You can find them all here.

CER Scan [Epub ahead of print]

Bias in Observational Studies of Prevalent Users: [...]]]> The DEcIDE Methods Center publishes a monthly literature scan of current articles of interest to the field of comparative effectiveness research.

You can find them all here.

February 2012

CER Scan [Epub ahead of print]

1. Am J Epidemiol. 2012 Jan 5. [Epub ahead of print]

Bias in Observational Studies of Prevalent Users: Lessons for Comparative Effectiveness Research From a Meta-Analysis of Statins. Danaei G, Tavakkoli M, Hernán MA.

Randomized clinical trials (RCTs) are usually the preferred strategy with which to generate evidence of comparative effectiveness, but conducting an RCT is not always feasible. Though observational studies and RCTs often provide comparable estimates, the questioning of observational analyses has recently intensified because of randomized-observational discrepancies regarding the effect of postmenopausal hormone replacement therapy on coronary heart disease. Reanalyses of observational data that excluded prevalent users of hormone replacement therapy led to attenuated discrepancies, which begs the question of whether exclusion of prevalent users should be generally recommended. In the current study, the authors evaluated the effect of excluding prevalent users of statins in a meta-analysis of observational studies of persons with cardiovascular disease. The pooled, multivariate-adjusted mortality hazard ratio for statin use was 0.77 (95% confidence interval (CI): 0.65, 0.91) in 4 studies that compared incident users with nonusers, 0.70 (95% CI: 0.64, 0.78) in 13 studies that compared a combination of prevalent and incident users with nonusers, and 0.54 (95% CI: 0.45, 0.66) in 13 studies that compared prevalent users with nonusers. The corresponding hazard ratio from 18 RCTs was 0.84 (95% CI: 0.77, 0.91). It appears that the greater the proportion of prevalent statin users in observational studies, the larger the discrepancy between observational and randomized estimates.
PMID:22223710

CER Scan [published within the last 30 days]

1. J Clin Epidemiol. 2012 Feb;65(2):132-7. Epub 2011 Aug 12.

The "best balance" allocation led to optimal balance in cluster-controlled trials. de Hoop E, Teerenstra S, van Gaal BG, Moerbeek M, Borm GF. Department of Epidemiology, Biostatistics and HTA, 133, Radboud University Nijmegen Medical Centre, PO Box 9101, 6500 HB Nijmegen, The Netherlands.

OBJECTIVE: Balance of prognostic factors between treatment groups is desirable because it improves the accuracy, precision, and credibility of the results. In cluster-controlled trials, imbalance can easily occur by chance when the number of cluster is small. If all clusters are known at the start of the study, the "best balance" allocation method (BB) can be used to obtain optimal balance. This method will be compared with other allocation methods.
STUDY DESIGN AND SETTING: We carried out a simulation study to compare the balance obtained with BB, minimization, unrestricted randomization, and matching for four to 20 clusters and one to five categorical prognostic factors at cluster level.
RESULTS: BB resulted in a better balance than randomization in 13-100% of the situations, in 0-61% for minimization, and in 0-88% for matching. The superior performance of BB increased as the number of clusters and/or the number of factors increased.
CONCLUSION: BB results in a better balance of prognostic factors than randomization, minimization, stratification, and matching in most situations. Furthermore, BB cannot result in a worse balance of prognostic factors than the other methods. Copyright © 2012 Elsevier Inc. All rights reserved.
PMID: 21840173 

2. Clin Pharmacol Ther. 2012 Feb;91(2):165-7. doi: 10.1038/clpt.2011.208.

Challenges in designing comparative-effectiveness trials for antidepressants. Leon AC. Departments of Psychiatry and Public Health, Weill Cornell Medical College, New York, New York, USA.

Comparative-effectiveness antidepressant trials offer promise to provide empirical evidence for clinicians choosing among interventions. Whether such trials posit superiority or noninferiority (NI) hypotheses, they pose formidable challenges. For instance, if meaningful antidepressant differences are seen in comparative-superiority trials, they will be small. NI hypothesis testing, on the other hand, requires an a priori NI margin and evidence of trial assay sensitivity. Either design demands unusually large samples, which could render such trials infeasible.
PMID: 22261683  [PubMed - in process]

FEBRUARY THEME: Selected Methods Manuscripts from the Pharmacoepidemiology and Drug Safety Mini-Sentinel Supplement

1. The U.S. Food and Drug Administration’s Mini-Sentinel program: status and direction (pages 1–8). Platt R, Carnahan RM, Brown JS, Chrischilles E, Curtis LH, Hennessy S, Nelson JC, Racoosin JA, Robb M, Schneeweiss S, Toh S, Weiner MG. Article first published online: 19 JAN 2012 | DOI: 10.1002/pds.2343

The Mini-Sentinel is a pilot program that is developing methods, tools, resources, policies, and procedures to facilitate the use of routinely collected electronic healthcare data to perform active surveillance of the safety of marketed medical products, including drugs, biologics, and medical devices. The U.S. Food and Drug Administration (FDA) initiated the program in 2009 as part of its Sentinel Initiative, in response to a Congressional mandate in the FDA Amendments Act of 2007. After two years, Mini-Sentinel includes 31 academic and private organizations. It has developed policies, procedures, and technical specifications for developing and operating a secure distributed data system comprised of separate data sets that conform to a common data model covering enrollment, demographics, encounters, diagnoses, procedures, and ambulatory dispensing of prescription drugs. The distributed data sets currently include administrative and claims data from 2000 to 2011 for over 300 million person-years, 2.4 billion encounters, 38 million inpatient hospitalizations, and 2.9 billion dispensings. Selected laboratory results and vital signs data recorded after 2005 are also available. There is an active data quality assessment and characterization program, and eligibility for medical care and pharmacy benefits is known. Systematic reviews of the literature have assessed the ability of administrative data to identify health outcomes of interest, and procedures have been developed and tested to obtain, abstract, and adjudicate full-text medical records to validate coded diagnoses. Mini-Sentinel has also created a taxonomy of study designs and analytical approaches for many commonly occurring situations, and it is developing new statistical and epidemiologic methods to address certain gaps in analytic capabilities. Assessments are performed by distributing computer programs that are executed locally by each data partner. The system is in active use by FDA, with the majority of assessments performed using customizable, reusable queries (programs). Prospective and retrospective assessments that use customized protocols are conducted as well. To date, several hundred unique programs have been distributed and executed. Current activities include active surveillance of several drugs and vaccines, expansion of the population, enhancement of the common data model to include additional types of data from electronic health records and registries, development of new methodologic capabilities, and assessment of methods to identify and validate additional health outcomes of interest. Copyright © 2012 John Wiley & Sons, Ltd.

Link to Free PDF: http://onlinelibrary.wiley.com/doi/10.1002/pds.2343/pdf

2. A policy framework for public health uses of electronic health data (pages 18–22). McGraw D, Rosati K, Evans B. Article first published online: 19 JAN 2012 | DOI: 10.1002/pds.2319

Successful implementation of a program of active safety surveillance of drugs and medical products depends on public trust. This article summarizes how the initial pilot phase of the FDA’s Sentinel Initiative, Mini-Sentinel, is being conducted in compliance with applicable federal and state laws. The article also sets forth the attributes of Mini-Sentinel that enhance privacy and public trust, including the use of a distributed data system (where identifiable information remains at the data partners) and the adoption by participants of additional mandatory policies and procedures implementing fair information practices. The authors conclude by discussing the implications of this model for other types of secondary health data uses. Copyright © 2012 John Wiley & Sons, Ltd.

Link to Free PDF: http://onlinelibrary.wiley.com/doi/10.1002/pds.2319/pdf

3. Design considerations, architecture, and use of the Mini-Sentinel distributed data system (pages 23–31). Curtis LH,Weiner MG, Boudreau DM, Cooper WO, Daniel GW, Nair VP, Raebel MA, Beaulieu NU, Rosofsky R, Woodworth TS, Brown JS. Article first published online: 19 JAN 2012 | DOI: 10.1002/pds.2336

Purpose: We describe the design, implementation, and use of a large, multiorganizational distributed database developed to support the Mini-Sentinel Pilot Program of the US Food and Drug Administration (FDA). As envisioned by the US FDA, this implementation will inform and facilitate the development of an active surveillance system for monitoring the safety of medical products (drugs, biologics, and devices) in the USA.
Methods: A common data model was designed to address the priorities of the Mini-Sentinel Pilot and to leverage the experience and data of participating organizations and data partners. A review of existing common data models informed the process. Each participating organization designed a process to extract, transform, and load its source data, applying the common data model to create the Mini-Sentinel Distributed Database. Transformed data were characterized and evaluated using a series of programs developed centrally and executed locally by participating organizations. A secure communications portal was designed to facilitate queries of the Mini-Sentinel Distributed Database and transfer of confidential data, analytic tools were developed to facilitate rapid response to common questions, and distributed querying software was implemented to facilitate rapid querying of summary data.
Results: As of July 2011, information on 99 260 976 health plan members was included in the Mini-Sentinel Distributed Database. The database includes 316 009 067 person-years of observation time, with members contributing, on average, 27.0 months of observation time. All data partners have successfully executed distributed code and returned findings to the Mini-Sentinel Operations Center.
Conclusion: This work demonstrates the feasibility of building a large, multiorganizational distributed data system in which organizations retain possession of their data that are used in an active surveillance system. Copyright © 2012 John Wiley & Sons, Ltd.

