real data of one of the experimental arms and the control arm of the MORPHEUS-UC Phase Ib/II study and two Roche HTD for atezolizumab monotherapy. We consider potential complications such as model misspecification, unmeasured confounding, different sample sizes of current treatment groups, and heterogeneity among the three trials. We evaluate two frequentist methods (with both Cox and Weibull accelerated failure time [AFT] models) and three different commensurate priors in Bayesian dynamic borrowing (with a Weibull AFT model), and modifications within each of those, when estimating the effect of treatment on survival outcomes and measures of effect such as marginal hazard ratios. We assess the performance of these methods in different settings and the potential of generalizations to supplement decisions in early-phase oncology trials. The results show that the proposed joint frequentist methods and noninformative priors within Bayesian dynamic borrowing with no adjustment on covariates are preferred, especially when treatment effects across the three trials are heterogeneous. For generalization of hybrid control methods in such settings, we recommend more simulation studies.

Precision medicine analysis of heterogeneity in individual-level treatment response to amyloid beta removal in early Alzheimer's disease.

INTRODUCTION: Alzheimer's disease (AD) is a neurological disorder with variability in pathology and clinical progression. AD patients may differ in individual-level benefit from amyloid beta removal therapy. METHODS: Random forest models were applied to the EMERGE trial to create an individual-level treatment response (ITR) score which represents individual-level benefit of high-dose aducanumab relative to the placebo. This ITR score was used to test the existence of heterogeneity in treatment effect (HTE). RESULTS: We found statistical evidence of HTE in the Clinical Dementia Rating-Sum of Boxes (CDR-SB;P =  0.034). The observed CDR-SB benefit was 0.79 points greater in the group with the top 25% of ITR score compared to the remaining 75% (P = 0.020). Of note, the highest treatment responders had lower hippocampal volume, higher plasma phosphorylated tau 181 and a shorter duration of clinical AD at baseline. DISCUSSION: This ITR analysis provides a proof of concept for precision medicine in future AD research and drug development. HIGHLIGHTS: Emerging trials have shown a population-level benefit from amyloid beta (Aß) removal in slowing cognitive decline in early Alzheimer's disease (AD). This work demonstrates significant heterogeneity of individual-level treatment effect of aducanumab in early AD. The greatest clinical responders to Aß removal therapy have a pattern of more severe neurodegenerative process.

Psychometric properties of the Alzheimer's Disease Cooperative Study - Activities of Daily Living for Mild Cognitive Impairment (ADCS-MCI-ADL) scale: a post hoc analysis of the ADCS ADC-008 trial.

BACKGROUND: The Alzheimer's Disease Cooperative Study - Activities of Daily Living Scale for use in Mild Cognitive Impairment (MCI), the ADCS-ADL-MCI, is an evaluation scale with information provided by an informant/caregiver to describe the functional impairment of patients with MCI. As the ADCS-ADL-MCI has yet to undergo a full psychometric evaluation, this study aimed to evaluate the measurement properties of the ADCS-ADL-MCI in subjects with amnestic MCI. METHODS: Measurement properties, including item-level analysis, internal consistency reliability, test-retest reliability, construct validity (convergent/discriminant, known-groups validity), and responsiveness were evaluated using data from the ADCS ADC-008 trial, a 36-month, multicenter, placebo-controlled study in 769 subjects with amnestic MCI (defined by clinical criteria and a global clinical dementia rating, CDR, score of 0.5). Due to most subjects' mild condition at baseline and resulting low variance in scores, psychometric properties were assessed using both baseline and 36-month data. RESULTS: Ceiling effects were not apparent at the total score level, with 3% of the cohort reaching the maximum score of 53, despite most subjects having a relatively high score at baseline (mean score = 46.0 [standard deviation = 4.8]). Item-total correlations were overall weak at baseline, most likely due to low variability in responses; however, at month 36, good item homogeneity was found. Cronbach's alpha values ranged from acceptable (0.64 at baseline) to good (0.87 at month 36), indicating overall very good internal consistency reliability. Further, moderate to good test-retest reliability was found (intraclass correlation coefficients ranging from 0.62-0.73). The analyses also largely supported convergent/discriminant validity, especially at month 36. Finally, the ADCS-ADL-MCI discriminated well between groups showing good known-groups validity, and was responsive in patients who indicated a longitudinal change in other instruments. CONCLUSIONS: This study provides a comprehensive psychometric evaluation of the ADCS-ADL-MCI. Findings suggest that the ADCS-ADL-MCI is a reliable, valid and responsive measure capable of capturing functional abilities in patients with amnestic MCI. TRIAL REGISTRATION: ClinicalTrials.gov Identifier: NCT00000173.

