Jessica A. Myers

One-to-many propensity score matching in cohort studies.

Among the large number of cohort studies that employ propensity score matching, most match patients 1:1. Increasing the matching ratio is thought to improve precision but may come with a trade‐off with respect to bias.

Characteristics of Clinical Trials to Support Approval of Orphan vs Nonorphan Drugs for Cancer

CONTEXT The Orphan Drug Act incentivizes medication development for rare diseases, offering substantial financial benefits to the manufacturer. Orphan products constitute most new drug approvals in oncology, but safety and efficacy questions have emerged about some of these agents. OBJECTIVES To define characteristics of orphan cancer drugs and their pivotal clinical trials and to compare these with nonorphan drugs. DESIGN AND SETTING We identified all new orphan and nonorphan drugs approved between 2004 and 2010 to treat cancer. We then collected data on key development variables from publicly available information on the US Food and Drug Administrations Web site and in the Code of Federal Regulations. MAIN OUTCOME MEASURES We assessed clinical testing dates, approved indications, and regulatory characteristics (regular vs accelerated review, advisory committee review, postmarketing commitments). We then compared design features (randomization, blinding, primary end point) of pivotal trials supporting approval of orphan and nonorphan drugs and rates of adverse safety outcomes (deaths not attributed to disease progression, serious adverse events, dropouts) in pivotal trials. RESULTS Fifteen orphan and 12 nonorphan drugs were approved between January 1, 2004, and December 31, 2010. Pivotal trials of orphan drugs had smaller participant numbers (median, 96 [interquartile range {IQR}, 66-152] vs 290 [IQR, 185-394] patients exposed to the drug; P < .001) and were less likely to be randomized (30% vs 80%; P = .007). Orphan and nonorphan pivotal trials varied in their blinding (P = .04), with orphan trials less likely to be double-blind (4% vs 33%). Primary study outcomes also varied (P = .04), with orphan trials more likely to assess disease response (68% vs 27%) rather than overall survival (8% vs 27%). More treated patients had serious adverse events in trials of orphan drugs vs trials of nonorphan drugs (48% vs 36%; odds ratio, 1.72; 95% confidence interval, 1.02-2.92; P = .04). CONCLUSION Compared with pivotal trials used to approve nonorphan cancer drugs, pivotal trials for recently approved orphan drugs for cancer were more likely to be smaller and to use nonrandomized, unblinded trial designs and surrogate end points to assess efficacy.

Effects of Adjusting for Instrumental Variables on Bias and Precision of Effect Estimates

Recent theoretical studies have shown that conditioning on an instrumental variable (IV), a variable that is associated with exposure but not associated with outcome except through exposure, can increase both bias and variance of exposure effect estimates. Although these findings have obvious implications in cases of known IVs, their meaning remains unclear in the more common scenario where investigators are uncertain whether a measured covariate meets the criteria for an IV or rather a confounder. The authors present results from two simulation studies designed to provide insight into the problem of conditioning on potential IVs in routine epidemiologic practice. The simulations explored the effects of conditioning on IVs, near-IVs (predictors of exposure that are weakly associated with outcome), and confounders on the bias and variance of a binary exposure effect estimate. The results indicate that effect estimates which are conditional on a perfect IV or near-IV may have larger bias and variance than the unconditional estimate. However, in most scenarios considered, the increases in error due to conditioning were small compared with the total estimation error. In these cases, minimizing unmeasured confounding should be the priority when selecting variables for adjustment, even at the risk of conditioning on IVs.

The Prevalence and Cost of Unapproved Uses of Top-Selling Orphan Drugs

Introduction The Orphan Drug Act encourages drug development for rare conditions. However, some orphan drugs become top sellers for unclear reasons. We sought to evaluate the extent and cost of approved and unapproved uses of orphan drugs with the highest unit sales. Methods We assessed prescription patterns for four top-selling orphan drugs: lidocaine patch (Lidoderm) approved for post-herpetic neuralgia, modafinil (Provigil) approved for narcolepsy, cinacalcet (Sensipar) approved for hypercalcemia of parathyroid carcinoma, and imatinib (Gleevec) approved for chronic myelogenous leukemia and gastrointestinal stromal tumor. We pooled patient-specific diagnosis and prescription data from two large US state pharmaceutical benefit programs for the elderly. We analyzed the number of new and total patients using each drug and patterns of reimbursement for approved and unapproved uses. For lidocaine patch, we subcategorized approved prescriptions into two subtypes of unapproved uses: neuropathic pain, for which some evidence of efficacy exists, and non-neuropathic pain. Results We found that prescriptions for lidocaine patch, modafinil, and cinacalcet associated with non-orphan diagnoses rose at substantially higher rates (average monthly increases in number of patients of 14.6, 1.45, and 1.58) than prescriptions associated with their orphan diagnoses (3.12, 0.24, and 0.03, respectively (p<0.001 for all)). By contrast, for imatinib, approved uses increased significantly over off-label (0.97 vs. 0.47 patients, p<0.001). Spending on off-label uses was highest for lidocaine patch and modafinil (>75%). Increases in lidocaine patch use for non-neuropathic pain far exceeded neuropathic pain (10.2 vs. 3.6 patients, p<0.001). Discussion In our sample, three of four top-selling orphan drugs were used more commonly for non-orphan indications. These orphan drugs treated common clinical symptoms (pain and fatigue) or laboratory abnormalities. We should continue to monitor orphan drug use after approval to identify products that come to be widely used for non-FDA approved indications, particularly those without adequate evidence of efficacy.

