Increasing the efficiency of randomized trial estimates via linear adjustment for a prognostic score

Estimating causal effects from randomized experiments is central to clinical research. Reducing the statistical uncertainty in these analyses is an important objective for statisticians. Registries, prior trials, and health records constitute a growing compendium of historical data on patients under standard-of-care that may be exploitable to this end. However, most methods for historical borrowing achieve reductions in variance by sacrificing strict type-I error rate control. Here, we propose a use of historical data that exploits linear covariate adjustment to improve the efficiency of trial analyses without incurring bias. Specifically, we train a prognostic model on the historical data, then estimate the treatment effect using a linear regression while adjusting for the trial subjects’ predicted outcomes (their prognostic scores). We prove that, under certain conditions, this prognostic covariate adjustment procedure attains the minimum variance possible among a large class of estimators. When those conditions are not met, prognostic covariate adjustment is still more efficient than raw covariate adjustment and the gain in efficiency is proportional to a measure of the predictive accuracy of the prognostic model above and beyond the linear relationship with the raw covariates. We demonstrate the approach using simulations and a reanalysis of an Alzheimer’s disease clinical trial and observe meaningful reductions in mean-squared error and the estimated variance. Lastly, we provide a simplified formula for asymptotic variance that enables power calculations that account for these gains. Sample size reductions between 10% and 30% are attainable when using prognostic models that explain a clinically realistic percentage of the outcome variance.

Identifying miRNA-mRNA Networks Associated With COPD Phenotypes

Chronic obstructive pulmonary disease (COPD) is characterized by expiratory airflow limitation and symptoms such as shortness of breath. Although many studies have demonstrated dysregulated microRNA (miRNA) and gene (mRNA) expression in the pathogenesis of COPD, how miRNAs and mRNAs systematically interact and contribute to COPD development is still not clear. To gain a deeper understanding of the gene regulatory network underlying COPD pathogenesis, we used Sparse Multiple Canonical Correlation Network (SmCCNet) to integrate whole blood miRNA and RNA-sequencing data from 404 participants in the COPDGene study to identify novel miRNA–mRNA networks associated with COPD-related phenotypes including lung function and emphysema. We hypothesized that phenotype-directed interpretable miRNA–mRNA networks from SmCCNet would assist in the discovery of novel biomarkers that traditional single biomarker discovery methods (such as differential expression) might fail to discover. Additionally, we investigated whether adjusting -omics and clinical phenotypes data for covariates prior to integration would increase the statistical power for network identification. Our study demonstrated that partial covariate adjustment for age, sex, race, and CT scanner model (in the quantitative emphysema networks) improved network identification when compared with no covariate adjustment. However, further adjustment for current smoking status and relative white blood cell (WBC) proportions sometimes weakened the power for identifying lung function and emphysema networks, a phenomenon which may be due to the correlation of smoking status and WBC counts with the COPD-related phenotypes. With partial covariate adjustment, we found six miRNA–mRNA networks associated with COPD-related phenotypes. One network consists of 2 miRNAs and 28 mRNAs which had a 0.33 correlation (p = 5.40E-12) to forced expiratory volume in 1 s (FEV1) percent predicted. We also found a network of 5 miRNAs and 81 mRNAs that had a 0.45 correlation (p = 8.80E-22) to percent emphysema. The miRNA–mRNA networks associated with COPD traits provide a systems view of COPD pathogenesis and complements biomarker identification with individual miRNA or mRNA expression data.

Covariate Adjustment in Randomized Controlled Trials

Introduction & Objective: Unadjusted analyses, fully adjusted analyses, or adjusted analyses based on tests of significance on covariate imbalance are recommended for covariate adjustment in randomized controlled trials. It has been indicated that the tests of significance on baseline comparability is inappropriate, rather it is important to indicate the strength of relationship with outcomes. Our goal is to understand when the adjustment should be used in randomized controlled trials. Methods: Unadjusted analysis, fully adjusted analysis, and adjusted analysis based on baseline comparability were examined under null and alternative hypothesis by simulation studies. Each data set was simulated 3000 times for a total of 9 scenarios for sample sizes of 20, 40, and 100 each with baseline thresholds of 0.05, 0.1, and 0.2. Each scenario was examined by the change in magnitude of correlation from 0.1 to 0.9. Results: Power of fully adjusted analysis under alternative hypothesis was increased as the correlation increased while adjusted analysis based on the covariate imbalance did not compare favorably to the unadjusted analysis. Type 1 error was decreased in adjusted analysis based on the covariate imbalance under null hypothesis. It was then observed that p-value does not follow a uniform distribution under the null hypothesis. Conclusion: Unadjusted and fully adjusted analyses were valid analyses. Full adjustment could potentially increase the power if adjustment is known. However, adjusted analysis based on the test of significance on covariate imbalance may not be a valid analysis. Tests of significance should not be used for comparing baseline comparability.