Effect of reduction in brain amyloid levels on change in cognitive and functional decline in randomized clinical trials: An instrumental variable meta-analysis.

INTRODUCTION: Whether the reduction in brain amyloid beta (Aß) plaque alone may substantially slow cognitive and functional decline in patients with dementia or mild cognitive impairment due to Alzheimer's disease (AD) remains debated. METHODS: An instrumental variable meta-analysis was performed to infer the effect of change in positron emission tomography (PET)-measured Aß standardized uptake value ratio (SUVR) on cognitive and functional decline. RESULTS: Pooling data from 16 randomized trials demonstrates that each 0.1-unit decrease in PET Aß SUVR is associated with a reduction (95% confidence interval) by 0.09 (0.034-0.15), 0.33 (0.12-0.55), and 0.13 (0.017-0.24) point in the average change of the Clinical Dementia Rating-Sum of Boxes, the Alzheimer's Disease Assessment Scale-Cognitive Subscale, and the Mini-Mental State Examination, respectively. DISCUSSION: This meta-analysis provides statistically significant evidence of a likely causal relationship between a reduction in Aß plaque and a reduction in cognitive and functional decline in patients with AD. HIGHLIGHTS: A widely cited meta-analysis article concluded amyloid beta reduction does not substantially improve cognition. We identified data inconsistencies in the initial publication and found new trial data. We repeated the meta-analysis after correcting data inconsistencies and adding new trial data. Updated results suggested statistically significant clinical benefit of amyloid beta reduction. Amyloid beta is a viable biological target for the treatment and prevention of AD.

"Translating" All-Cause Mortality Rate Ratios or Hazard Ratios to Age-, Longevity-, and Probability-Based Measures.

Epidemiologists commonly use an adjusted hazard ratio or incidence density ratio, or a standardized mortality ratio, to measure a difference in all-cause mortality rates. They seldom translate it into an age-, time-, or probability-based measure that would be easier to communicate and to relate to. Several articles have shown how to translate from a standardized mortality ratio or hazard ratio to a longevity difference, a difference in actuarial ages, or a probability of being outlived. In this paper, we describe the settings where these translations are and are not appropriate and provide some of the heuristics behind the formulae. The tools that yield differences in "effective age" and in longevity are applicable when both 1) the mortality rate ratio (hazard ratio) is constant over age and 2) the rates themselves are log-linear in age. The "probability/odds of being outlived" metric is applicable whenever the first condition holds, and thus it provides no direct information on the magnitude of the effective age/longevity difference.

Spline-based accelerated failure time model.

The accelerated failure time (AFT) model has been suggested as an alternative to the Cox proportional hazards model. However, a parametric AFT model requires the specification of an appropriate distribution for the event time, which is often difficult to identify in real-life studies and may limit applications. A semiparametric AFT model was developed by Komárek et al based on smoothed error distribution that does not require such specification. In this article, we develop a spline-based AFT model that also does not require specification of the parametric family of event time distribution. The baseline hazard function is modeled by regression B-splines, allowing for the estimation of a variety of smooth and flexible shapes. In comprehensive simulations, we validate the performance of our approach and compare with the results from parametric AFT models and the approach of Komárek. Both the proposed spline-based AFT model and the approach of Komárek provided unbiased estimates of covariate effects and survival curves for a variety of scenarios in which the event time followed different distributions, including both simple and complex cases. Spline-based estimates of the baseline hazard showed also a satisfactory numerical stability. As expected, the baseline hazard and survival probabilities estimated by the misspecified parametric AFT models deviated from the truth. We illustrated the application of the proposed model in a study of colon cancer.

