Propensity score matching for estimating a marginal hazard ratio.

Propensity score matching is commonly used to draw causal inference from observational survival data. However, its asymptotic properties have yet to be established, and variance estimation is still open to debate. We derive the statistical properties of the propensity score matching estimator of the marginal causal hazard ratio based on matching with replacement and a fixed number of matches. We also propose a double-resampling technique for variance estimation that takes into account the uncertainty due to propensity score estimation prior to matching.

Predicting Response to Bolus Insulin Therapy in Patients With Type 2 Diabetes.

BACKGROUND: The aim of this study was to develop a predictive model to classify people with type 2 diabetes (T2D) into expected levels of success upon bolus insulin initiation. METHODS: Machine learning methods were applied to a large nationally representative insurance claims database from the United States (dNHI database; data from 2007 to 2017). We trained boosted decision tree ensembles (XGBoost) to assign people into Class 0 (never meeting HbA1c goal), Class 1 (meeting but not maintaining HbA1c goal), or Class 2 (meeting and maintaining HbA1c goal) based on the demographic and clinical data available prior to initiating bolus insulin. The primary objective of the study was to develop a model capable of determining at an individual level, whether people with T2D are likely to achieve and maintain HbA1c goals. HbA1c goal was defined at <8.0% or reduction of baseline HbA1c by >1.0%. RESULTS: Of 15 331 people with T2D (mean age, 53.0 years; SD, 8.7), 7800 (50.9%) people met HbA1c goal but failed to maintain that goal (Class 1), 4510 (29.4%) never attained this goal (Class 0), and 3021 (19.7%) people met and maintained this goal (Class 2). Overall, the model's receiver operating characteristic (ROC) was 0.79 with greater performance on predicting those in Class 2 (ROC = 0.92) than those in Classes 0 and 1 (ROC = 0.71 and 0.62, respectively). The model achieved high area under the precision-recall curves for the individual classes (Class 0, 0.46; Class 1, 0.58; Class 2, 0.71). CONCLUSIONS: Predictive modeling using routine health care data reasonably accurately classified patients initiating bolus insulin who would achieve and maintain HbA1c goals, but less so for differentiation between patients who never met and who did not maintain goals. Prior HbA1c was a major contributing parameter for the predictions.

Causal analyses with target trial emulation for real-world evidence removed large self-inflicted biases: systematic bias assessment of ovarian cancer treatment effectiveness.

BACKGROUND AND OBJECTIVES: Drawing causal conclusions from real-world data (RWD) poses methodological challenges and risk of bias. We aimed to systematically assess the type and impact of potential biases that may occur when analyzing RWD using the case of progressive ovarian cancer. METHODS: We retrospectively compared overall survival with and without second-line chemotherapy (LOT2) using electronic medical records. Potential biases were determined using directed acyclic graphs. We followed a stepwise analytic approach ranging from crude analysis and multivariable-adjusted Cox model up to a full causal analysis using a marginal structural Cox model with replicates emulating a reference randomized controlled trial (RCT). To assess biases, we compared effect estimates (hazard ratios [HRs]) of each approach to the HR of the reference trial. RESULTS: The reference trial showed an HR for second line vs. delayed therapy of 1.01 (95% confidence interval [95% CI]: 0.82-1.25). The corresponding HRs from the RWD analysis ranged from 0.51 for simple baseline adjustments to 1.41 (95% CI: 1.22-1.64) accounting for immortal time bias with time-varying covariates. Causal trial emulation yielded an HR of 1.12 (95% CI: 0.96-1.28). CONCLUSION: Our study, using ovarian cancer as an example, shows the importance of a thorough causal design and analysis if one is expecting RWD to emulate clinical trial results.

Diagnosis, consultation, treatment, and impact of migraine in the US: Results of the OVERCOME (US) study.

