K-means for shared frailty models

Background The purpose of this research was to see how the k-means algorithm can be applied to survival analysis with single events per subject for defining groups, which can then be modeled in a shared frailty model to further allow the capturing the unmeasured confounding not already explained by the covariates in the model. Methods For this purpose we developed our own k-means survival grouping algorithm to handle this approach. We compared a regular shared frailty model with a regular grouping variable and a shared frailty model with a k-means grouping variable in simulations as well as analysis on a real dataset. Results We found that in both simulations as well as real data showed that our k-means clustering is no different than the typical frailty clustering even under different situations of varied case rates and censoring. It appeared our k-means algorithm could be a trustworthy mechanism of creating groups from data when no grouping term exists for including in a frailty term in a survival model or comparing to an existing grouping variable available in the current data to use in a frailty model.

Assessing The Performance of Physician’s Prescribing Preference As An Instrumental Variable in Comparative Effectiveness Research With Moderate and Small Sample Sizes: A Simulation Study

BackgroundInstrumental variable (IV) analyses are used to account for unmeasured confounding in Comparative Effectiveness Research (CER) in pharmacoepidemiology. To date, simulation studies assessing the performance of IV analyses have been based on large samples. However, in many settings, sample sizes are not large.Objective In this simulation study, we assess the utility of Physician’s Prescribing Preference (PPP) as an IV for moderate and smaller sample sizes.MethodsWe designed a simulation study in a CER setting with moderate (around 2500) and small (around 600) sample sizes. The outcome and treatment variables were binary and three variables were used to represent confounding (a binary and a continuous variable representing measured confounding, and a further continuous variable representing unmeasured confounding). We compare the performance of IV and non-IV approaches using two-stage least squares (2SLS) and ordinary least squares (OLS) methods, respectively. Further, we test the performance of different forms of proxies for PPP as an IV.ResultsThe PPP IV approach results in a percent bias of approximately 20%, while the percent bias of OLS is close to 60%. The sample size is not associated with the level of bias for the PPP IV approach. However, smaller sample sizes led to lower statistical power for the PPP IV. Using proxies for PPP based on longer prescription histories result in stronger IVs, partly offsetting the effect on power of smaller sample sizes.Conclusion Irrespective of sample size, the PPP IV approach leads to less biased estimates of treatment effectiveness than conventional multivariable regression adjusting for known confounding only. Particularly for smaller sample sizes, we recommend constructing PPP from long prescribing histories to improve statistical power.

Clinical Features of Extragastrointestinal Stromal Tumor Compared with Gastrointestinal Stromal Tumor: A Retrospective, Multicenter, Real-World Study

Importance. Extragastrointestinal stromal tumor (EGIST) is a rare tumor, and its diagnosis and treatment strategy lack clinical guideline and relative literature evidence. In clinical practice, EGIST only misuses the pattern of GIST of diagnosis and treatment. The study hopes to find evidence of the treatment pattern for EGIST. Objective. This study aimed to compare the tumor characteristics and long-term outcomes between EGIST and GIST. The confounding function was applied to improve the result credibility in the case of small sample size. Design, Setting, and Participants. This cohort study enrolled 55 patients with EGIST who underwent surgery and were selected from four high-volume hospitals in China and 221 GIST patients who were collected from one of the four hospitals between January 2006 and September 2017. We used propensity score matching (PSM) and subgroup analysis to compare EGIST with GIST in terms of prognosis. The confounding function was used for sensitivity analysis to reduce unmeasured confounding. Results. We matched 43 patients in each of the GIST and EGIST groups by PSM. We compared EGIST data with GIST data to explore the prognostic factors between them. In the multivariate Cox regression model, tumor location of EGIST was negatively correlated with overall survival (after PSM: HR, 4.32; 95% CI, 1.22–15.26) or disease-free survival (after PSM: HR, 9.79; 95% CI, 2.22–43.31), which was also intuitively shown in the Kaplan–Meier survival curves (all P values < 0.05). In the subgroup analysis, EGIST with high risk factors had a worse prognosis than GIST. In unmeasured confounding analysis, the overall curve tends to show all combinations of c(0) of c(1) up to 2.0, none of which would bring the corrected relative risk to 1 for OS and DFS. Conclusions and Relevance. EGIST was associated with worse prognosis compared with GIST patients, particularly in EGIST patients with high risk factors, while there was a similar prognosis without those high risk factors.

Instrumental variable estimation of the marginal structural Cox model for time-varying treatments

Summary Robins (1998) introduced marginal structural models, a general class of counterfactual models for the joint effects of time-varying treatments in complex longitudinal studies subject to time-varying confounding. Robins (1998) established the identification of marginal structural model parameters under a sequential randomization assumption, which rules out unmeasured confounding of treatment assignment over time. The marginal structural Cox model is one of the most popular marginal structural models to evaluate the causal effect of time-varying treatments on a censored failure time outcome. In this paper, we establish sufficient conditions for identification of marginal structural Cox model parameters with the aid of a time-varying instrumental variable, when sequential randomization fails to hold due to unmeasured confounding. Our instrumental variable identification condition rules out any interaction between an unmeasured confounder and the instrumental variable in its additive effects on the treatment process, the longitudinal generalization of the identifying condition of Wang & Tchetgen Tchetgen (2018). We describe a large class of weighted estimating equations that give rise to consistent and asymptotically normal estimators of the marginal structural Cox model, thereby extending the standard inverse probability of treatment weighted estimation of marginal structural models to the instrumental variable setting. Our approach is illustrated via extensive simulation studies and an application to estimate the effect of community antiretroviral therapy coverage on HIV incidence.