Effectiveness and safety among direct oral anticoagulants in nonvalvular atrial fibrillation: A multi-database cohort study with meta-analysis.

AIMS: There are conflicting signals in the literature about comparative safety and effectiveness of direct oral anticoagulants (DOACs) for nonvalvular atrial fibrillation (NVAF). METHODS: We conducted multicentre matched cohort studies with secondary meta-analysis to assess safety and effectiveness of dabigatran, rivaroxaban and apixaban across 9 administrative healthcare databases. We included adults with NVAF initiating anticoagulation therapy (dabigatran, rivaroxaban or apixaban), and constructed 3 cohorts to compare DOACs pairwise. The primary outcome was pooled hazard ratio (pHR) of ischaemic stroke or systemic thromboembolism. Secondary outcomes included pHR of major bleeding, and a composite of stroke, major bleeding, or all-cause mortality. We used proportional hazard Cox regressions models, and pooled estimates were obtained with random effect meta-analyses. RESULTS: The cohorts included 73 414 new users of dabigatran, 92 881 of rivaroxaban, and 61 284 of apixaban. After matching, the pHRs (95% confidence intervals) comparing rivaroxaban initiation to dabigatran were: 1.11 (0.93, 1.32) for ischaemic stroke or systemic thromboembolism, 1.26 (1.09, 1.46) for major bleeding, and 1.17 (1.05, 1.30) for the composite endpoint. For apixaban vs dabigatran, they were: 0.91 (0.74, 1.12) for ischaemic stroke or systemic thromboembolism, 0.89 (0.75, 1.05) for major bleeding, and 0.94 (0.78 to 1.14) for the composite endpoint. For apixaban vs rivaroxaban, they were: 0.85 (0.74, 0.99) for ischaemic stroke or systemic thromboembolism, 0.61 (0.53, 0.70) for major bleeding, and 0.82 (0.76, 0.88) for the composite endpoint. CONCLUSION: We found that apixaban use is associated with lower risks of stroke and bleeding compared with rivaroxaban, and similar risks compared with dabigatran.

Can We Train Machine Learning Methods to Outperform the High-dimensional Propensity Score Algorithm?

The use of retrospective health care claims datasets is frequently criticized for the lack of complete information on potential confounders. Utilizing patient's health status-related information from claims datasets as surrogates or proxies for mismeasured and unobserved confounders, the high-dimensional propensity score algorithm enables us to reduce bias. Using a previously published cohort study of postmyocardial infarction statin use (1998-2012), we compare the performance of the algorithm with a number of popular machine learning approaches for confounder selection in high-dimensional covariate spaces: random forest, least absolute shrinkage and selection operator, and elastic net. Our results suggest that, when the data analysis is done with epidemiologic principles in mind, machine learning methods perform as well as the high-dimensional propensity score algorithm. Using a plasmode framework that mimicked the empirical data, we also showed that a hybrid of machine learning and high-dimensional propensity score algorithms generally perform slightly better than both in terms of mean squared error, when a bias-based analysis is used.

Targeted Maximum Likelihood Estimation for Pharmacoepidemiologic Research.

