Search

To compare the risk of incident hypertension between initiators of tumor necrosis factor (TNF)-α inhibitors and initiators of nonbiologic disease modifying antirheumatic drugs (hereafter referred to as nonbiologics) in rheumatoid arthritis patients taking methotrexate monotherapy.We conducted a cohort study using insurance claims data (2001-2012) from the US. We identified initiators of use of either TNF-α inhibitors or nonbiologics. Subsequent exposure to these agents was measured monthly in a time-varying manner. The outcome of interest was incident hypertension, defined by a diagnosis and a prescription for an antihypertensive drug. Marginal structural models estimated hazard ratios (HRs) adjusted for both baseline and time-varying confounders. To validate the primary analysis, we designed a verification analysis to evaluate a known association between leflunomide (a nonbiologic disease modifying agent) and hypertension.We identified 4,822 initiations of TNF-α inhibitor use and 2,400 of nonbiologic use. Crude incidence rates of hypertension per 1,000 person-years of follow-up were 36 (95% CI [confidence interval]: 32, 41) for the TNF-α inhibitor group and 42 (95% CI: 34, 51) for the nonbiologics group. The crude HR of TNF-α inhibitors versus nonbiologics for the risk of incident hypertension was 0.85 (95% CI: 0.67, 1.1). After adjusting for both baseline and time-varying covariates using marginal structural models, the HR was 0.95 (95% CI: 0.74, 1.2). In the verification analysis, the adjusted HR of incident hypertension was 2.3 (95% CI: 1.7, 3.0) in leflunomide initiators compared with methotrexate initiators.Treatment with TNF-α inhibitors was not associated with a reduced risk of incident hypertension compared with nonbiologics in rheumatoid arthritis patients.See Video Abstract at http://links.lww.com/EDE/B36.

In nonrandomized studies of comparative effectiveness of medications, the prescriber may be the most important determinant of treatment assignment, yet the majority of analyses ignore the prescriber. Via Monte Carlo simulation, we evaluated the bias of 3 approaches that utilize the prescriber in analysis compared against the default approach that ignores the prescriber. Prescriber preference instrumental variable (IV) analyses were unbiased when IV criteria were met, which required no clustering of unmeasured patient characteristics within prescriber. In all other scenarios, IV analyses were highly biased, and stratification on the prescriber reduced confounding bias at the patient or prescriber levels. Including a prescriber random intercept in the propensity score model reversed the direction of confounding from measured patient factors and resulted in unpredictable changes in bias. Therefore, we recommend caution when using the IV approach, particularly when the instrument is weak. Stratification on the prescriber may be more robust; this approach warrants additional research.

Background Propensity scores are useful for confounding adjustment in the commonly observed setting of many potential confounders, frequent exposure, and rare events. However, with few exposed outcomes to inform covariate selection and many candidate confounders, optimal approaches to construct and implement propensity-score-based confounding adjustment remain unclear. Methods In a cohort study on the effect of anticonvulsant drugs on cardiovascular risk among adult patients from the HealthCore Integrated Research Database, we compared the performance for confounding control of various covariate-selection strategies for propensity-score estimation (expert knowledge only, expert knowledge informed by empirical covariate selection via high-dimensional propensity-score, and high-dimensional propensity-score empirical specification only) and propensity-score-based adjustment methods (propensity-score-matching and propensity-score-decile stratification). This article focuses on the first 90 days of follow-up because any treatment effect identified in this temporal window almost certainty originates from residual confounding rather than pharmacologic action. Results We identified 166,031 new users and 564 ischemic cardiovascular events. Among those, 12,580 patients initiated anticonvulsants that strongly induce cytochrome P450 enzymes and experienced 68 events. The unadjusted hazard ratio was 1.72 (95% confidence interval = 1.34-2.22). Adjustment for investigator-identified covariates led to 41% to 59% reductions in the hazard ratio; adjustment for both investigator-identified and high-dimensional propensity-score empirically identified covariates led to larger reductions (54% to 72%). A selection strategy based on high-dimensional propensity-score empirical specification alone produced less-attenuated and more-volatile hazard ratio estimates. This volatility seemed to be slightly attenuated in a trimmed propensity-score-stratified analysis. Conclusions The high-dimensional propensity-score algorithm complements expert knowledge for confounding adjustment, but in settings with few exposed outcomes, its performance without investigator-specified covariates is less clear and may be associated with an increased likelihood of bias. In our example, investigator specification of variables combined with high-dimensional propensity-score empirical selection and the use of trimmed propensity-score-stratified analysis seem to improve effect estimation. Plotting the relation of effect estimates to the increasing number of empirical covariates is a useful diagnostic.