Applying Propensity Score Methods in Clinical Research in Neurology

Propensity score-based analysis is increasingly being used in observational studies to estimate the effects of treatments, interventions, and exposures. We introduce the concept of the propensity score and how it can be used in observational research. We describe four different ways of using the propensity score: matching on the propensity score, inverse probability of treatment weighting using the propensity score, stratification on the propensity score, and covariate adjustment on the propensity score (with a focus on the first two). We provide recommendations for the use and reporting of propensity score methods for the conduct of observational studies in neurological research.

Comparisons of Propensity Score Methods for Time to Event Outcomes: Evaluation through Simulations and Oral Squamous Cell Carcinoma Case Study

Introduction & Objective: In observational studies, it is recommended to use propensity score (PS) methods or covariate adjustment for confounding effect adjustment. However, few guidelines are available regarding the choice of PS approaches or covariate adjustment for the best performance in a particular data. In this study, we compared different PS methods and conventional covariate adjustment to investigate the treatment effect for the overall population on time-to-event outcomes. Methods: In the Monte Carlo simulations, we compared the hazard ratio (HR) and precision estimated using covariate adjustment and eight different PS approaches, including matching, stratification, and inverse probability of treatment weighting (IPTW). In the Oral Squamous-Cell Carcinoma Cancer case study, we applied the aforementioned PS approaches to compare the effect of receiving post-operative radiation therapy (PORT) and having engraftable tumors on different time-to-event clinical outcomes. Results: In the simulations, both IPTW and covariate adjustment produced unbiased HR estimates with small uncertainty. In the case study, covariate adjustment showed that patients with engraftable tumors were twice as likely to have local/regional recurrence (HR 1.98 [1.23, 3.18], p-value<0.005) and any recurrence or death (HR 2.02 [1.38, 2.96], p-value<0.001); patients received PORT were twice as likely to develop either local, regional, or distance recurrence (HR 2.12 [1.32, 3.41], p-value<0.005). Results produced by IPTW were consistent with covariate adjustment method (within ± 0.1 differences). Conclusion: Covariate adjustment and the IPTW method performed well across simulations and the case study. In practice, care should be taken to select the most suitable method when estimating the treatment, exposure or intervention effect on time-to-event outcomes.

Mid-term outcomes of off-pump versus on-pump coronary artery bypass graft surgery; statistical challenges in comparison

Background Despite several studies comparing off- and on-pump coronary artery bypass grafting (CABG), the effectiveness and outcomes of off-pump CABG still remain uncertain. Methods In this registry-based study, we assessed 8163 patients who underwent isolated CABG between 2014 and 2016. Propensity score matching (PSM), inverse probability of weighting (IPW) and covariate adjustment were performed to correct for and minimize selection bias. Results The overall mean age of the patients was 62 years, and 25.7% were women. Patients who underwent off-pump CABG had shorter length of hospitalization (p < 0.001), intubation time (p = 0.003) and length of ICU admission (p < 0.001). Off-pump CABG was associated with higher risk of 30-days mortality (OR: 1.7; 95% CI 1.09–2.65; p = 0.019) in unadjusted analysis. After covariate adjustment and matching (PSM and IPW), this difference was not statistically significant. After an average of 36.1 months follow-up, risk of MACCE and all-cause mortality didn’t have significant differences in both surgical methods by adjusting with IPW (HR: 1.03; 95% CI 0.87–1.24; p = 0.714; HR: 0.91; 95% CI 0.73–1.14; p = 578, respectively). Conclusion Off-pump and on-pump techniques have similar 30-day mortality (adjusted, PSM and IPW). Off-pump surgery is probably more cost-effective in short term; however, mid-term survival and MACCE trends in both surgical methods are comparable.