OP208 Did Health Technology Assessments Make the Wrong Call? Quantitative Bias Analysis: Alectinib versus Ceritinib in Non-Small Cell Lung Cancer

IntroductionThe German health technology assessment (HTA) rejected additional benefit of alectinib for second line (2L) ALK+ NSCLC, citing possible biases from missing ECOG performance status data and unmeasured confounding in real-world evidence (RWE) for 2L ceritinib that was submitted as a comparator to the single arm alectinib trial. Alectinib was approved in the US and therefore US post-launch RWE can be used to evaluate this HTA decision.MethodsWe compared the real-world effectiveness of alectinib with ceritinib in 2L post-crizotinib ALK+ NSCLC using the nationwide Flatiron Health electronic health record (EHR)-derived de-identified database. Using quantitative bias analysis (QBA), we estimated the strength of (i) unmeasured confounding and (ii) deviation from missing-at-random (MAR) assumptions needed to nullify any overall survival (OS) benefit.ResultsAlectinib had significantly longer median OS than ceritinib in complete case analysis. The estimated effect size (Hazard Ratio: 0.55) was robust to risk ratios of unmeasured confounder-outcome and confounder-exposure associations of <2.4.Based on tipping point analysis, missing baseline ECOG performance status for ceritinib-treated patients (49% missing) would need to be more than 3.4-times worse than expected under MAR to nullify the OS benefit observed for alectinib.ConclusionsOnly implausible levels of bias reversed our conclusions. These methods could provide a framework to explore uncertainty and aid decision-making for HTAs to enable patient access to innovative therapies.

Comparing the impact on the prognosis of acute myocardial infarction critical patients of using midazolam, propofol, and dexmedetomidine for sedation

Background There are less studies focusing on the sedative therapy of acute myocardial infarction (AMI) critical patients. This study aim to compare the impact on the prognosis of AMI critical patients of using midazolam, propofol and dexmedetomidine. Methods We collected clinical data from the Medical Information Mart for Intensive Care III (MIMIC III) database. Data on 427 AMI patients with sedatives using were recruited from in Coronary Heart Disease Intensive Care unit (CCU). Results There were 143 patients in midazolam using, 272 in propofol using and 28 in dexmedetomidine using. The rate of 28-days mortality was 23.9% in overall patients. Through logistic regression analysis, only midazolam using was significant association with increased 28-days mortality when compared with propofol or dexmedetomidine using. In the subgroup analysis of age, gender, body mass index (BMI), white blood cell (WBC), beta-block, and revascularization, the association between midazolam using and increased 28-days mortality remained significantly. Through propensity score matching, 140 patients using midazolam and 192 using non-midazolam were successfully matched, the midazolam using presented with higher rate of CCU mortality, hospital mortality and 28-days mortality, longer of mechanical ventilation time and CCU duration. E-value analysis suggested robustness to unmeasured confounding. Conclusion Propofol or dexmedetomidine are preferred to be used in AMI critical patients for sedative therapy.

Causal simulation experiments: Lessons from bias amplification

Recent theoretical work in causal inference has explored an important class of variables which, when conditioned on, may further amplify existing unmeasured confounding bias (bias amplification). Despite this theoretical work, existing simulations of bias amplification in clinical settings have suggested bias amplification may not be as important in many practical cases as suggested in the theoretical literature. We resolve this tension by using tools from the semi-parametric regression literature leading to a general characterization in terms of the geometry of OLS estimators which allows us to extend current results to a larger class of DAGs, functional forms, and distributional assumptions. We further use these results to understand the limitations of current simulation approaches and to propose a new framework for performing causal simulation experiments to compare estimators. We then evaluate the challenges and benefits of extending this simulation approach to the context of a real clinical data set with a binary treatment, laying the groundwork for a principled approach to sensitivity analysis for bias amplification in the presence of unmeasured confounding.

Assessing The Effectiveness of Empirical Calibration Under Different Bias Scenarios

Background: Estimations of causal effects from observational data are subject to various sources of bias. These biases can be adjusted by using negative control outcomes not affected by the treatment. The empirical calibration procedure uses negative controls to calibrate p-values and both negative and positive controls to calibrate coverage of the 95% confidence interval of the outcome of interest. Although empirical calibration has been used in several large observational studies, there is no systematic examination of its effect under different bias scenarios. Methods: The effect of empirical calibration of confidence intervals was analyzed using simulated datasets with known treatment effects. The simulations were for binary treatment and binary outcome, with simulated biases resulting from unmeasured confounder, model misspecification, measurement error, and lack of positivity. The performance of empirical calibration was evaluated by determining the change of the confidence interval coverage and bias of the outcome of interest. Results: Empirical calibration increased coverage of the outcome of interest by the 95% confidence interval under most settings but was inconsistent in adjusting the bias of the outcome of interest. Empirical calibration was most effective when adjusting for unmeasured confounding bias. Suitable negative controls had a large impact on the adjustment made by empirical calibration, but small improvements in the coverage of the outcome of interest was also observable when using unsuitable negative controls. Conclusions: This work adds evidence to the efficacy of empirical calibration on calibrating the confidence intervals of treatment effects in observational studies. We recommend empirical calibration of confidence intervals, especially when there is a risk of unmeasured confounding.