OBJECTIVE: The ObserVational survey of the Epidemiology, tReatment and Care of MigrainE (OVERCOME; United States) study is a multicohort, longitudinal web survey that assesses symptomatology, consulting, diagnosis, treatment, and impact of migraine in the United States. BACKGROUND: Regularly updating population-based views of migraine in the United States provides a method for assessing the quality of ongoing migraine care and identifying unmet needs. METHODS: The OVERCOME (US) 2018 migraine cohort involved: (I) creating a demographically representative sample of US adults using quota sampling (n = 97,478), (II) identifying people with active migraine in the past year via a validated migraine diagnostic questionnaire and/or self-reported medical diagnosis of migraine (n = 24,272), and (III) assessing consultation, diagnosis, and treatment of migraine (n = 21,143). The current manuscript evaluated whether those with low frequency episodic migraine (LFEM; 0-3 monthly headache days) differed from other categories on outcomes of interest. RESULTS: Among the migraine cohort (n = 21,143), 19,888 (94.1%) met our International Classification of Headache Disorders, 3rd edition-based case definition of migraine and 12,905 (61.0%) self-reported a medical diagnosis of migraine. Respondents' mean (SD) age was 42.2 (15.0) years; 15,697 (74.2%) were women. Having at least moderate disability was common (n = 8965; 42.4%) and around half (n = 10,783; 51.0%) had consulted a medical professional for migraine care in the past year. Only 4792 (22.7%) of respondents were currently using a triptan. Overall, 8539 (40.4%) were eligible for migraine preventive medication and 3555 (16.8%) were currently using migraine preventive medication. Those with LFEM differed from moderate and high frequency episodic migraine and chronic migraine on nearly all measures of consulting, diagnosis, and treatment. CONCLUSION: The OVERCOME (US) 2018 cohort revealed slow but steady progress in diagnosis and preventive treatment of migraine. However, despite significant impact among the population, many with migraine have unmet needs related to consulting for migraine, migraine diagnosis, and getting potentially beneficial migraine treatment. Moreover, it demonstrated the heterogeneity and varying unmet needs within episodic migraine.

Practical recommendations on double score matching for estimating causal effects.

Unlike in randomized clinical trials (RCTs), confounding control is critical for estimating the causal effects from observational studies due to the lack of treatment randomization. Under the unconfoundedness assumption, matching methods are popular because they can be used to emulate an RCT that is hidden in the observational study. To ensure the key assumption hold, the effort is often made to collect a large number of possible confounders, rendering dimension reduction imperative in matching. Three matching schemes based on the propensity score (PSM), prognostic score (PGM), and double score (DSM, ie, the collection of the first two scores) have been proposed in the literature. However, a comprehensive comparison is lacking among the three matching schemes and has not made inroads into the best practices including variable selection, choice of caliper, and replacement. In this article, we explore the statistical and numerical properties of PSM, PGM, and DSM via extensive simulations. Our study supports that DSM performs favorably with, if not better than, the two single score matching in terms of bias and variance. In particular, DSM is doubly robust in the sense that the matching estimator is consistent requiring either the propensity score model or the prognostic score model is correctly specified. Variable selection on the propensity score model and matching with replacement is suggested for DSM, and we illustrate the recommendations with comprehensive simulation studies. An R package is available at https://github.com/Yunshu7/dsmatch.

Predicting optimal treatment regimens for patients with HR+/HER2- breast cancer using machine learning based on electronic health records.

Aim: To predict optimal treatments maximizing overall survival (OS) and time to treatment discontinuation (TTD) for patients with metastatic breast cancer (MBC) using machine learning methods on electronic health records. Patients/methods: Adult females with HR+/HER2- MBC on first- or second-line systemic therapy were eligible. Random survival forest (RSF) models were used to predict optimal regimen classes for individual patients and each line of therapy based on baseline characteristics. Results: RSF models suggested greater use of CDK4 & 6 inhibitor-based therapies may maximize OS and TTD. RSF-predicted optimal treatments demonstrated longer OS and TTD compared with nonoptimal treatments across line of therapy (hazard ratios = 0.44∼0.79). Conclusion: RSF may help inform optimal treatment choices and improve outcomes for patients with HR+/HER2- MBC.

Replication of randomized clinical trial results using real-world data: paving the way for effectiveness decisions.