BACKGROUND: Targeted maximum likelihood estimation has been proposed for estimating marginal causal effects, and is robust to misspecification of either the treatment or outcome model. However, due perhaps to its novelty, targeted maximum likelihood estimation has not been widely used in pharmacoepidemiology. The objective of this study was to demonstrate targeted maximum likelihood estimation in a pharmacoepidemiological study with a high-dimensional covariate space, to incorporate the use of high-dimensional propensity scores into this method, and to compare the results to those of inverse probability weighting. METHODS: We implemented the targeted maximum likelihood estimation procedure in a single-point exposure study of the use of statins and the 1-year risk of all-cause mortality postmyocardial infarction using data from the UK Clinical Practice Research Datalink. A range of known potential confounders were considered, and empirical covariates were selected using the high-dimensional propensity scores algorithm. We estimated odds ratios using targeted maximum likelihood estimation and inverse probability weighting with a variety of covariate selection strategies. RESULTS: Through a real example, we demonstrated the double robustness of targeted maximum likelihood estimation. We showed that results with this method and inverse probability weighting differed when a large number of covariates were included in the treatment model. CONCLUSIONS: Targeted maximum likelihood can be used in high-dimensional covariate settings. In high-dimensional covariate settings, differences in results between targeted maximum likelihood and inverse probability weighted estimation are likely due to sensitivity to (near) positivity violations. Further investigations are needed to gain better understanding of the advantages and limitations of this method in pharmacoepidemiological studies.

On the role of marginal confounder prevalence - implications for the high-dimensional propensity score algorithm.

PURPOSE: The high-dimensional propensity score algorithm attempts to improve control of confounding in typical treatment effect studies in pharmacoepidemiology and is increasingly being used for the analysis of large administrative databases. Within this multi-step variable selection algorithm, the marginal prevalence of non-zero covariate values is considered to be an indicator for a count variable's potential confounding impact. We investigate the role of the marginal prevalence of confounder variables on potentially caused bias magnitudes when estimating risk ratios in point exposure studies with binary outcomes. METHODS: We apply the law of total probability in conjunction with an established bias formula to derive and illustrate relative bias boundaries with respect to marginal confounder prevalence. RESULTS: We show that maximum possible bias magnitudes can occur at any marginal prevalence level of a binary confounder variable. In particular, we demonstrate that, in case of rare or very common exposures, low and high prevalent confounder variables can still have large confounding impact on estimated risk ratios. CONCLUSIONS: Covariate pre-selection by prevalence may lead to sub-optimal confounder sampling within the high-dimensional propensity score algorithm. While we believe that the high-dimensional propensity score has important benefits in large-scale pharmacoepidemiologic studies, we recommend omitting the prevalence-based empirical identification of candidate covariates.

Comparison of time to clinically meaningful improvement in quality of life in neurological disorders in patients treated with natalizumab versus ocrelizumab.

Aim: To assess time to improvement in Quality of Life in Neurological Disorders (Neuro-QoL) domains for patients treated with natalizumab versus ocrelizumab. Methods: Patients enrolled in the MS PATHS network who initiated treatment with either natalizumab or ocrelizumab rated the Neuro-QoL domains of physical function, symptoms, emotional health, cognitive function and social ability. Results: Time to clinically meaningful improvement was significantly shorter with natalizumab versus ocrelizumab for cognitive function (event time ratio [95% CI]: 0.37 [0.24-0.57]; p < 0.001), sleep disturbance (0.45 [0.28-0.72]; p = 0.001), social role participation (0.37 [0.21-0.66]; p = 0.001) and social role satisfaction (0.5 [0.31-0.8]; p = 0.004). Conclusion: Natalizumab had shorter time to clinically meaningful improvement in cognitive, sleep, and social role Neuro-QoL domains versus ocrelizumab.