Link to Free PDF: http://onlinelibrary.wiley.com/doi/10.1002/pds.2336/pdf

4. Using high-dimensional propensity scores to automate confounding control in a distributed medical product safety surveillance system (pages 41–49). Rassen JA, Schneeweiss S. Article first published online: 19 JAN 2012 | DOI: 10.1002/pds.2328

Distributed medical product safety monitoring systems such as the Sentinel System, to be developed as a part of Food and Drug Administration’s Sentinel Initiative, will require automation of large parts of the safety evaluation process to achieve the necessary speed and scale at reasonable cost without sacrificing validity. Although certain functions will require investigator intervention, confounding control is one area that can largely be automated. The high-dimensional propensity score (hd-PS) algorithm is one option for automated confounding control in longitudinal healthcare databases. In this article, we discuss the use of hd-PS for automating confounding control in sequential database cohort studies, as applied to safety monitoring systems. In particular, we discuss the robustness of the covariate selection process, the potential for over- or under-selection of variables including the possibilities of M-bias and Z-bias, the computation requirements, the practical considerations in a federated database network, and the cases where automated confounding adjustment may not function optimally. We also outline recent improvements to the algorithm and show how the algorithm has performed in several published studies. We conclude that despite certain limitations, hd-PS offers substantial advantages over non-automated alternatives in active product safety monitoring systems. Copyright © 2012 John Wiley & Sons, Ltd.

Link to Free PDF: http://onlinelibrary.wiley.com/doi/10.1002/pds.2328/pdf

5. When should case-only designs be used for safety monitoring of medical products? (pages 50–61). Maclure M, Fireman B, Nelson JC, Hua W, Shoaibi A, Paredes A, Madigan D. Article first published online: 19 JAN 2012 | DOI: 10.1002/pds.2330

Purpose: To assess case-only designs for surveillance with administrative databases.
Methods: We reviewed literature on two designs that are observational analogs to crossover experiments: the self-controlled case series (SCCS) and the case-crossover (CCO) design.
Results: SCCS views the ‘experiment’ prospectively, comparing outcome risks in windows with different exposures. CCO retrospectively compares exposure frequencies in case and control windows. The main strength of case-only designs is they entail self-controlled analyses that eliminate confounding and selection bias by time-invariant characteristics not recorded in healthcare databases. They also protect privacy and are computationally efficient, as they require fewer subjects and variables. They are better than cohort designs for investigating transient effects of accurately recorded preventive agents, for example, vaccines. They are problematic if timing of self-administration is sporadic and dissociated from dispensing times, for example, analgesics. They tend to have less exposure misclassification bias and time-varying confounding if exposures are brief. Standard SCCS designs are bidirectional (using time both before and after the first exposure event), so they are more susceptible than CCOs to reverse-causality bias, including immortal-time bias. This is true also for sequence symmetry analysis, a simplified SCCS. Unidirectional CCOs use only time before the outcome, so they are less affected by reverse causality but susceptible to exposure-trend bias. Modifications of SCCS and CCO partially deal with these biases. The head-to-head comparison of multiple products helps to control residual biases.
Conclusion: The case-only analyses of intermittent users complement the cohort analyses of prolonged users because their different biases compensate for one another. Copyright © 2012 John Wiley & Sons, Ltd.

Link to Free PDF: http://onlinelibrary.wiley.com/doi/10.1002/pds.2330/pdf

6. Challenges in the design and analysis of sequentially monitored postmarket safety surveillance evaluations using electronic observational health care data (pages 62–71). Nelson JC, Cook AJ, Yu O, Dominguez C, Zhao S, Greene SK, Fireman BH, Jacobsen SJ, Weintraub ES, Jackson LA. Article first published online: 19 JAN 2012 | DOI: 10.1002/pds.2324

Purpose: Many challenges arise when conducting a sequentially monitored medical product safety surveillance evaluation using observational electronic data captured during routine care. We review existing sequential approaches for potential use in this setting, including a continuous sequential testing method that has been utilized within the Vaccine Safety Datalink (VSD) and group sequential methods, which are used widely in randomized clinical trials.
Methods: Using both simulated data and preliminary data from an ongoing VSD evaluation, we discuss key sequential design considerations, including sample size and duration of surveillance, shape of the signaling threshold, and frequency of interim testing.
Results and Conclusions: All designs control the overall Type 1 error rate across all tests performed, but each yields different tradeoffs between the probability and timing of true and false positive signals. Designs tailored to monitor efficacy outcomes in clinical trials have been well studied, but less consideration has been given to optimizing design choices for observational safety settings, where the hypotheses, population, prevalence and severity of the outcomes, implications of signaling, and costs of false positive and negative findings are very different. Analytic challenges include confounding, missing and partially accrued data, high misclassification rates for outcomes presumptively defined using diagnostic codes, and unpredictable changes in dynamically accessed data over time (e.g., differential product uptake). Many of these factors influence the variability of the adverse events under evaluation and, in turn, the probability of committing a Type 1 error. Thus, to ensure proper Type 1 error control, planned sequential thresholds should be adjusted over time to account for these issues. Copyright © 2012 John Wiley & Sons, Ltd.

Link to Free PDF: http://onlinelibrary.wiley.com/doi/10.1002/pds.2324/pdf

7. Statistical approaches to group sequential monitoring of postmarket safety surveillance data: current state of the art for use in the Mini-Sentinel pilot (pages 72–81). Cook AJ, Tiwari RC, Wellman RD, Heckbert SR, Li L, Heagerty P, Marsh T, Nelson JC. Article first published online: 19 JAN 2012 | DOI: 10.1002/pds.2320

Purpose: This manuscript describes the current statistical methodology available for active postmarket surveillance of pre-specified safety outcomes using a prospective incident user concurrent control cohort design with existing electronic healthcare data.
Methods: Motivation of the active postmarket surveillance setting is provided using the Food and Drug Administration’s Mini-Sentinel Pilot as an example. Four sequential monitoring statistical methods are presented including the Lan–Demets error spending approach, a matched likelihood ratio test statistic approach with the binomial MaxSPRT as a special case, the conditional sequential sampling procedure with stratification, and a generalized estimating equation regression approach using permutation. Information on the assumptions, limitations, and advantages of each approach is provided, including how each method defines sequential monitoring boundaries, what test statistic is used, and how robust it is to settings of rare events or frequent testing.
Results: A hypothetical example of how the approaches could be applied to data comparing a medical product of interest, drug A, to a concurrent control drug, drug B, is presented including providing the type of information one would have available for monitoring such drugs. Copyright © 2012 John Wiley & Sons, Ltd.

Link to Free PDF: http://onlinelibrary.wiley.com/doi/10.1002/pds.2320/pdf

8. A protocol for active surveillance of acute myocardial infarction in association with the use of a new antidiabetic pharmaceutical agent (pages 282–290). Fireman B, Toh S, Butler MG, Go AS, Joffe HV, Graham DJ, Nelson JC, Daniel GW, Selby JV. Article first published online: 19 JAN 2012 | DOI: 10.1002/pds.2337

Purpose: To describe a protocol for active surveillance of acute myocardial infarction (AMI) in users of a recently approved oral antidiabetic medication, saxagliptin, and to provide the rationale for decisions made in drafting the protocol.
Methods: A new-user cohort design is planned for evaluating data from at least four Mini-Sentinel data partners from 1 August 2009 (following US Food and Drug Administration’s approval of saxagliptin) through mid-2013. New users of saxagliptin will be compared in separate analyses with new users of sitagliptin, pioglitazone, long-acting insulins, and second-generation sulfonylureas. Two approaches to controlling for confounding will be evaluated: matching by exposure propensity score and stratification by AMI risk score. The primary analyses will use Cox regression models specified in a way that does not require pooling of patient-level data from the data partners. The Cox models are fit to summarized data on risk sets composed of saxagliptin users and similar comparator users at the time of an AMI. Secondary analyses will use alternative methods including Poisson regression and will explore whether further adjustment for covariates available only at some data partners (e.g., blood pressure) modifies results.
Results: The results of this study are pending.
Conclusions: The proposed protocol describes a design for surveillance to evaluate the safety of a newly marketed agent as postmarket experience accrues. It uses data from multiple partner organizations without requiring sharing of patient-level data and compares alternative approaches to controlling for confounding. It is hoped that this initial active surveillance project of the Mini-Sentinel will provide insights that inform future population-based surveillance of medical product safety. Copyright © 2012 John Wiley & Sons, Ltd.