The FDA is preparing guidance about using real-world evidence (RWE) to support decisions about product effectiveness. Several ongoing efforts aim to replicate randomized clinical trial (RCT) results using RWE with the intent of identifying circumstances and methods that provide valid evidence of drug effects. Lack of agreement may not be due to faulty methods but rather to the challenges with emulating RCTs, differences in healthcare settings and patient populations, differences in effect measures and data analysis, bias, and/or the efficacy-effectiveness gap. In fact, for some decisions, RWE may lead to better understanding of how treatments work in usual care settings than a more constrained view from RCTs. Efforts to reconcile the role and opportunities for generating complementary evidence from RWE and RCTs will advance regulatory science.

The Association Between Adherence to Insulin Therapy and Health Care Costs for Adults with Type 2 Diabetes: Evidence from a U.S. Retrospective Claims Database.

BACKGROUND: Research has shown that many patients with type 2 diabetes (T2D) are not adherent to their medication regimen. OBJECTIVE: To examine the association between adherence to insulin therapy and all-cause health care costs for patients with T2D. METHODS: This study used the IQVIA PharMetrics Plus Linkable to Ambulatory Electronic Medical Record data from January 1, 2012, through September 30, 2017. Patients were included if they were identified with T2D and initiated therapy on basal insulin (BAS) or basal-bolus (BAS-BOL) combination at any time from January 1, 2013, through October 1, 2016. Patients aged < 18 years, who used an insulin pump, identified as pregnant, or did not have continuous insurance coverage from 1 year before initiation on insulin therapy through 1 year after initiation were excluded. Descriptive statistics compared patient characteristics and costs (in U.S. 2017 dollars) between patients who were adherent or nonadherent to their insulin therapy in the 1-year postperiod, where adherence was defined as having proportion of days covered (PDC) of at least 80%. In addition, generalized linear models were used to compare costs between adherent and nonadherent patients, while controlling for patient characteristics, previous general health and comorbidities, resource utilization, medication use and type of insulin. RESULTS: 13,296 patients were included in the BAS cohort (5,502 adherent; 7,794 nonadherent) and 10,069 in the BAS-BOL cohort (2,006 adherent; 8,063 nonadherent). Adherent patients had significantly lower all-cause total unadjusted costs following initiation on BAS ($29,322 vs. $31,888, P = 0.0134) and BAS-BOL combination ($36,229 vs. $40,147, P = 0.0078). Drug costs comprised 39.5%-45.4% of costs among adherent patients and 23.0%-25.9% of costs among nonadherent patients. Multivariable analyses revealed that adherent patients had significantly lower adjusted all-cause total costs than nonadherent patients in the BAS cohort ($30,127 vs. $37,049, 95% CI for difference -$8,460 to -$5,384) and the BAS-BOL cohort ($36,603 vs. $44,702, 95% CI for difference -$9,129 to -$6,980). CONCLUSIONS: In patients with T2D who initiated BAS or BAS-BOL combination therapy, adherence was associated with significantly lower all-cause total health care costs, despite significantly higher drug costs. These results illustrate the potential economic benefits associated with adherence to insulin therapy. DISCLOSURES": Eli Lilly and Company funded this study and was responsible for study design and execution. Bajpai, Eby, Faries, and Haynes are employees and own stock in Eli Lilly and Company. Lage received compensation from Eli Lilly and Company for her work on this research project.

Digital Tracking of Rheumatoid Arthritis Longitudinally (DIGITAL) Using Biosensor and Patient-Reported Outcome Data: Protocol for a Real-World Study.