BACKGROUND Nonrandomized safety and effectiveness studies are often initiated immediately after the approval of a new medication, but patients prescribed the new medication during this period may be substantially different from those receiving an existing comparator treatment. Restricting the study to comparable patients after data have been collected is inefficient in prospective studies with primary collection of outcomes. METHODS We discuss design and methods for evaluating covariate data to assess the comparability of treatment groups, identify patient subgroups that are not comparable, and decide when to transition to a large-scale comparative study. We demonstrate methods in an example study comparing Cox-2 inhibitors during their postmarketing period (1999-2005) with nonselective nonsteroidal anti-inflammatory drugs (NSAIDs). RESULTS Graphical checks of propensity score distributions in each treatment group showed substantial problems with overlap in the initial cohorts. In the first half of 1999, >40% of patients were in the region of nonoverlap on the propensity score, and across the study period this fraction never dropped below 10% (the a priori decision threshold for transitioning to the large-scale study). After restricting to patients with no prior NSAID use

Adjusting for large numbers of covariates ascertained from patients' health care claims data may improve control of confounding, as these variables may collectively be proxies for unobserved factors. Here, we develop and test an algorithm that empirically identifies candidate covariates, prioritizes covariates, and integrates them into a propensity-score-based confounder adjustment model.We developed a multistep algorithm to implement high-dimensional proxy adjustment in claims data. Steps include (1) identifying data dimensions, eg, diagnoses, procedures, and medications; (2) empirically identifying candidate covariates; (3) assessing recurrence of codes; (4) prioritizing covariates; (5) selecting covariates for adjustment; (6) estimating the exposure propensity score; and (7) estimating an outcome model. This algorithm was tested in Medicare claims data, including a study on the effect of Cox-2 inhibitors on reduced gastric toxicity compared with nonselective nonsteroidal anti-inflammatory drugs (NSAIDs).In a population of 49,653 new users of Cox-2 inhibitors or nonselective NSAIDs, a crude relative risk (RR) for upper GI toxicity (RR = 1.09 [95% confidence interval = 0.91-1.30]) was initially observed. Adjusting for 15 predefined covariates resulted in a possible gastroprotective effect (0.94 [0.78-1.12]). A gastroprotective effect became stronger when adjusting for an additional 500 algorithm-derived covariates (0.88 [0.73-1.06]). Results of a study on the effect of statin on reduced mortality were similar. Using the algorithm adjustment confirmed a null finding between influenza vaccination and hip fracture (1.02 [0.85-1.21]).In typical pharmacoepidemiologic studies, the proposed high-dimensional propensity score resulted in improved effect estimates compared with adjustment limited to predefined covariates, when benchmarked against results expected from randomized trials.

A risk ratio or difference from a meta-analysis is as many as ten steps away from the unobservable causal risk ratios and differences in target populations. The steps are like lenses, filters, or other fallible components of the epidemiologist's "telescope" for observing populations. Each step is another domain where different biases can be caused. How biases combine across domains in the production of epidemiologic evidence can be quickly explained to nonepidemiologists by using a sequence of causal arrow diagrams with easy notation: (a) agent of interest, (b) background risk factors, (c) correlated causes, (d) diagnosis, (e) exposure measurement, (f) filing of data, (g) grouping of cohorts, (h) harvesting of cases and controls, (i) interpretations of investigators, (j) judgments of journals, and (k) knowledge of meta-analysts. For epidemiologists, this article serves as a review of ideas about confounding, information bias, and selection bias and underscores the need for routinely analyzing the sensitivity of study findings to multiple hypothesized biases.