Empirical Power and Sample Size Calculations for Cluster-Randomized and Cluster-Randomized Crossover Studies

In recent years, the number of studies using a cluster-randomized design has grown dramatically. In addition, the cluster-randomized crossover design has been touted as a methodological advance that can increase efficiency of cluster-randomized studies in certain situations. While the cluster-randomized crossover trial has become a popular tool, standards of design, analysis, reporting and implementation have not been established for this emergent design. We address one particular aspect of cluster-randomized and cluster-randomized crossover trial design: estimating statistical power. We present a general framework for estimating power via simulation in cluster-randomized studies with or without one or more crossover periods. We have implemented this framework in the clusterPower software package for R, freely available online from the Comprehensive R Archive Network. Our simulation framework is easy to implement and users may customize the methods used for data analysis. We give four examples of using the software in practice. The clusterPower package could play an important role in the design of future cluster-randomized and cluster-randomized crossover studies. This work is the first to establish a universal method for calculating power for both cluster-randomized and cluster-randomized clinical trials. More research is needed to develop standardized and recommended methodology for cluster-randomized crossover studies.

No differences in cancer screening rates in patients with rheumatoid arthritis compared to the general population

OBJECTIVE Previous study findings have suggested that patients with chronic diseases such as rheumatoid arthritis (RA) do not receive optimal preventive medical services, including cancer screening tests. This study was undertaken to evaluate cancer screening rates in RA patients compared to non-RA control populations. METHODS Using data from a large US commercial insurance plan, we examined rates of screening tests for cervical, breast, and colon cancer in patients with RA compared to control subjects without RA (non-RA controls) or control subjects with hypertension. Individuals were included in the RA cohort if they had at least 2 visits coded for a diagnosis of RA and had received at least 1 prescription for a disease-modifying antirheumatic drug during the study period. Multivariable Cox proportional hazards models were used to compare the rates of different cancer screening tests between RA patients and non-RA controls. RESULTS RA patients (n = 13,314) and control subjects (non-RA and hypertension controls) (n = 212,324) were screened, on average, once every 3 years for cervical cancer and once every 2 years for breast cancer during the followup period (mean 2.3 years of followup). In the age-adjusted Cox regression model, women with RA were more likely to receive ≥ 1 Papanicolaou smear (hazard ratio [HR] 1.21, 95% confidence interval [95% CI] 1.17-1.24), ≥ 1 mammogram (HR 1.49, 95% CI 1.45-1.53), and ≥ 1 colonoscopy (HR 1.69, 95% CI 1.61-1.77) compared to female non-RA control subjects. Men with RA were also more likely to receive at least 1 colonoscopy (HR 1.52, 95% CI 1.40-1.64) than were male non-RA control subjects. These results were robust in multivariable analyses adjusted for age, number of physician visits, percentage of visits made to primary care physicians, and the Charlson Comorbidity Index. CONCLUSION Patients with RA did not appear to be at risk for receiving fewer cancer screening tests when compared to individuals without RA. The majority of both RA patients and non-RA control subjects were screened regularly for cervical, breast, and colon cancer, in accordance with current recommendations.

Supplementing claims data with outpatient laboratory test results to improve confounding adjustment in effectiveness studies of lipid-lowering treatments

BackgroundAdjusting for laboratory test results may result in better confounding control when added to administrative claims data in the study of treatment effects. However, missing values can arise through several mechanisms.MethodsWe studied the relationship between availability of outpatient lab test results, lab values, and patient and system characteristics in a large healthcare database using LDL, HDL, and HbA1c in a cohort of initiators of statins or Vytorin (ezetimibe & simvastatin) as examples.ResultsAmong 703,484 patients 68% had at least one lab test performed in the 6 months before treatment. Performing an LDL test was negatively associated with several patient characteristics, including recent hospitalization (OR = 0.32, 95% CI: 0.29-0.34), MI (OR = 0.77, 95% CI: 0.69-0.85), or carotid revascularization (OR = 0.37, 95% CI: 0.25-0.53). Patient demographics, diagnoses, and procedures predicted well who would have a lab test performed (AUC = 0.89 to 0.93). Among those with test results available claims data explained only 14% of variation.ConclusionsIn a claims database linked with outpatient lab test results, we found that lab tests are performed selectively corresponding to current treatment guidelines. Poor ability to predict lab values and the high proportion of missingness reduces the added value of lab tests for effectiveness research in this setting.