BACKGROUND: Rheumatoid arthritis (RA) is a condition with symptoms that vary over time. The typical 3- to 6-month interval between physician visits may lead to patients failing to recall or underreporting symptoms experienced during the interim. Wearable digital technology enables the regular passive collection of patients' biometric and activity data. If it is shown to be strongly related to data captured by patient-reported outcome (PRO) measures, information collected passively from wearable digital technology could serve as an objective proxy or be complementary to patients' subjective experience of RA symptoms. OBJECTIVE: The goal of this study is to characterize the extent to which digital measures collected from a consumer-grade smartwatch agree with measures of RA disease activity and other PROs collected via a smartphone app. METHODS: This observational study will last 6 months for each participant. We aim to recruit 250 members of the ArthritisPower registry with an RA diagnosis who will receive a smartwatch to wear for the period of the study. From the ArthritisPower mobile app on their own smartphone device, participants will be prompted to answer daily and weekly electronic PRO (ePRO) measures for the first 3 months. RESULTS: The study was launched in December 2018 and will require up to 18 months to complete. Study results are expected to be published by the end of 2021. CONCLUSIONS: The completion of this study will provide important data regarding the following: (1) the relationship between passively collected digital measures related to activity, heart rate, and sleep collected from a smartwatch with ePROs related to pain, fatigue, physical function, and RA flare entered via smartphone app; (2) determine predictors of adherence with smartwatch and smartphone app technology; and (3) assess the effect of study-specific reminders on adherence with the smartwatch. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): DERR1-10.2196/14665.

The Effectiveness of a Private Orthopaedic Practice-Based Osteoporosis Management Service to Reduce the Risk of Subsequent Fractures.

BACKGROUND: Osteoporosis is prevalent in the United States, with an increasing need for management. In this study, we evaluated the effectiveness of a private orthopaedic practice-based osteoporosis management service (OP MS) in reducing subsequent fracture risk and improving other aspects of osteoporosis management of patients who had sustained fractures. METHODS: This was a retrospective cohort study using the 100% Medicare data set for Michigan residents with any vertebral; hip, pelvic or femoral; or other nonvertebral fracture during the period of April 1, 2010 to September 30, 2014. Patients who received OP MS care with a follow-up visit within 90 days of the first fracture, and those who did not seek OP MS care but had a physician visit within 90 days of the first fracture, were considered as exposed and unexposed, respectively (first follow-up visit = index date). Eligible patients with continuous enrollment in Medicare Parts A and B for the 90-day pre-index period were followed until the earliest of death, health-plan disenrollment, or study end (December 31, 2014) to evaluate rates of subsequent fracture, osteoporosis medication prescriptions filled, and bone mineral density (BMD) assessments. Health-care costs were evaluated among patients with 12 months of post-index continuous enrollment. Propensity-score matching was used to balance differences in baseline characteristics. Each exposed patient was matched to an unexposed patient within ± 0.01 units of the propensity score. After propensity-score matching, Cox regression examined the hazard ratio (HR) of clinical and economic outcomes in the exposed and unexposed cohorts. RESULTS: Two well-matched cohorts of 1,304 patients each were produced. The exposed cohort had a longer median time to subsequent fracture (998 compared with 743 days; log-rank p = 0.001), a lower risk of subsequent fracture (HR = 0.8; 95% confidence interval [CI] = 0.7 to 0.9), and a higher likelihood of having osteoporosis medication prescriptions filled (HR = 1.7; 95% CI = 1.4 to 2.0) and BMD assessments (HR = 4.3; 95% CI = 3.7 to 5.0). The total 12-month costs ($25,306 compared with $22,896 [USD]; p = 0.082) did not differ significantly between the cohorts. CONCLUSIONS: A private orthopaedic practice-based OP MS effectively reduced subsequent fracture risk, likely through coordinated and ongoing comprehensive patient care, without a significant overall higher cost. LEVEL OF EVIDENCE: Therapeutic Level III. See Instructions for Authors for a complete description of levels of evidence.

Addressing unmeasured confounding in comparative observational research.