Knowledge of treatment-related benefits associated with medication choices, including improvement of quality of life (QoL), are strong influential factors for patients to start and continue their therapies. Little is known about patient-reported time to onset of functional improvement upon the initiation of medications for multiple sclerosis (MS). The Multiple Sclerosis Partners Advancing Technology and Health Solutions (MS PATHS) network, a repository of collaborative international data on routine MS management, includes patient-reported information on the health-related QoL using the Quality of Life in Neurological Disorders (Neuro-QoL) measure. This study included data from 883 eligible patients enrolled in MS PATHS, with the aim of assessing and comparing the time to improvement in physical, mental and social health for patients treated with natalizumab versus ocrelizumab using Neuro-QoL. Natalizumab and ocrelizumab are both high-efficacy treatment options for relapsing forms of MS. The results demonstrated that, compared with ocrelizumab, natalizumab treatment led to faster effect on mental and social health, as well as quicker improvements in physical functioning in the arms and hands. Overall, it took shorter time for natalizumab-treated patients to achieve better QoL compared with ocrelizumab. These findings highlight the importance of QoL in disease management and provide a patient perspective for healthcare providers when making decisions about high-efficacy treatments for their patients with MS.

Extracellular vesicles derived from Akkermansia muciniphila promote placentation and mitigate preeclampsia in a mouse model.

Preeclampsia (PE) is a multisystem disorder with high maternal morbidity and mortality rates. Currently, no practical therapeutic approach is available to prevent PE progression, except for early delivery. Gut dysbiosis is associated with PE development. Previous data showed that the abundance of Akkermansia muciniphila (Am) was lower in patients with PE than in normotensive pregnant women. Here, in this study, decreased abundance of Am was observed in a PE mouse model. Also, we found that administration with Am could significantly attenuate systolic blood pressure, promote foetal growth and improve the placental pathology in mice with PE. Moreover, Am-derived extracellular vesicles (AmEVs) were transferred from the gastrointestinal (GI) tract to the placenta and mitigated pre-eclamptic symptoms in PE mice. These beneficial effects of AmEVs were mediated by enhanced trophoblast invasion of the spiral artery (SpA) and SpA remodelling through activation of the epidermal growth factor receptor (EGFR)-phosphatidylinositol-3-kinase (PI3K)-protein kinase B (AKT) signalling pathway. Collectively, our findings revealed the potential benefit of using AmEVs for PE treatment and highlighted important host-microbiota interactions.

Knockdown of ENTPD5 inhibits tumor metastasis and growth via regulating the GRP78/p-eIF-2α/CHOP pathway in serous ovarian cancer.

BACKGROUND: Dysregulation of Ectonucleoside Triphospahate Diphosphohydrolase 5 (ENTPD5) in tumors might be associated with tumor progression, while the role of ENTPD5 in the growth and metastasis of serous ovarian cancer (SOC) is still unclear. METHODS: ENTPD5 expression patterns in ovarian cancer tissues were analyzed by qRT-PCR and immunohistochemistry assay (IHC). Two SOC cell lines, SKOV3 and OVCAR8, were stably transfected with lentivirus to build knockdown and overexpression cell lines. Clone formation assay, collagen gel droplet culture technology, wound healing assay and flow cytometry were used to assess the migration and growth traits of SOC cells. Expression levels of ENTPD5, glucose regulated protein 78 (GRP78), eukaryotic translation initiation factor 2 alpha (eIF-2α), phosphorylated -eIF-2α and, C/EBP homologous protein (CHOP) in SOC cells were detected by Western blot. RESULTS: Compared to fallopian tube tissues, the expression of ENTPD5 was significantly higher in tumor tissues obtained from SOC patients, and positively correlated with clinical stage and metastasis. ENTPD5 knockdown robustly inhibited cell proliferation, migration, whereas ENTPD5 overexpression elicited the opposite effect on SOC cells. ENTPD5 knockdown arrested cell cycle in G0/G1 phase and increased apoptosis. Importantly, ENTPD5 knockdown was associated with significantly decreased protein levels for GRP78, CHOP, and p-eIF-2α, suggesting possible involvement of ENTPD5 in endoplasmic reticulum stress (ERS). CONCLUSIONS: Our study demonstrates that ENTPD5 knockdown inhibited SOC cell proliferation, migration and restrained the activation of the GRP78/p-eIF-2α/CHOP pathway, which provides a potentially effective therapeutic target for the treatment of SOC.