Link to Free PDF: http://onlinelibrary.wiley.com/doi/10.1002/pds.2337/pdf

Investigator Examines Path to More Affordable and Effective Drugs

You can find them all here.

January 2012

CER Scan [Epub ahead of print]

Applying propensity scores estimated in [...]]]> The DEcIDE Methods Center publishes a monthly literature scan of current articles of interest to the field of comparative effectiveness research.

You can find them all here.

January 2012

CER Scan [Epub ahead of print]

1. Pharmacoepidemiol Drug Saf. 2011 Dec 8. doi: 10.1002/pds.2256. [Epub ahead of print]

Applying propensity scores estimated in a full cohort to adjust for confounding in subgroup analyses. Rassen JA, Glynn RJ, Rothman KJ, Setoguchi S, Schneeweiss S. Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, USA. jrassen@post.harvard.edu.

BACKGROUND: A correctly specified propensity score (PS) estimated in a cohort ("cohort PS") should, in expectation, remain valid in a subgroup population.
OBJECTIVE: We sought to determine whether using a cohort PS can be validly applied to subgroup analyses and, thus, add efficiency to studies with many subgroups or restricted data. METHODS: In each of three cohort studies, we estimated a cohort PS, defined five subgroups, and then estimated subgroup-specific PSs. We compared difference in treatment effect estimates for subgroup analyses adjusted by cohort PSs versus subgroup-specific PSs. Then, over 10 million times, we simulated a population with known characteristics of confounding, subgroup size, treatment interactions, and treatment effect and again assessed difference in point estimates. RESULTS: We observed that point estimates in most subgroups were substantially similar with the two methods of adjustment. In simulations, the effect estimates differed by a median of 3.4% (interquartile (IQ) range 1.3-10.0%). The IQ range exceeded 10% only in cases where the subgroup had < 1000 patients or few outcome events. CONCLUSIONS: Our empirical and simulation results indicated that using a cohort PS in subgroup analyses was a feasible approach, particularly in larger subgroups. Copyright © 2011 John Wiley & Sons, Ltd.
PMID: 22162077  [PubMed - as supplied by publisher]

2. Stat Methods Med Res. 2011 Nov 8. [Epub ahead of print]

Extension of the modified Poisson regression model to prospective studies with correlated binary data. Zou GY, Donner A. Department of Epidemiology & Biostatistics, and Robarts Clinical Trials of Robarts Research Institute, Schulich School of Medicine & Dentistry, Canada.

The Poisson regression model using a sandwich variance estimator has become a viable alternative to the logistic regression model for the analysis of prospective studies with independent binary outcomes. The primary advantage of this approach is that it readily provides covariate-adjusted risk ratios and associated standard errors. In this article, the model is extended to studies with correlated binary outcomes as arise in longitudinal or cluster randomization studies. The key step involves a cluster-level grouping strategy for the computation of the middle term in the sandwich estimator. For a single binary exposure variable without covariate adjustment, this approach results in risk ratio estimates and standard errors that are identical to those found in the survey sampling literature. Simulation results suggest that it is reliable for studies with correlated binary data, provided the total number of clusters is at least 50. Data from observational and cluster randomized studies are used to illustrate the methods.
PMID: 22072596  [PubMed - as supplied by publisher]

3. J Clin Psychopharmacol. 2011 Dec 22. [Epub ahead of print]

Treating Depression After Initial Treatment Failure: Directly Comparing Switch and Augmenting Strategies in STAR*D. Gaynes BN, Dusetzina SB, Ellis AR, Hansen RA, Farley JF, Miller WC, Stürmer T. Department of Psychiatry, School of Medicine, UNC at Chapel Hill, Chapel Hill, NC; Department of Health Care Policy, Harvard Medical School, Boston, MA; Harrison School of Pharmacy, Auburn University, Auburn, AL; Division of Pharmaceutical Outcomes and Policy, Eshelman School of Pharmacy, and Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC.

OBJECTIVE: Augmenting and switching antidepressant medications are the 2 most common next-step strategies for depressed patients failing initial medication treatment. These approaches have not been directly compared; thus, our objectives are to compare outcomes for medication augmentation versus switching for patients with major depressive disorder in the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) clinical trial. METHODS: We conducted a retrospective analysis of participants aged 18 to 75 years with DSM-IV nonpsychotic depression who failed to remit with initial treatment in the STAR*D clinical trial (N =1292). We compared depressive symptom remission, response, and quality of life among participants in each study arm using propensity score matching to minimize selection bias. RESULTS: The propensity-score-matched augment (N = 269) and switch (N = 269) groups were well balanced on measured characteristics. Neither the likelihood of remission (risk ratio, 1.14; 95% confidence level, 0.82-1.58) or response (risk ratio, 1.14; 95% confidence level, 0.82-1.58), nor the time to remission (log-rank test, P = 0.946) or response (log-rank test, P = 0.243) differed by treatment strategy. Similarly, quality of life did not differ. Post hoc analyses suggested that augmentation improved outcomes for patients tolerating 12 or more weeks of initial treatment and those with partial initial treatment response. CONCLUSIONS: For patients receiving and tolerating aggressive initial antidepressant trials, there is no clear preference for next-step augmentation versus switching. Findings tentatively suggest that those who complete an initial treatment of 12 weeks or more and have a partial response with residual mild depressive severity may benefit more from augmentation relative to switching.
PMID: 22198447  [PubMed - as supplied by publisher]

4. J Clin Psychopharmacol. 2011 Dec 22. [Epub ahead of print]

Variation in Antipsychotic Treatment Choice Across US Nursing Homes. Huybrechts KF, Rothman KJ, Brookhart MA, Silliman RA, Crystal S, Gerhard T, Schneeweiss S. Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School; Department of Epidemiology, Boston University School of Public Health, Boston, MA; RTI Health Solutions, Research Triangle Park; UNC, Gillings School of Global Public Health, Chapel Hill, NC; Department of Medicine, Boston University School of Medicine, Boston, MA; and Rutgers University, New Brunswick, NJ.

OBJECTIVE: Despite serious safety concerns, antipsychotic medications continue to be used widely in US nursing homes. The objective of this study was to quantify the variation in antipsychotic treatment choice across US nursing homes, and to characterize its correlates.
METHODS: Prescribing practices were assessed in a cohort of 65,618 patients 65 years or older in 45 states who initiated treatment with an antipsychotic medication after nursing home admission between 2001 and 2005, using merged Medicaid; Medicare; Minimum Data Set; and Online Survey, Certification, and Reporting data. We fit mixed-effects logistic regression models to examine how antipsychotic treatment choice at the patient-level depends on patient and nursing home fixed and random effects. RESULTS: Among antipsychotic medication users, 9% of patients initiated treatment with a conventional agent. After adjustment for case-mix and facility characteristics, 95% of nursing homes had a predicted conventional antipsychotic prescribing rate between 2% and 20%. Individually, patient characteristics accounted for 36% of the explained variation, facility characteristics for 23%, and nursing home prescribing tendency for 81%. Results were consistent in the subgroup of nursing home patients with a diagnosis of dementia. The prescribing physician was not considered as a determinant of treatment choice owing to data limitations.
CONCLUSION: These findings indicate that antipsychotic treatment choice is to some extent influenced by a nursing home’s underling prescribing "culture." This culture may reveal strategies for targeting quality improvement interventions. In addition, these findings suggest that a nursing home’s tendency for specific antipsychotics merits further exploration as an instrumental variable for improved confounding adjustment in comparative effectiveness studies.
PMID: 22198446  [PubMed - as supplied by publisher]

5. Stat Med. 2011 Dec 4. doi: 10.1002/sim.4413. [Epub ahead of print]

Diagnosing imputation models by applying target analyses to posterior replicates of completed data. He Y, Zaslavsky AM. Department of Health Care Policy, Harvard Medical School, Boston, MA, 02115, USA. he@hcp.med.harvard.edu.