PURPOSE: Observational pharmacoepidemiological studies can provide valuable information on the effectiveness or safety of interventions in the real world, but one major challenge is the existence of unmeasured confounder(s). While many analytical methods have been developed for dealing with this challenge, they appear under-utilized, perhaps due to the complexity and varied requirements for implementation. Thus, there is an unmet need to improve understanding the appropriate course of action to address unmeasured confounding under a variety of research scenarios. METHODS: We implemented a stepwise search strategy to find articles discussing the assessment of unmeasured confounding in electronic literature databases. Identified publications were reviewed and characterized by the applicable research settings and information requirements required for implementing each method. We further used this information to develop a best practice recommendation to help guide the selection of appropriate analytical methods for assessing the potential impact of unmeasured confounding. RESULTS: Over 100 papers were reviewed, and 15 methods were identified. We used a flowchart to illustrate the best practice recommendation which was driven by 2 critical components: (1) availability of information on the unmeasured confounders; and (2) goals of the unmeasured confounding assessment. Key factors for implementation of each method were summarized in a checklist to provide further assistance to researchers for implementing these methods. CONCLUSION: When assessing comparative effectiveness or safety in observational research, the impact of unmeasured confounding should not be ignored. Instead, we suggest quantitatively evaluating the impact of unmeasured confounding and provided a best practice recommendation for selecting appropriate analytical methods.

Healthcare utilization and costs of children with attention deficit/hyperactivity disorder initiating atomoxetine versus extended-release guanfacine.

OBJECTIVES: To compare 1-year direct healthcare costs and utilization among children and adolescents initiating non-stimulant medications atomoxetine (ATX) or extended-release guanfacine (GXR). METHODS: In this retrospective, observational cohort study, children and adolescents aged 6-17 years with attention deficit/hyperactivity disorder (ADHD) who had ≥1 prescription claim for ATX or GXR between December 31, 2009 and January 1, 2011 were identified in the MarketScan Commercial or Multi-State Medicaid claims databases. The first claim was set as the index. Patients with no claims for other ADHD medications that overlapped with the days' supply for the index therapy during the post-period were classified as initiating monotherapy. All-cause and ADHD-related utilization and costs (2011 US$) and treatment patterns (adherence and persistence) were evaluated during the 12 months following index. Propensity score adjustment accounted for differences in patient characteristics, and bootstrapping was used for comparisons. RESULTS: A total of 13,239 children and adolescents with ADHD met the study criteria (4,411 ATX initiators and 8,828 GXR initiators). There were 2,699 ATX monotherapy patients. In propensity-score-adjusted analyses, mean all-cause total costs were significantly less for monotherapy ATX initiators than for GXR initiators ($7,553 vs $10,639; difference = -$3,086, p < .0001), as were mean ADHD-related total costs ($3,213 vs $4,544; difference = -$1,330, p < .0001). Monotherapy ATX initiators had significantly fewer all-cause and ADHD-related total medical visits and ∼22 days shorter persistence to index therapy (p < .0001). Results were similar for secondary analyses comparing all ATX with all GXR initiators, regardless of monotherapy or combination regimen, and comparing only monotherapy initiators. CONCLUSIONS: Children and adolescents with ADHD who initiated ATX monotherapy incurred lower all-cause and ADHD-related total healthcare costs than patients who initiated GXR. This was due in part to less healthcare resource utilization and slightly shorter persistence for ATX patients. These findings may aid decision-making and inform future studies, but must be tempered due to inherent observational research limitations.

Evaluating different strategies for estimating treatment effects in observational studies.

Since the introduction of the propensity score (PS), methods for estimating treatment effects with observational data have received growing attention in the literature. Recent research has added substantially to the number of available statistical approaches for controlling confounding in such analyses. However, researchers need guidance to decide on the optimal analytic strategy for any given scenario. To address this gap, we conducted simulations evaluating both well-established methods (regression, PS weighting, stratification, and matching) and more recently proposed approaches (tree-based methods, local control, entropy balancing, genetic matching, prognostic scoring). The simulation scenarios included tree-based and smooth regression models as true data-generation mechanisms. We evaluated an extensive number of analysis strategies combining different treatment choices and outcome models. Key findings include 1) the lack of a single best strategy across all potential scenarios; 2) the importance of appropriately addressing interactions in the treatment choice model and/or outcome model; and 3) a tree-structured treatment choice model and a polynomial outcome model with second-order interactions performed well. One limitation to this initial assessment is the lack of heterogeneous simulation scenarios allowing treatment effects to vary by patient.