Comparative Effectiveness of Preventative Therapy for Venous Thromboembolism After Coronary Artery Bypass Graft Surgery

Background— Controversy exists regarding the optimal preventative therapy for venous thromboembolism (VTE) after coronary artery bypass graft (CABG) surgery. We sought to compare the effectiveness and safety of the most commonly used regimens. Methods and Results— We assembled a cohort of 92 699 patients who underwent CABG between 2004 and 2008, using the Premier database. Patients were categorized by method of VTE prevention initiated within 48 hours of surgery, including no preventative therapy (n=55 400), mechanical preventative therapy (n=21 162), subcutaneous unfractio--nated or low-molecular-weight heparin (n=10 718), subcutaneous fondaparinux (n=88), and concurrent mechanical-chemical therapy (n=5331). The incidence of VTE and major bleeding events within 6 weeks of CABG were compared, using multivariable and propensity score adjustment. The overall incidence of VTE for the entire cohort was 0.74%, and the incidence of major bleeding was 1.43%. VTE and bleeding events occurred with similar incidence in each of the patient categories (VTE: 0.70%, 0.79%, 0.81%, 1.14%, and 0.73%; major bleeding: 1.36%, 1.45%, 1.69%, 3.41%, 1.50%; no prevention, mechanical prevention, subcutaneous heparin, subcutaneous fondaparinux, concurrent mechanical-chemical prevention, respectively). Compared with receiving no prevention, the use of mechanical prevention or subcutaneous heparin did not significantly reduce the risk of VTE or change the risk of major bleeding (P=NS). Conclusion— Venous thromboembolism occurs infrequently after CABG. Compared with the use of no prevention, the administration of chemical or mechanical preventative therapies to CABG patients does not appreciably lower the risk of VTE. These data provide support for the common practice of administering no VTE preventative therapy after CABG, used for nearly 60% of patients within this cohort.

Warnings without guidance: patient responses to an FDA warning about ezetimibe.

Background:In January 2008, the Food and Drug Administration (FDA) communicated concerns about the efficacy of ezetimibe, but did not provide clear clinical guidance, and substantial media attention ensued. We investigated the proportion of patients who discontinued therapy and switched to a clinically appropriate alternative after the FDA communication. Methods:Using claims data from a national pharmacy benefits manager, we created a rolling cohort of new users of ezetimibe between January 2006 and August 2008 and created a supply diary for each patient in the year after cohort entry. A patient was identified as nonpersistent if a gap of 90 days was seen in the diary. Using segmented linear regression, we compared rates of nonpersistence before and after the FDA communication and assessed patient-level characteristics associated with discontinuation. Among nonpersistent patients, we determined whether a patient made a clinically appropriate switch in the subsequent 90 days by adding a new cholesterol-lowering medication or by increasing the dose of an existing one. We used a weighted t test to compare the rates of appropriate switching before and after the communication. Results:Among 867,027 new ezetimibe users, 407,006 (46.9%) were nonpersistent in the first year. After the FDA communication, the monthly level of ezetimibe nonpersistence increased by 5.7 percentage points (P<0.0001). Younger patients, those who lived in low-income zip codes, and female patients were less likely to discontinue therapy (P<0.0001 for all). Among nonpersistent patients, rates of clinically appropriate switching increased from 10.8% before to 16.5% after the FDA warning (P=0.004). Conclusions:A substantial increase in ezetimibe nonpersistence rates was seen after an FDA communication regarding its efficacy and following associated media attention, and a small proportion of patients made a clinically appropriate switch after discontinuation. Further consideration is needed to deliver messages that promote appropriate use of chronic therapy rather than simply reduce use.

Comparing treatments via the propensity score: stratification or modeling?

In observational studies of treatments or interventions, propensity score (PS) adjustment is often useful for controlling bias in estimation of treatment effects. Regression on PS is used most often and can be highly efficient, but it can lead to biased results when model assumptions are violated. The validity of stratification on PS depends on fewer model assumptions, but this approach is less efficient than regression adjustment when the regression assumptions hold. To investigate these issues, we compare stratification and regression adjustments in a Monte Carlo simulation study. We consider two stratification approaches: equal frequency strata and an approach that attempts to choose strata that minimize the mean squared error (MSE) of the treatment effect estimate. The regression approach that we consider is a generalized additive model (GAM) that estimates treatment effect controlling for a potentially nonlinear association between PS and outcome. We find that under a wide range of plausible data generating distributions the GAM approach outperforms stratification in treatment effect estimation with respect to bias, variance, and thereby MSE. We illustrate each approach in an analysis of insurance plan choice and its relation to satisfaction with asthma care.