Multiple imputation fills in missing data with posterior predictive draws from imputation models. To assess the adequacy of imputation models, we can compare completed data with their replicates simulated under the imputation model. We apply analyses of substantive interest to both datasets and use posterior predictive checks of the differences of these estimates to quantify the evidence of model inadequacy. We can further integrate out the imputed missing data and their replicates over the completed-data analyses to reduce variance in the comparison. In many cases, the checking procedure can be easily implemented using standard imputation software by treating re-imputations under the model as posterior predictive replicates. Thus, it can be applied for non-Bayesian imputation methods. We also sketch several strategies for applying the method in the context of practical imputation analyses. We illustrate the method using two real data applications and study its property using a simulation. Copyright © 2011 John Wiley & Sons, Ltd.
PMID: 22139814  [PubMed - as supplied by publisher]

CER Scan [published within the last 30 days]

1. Epidemiology. 2012 Jan;23(1):151-8.

Is probabilistic bias analysis approximately bayesian? Maclehose RF, Gustafson P. From the Divisions of Biostatistics, and Epidemiology and Community Health, University of Minnesota, Minneapolis, MN; and Department of Statistics, University of British Columbia, Vancouver, British Columbia, Canada.

Case-control studies are particularly susceptible to differential exposure misclassification when exposure status is determined following incident case status. Probabilistic bias analysis methods have been developed as ways to adjust standard effect estimates based on the sensitivity and specificity of exposure misclassification. The iterative sampling method advocated in probabilistic bias analysis bears a distinct resemblance to a Bayesian adjustment; however, it is not identical. Furthermore, without a formal theoretical framework (Bayesian or frequentist), the results of a probabilistic bias analysis remain somewhat difficult to interpret. We describe, both theoretically and empirically, the extent to which probabilistic bias analysis can be viewed as approximately Bayesian. Although the differences between probabilistic bias analysis and Bayesian approaches to misclassification can be substantial, these situations often involve unrealistic prior specifications and are relatively easy to detect. Outside of these special cases, probabilistic bias analysis and Bayesian approaches to exposure misclassification in case-control studies appear to perform equally well.
PMID: 22157311  [PubMed - in process]

2. BMC Med Inform Decis Mak. 2011 Dec 14;11(1):75. [Epub ahead of print]

Evaluation of an automated safety surveillance system using risk adjusted Sequential Probability Ratio Testing. Matheny ME, Normand SL, Gross TP, Marinac-Dabic D, Loyo-Berrios N, Vidi VD, Donnelly S, Resnic FS.

BACKGROUND: Automated adverse outcome surveillance tools and methods have potential utility in quality improvement and medical product surveillance activities. Their use for assessing hospital performance on the basis of patient outcomes has received little attention. We compared risk-adjusted sequential probability ratio testing (RA-SPRT) implemented in an automated tool to Massachusetts public reports of 30-day mortality after isolated coronary artery bypass graft surgery. METHODS: A total of 23,020 isolated adult coronary artery bypass surgery admissions performed in Massachusetts hospitals between January 1, 2002 and September 30, 2007 were retrospectively re-evaluated. The RA-SPRT method was implemented within an automated surveillance tool to identify hospital outliers in yearly increments. We used an overall type I error rate of 0.05, an overall type II error rate of 0.10, and a threshold that signaled if the odds of dying 30-days after surgery was at least twice than expected. Annual hospital outlier status, based on the state-reported classification, was considered the gold standard. An event was defined as at least one occurrence of a higher-than-expected hospital mortality rate during a given year. RESULTS: We examined a total of 83 hospital-year observations. The RA-SPRT method alerted 6 events among three hospitals for 30-day mortality compared with 5 events among two hospitals using the state public reports, yielding a sensitivity of 100% (5/5) and specificity of 98.8% (79/80). CONCLUSIONS: The automated RA-SPRT method performed well, detecting all of the true institutional outliers with a small false positive alerting rate. Such a system could provide confidential automated notification to local institutions in advance of public reporting providing opportunities for earlier quality improvement interventions.
PMID: 22168892  [PubMed - as supplied by publisher]

Free Full Text: http://www.biomedcentral.com/content/pdf/1472-6947-11-75.pdf

3. Stat Med. 2011 Dec 20;30(29):3447-60. doi: 10.1002/sim.4355.

Gaussian-based routines to impute categorical variables in health surveys. Yucel RM, He Y, Zaslavsky AM. Department of Epidemiology and Biostatistics, School of Public Health, University at Albany, SUNY, One University Place, Rensselaer, NY 12144-3456, USA. ryucel@albany.edu

The multivariate normal (MVN) distribution is arguably the most popular parametric model used in imputation and is available in most software packages (e.g., SAS PROC MI, R package norm). When it is applied to categorical variables as an approximation, practitioners often either apply simple rounding techniques for ordinal variables or create a distinct ‘missing’ category and/or disregard the nominal variable from the imputation phase. All of these practices can potentially lead to biased and/or uninterpretable inferences. In this work, we develop a new rounding methodology calibrated to preserve observed distributions to multiply impute missing categorical covariates. The major attractiveness of this method is its flexibility to use any ‘working’ imputation software, particularly those based on MVN, allowing practitioners to obtain usable imputations with small biases. A simulation study demonstrates the clear advantage of the proposed method in rounding ordinal variables and, in some scenarios, its plausibility in imputing nominal variables. We illustrate our methods on a widely used National Survey of Children with Special Health Care Needs where incomplete values on race posed a valid threat on inferences pertaining to disparities. Copyright © 2011 John Wiley & Sons, Ltd.
PMID: 21976366  [PubMed - in process]

JANUARY THEME: Applications of MSMs for Dealing with Time-varying Exposure

1. Int J Biostat. 2011;7(1):Article 34. Epub 2011 Sep 8.

Antihypertensive medication use and change in kidney function in elderly adults: a marginal structural model analysis. Odden MC, Tager IB, van der Laan MJ, Delaney JA, Peralta CA, Katz R, Sarnak MJ, Psaty BM, Shlipak MG. Oregon State University, USA.

BACKGROUND: The evidence for the effectiveness of antihypertensive medication use for slowing decline in kidney function in older persons is sparse. We addressed this research question by the application of novel methods in a marginal structural model.
METHODS: Change in kidney function was measured by two or more measures of cystatin C in 1,576 hypertensive participants in the Cardiovascular Health Study over 7 years of follow-up (1989-1997 in four U.S. communities). The exposure of interest was antihypertensive medication use. We used a novel estimator in a marginal structural model to account for bias due to confounding and informative censoring.
RESULTS: The mean annual decline in eGFR was 2.41 ± 4.91 mL/min/1.73 m(2). In unadjusted analysis, antihypertensive medication use was not associated with annual change in kidney function. Traditional multivariable regression did not substantially change these estimates. Based on a marginal structural analysis, persons on antihypertensives had slower declines in kidney function; participants had an estimated 0.88 (0.13, 1.63) ml/min/1.73 m(2) per year slower decline in eGFR compared with persons on no treatment. In a model that also accounted for bias due to informative censoring, the estimate for the treatment effect was 2.23
(-0.13, 4.59) ml/min/1.73 m(2) per year slower decline in eGFR.
CONCLUSION: In summary, estimates from a marginal structural model suggested that antihypertensive therapy was associated with preserved kidney function in hypertensive elderly adults. Confirmatory studies may provide power to determine the strength and validity of the findings.
PMCID: PMC3204667 [Available on 2012/9/8]
PMID: 22049266  [PubMed - in process]

2. Epidemiology. 2011 Nov;22(6):877-8.

Hormonal contraception and HIV risk: evaluating marginal-structural-model assumptions. Chen PL, Cole SR, Morrison CS.

Letter to the editor

PMID: 21968782  [PubMed - in process]

3. Pharmacoepidemiol Drug Saf. 2011 Jul 22. doi: 10.1002/pds.2175. [Epub ahead of print] Comparative effectiveness of individual angiotensin receptor blockers on risk of mortality in patients with chronic heart failure. Desai RJ, Ashton CM, Deswal A, Morgan RO, Mehta HB, Chen H, Aparasu RR, Johnson ML. Eshelman School of Pharmacy, University of North Carolina, Chapel Hill, NC, USA.