Impact of Proton Pump Inhibitor Use on the Comparative Effectiveness and Safety of Prasugrel Versus Clopidogrel: Insights From the Treatment With Adenosine Diphosphate Receptor Inhibitors: Longitudinal Assessment of Treatment Patterns and Events After Acute Coronary Syndrome (TRANSLATE-ACS) Study.

BACKGROUND: Proton pump inhibitors (PPIs) reduce gastrointestinal bleeding events but may alter clopidogrel metabolism. We sought to understand the comparative effectiveness and safety of prasugrel versus clopidogrel in the context of proton pump inhibitor (PPI) use. METHODS AND RESULTS: Using data on 11 955 acute myocardial infarction (MI) patients treated with percutaneous coronary intervention at 233 hospitals and enrolled in the TRANSLATE-ACS study, we compared whether discharge PPI use altered the association of 1-year adjusted risks of major adverse cardiovascular events (MACE; death, MI, stroke, or unplanned revascularization) and Global Use of Strategies To Open Occluded Arteries (GUSTO) moderate/severe bleeding between prasugrel- and clopidogrel-treated patients. Overall, 17% of prasugrel-treated and 19% of clopidogrel-treated patients received a PPI at hospital discharge. At 1 year, patients discharged on a PPI versus no PPI had higher risks of MACE (adjusted hazard ratio [HR] 1.38, 95% confidence interval [CI] 1.21-1.58) and GUSTO moderate/severe bleeding (adjusted HR 1.55, 95% CI 1.15-2.09). Risk of MACE was similar between prasugrel and clopidogrel regardless of PPI use (adjusted HR 0.88, 95% CI 0.62-1.26 with PPI, adjusted HR 1.07, 95% CI 0.90-1.28 without PPI, interaction P=0.31). Comparative bleeding risk associated with prasugrel versus clopidogrel use differed based on PPI use but did not reach statistical significance (adjusted HR 0.73, 95% CI 0.36-1.48 with PPI, adjusted HR 1.34, 95% CI 0.79-2.27 without PPI, interaction P=0.17). CONCLUSIONS: PPIs did not significantly affect the MACE and bleeding risk associated with prasugrel use, relative to clopidogrel. CLINICAL TRIAL REGISTRATION: URL: https://www.clinicaltrials.gov. Unique identifier: NCT01088503.

Factors Associated With Initial Prasugrel Versus Clopidogrel Selection for Patients With Acute Myocardial Infarction Undergoing Percutaneous Coronary Intervention: Insights From the Treatment With ADP Receptor Inhibitors: Longitudinal Assessment of Treatment Patterns and Events After Acute Coronary Syndrome (TRANSLATE-ACS) Study.

BACKGROUND: Few studies have examined how antiplatelet therapies are selected during the routine care of acute myocardial infarction patients, particularly relative to the patient's estimated mortality and bleeding risks. METHODS AND RESULTS: We examined patients presenting with acute myocardial infarction treated with percutaneous coronary intervention at 233 US hospitals in the TRANSLATE-ACS observational study from April 2010 to October 2012. We developed a multivariable logistic regression model to identify factors associated with prasugrel selection. Prasugrel use rates and associated 1-year risk-adjusted major adverse cardiovascular events and Global Utilization of Streptokinase and t-PA for Occluded Coronary Arteries (GUSTO) moderate/severe bleeding outcomes were also examined in relation to predicted mortality and bleeding using the validated Acute Coronary Treatment and Intervention Outcomes (ACTION) risk prediction scores. Among 11 969 patients, 3123 (26%) received prasugrel at the time of percutaneous coronary intervention. The strongest factors associated with prasugrel use included cardiogenic shock (odds ratio [OR] 1.68, 95% CI 1.25-2.26), drug-eluting stent use (OR 1.45, 95% CI 1.31-1.62), and ST-segment elevation myocardial infarction presentation (OR 1.23, 95% CI 1.12-1.35). Older age (OR 0.57, 95% CI 0.0.53-0.61), dialysis (OR 0.56, 95% CI 0.32-0.96), prior history of stroke/transient ischemic attack (OR 0.52, 95% CI 0.38-0.73), and interhospital transfer (OR 0.50, 95% CI 0.46-0.55) were associated with lowest prasugrel selection. Prasugrel was used less often than clopidogrel in patients at higher predicted bleeding risk (21.9% versus 29.7%, P<0.001). Yet paradoxically, prasugrel was also less likely than clopidogrel to be used in patients with higher predicted mortality risk (21.1% versus 30.2%, P<0.001). Adjusted bleeding and outcomes events were similar among those receiving prasugrel and clopidogrel in the 4 subgroups of patients based on bleeding risk and ischemic benefits. CONCLUSIONS: In community practice, prasugrel use may be driven more by bleeding risk rather than ischemic benefit. This may result in underutilization of higher potency ADP receptor inhibitor among patients more likely to derive ischemic benefit.