Flexible extension of the accelerated failure time model to account for nonlinear and time-dependent effects of covariates on the hazard.

The accelerated failure time model is an alternative to the Cox proportional hazards model in survival analysis. However, conclusions regarding the associations of prognostic factors with event times are valid only if the underlying modeling assumptions are met. In contrast to several flexible methods for relaxing the proportional hazards and linearity assumptions in the Cox model, formal investigation of the constant-over-time time ratio and linearity assumptions in the accelerated failure time model has been limited. Yet, in practice, prognostic factors may have time-dependent and/or nonlinear effects. Furthermore, parametric accelerated failure time models require correct specification of the baseline hazard function, which is treated as a nuisance parameter in the Cox proportional hazards model, and is rarely known in practice. To address these challenges, we propose a flexible extension of the accelerated failure time model where unpenalized regression B-splines are used to model (i) the baseline hazard function of arbitrary shape, (ii) the time-dependent covariate effects on the hazard, and (iii) nonlinear effects for continuous covariates. Simulations evaluate the accuracy of the time-dependent and/or nonlinear estimates, and of the resulting survival functions, in multivariable settings. The proposed flexible extension of the accelerated failure time model is applied to re-assess the effects of prognostic factors on mortality after septic shock.

Comparative effectiveness and safety of direct oral anticoagulants versus vitamin K antagonists in nonvalvular atrial fibrillation: a Canadian multicentre observational cohort study.

BACKGROUND: Direct oral anticoagulants (DOACs) have widely replaced warfarin for stroke prevention in nonvalvular atrial fibrillation. Our objective was to compare the safety and effectiveness of DOACs (dabigatran, rivaroxaban, apixaban) versus warfarin for stroke prevention in nonvalvular atrial fibrillation in the Canadian setting. METHODS: We conducted a population-based observational multicentre cohort study with propensity score matching and subsequent meta-analysis. We used health care databases from 7 Canadian provinces (British Columbia, Alberta, Saskatchewan, Manitoba, Ontario, Quebec and Nova Scotia). Patients with nonvalvular atrial fibrillation who initiated anticoagulation therapy in 2009-2017 were matched to an equal number who initiated warfarin. The primary outcome was the pooled hazard ratio (HR) for ischemic stroke or systemic embolization. Secondary outcomes included pooled HRs for major bleeding; a composite outcome of stroke, systemic embolization, major bleeding and all-cause mortality; and myocardial infarction. We modelled HRs using proportional hazard Cox regression with inverse probability of censoring weights, and estimated pooled HRs with random-effect meta-analyses. RESULTS: We included 128 273 patients who initiated anticoagulation with a DOAC (40 503 dabigatran, 49 498 rivaroxaban and 38 272 apixaban) and 128 273 patients who initiated anticoagulation with warfarin. The pooled HR for ischemic stroke or systemic embolization comparing DOACs to warfarin was 1.02 (95% confidence interval [CI] 0.87 to 1.19). Direct oral anticoagulants were associated with lower rates of major bleeding (pooled HR 0.81, 95% CI 0.69 to 0.97), the composite outcome (pooled HR 0.81, 95% CI 0.74 to 0.89) and all-cause mortality (pooled HR 0.81, 95% CI 0.78 to 0.85). INTERPRETATION: In this real-world study, DOACs were associated with similar risks of ischemic stroke or systemic embolization, and lower risks of bleeding and total mortality compared to warfarin. These findings support the use of DOACs for anticoagulation in nonvalvular atrial fibrillation. TRIAL REGISTRATION: ClinicalTrials.gov, no. NCT03596502.

Graphic report of the results from propensity score method analyses.