 

OBJECTIVE: There is little evidence on comparative effectiveness of individual angiotensin receptor blockers (ARBs) in patients with chronic heart failure (CHF). This study compared four ARBs in reducing risk of mortality in clinical practice. METHODS: A retrospective analysis was conducted on a national sample of patients diagnosed with CHF from 1 October 1996 to 30 September 2002 identified from Veterans Affairs electronic medical records, with supplemental clinical data obtained from chart review. After excluding patients with exposure to ARBs within the previous 6 months, four treatment groups were defined based on initial use of candesartan, valsartan, losartan, and irbesartan between the index date (1 October 2000) and the study end date (30 September 2002). Time to death was measured concurrently during that period. A marginal structural model controlled for sociodemographic factors, comorbidities, comedications, disease severity (left ventricular ejection fraction), and potential time-varying confounding affected by previous treatment (hospitalization). Propensity scores derived from a multinomial logistic regression were used as inverse probability of treatment weights in a generalized estimating equation to estimate causal effects. RESULTS: Among the 1536 patients identified on ARB therapy, irbesartan was most frequently used (55.21%), followed by losartan (21.74%), candesartan (15.23%), and valsartan (7.81%). When compared with losartan, after adjusting for time-varying hospitalization in marginal structural model, candesartan (OR = 0.79, 95%CI = 0.42-1.50), irbesartan (OR = 1.17, 95%CI = 0.72-1.90), and valsartan (OR = 0.98, 95%CI = 0.45-2.14) were found to have similar effectiveness in reducing mortality in CHF patients. CONCLUSION: Effectiveness of ARBs in reducing mortality is similar in patients with CHF in everyday clinical practice. Copyright © 2011 John Wiley & Sons, Ltd.
PMID: 21786364  [PubMed - as supplied by publisher]

4. Clin Trials. 2011 Jun;8(3):277-87. doi: 10.1177/1740774511402526.

How to use marginal structural models in randomized trials to estimate the natural direct and indirect effects of therapies mediated by causal intermediates. Oba K, Sato T, Ogihara T, Saruta T, Nakao K. Translational Research and Clinical Trial Center, Hokkaido University Hospital, Hokkaido University, Japan. k.oba@huhp.hokudai.ac.jp

Erratum in
    Clin Trials. 2011;8(5):680.

BACKGROUND: Although intention-to-treat analysis is a standard approach, additional supplemental analyses are often required to evaluate the biological relationship among interventions, intermediates, and outcomes. Therefore, we need to evaluate whether the effect of an intervention on a particular outcome is mediated by a hypothesized intermediate variable.
PURPOSE: To evaluate the size of the direct effect in the total effect, we applied the marginal structural model to estimate the average natural direct and indirect effects in a large-scale randomized controlled trial (RCT). Method The average natural direct effect is defined as the difference in the probability of a counterfactual outcome between the experimental and control arms, with the intermediate set to what it would have been, had the intervention been a control treatment. We considered two marginal structural models to estimate the average natural direct and indirect effects introduced by VanderWeele (Epidemiology 2009) and applied them in a large-scale RCT – the Candesartan Antihypertensive Survival
Evaluation in Japan (CASE-J trial) – that compared angiotensin receptor blockers and calcium-channel blockers in high-risk hypertensive patients.
RESULTS: There were no strong blood pressure-independent or dependent effects; however, a systolic blood pressure reduction of about 1.9  mmHg suppressed all events. Compared to the blood pressure-independent effects of calcium channel blockers, those of angiotensin receptor blockers contributed positively to cardiovascular and cardiac events, but negatively to cerebrovascular events.
LIMITATIONS: There is a particular condition for estimating the average natural direct effect. It is impossible to check whether this condition is satisfied with the available data.
CONCLUSION: We estimated the average natural direct and indirect effects through the achieved systolic blood pressure in the CASE-J trial. This first application of estimating the average natural effects in an RCT can be useful for obtaining an in-depth understanding of the results and further development of similar interventions.
PMID: 21730076  [PubMed - indexed for MEDLINE]

5. J Consult Clin Psychol. 2011 Apr;79(2):225-35. A marginal structural model analysis for loneliness: implications for intervention trials and clinical practice. VanderWeele TJ, Hawkley LC, Thisted RA, Cacioppo JT. Harvard University, Department of Epidemiology, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115, USA. tvanderw@hsph.harvard.edu

 

OBJECTIVE: Clinical scientists, policymakers, and individuals must make decisions concerning effective interventions that address health-related issues. We use longitudinal data on loneliness and depressive symptoms and a new class of causal models to illustrate how empirical evidence can be used to inform intervention trial design and clinical practice.
METHOD: Data were obtained from a population-based study of non-Hispanic Caucasians, African Americans, and Latino Americans (N = 229) born between 1935 and 1952. Loneliness and depressive symptoms were measured with the UCLA Loneliness Scale-Revised and Center for Epidemiologic Studies Depression Scale, respectively. Marginal structural causal models were employed to evaluate the extent to which depressive symptoms depend not only on loneliness measured at a single point in time (as in prior studies of the effect of loneliness) but also on an individual’s entire loneliness history.
RESULTS: Our results indicate that if interventions to reduce loneliness by 1 standard deviation were made 1 and 2 years prior to assessing depressive symptoms, both would have an effect; together, they would result in an average reduction in depressive symptoms of 0.33 standard deviations, 95% CI [0.21,
0.44], p < .0001.
CONCLUSIONS: The magnitude and persistence of these effects suggest that greater effort should be devoted to developing practical interventions on alleviating loneliness and that doing so could be useful in the treatment and prevention of depressive symptoms. In light of the persistence of the effects of loneliness, our results also suggest that, in the evaluation of interventions on loneliness, it may be important to allow for a considerable follow-up period in assessing outcomes.
(c) 2011 APA, all rights reserved.
PMCID: PMC3079447 [Available on 2012/4/1]
PMID: 21443322  [PubMed - indexed for MEDLINE]

6. J Clin Psychopharmacol. 2011 Apr;31(2):226-30.

Differential 3-year effects of first- versus second-generation antipsychotics on subjective well-being in schizophrenia using marginal structural models. Lambert M, Schimmelmann BG, Schacht A, Suarez D, Haro JM, Novick D, Wagner T, Wehmeier PM, Huber CG, Hundemer HP, Dittmann RW, Naber D. Psychosis Centre, Department of Psychiatry and Psychotherapy, Centre for Psychosocial Medicine, University Medical Centre Hamburg-Eppendorf, Germany.

OBJECTIVE: This study examined the differential effects of first- (FGAs) versus second-generation antipsychotics (SGAs) on subjective well-being in patients with schizophrenia.
METHOD: Data were collected in an observational 3-year follow-up study of 2224 patients with schizophrenia. Subjective well-being was assessed with the Subjective Well-being under Neuroleptic Treatment Scale (SWN-K). Differential effects of FGAs versus SGAs were analyzed using marginal structural models in those patients taking antipsychotic monotherapy.
RESULTS: The marginal structural model, which analyzed the differential effect on the SWN-K total score, revealed that patients on SGAs had significantly higher SWN-K total scores, starting at 6 months (3.02 points; P = 0.0061, d = 0.20) and remaining significant thereafter (end point: 5.35 points; P = 0.0074, d = 0.36).
CONCLUSIONS: Results of this large observational study are consistent with a small but clinically relevant superiority of SGAs over FGAs in subjective well-being extending previous positive findings of differential effects on quality of life.
PMID: 21346606  [PubMed - indexed for MEDLINE]

7. Arch Intern Med. 2011 Jan 24;171(2):110-8. Epub 2010 Sep 27.

Similar outcomes with hemodialysis and peritoneal dialysis in patients with end-stage renal disease. Mehrotra R, Chiu YW, Kalantar-Zadeh K, Bargman J, Vonesh E. Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center, Torrance, CA 90502, USA. rmehrotra@labiomed.org

Comment in
    Arch Intern Med. 2011 Jan 24;171(2):107-9.

BACKGROUND: The annual payer costs for patients treated with peritoneal dialysis (PD) are lower than with hemodialysis (HD), but in 2007, only 7% of dialysis patients in the United States were treated with PD. Since 1996, there has been no change in the first-year mortality of HD patients, but both short- and long-term outcomes of PD patients have improved.
METHODS: Data from the US Renal Data System were examined for secular trends in survival among patients treated with HD and PD on day 90 of end-stage renal disease (HD, 620 020 patients; PD, 64 406 patients) in three 3-year cohorts (1996-1998, 1999-2001, and 2002-2004) for up to 5 years of follow-up using a nonproportional hazards marginal structural model with inverse probability of treatment and censoring weighting.
RESULTS: There was a progressive attenuation in the higher risk for death seen in patients treated with PD in earlier cohorts; for the 2002-2004 cohort, there was no significant difference in the risk of death for HD and PD patients through 5 years of follow-up. The median life expectancy of HD and PD patients was 38.4 and 36.6 months, respectively. Analyses in 8 subgroups based on age (<65 and ≥65 years), diabetic status, and baseline comorbidity (none and ≥1) showed greater improvement in survival among patients treated with PD relative to HD at all follow-up periods.
CONCLUSION: In the most recent cohorts, patients who began treatment with HD or PD have similar outcomes.
PMID: 20876398  [PubMed - indexed for MEDLINE]

8. Epidemiology. 2010 Jul;21(4):528-39.

Estimating absolute risks in the presence of nonadherence: an application to a follow-up study with baseline randomization. Toh S, Hernández-Díaz S, Logan R, Robins JM, Hernán MA. Department of Epidemiology, Harvard School of Public Health, Boston, MA 02215