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

IMPORTANCE: There is increasing interest in performing comparative effectiveness analyses in large observational databases, yet these analyses must adjust for treatment selection issues. OBJECTIVES: To conduct comparative safety and efficacy analyses of prasugrel vs clopidogrel bisulfate after percutaneous coronary intervention and to evaluate inverse probability of treatment weighting (a propensity score method) and instrumental variable methods. DESIGN, SETTING, AND PARTICIPANTS: This study used data from the Treatment With Adenosine Diphosphate Receptor Inhibitors-Longitudinal Assessment of Treatment Patterns and Events After Acute Coronary Syndrome (TRANSLATE-ACS) study. Included in the study were patients undergoing percutaneous coronary intervention for myocardial infarction, 26.0% of whom received prasugrel. The study dates were April 4, 2010, to October 31, 2012. EXPOSURES: Choice of initial antiplatelet agent (prasugrel or clopidogrel). MAIN OUTCOMES AND MEASURES: Safety and efficacy outcomes included 1-year composite major adverse cardiovascular events, moderate to severe bleeding, and stent thrombosis. Hospitalizations for pneumonia, bone fractures, and planned percutaneous coronary intervention were used as the falsification end points. RESULTS: The study cohort comprised 11 784 participants (mean [SD] age, 60.0 [11.6] years, and 28.0% were female). Using inverse probability of treatment weighting adjustment, prasugrel and clopidogrel had similar major adverse cardiovascular events (hazard ratio [HR], 0.98; 95% CI, 0.83-1.16) and bleeding outcomes (1.18; 0.77-1.80), but prasugrel had a lower rate of stent thrombosis (0.51; 0.31-0.85). Using instrumental variable methods, prasugrel use was associated with a lower rate of the major adverse cardiovascular event end point (HR, 0.68; 95% CI, 0.47-1.00) but nonsignificant differences in the rates of bleeding (0.95; 0.41-2.08) and stent thrombosis (0.67; 0.16-2.00). There was no significant treatment difference noted in any of the falsification end-point rates when analyses were performed using inverse probability of treatment weighting, although the bone fracture end point approached statistical significance. Nevertheless, a lower rate of pneumonia-related hospitalizations was noted in the prasugrel-treated patients when analyses were performed using instrumental variable methods. CONCLUSIONS AND RELEVANCE: Conclusions regarding the safety and efficacy of antiplatelet therapy varied depending on analytic technique, and none were concordant with the results from randomized trials. In addition, both statistical strategies demonstrated concerning associations when tested in the falsification analyses. A high level of scrutiny and careful attention to assumptions and validity are required when interpreting complex analyses of observational data.

"A Bayesian sensitivity analysis to evaluate the impact of unmeasured confounding with external data: a real world comparative effectiveness study in osteoporosis".