OBJECTIVES: To increase transparency in studies reporting propensity scores by using graphical methods that clearly illustrate (1) the number of participant exclusions that occur as a consequence of the analytic strategy and (2) whether treatment effects are constant or heterogeneous across propensity scores. STUDY DESIGN AND SETTING: We applied graphical methods to a real-world pharmacoepidemiologic study that evaluated the effect of initiating statin medication on the 1-year all-cause mortality post-myocardial infarction. We propose graphical methods to show the consequences of trimming and matching on the exclusion of participants from the analysis. We also propose the use of meta-analytical forest plots to show the magnitude of effect heterogeneity. RESULTS: A density plot with vertical lines demonstrated the proportion of subjects excluded because of trimming. A frequency plot with horizontal lines demonstrated the proportion of subjects excluded because of matching. An augmented forest plot illustrates the amount of effect heterogeneity present in the data. CONCLUSION: Our proposed techniques present additional and useful information that helps readers understand the sample that is analyzed with propensity score methods and whether effect heterogeneity is present.

Studying noncollapsibility of the odds ratio with marginal structural and logistic regression models.

One approach to quantifying the magnitude of confounding in observational studies is to compare estimates with and without adjustment for a covariate, but this strategy is known to be defective for noncollapsible measures such as the odds ratio. Comparing estimates from marginal structural and standard logistic regression models, the total difference between crude and conditional effects can be decomposed into the sum of a noncollapsibility effect and confounding bias. We provide an analytic approach to assess the noncollapsibility effect in a point-exposure study and provide a general formula for expressing the noncollapsibility effect. Next, we provide a graphical approach that illustrates the relationship between the noncollapsibility effect and the baseline risk, and reveals the behavior of the noncollapsibility effect for a range of different exposure and covariate effects. Various observations about noncollapsibility can be made from the different scenarios with or without confounding; for example, the magnitude of effect of the covariate plays a more important role in the noncollapsibility effect than does that of the effect of the exposure. In order to explore the noncollapsibility effect of the odds ratio in the presence of time-varying confounding, we simulated an observational cohort study. The magnitude of noncollapsibility was generally comparable to the effect in the point-exposure study in our simulation settings. Finally, in an applied example we demonstrate that collapsibility can have an important impact on estimation in practice.

Effect Estimation in Point-Exposure Studies with Binary Outcomes and High-Dimensional Covariate Data - A Comparison of Targeted Maximum Likelihood Estimation and Inverse Probability of Treatment Weighting.

Inverse probability of treatment weighting (IPW) and targeted maximum likelihood estimation (TMLE) are relatively new methods proposed for estimating marginal causal effects. TMLE is doubly robust, yielding consistent estimators even under misspecification of either the treatment or the outcome model. While IPW methods are known to be sensitive to near violations of the practical positivity assumption (e. g., in the case of data sparsity), the consequences of this violation in the TMLE framework for binary outcomes have been less widely investigated. As near practical positivity violations are particularly likely in high-dimensional covariate settings, a better understanding of the performance of TMLE is of particular interest for pharmcoepidemiological studies using large databases. Using plasmode and Monte-Carlo simulation studies, we evaluated the performance of TMLE compared to that of IPW estimators based on a point-exposure cohort study of the marginal causal effect of post-myocardial infarction statin use on the 1-year risk of all-cause mortality from the Clinical Practice Research Datalink. A variety of treatment model specifications were considered, inducing different degrees of near practical non-positivity. Our simulation study showed that the performance of the TMLE and IPW estimators were comparable when the dimension of the fitted treatment model was small to moderate; however, they differed when a large number of covariates was considered. When a rich outcome model was included in the TMLE, estimators were unbiased. In some cases, we found irregular bias and large standard errors with both methods even with a correctly specified high-dimensional treatment model. The IPW estimator showed a slightly better root MSE with high-dimensional treatment model specifications in our simulation setting. In conclusion, for estimation of the marginal expectation of the outcome under a fixed treatment, TMLE and IPW estimators employing the same treatment model specification may perform differently due to differential sensitivity to practical positivity violations; however, TMLE, being doubly robust, shows improved performance with richer specifications of the outcome model. Although TMLE is appealing for its double robustness property, such violations in a high-dimensional covariate setting are problematic for both methods.