The intention-to-treat (ITT) analysis provides a valid test of the null hypothesis and naturally results in both absolute and relative measures of risk. However, this analytic approach may miss the occurrence of serious adverse effects that would have been detected under full adherence to the assigned treatment. Inverse probability weighting of marginal structural models has been used to adjust for nonadherence, but most studies have provided only relative measures of risk. In this study, we used inverse probability weighting to estimate both absolute and relative measures of risk of invasive breast cancer under full adherence to the assigned treatment in the Women’s Health Initiative estrogen-plus-progestin trial. In contrast to an ITT hazard ratio (HR) of 1.25 (95% confidence interval [CI] = 1.01 to 1.54), the HR for 8-year continuous estrogen-plus-progestin use versus no use was 1.68 (1.24 to 2.28). The estimated risk difference (cases/100 women) at year 8 was 0.83 (-0.03 to 1.69) in the ITT analysis, compared with 1.44 (0.52 to 2.37) in the adherence-adjusted analysis. Results were robust across various dose-response models. We also compared the dynamic treatment regimen "take hormone therapy until certain adverse events become apparent, then stop taking hormone therapy" with no use (HR = 1.64; 95% CI
= 1.24 to 2.18). The methods described here are also applicable to observational studies with time-varying treatments.
PMID: 20526200  [PubMed - indexed for MEDLINE]

9. Lifetime Data Anal. 2010 Jan;16(1):71-84. Epub 2009 Nov 6.

Relation between three classes of structural models for the effect of a time-varying exposure on survival. Young JG, Hernán MA, Picciotto S, Robins JM. Department of Epidemiology, Harvard School of Public Health, 677 Huntington Avenue, Kresge Bldg Suite 820, Boston, MA 02115, USA. jyoung@hsph.harvard.edu

Standard methods for estimating the effect of a time-varying exposure on survival may be biased in the presence of time-dependent confounders themselves affected by prior exposure. This problem can be overcome by inverse probability weighted estimation of Marginal Structural Cox Models (Cox MSM), g-estimation of Structural Nested Accelerated Failure Time Models (SNAFTM) and g-estimation of
Structural Nested Cumulative Failure Time Models (SNCFTM). In this paper, we describe a data generation mechanism that approximately satisfies a Cox MSM, an SNAFTM and an SNCFTM. Besides providing a procedure for data simulation, our formal description of a data generation mechanism that satisfies all three models allows one to assess the relative advantages and disadvantages of each modeling approach. A simulation study is also presented to compare effect estimates across the three models.
PMID: 19894116  [PubMed - indexed for MEDLINE]

10. J Rheumatol. 2009 Mar;36(3):560-4. Epub 2009 Feb 4.

Prednisone, lupus activity, and permanent organ damage. Thamer M, Hernán MA, Zhang Y, Cotter D, Petri M. Medical Technology and Practice Patterns Institute, Bethesda, MD 20814

OBJECTIVE: To estimate the effect of corticosteroids (prednisone dose) on permanent organ damage among persons with systemic lupus erythematosus (SLE). METHODS: We identified 525 patients with incident SLE in the Hopkins Lupus Cohort. At each visit, clinical activity indices, laboratory data, and treatment were recorded. The study population was followed from the month after the first visit until June 29, 2006, or attainment of irreversible organ damage, death, loss to follow-up, or receipt of pulse methylprednisolone therapy. We estimated the effect of cumulative average dose of prednisone on organ damage using a marginal structural model to adjust for time-dependent confounding by indication due to SLE disease activity.
RESULTS: Compared with non-prednisone use, the hazard ratio of organ damage for prednisone was 1.16 (95% CI 0.54, 2.50) for cumulative average doses > 0-180 mg/month, 1.50 (95% CI 0.58, 3.88) for > 180-360 mg/month, 1.64 (95% CI 0.58, 4.69) for > 360-540 mg/month, and 2.51 (95% CI 0.87, 7.27) for > 540 mg/month. In contrast, standard Cox regression models estimated higher hazard ratios at all dose levels.
CONCLUSION: Our results suggest that low doses of prednisone do not result in a substantially increased risk of irreversible organ damage.
PMID: 19208608  [PubMed - indexed for MEDLINE]

Learning about the Safety [...]]]> Jerry Avorn, MD, offers his perspective on the evolution of drug safety monitoring in the US in the December 8th, 2011 issue of the New England Journal of Medicine. In addition to his written perspective, Dr. Avorn also gave an interview with NEJM on the same topic.

Learning about the Safety of Drugs — A Half-Century of Evolution

You can find them all here.

CER Scan [Epub ahead of print]

Identifying Adverse Events of Vaccines Using a [...]]]> The DEcIDE Methods Center publishes a monthly literature scan of current articles of interest to the field of comparative effectiveness research.

You can find them all here.

December 2011

CER Scan [Epub ahead of print]

1. Drug Saf. 2011 Jan 2012; 35(1):61-78 [Epub ahead of print]

Identifying Adverse Events of Vaccines Using a Bayesian Method of Medically Guided Information Sharing. Crooks CJ, Prieto-Merino D, Evans SJ. Division of Epidemiology and Public Health, University of Nottingham, Nottingham, UK.

Background: The detection of adverse events following immunization (AEFI) fundamentally depends on how these events are classified. Standard methods impose a choice between either grouping similar events together to gain power or splitting them into more specific definitions. We demonstrate a method of medically guided Bayesian information sharing that avoids grouping or splitting the data, and we further combine this with the standard epidemiological tools of stratification and multivariate regression. Objective: The aim of this study was to assess the ability of a Bayesian hierarchical model to identify gastrointestinal AEFI in children, and then combine this with testing for effect modification and adjustments for confounding. Study Design: Reporting odds ratios were calculated for each gastrointestinal AEFI and vaccine combination. After testing for effect modification, these were then re-estimated using multivariable logistic regression adjusting for age, sex, year and country of report. A medically guided hierarchy of AEFI terms was then derived to allow information sharing in a Bayesian model. Setting: All spontaneous reports of AEFI in children under 18 years of age in the WHO VigiBase™ (Uppsala Monitoring Centre, Uppsala, Sweden) before June 2010. Reports with missing age were included in the main analysis in a separate category and excluded in a subsequent sensitivity analysis. Exposures: The 15 most commonly prescribed childhood vaccinations, excluding influenza vaccines. Main Outcome Measures: All gastrointestinal AEFI coded by WHO Adverse Reaction Terminology. Results: A crude analysis identified 132 signals from 655 reported combinations of gastrointestinal AEFI. Adjusting for confounding by age, sex, year of report and country of report, where appropriate, reduced the number of signals identified to 88. The addition of a Bayesian hierarchical model identified four further signals and removed three. Effect modification by age and sex was identified for six vaccines for the outcomes of vomiting, nausea, diarrhoea and salivary gland enlargement.
Conclusion: This study demonstrated a sequence of methods for routinely analysing spontaneous report databases that was easily understandable and reproducible. The combination of classical and Bayesian methods in this study help to focus the limited resources for hypothesis testing studies towards adverse events with the strongest support from the data.
PMID: 22136183 [PubMed - as supplied by publisher]

CER Scan [published within the last 30 days]

1. Am J Epidemiol. 2011 Dec 1;174(11):1213-22. Epub 2011 Oct 24.

Effects of adjusting for instrumental variables on bias and precision of effect estimates.
Myers JA, Rassen JA, Gagne JJ, Huybrechts KF, Schneeweiss S, Rothman KJ, Joffe MM, Glynn RJ.

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.
PMID: 22025356  [PubMed - in process]

2. Am J Epidemiol. 2011 Dec 1;174(11):1223-7. Epub 2011 Oct 27.

Invited commentary: understanding bias amplification. Pearl J.

In choosing covariates for adjustment or inclusion in propensity score analysis, researchers must weigh the benefit of reducing confounding bias carried by those covariates against the risk of amplifying residual bias carried by unmeasured confounders. The latter is characteristic of covariates that act like instrumental variables-that is, variables that are more strongly associated with the exposure than with the outcome. In this issue of the Journal (Am J Epidemiol. 2011;174(11):1213-1222), Myers et al. compare the bias amplification of a near-instrumental variable with its bias-reducing potential and suggest that, in practice, the latter outweighs the former. The author of this commentary sheds broader light on this comparison by considering the cumulative effects of conditioning on multiple covariates and showing that bias amplification may build up at a faster rate than bias reduction. The author further derives a partial order on sets of covariates which reveals preference for conditioning on outcome-related, rather than exposure-related, confounders.
PMCID: PMC3224255 [Available on 2012/12/1] PMID: 22034488 [PubMed - in process]

3. Am J Epidemiol. 2011 Dec 1;174(11):1228-9. Epub 2011 Oct 24. Myers et Al. Response to "understanding bias amplification". Myers JA, Rassen JA, Gagne JJ, Huybrechts KF, Schneeweiss S, Rothman KJ, Glynn RJ.