PURPOSE: Observational studies are frequently used to assess the effectiveness of medical interventions in routine clinical practice. However, the use of observational data for comparative effectiveness is challenged by selection bias and the potential of unmeasured confounding. This is especially problematic for analyses using a health care administrative database, in which key clinical measures are often not available. This paper provides an approach to conducting a sensitivity analyses to investigate the impact of unmeasured confounding in observational studies. METHODS: In a real world osteoporosis comparative effectiveness study, the bone mineral density (BMD) score, an important predictor of fracture risk and a factor in the selection of osteoporosis treatments, is unavailable in the data base and lack of baseline BMD could potentially lead to significant selection bias. We implemented Bayesian twin-regression models, which simultaneously model both the observed outcome and the unobserved unmeasured confounder, using information from external sources. A sensitivity analysis was also conducted to assess the robustness of our conclusions to changes in such external data. RESULTS: The use of Bayesian modeling in this study suggests that the lack of baseline BMD did have a strong impact on the analysis, reversing the direction of the estimated effect (odds ratio of fracture incidence at 24 months: 0.40 vs. 1.36, with/without adjusting for unmeasured baseline BMD). CONCLUSIONS: The Bayesian twin-regression models provide a flexible sensitivity analysis tool to quantitatively assess the impact of unmeasured confounding in observational studies. Copyright © 2016 John Wiley & Sons, Ltd.

Comparative resource utilization and costs for patients with acute coronary syndrome managed with percutaneous coronary intervention and treated with clopidogrel or prasugrel.

PURPOSE: Results of a study of bleeding events and other inhospital outcomes with the use of clopidogrel versus prasugrel in patients with acute coronary syndrome (ACS) managed with percutaneous coronary intervention (PCI) are reported. METHODS: Demographic and clinical data on adults hospitalized for ACS, managed with PCI, and treated with clopidogrel or prasugrel during a two-year period were extracted from a large hospital claims database. Bleeding rates, hospital length of stay (LOS), and total hospital costs during the index hospitalization were evaluated. RESULTS: The study sample consisted of 75,297 patients who received clopidogrel and 9,477 who received prasugrel. The unadjusted bleeding rates were 5.7% with clopidogrel use and 3.2% with prasugrel use (p < 0.0001). After propensity score stratification to adjust for selection bias, rates of bleeding events were not significantly different between clopidogrel- and prasugrel-treated patients (odds ratio, 0.90; 95% confidence interval [CI], 0.80-1.02; p = 0.0949). The adjusted mean ± S.D. hospital LOS was 0.22 day lower (95% CI, 0.15-0.28; p < 0.001) with the use of prasugrel versus clopidogrel, and adjusted total mean hospital costs were $375 less for prasugrel-treated patients (p = 0.003). CONCLUSION: After adjustments for demographic and clinical characteristics, rates of inhospital bleeding in patients who received prasugrel and those who received clopidogrel were not significantly different. The adjusted analyses showed that the mean hospital LOS was shorter and total mean hospital costs were lower for patients treated with prasugrel.

Propensity score matching and subclassification in observational studies with multi-level treatments.

In this article, we develop new methods for estimating average treatment effects in observational studies, in settings with more than two treatment levels, assuming unconfoundedness given pretreatment variables. We emphasize propensity score subclassification and matching methods which have been among the most popular methods in the binary treatment literature. Whereas the literature has suggested that these particular propensity-based methods do not naturally extend to the multi-level treatment case, we show, using the concept of weak unconfoundedness and the notion of the generalized propensity score, that adjusting for a scalar function of the pretreatment variables removes all biases associated with observed pretreatment variables. We apply the proposed methods to an analysis of the effect of treatments for fibromyalgia. We also carry out a simulation study to assess the finite sample performance of the methods relative to previously proposed methods.

Estimating optimal treatment regimes via subgroup identification in randomized control trials and observational studies.

With new treatments and novel technology available, personalized medicine has become an important piece in the new era of medical product development. Traditional statistics methods for personalized medicine and subgroup identification primarily focus on single treatment or two arm randomized control trials. Motivated by the recent development of outcome weighted learning framework, we propose an alternative algorithm to search treatment assignments which has a connection with subgroup identification problems. Our method focuses on applications from clinical trials to generate easy to interpret results. This framework is able to handle two or more than two treatments from both randomized control trials and observational studies. We implement our algorithm in C++ and connect it with R. Its performance is evaluated by simulations, and we apply our method to a dataset from a diabetes study. Copyright © 2016 John Wiley & Sons, Ltd.