Epidemiologic studies of the effects of drugs are essential. Increasingly, these studies are based on large administra tive data files containing longitudinal drug exposure and clinical outcome data for entire populations. These data allow for rapid, retrospective evaluation of the intended and adverse effects of prescription medications as they are used in routine practice. Health care claims data, however, do not contain measurements of many important risk factors that physicians use to make prescribing decisions. Confounding by unmea sured indication is a threat to the validity of most pharmaco epidemiologic studies but is likely to be particularly severe for studies based on administrative data.' Instrumental variable (IV) methods present a possible solution to the problem of residual confounding, provided suitable instruments can be identified. For an IV to reduce confounding in the initial treatment assignment, it should influence the prescribing decision but have no direct effect on the health outcome under study. We have proposed an instru ment related to physician prescribing preference for use in database studies of short-term drug effects.2 In this issue of EPIDEMIOLOGY, Hernan and Robins3 give a comprehensive overview of IV methods and provide a detailed commentary on our analysis. We would like to review briefly the central assumptions of our study, discuss an additional concern raised by Hernan and Robins, and elaborate on further aspects of this analysis and the IV method in the context of empirical pharmacoepidemiologic research with secondary data. In our study, we assessed the risk of gastrointestinal (GI) toxicity associated with selective versus nonselective nonsteroidal anti-inflammatory drug (NSAID) exposure using a physician's NSAID preference (as measured by the last NSAID prescription written by a physician) as an instrument. When treatment effects are homogeneous, the internal valid ity of our approach relies principally on 3 assumptions: (1) physicians differ in their NSAID preference; (2) a physician's NSAID prescribing preference as reflected in the last NSAID prescription he has written is unrelated to unmeasured risk factors in his current patient; and (3) a physician's NSAID prescribing preference as reflected in the last prescription written has no direct effect on the outcome in the current patient (known as the exclusion restriction). As stated,2 these assumptions are not verifiable with data, but their plausibility can be evaluated in a variety of ways. Hernan and Robins3 raise the important point that if the effect of treatment is heterogeneous (eg, if selective NSAIDs are protective in some patients and hazardous in others), then an additional assumption is required to justify the use of the usual IV estimator. One such assumption states that there should be no additive effect modification by the instrument among the treated patients (and similarly among the untreated patients). Although the possibility of effect modification by the instrument seems remote, Hema'n and Robins make the interesting observation that this assumption can be violated if the instrument interacts with important unmeasured GI risk factors (eg, tobacco use, alcohol consumption, obesity) to determine treatment. For example, this could happen if phy sicians who prefer nonselective NSAIDs rely strongly on a patient's smoking status to make treatment decisions, whereas selective NSAID preferring physicians decide on treatment without much consideration of smoking status. Such interactions could also lead to a violation of the "mono tonicity" assumption of Imbens and Angrist4 5-an alterna tive assumption that can be used to justify the use of the usual IV estimator when there is suspected treatment effect heter ogeneity. The assumption that the instrument does not interact with unmeasured GI risk factors to determine treatment is not empirically testable. However, its plausibility can be partly evaluated by exploring whether or not the instrument inter acts with measured GI risk factors to determine treatment. In our study, there were no such interactions, leading us to suspect that strong interactions with the unmeasured GI risk factors are unlikely. Although we do not think that treatment effect heterogeneity is a first-order source of bias in our study of NSAIDs, the issue is subtle and deserves further research. Our intention-to-treat approach has intrinsic limitations independent of the issues of IVs. Because many patients who initiate NSAIDs discontinue them by 60 days, the effect of treatment assignment might be different from the effect of uninterupted exposure. Furthermore, discontinuation may be more likely for nonselective NSAID users, leading to a bias against selective NSAIDs. We reduced this bias by focusing on short-term effects. Alternative "as-treated" longitudinal ap proaches suffer from confounding by informative treatment changes that occur during follow-up (for example, unrecorded GI symptoms that lead to the discontinuation of nonselective NSAIDs). The approach we adopted estimates the short-term effect of a treatment assignment that occurred in a routine care From the Division of Pharmacoepidemiology, Harvard Medical School, Brigham and Women's Hospital, Boston, MA. Correspondence: M. Alan Brookhart, Harvard Medical School, Brigham and Women's Hospital, 1620 Tremont St., Suite 3030, Boston, MA 02120. E-mail: abrookhart@rics.bwh.harvard.edu Copyright C 2006 by Lippincott Williams & Wilkins ISSN: 1044-3983/06/1704-0373 DOI: 10.1097/01.ede.0000222026.42077.ee