 

Response to Invited Commentary
PMID: 22025355  [PubMed - in process]

4. Epidemiology. 2011 Nov;22(6):815-22.

Estimating bias from loss to follow-up in the Danish National Birth Cohort. Greene N, Greenland S, Olsen J, Nohr EA. Department of Epidemiology, School of Public Health, University of California

Loss to follow-up in cohort studies may result in biased association estimates. Of 61,895 women entering the Danish National Birth Cohort and completing the first data-collection phase, 37,178 (60%) opted to be in the 7-year follow-up. Using national registry data to obtain end point information on all members of the cohort, we estimated associations in the baseline and the 7-year follow-up participant populations for 5 exposure-outcome associations: (a) size at birth and childhood asthma, (b) assisted reproductive treatment and childhood hospitalizations, (c) prepregnancy body mass index and childhood infections, (d) alcohol drinking in early pregnancy and childhood developmental disorders, and (e) maternal smoking in pregnancy and childhood attention-deficit hyperactivity disorder (ADHD). We estimated follow-up bias in the odds or rate ratios by calculating relative ratios. For all but one of the above analyses, the bias appeared to be small, between -10% and +8%. For maternal smoking in pregnancy and childhood ADHD, we estimated a positive bias of approximately 33% (95% bootstrap limits of -30% and +152%). The presence and magnitude of bias due to loss to follow-up depended on the nature of the factors or outcomes examined, with the most pronounced contribution in this study coming from maternal smoking. Our methods and results may inform bias analyses in future pregnancy cohort studies.
PMID: 21918455  [PubMed - in process]

DECEMBER THEME: Methods for Addressing Missing Data in CER

1. Stat Med. 2011 Dec 4. doi: 10.1002/sim.4413. [Epub ahead of print]

Diagnosing imputation models by applying target analyses to posterior replicates of completed data. He Y, Zaslavsky AM. Department of Health Care Policy, Harvard Medical School, Boston, MA, 02115, USA. he@hcp.med.harvard.edu.

Multiple imputation fills in missing data with posterior predictive draws from imputation models. To assess the adequacy of imputation models, we can compare completed data with their replicates simulated under the imputation model. We apply analyses of substantive interest to both datasets and use posterior predictive checks of the differences of these estimates to quantify the evidence of model inadequacy. We can further integrate out the imputed missing data and their replicates over the completed-data analyses to reduce variance in the comparison. In many cases, the checking procedure can be easily implemented using standard imputation software by treating re-imputations under the model as posterior predictive replicates. Thus, it can be applied for non-Bayesian imputation methods. We also sketch several strategies for applying the method in the context of practical imputation analyses. We illustrate the method using two real data applications and study its property using a simulation. Copyright © 2011 John Wiley & Sons, Ltd. Copyright © 2011 John Wiley & Sons, Ltd.
PMID: 22139814  [PubMed - as supplied by publisher]

2. Stat Methods Med Res. 2011 Mar 23. [Epub ahead of print]

Using causal diagrams to guide analysis in missing data problems. Daniel RM, Kenward MG, Cousens SN, De Stavola BL. Faculty of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, London WC1E 7HT, UK.

Estimating causal effects from incomplete data requires additional and inherently untestable assumptions regarding the mechanism giving rise to the missing data. We show that using causal diagrams to represent these additional assumptions both complements and clarifies some of the central issues in missing data theory, such as Rubin’s classification of missingness mechanisms (as missing completely at random (MCAR), missing at random (MAR) or missing not at random (MNAR)) and the circumstances in which causal effects can be estimated without bias by analysing only the subjects with complete data. In doing so, we formally extend the back-door criterion of Pearl and others for use in incomplete data examples. These ideas are illustrated with an example drawn from an occupational cohort study of the effect of cosmic radiation on skin cancer incidence.
PMID: 21389091  [PubMed - as supplied by publisher]

3. Stat Med. 2011 Mar 15;30(6):627-41. doi: 10.1002/sim.4124. Epub 2010 Dec 28.

Estimating propensity scores with missing covariate data using general location mixture models. Mitra R, Reiter JP. School of Mathematics, University of Southampton, Southampton, SO17 1BJ, U.K. R.Mitra@soton.ac.uk

In many observational studies, analysts estimate causal effects using propensity scores, e.g. by matching, sub-classifying, or inverse probability weighting based on the scores. Estimation of propensity scores is complicated when some values of the covariates are missing. Analysts can use multiple imputation to create completed data sets from which propensity scores can be estimated. We propose a general location mixture model for imputations that assumes that the control units are a latent mixture of (i) units whose covariates are drawn from the same distributions as the treated units’ covariates and (ii) units whose covariates are drawn from different distributions. This formulation reduces the influence of control units outside the treated units’ region of the covariate space on the estimation of parameters in the imputation model, which can result in more plausible imputations. In turn, this can result in more reliable estimates of propensity scores and better balance in the true covariate distributions when matching or sub-classifying. We illustrate the benefits of the latent class modeling approach with simulations and with an observational study of the effect of breast feeding on children’s cognitive abilities. Copyright © 2010 John Wiley & Sons, Ltd.
PMID: 21337358  [PubMed - indexed for MEDLINE]

4. Am J Epidemiol. 2010 Nov 1;172(9):1070-6. Epub 2010 Sep 14.

Multiple imputation for missing data via sequential regression trees. Burgette LF, Reiter JP. Department of Statistical Science, Duke University, Durham, North Carolina 27708.

Multiple imputation is particularly well suited to deal with missing data in large epidemiologic studies, because typically these studies support a wide range of analyses by many data users. Some of these analyses may involve complex modeling, including interactions and nonlinear relations. Identifying such relations and encoding them in imputation models, for example, in the conditional regressions for multiple imputation via chained equations, can be daunting tasks with large numbers of categorical and continuous variables. The authors present a nonparametric approach for implementing multiple imputation via chained equations by using sequential regression trees as the conditional models. This has the potential to capture complex relations with minimal tuning by the data imputer. Using simulations, the authors demonstrate that the method can result in more plausible imputations, and hence more reliable inferences, in complex settings than the naive application of standard sequential regression imputation techniques. They apply the approach to impute missing values in data on adverse birth outcomes with more than 100 clinical and survey variables. They evaluate the imputations using posterior predictive checks with several epidemiologic analyses of interest.
PMID: 20841346  [PubMed - indexed for MEDLINE]

Free Full Text: http://aje.oxfordjournals.org/content/172/9/1070.long

5. Artif Intell Med. 2010 Oct;50(2):105-15. Epub 2010 Jul 16.

Missing data imputation using statistical and machine learning methods in a real breast cancer problem. Jerez JM, Molina I, García-Laencina PJ, Alba E, Ribelles N, Martín M, Franco L. Departamento de Lenguajes y Ciencias de la Computación, Universidad de Málaga, E.T.S.I. Informática, Campus de Teatinos s/n, 29071 Málaga, Spain. jja@lcc.uma.es

OBJECTIVES: Missing data imputation is an important task in cases where it is crucial to use all available data and not discard records with missing values. This work evaluates the performance of several statistical and machine learning imputation methods that were used to predict recurrence in patients in an extensive real breast cancer data set.
MATERIALS AND METHODS: Imputation methods based on statistical techniques, e.g., mean, hot-deck and multiple imputation, and machine learning techniques, e.g., multi-layer perceptron (MLP), self-organisation maps (SOM) and k-nearest neighbour (KNN), were applied to data collected through the "El Álamo-I" project, and the results were then compared to those obtained from the listwise deletion
(LD) imputation method. The database includes demographic, therapeutic and recurrence-survival information from 3679 women with operable invasive breast cancer diagnosed in 32 different hospitals belonging to the Spanish Breast Cancer Research Group (GEICAM). The accuracies of predictions on early cancer relapse were measured using artificial neural networks (ANNs), in which different ANNs were estimated using the data sets with imputed missing values.
RESULTS: The imputation methods based on machine learning algorithms outperformed imputation statistical methods in the prediction of patient outcome. Friedman’s test revealed a significant difference (p=0.0091) in the observed area under the ROC curve (AUC) values, and the pairwise comparison test showed that the AUCs for MLP, KNN and SOM were significantly higher (p=0.0053, p=0.0048 and p=0.0071, respectively) than the AUC from the LD-based prognosis model.
CONCLUSION: The methods based on machine learning techniques were the most suited for the imputation of missing values and led to a significant enhancement of prognosis accuracy compared to imputation methods based on statistical procedures.
Copyright © 2010 Elsevier B.V. All rights reserved.
PMID: 20638252  [PubMed - indexed for MEDLINE]

6. J Clin Epidemiol. 2010 Jul;63(7):728-36. Epub 2010 Mar 25.

Unpredictable bias when using the missing indicator method or complete case analysis for missing confounder values: an empirical example. Knol MJ, Janssen KJ, Donders AR, Egberts AC, Heerdink ER, Grobbee DE, Moons KG, Geerlings MI. Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Str. 6.131, PO Box 85500, 3508 GA Utrecht, The Netherlands. m.j.knol@umcutrecht.nl

OBJECTIVE: Missing indicator method (MIM) and complete case analysis (CC) are frequently used to handle missing confounder data. Using empirical data, we demonstrated the degree and direction of bias in the effect estimate when using these methods compared with multiple imputation (MI).
STUDY DESIGN AND SETTING: From a cohort study, we selected an exposure (marital status), outcome (depression), and confounders (age, sex, and income). Missing values in "income" were created according to different patterns of missingness: missing values were created completely at random and depending on exposure and outcome values. Percentages of missing values ranged from 2.5% to 30%.
RESULTS: When missing values were completely random, MIM gave an overestimation of the odds ratio, whereas CC and MI gave unbiased results. MIM and CC gave under- or overestimations when missing values depended on observed values. Magnitude and direction of bias depended on how the missing values were related to exposure and outcome. Bias increased with increasing percentage of missing
values.
CONCLUSION: MIM should not be used in handling missing confounder data because it gives unpredictable bias of the odds ratio even with small percentages of missing values. CC can be used when missing values are completely random, but it gives loss of statistical power.
Copyright 2010 Elsevier Inc. All rights reserved.
PMID: 20346625  [PubMed - indexed for MEDLINE]

7. Pharmacoepidemiol Drug Saf. 2010 Jun;19(6):618-26.

Issues in multiple imputation of missing data for large general practice clinical databases. Marston L, Carpenter JR, Walters KR, Morris RW, Nazareth I, Petersen I. Department of Primary Care and Population Health, University College London, Rowland Hill Street, London NW32PF

PURPOSE: Missing data are a substantial problem in clinical databases. This paper aims to examine patterns of missing data in a primary care database, compare this to nationally representative datasets and explore the use of multiple imputation (MI) for these data.
METHODS: The patterns and extent of missing health indicators in a UK primary care database (THIN) were quantified using 488 384 patients aged 16 or over in their first year after registration with a GP from 354 General Practices. MI models were developed and the resulting data compared to that from nationally representative datasets (14 142 participants aged 16 or over from the Health Survey for England 2006 (HSE) and 4 252 men from the British Regional Heart Study (BRHS)).
RESULTS: Between 22% (smoking) and 38% (height) of health indicator data were missing in newly registered patients, 2004-2006. Distributions of height, weight and blood pressure were comparable to HSE and BRHS, but alcohol and smoking were not. After MI the percentage of smokers and non-drinkers was higher in THIN than the comparison datasets, while the percentage of ex-smokers and heavy drinkers was lower. Height, weight and blood pressure remained similar to the comparison datasets.
CONCLUSIONS: Given available data, the results are consistent with smoking and alcohol data missing not at random whereas height, weight and blood pressure missing at random. Further research is required on suitable imputation methods for smoking and alcohol in such databases.
PMID: 20306452  [PubMed - indexed for MEDLINE]

8. Circ Cardiovasc Qual Outcomes. 2010 Jan;3(1):98-105.

Missing data analysis using multiple imputation: getting to the heart of the matter. He Y. Department of Health Care Policy, Harvard Medical School

Missing data are a pervasive problem in health investigations. We describe some background of missing data analysis and criticize ad hoc methods that are prone to serious problems. We then focus on multiple imputation, in which missing cases are first filled in by several sets of plausible values to create multiple completed datasets, then standard complete-data procedures are applied to each completed dataset, and finally the multiple sets of results are combined to yield a single inference. We introduce the basic concepts and general methodology and provide some guidance for application. For illustration, we use a study assessing the effect of cardiovascular diseases on hospice discussion for late stage lung cancer patients.
PMCID: PMC2818781; PMID: 20123676 [PubMed - indexed for MEDLINE]

Free PDF: http://circoutcomes.ahajournals.org/content/3/1/98.full.pdf+html

9. Am J Epidemiol. 2010 Mar 1;171(5):624-32. Epub 2010 Jan 27.

Multiple imputation for missing data: fully conditional specification versus multivariate normal imputation. Lee KJ, Carlin JB. Clinical Epidemiology and Biostatistics Unit, Murdoch Childrens Research Institute, Royal Children’s Hospital, Flemington Road, Parkville, Victoria

Statistical analysis in epidemiologic studies is often hindered by missing data, and multiple imputation is increasingly being used to handle this problem. In a simulation study, the authors compared 2 methods for imputation that are widely available in standard software: fully conditional specification (FCS) or "chained equations" and multivariate normal imputation (MVNI). The authors created data sets of 1,000 observations to simulate a cohort study, and missing data were induced under 3 missing-data mechanisms. Imputations were performed using FCS (Royston’s "ice") and MVNI (Schafer’s NORM) in Stata (Stata Corporation, College Station, Texas), with transformations or prediction matching being used to manage nonnormality in the continuous variables. Inferences for a set of regression parameters were compared between these approaches and a complete-case analysis. As expected, both FCS and MVNI were generally less biased than complete-case analysis, and both produced similar results despite the presence of binary and ordinal variables that clearly did not follow a normal distribution. Ignoring
skewness in a continuous covariate led to large biases and poor coverage for the corresponding regression parameter under both approaches, although inferences for other parameters were largely unaffected. These results provide reassurance that similar results can be expected from FCS and MVNI in a standard regression analysis involving variously scaled variables.
PMID: 20106935  [PubMed - indexed for MEDLINE]

Free Full Text: http://aje.oxfordjournals.org/content/171/5/624.long

10. J Sch Psychol. 2010 Feb;48(1):5-37.

An introduction to modern missing data analyses. Baraldi AN, Enders CK. Arizona State University, USA. Amanda.Baraldi@asu.edu

A great deal of recent methodological research has focused on two modern missing data analysis methods: maximum likelihood and multiple imputation. These approaches are advantageous to traditional techniques (e.g. deletion and mean imputation techniques) because they require less stringent assumptions and mitigate the pitfalls of traditional techniques. This article explains the theoretical underpinnings of missing data analyses, gives an overview of traditional missing data techniques, and provides accessible descriptions of maximum likelihood and multiple imputation. In particular, this article focuses on maximum likelihood estimation and presents two analysis examples from the Longitudinal Study of American Youth data. One of these examples includes a description of the use of auxiliary variables. Finally, the paper illustrates ways that researchers can use intentional, or planned, missing data to enhance their research designs.
PMID: 20006986  [PubMed - indexed for MEDLINE]

11. Int J Epidemiol. 2010 Feb;39(1):118-28. Epub 2009 Oct 25.

Modelling relative survival in the presence of incomplete data: a tutorial. Nur U, Shack LG, Rachet B, Carpenter JR, Coleman MP. Cancer Research UK Cancer Survival Group, London School of Hygiene and Tropical Medicine, London, UK. ula.nur@lshtm.ac.uk

BACKGROUND: Missing data frequently create problems in the analysis of population-based data sets, such as those collected by cancer registries. Restriction of analysis to records with complete data may yield inferences that are substantially different from those that would have been obtained had no data been missing. ‘Naive’ methods for handling missing data, such as restriction of the analysis to complete records or creation of a ‘missing’ category, have drawbacks that can invalidate the conclusions from the analysis. We offer a tutorial on modern methods for handling missing data in relative survival analysis.
METHODS: We estimated relative survival for 29 563 colorectal cancer patients who were diagnosed between 1997 and 2004 and registered in the North West Cancer Intelligence Service. The method of multiple imputation (MI) was applied to account for the common example of incomplete stage at diagnosis, under the missing at random (MAR) assumption. Multivariable regression with a generalized linear model and Poisson error structure was then used to estimate the excess hazard of death of the colorectal cancer patients, over and above the background mortality, adjusting for significant predictors of mortality.
RESULTS: Incomplete information on stage, morphology and grade meant that only 55% of the data could be included in the ‘complete-case’ analysis. All cases could be included after indicator method (IM) or MI method. Handling missing data by MI produced a significantly lower estimate of the excess mortality for stage, morphology and grade, with the largest reductions occurring for late-stage and high-grade tumours, when compared with the results of complete-case analysis.
CONCLUSION: In complete-case analysis, almost 50% of the information could not be included, and with the IM, all records with missing values for stage were combined into a single ‘missing’ category. We show that MI methods greatly improved the results by exploiting all the information in the incomplete records. This method also helped to ensure efficient inferences about survival were made from the multivariate regression analyses.
PMID: 19858106  [PubMed - indexed for MEDLINE]

Free Full Text: http://ije.oxfordjournals.org/cgi/pmidlookup?view=long&pmid=19858106

Watch the interview with Dr. Avorn