Entertainment & Media

A protocol for active surveillance of acute myocardial infarction in association with the use of a new antidiabetic pharmaceutical agent

A protocol for active surveillance of acute myocardial infarction in association with the use of a new antidiabetic pharmaceutical agent
of 9
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  ORIGINAL REPORT A protocol for active surveillance of acute myocardial infarction inassociation with the use of a new antidiabetic pharmaceutical agent  Bruce Fireman 1 *, Sengwee Toh 2 , Melissa G. Butler  3 , Alan S. Go 1 , Hylton V. Joffe 4 , David J. Graham  4 ,Jennifer C. Nelson 5 , Gregory W. Daniel 6 and Joe V. Selby 1 1  Division of Research, Kaiser Permanente Northern California, Oakland, CA, USA 2  Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, Boston, MA, USA 3 Center for Health Research Southeast, Kaiser Permanente Georgia, Atlanta, GA, USA 4 Center for Drug Evaluation and Research, US Food and Drug Administration, Silver Spring, MD, USA 5 Group Health Research Institute, Group Health Cooperative, Seattle, WA, USA 6 Government and Academic Research, HealthCore, Inc., Alexandria, VA, USA ABSTRACT Purpose  To describe a protocol for active surveillance of acute myocardial infarction (AMI) in users of a recently approved oralantidiabetic 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 inseparate analyses with new users of sitagliptin, pioglitazone, long-acting insulins, and second-generation sulfonylureas. Two approachesto controlling for confounding will be evaluated: matching by exposure propensity score and strati fi cation by AMI risk score. The primaryanalyses will use Cox regression models speci fi ed in a way that does not require pooling of patient-level data from the data partners. The Coxmodels are  fi t to summarized data on risk sets composed of saxagliptin users and similar comparator users at the time of an AMI. Secondaryanalyses will use alternative methods including Poisson regression and will explore whether further adjustment for covariates available onlyat some data partners (e.g., blood pressure) modi fi es 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 alternativeapproaches to controlling for confounding. It is hoped that this initial active surveillance project of the Mini-Sentinel will provide insightsthat inform future population-based surveillance of medical product safety. Copyright © 2012 John Wiley & Sons, Ltd. key words — acute myocardial infarction; diabetes mellitus; drug safety; drug surveillance; saxagliptin INTRODUCTIONOne of the goals of the Sentinel Initiative of the USFood and Drug Administration (FDA) is to develop a capability for conducting active surveillance to evalu-ate the potential risks (or adverse outcomes) associatedwith medical products when in use in the generalpopulation. Acute myocardial infarction (AMI) is a good example of such an outcome and one that hasemerged repeatedly in the past decade as an area of concern. 1 – 3 Thus, in developing the Year 1 workplanfor the Mini-Sentinel pilot  — a component of theSentinel Initiative created to inform and facilitatethe development of a fully operational SentinelSystem  — a high priority was to create a protocolfor active surveillance for drug-induced AMI.Since 2008, FDA has requested that new medica-tions for treatment of Type 2 diabetes mellitus(T2DM) be systematically evaluated for risk of cardio-vascular disease (CVD). 4 The FDA approved a newdipeptidyl-peptidase-4 inhibitor, saxagliptin, for treat-ment of T2DM on 31 July 2009. Preapproval trialsprovided no signal of increased CVD risk with saxa-gliptin. In fact, pooled data suggest that it could bemodestly protective. 5 However, these trials were not prospectively designed to assess CVD risk, had lowreported CVD event rates, and excluded patients at high risk for CVD events from the trial. With helpfrom an advisory committee, FDA determined that  *Correspondence to: B. Fireman, Division of Research, Kaiser Permanente,2000 Broadway, Oakland, CA 94612, USA. E-mail: bruce. fi reman@kp.org Copyright © 2012 John Wiley & Sons, Ltd. pharmacoepidemiology and drug safety  2012;  21 (S1): 282 – 290 Published online in Wiley Online Library (wileyonlinelibrary.com)  DOI : 10.1002/pds.2337  there was adequate evidence of CVD safety to support approval of saxagliptin with a need for additionalassessment after approval. The FDA mandated a randomized, double-blind, controlled postmarket trialevaluating saxagliptin ’ s effect on the incidence of major CVD events. 6 The primary objective of the trialis to establish that the upper bound of the 2-sided 95%con fi dence interval for the estimated risk ratio compar-ing CVD incidence in saxagliptin users versus thecontrol arm is less than 1.3.When FDA reviewed potential new molecular entities for the  fi rst active surveillance activity inMini-Sentinel, saxagliptin provided a good option, inparticular because of the planned postmarket random-ized trial. Comparing results of active population-based surveillance with results of the trial could yieldimportant insights.This article describes the protocol developed by a group of Mini-Sentinel collaborators working together with the FDA in September 2010. 7 Given that this isthe  fi rst active surveillance project conducted withinthe Mini-Sentinel pilot, 8 we describe the deliberationsunderlying the protocol in hopes that they will beuseful in planning future surveillance projects withinthe Sentinel Initiative and elsewhere. In instanceswhere the best methodological approach to active sur-veillance was not apparent, the protocol incorporatedtwo distinct approaches, which we plan to comparefor feasibility and ef  fi ciency during the project. Thefull protocol, including a more detailed presentationof workgroup deliberations, is available online. 7 METHODSSurveillance of a parallel cohort is proposed in whichnew users of saxagliptin are identi fi ed, beginning inAugust of 2009, along with new users of four compar-ator antidiabetic agents and followed longitudinallyfor incidence of AMI. In separate analyses, the experi-ence of new saxagliptin users is compared with that of each comparator after removing prior users of that comparator from the saxagliptin cohort.The new-user design 9,10 is chosen over a mixed-user design (i.e., prevalent as well as new users) for severalreasons. Essentially all saxagliptin users would be newusers because identi fi cation of users begins on the daysaxagliptin became available in the USA. This designtherefore equalizes duration of use at the beginningof follow-up, removing risk of a   “ healthy user  ”  or   “ de-pletion of susceptibles ”  bias, wherein patients who havealready survived early exposure to therapy may be moreprevalent in comparator groups. This design also allowsdetection of early and delayed effects of the drug of in-terest, ensures that all persons included in the analyseswere considered eligible and appropriate for starting a new antidiabetic therapy when follow-up begins, andallows measurement of key patient characteristicsbefore therapy begins.  Data sources New users of saxagliptin and comparator agents, alongwith enrollment, covariate, and endpoint information,will be extracted from at least four data partners of the Mini-Sentinel pilot: HealthCore, Inc.; Humana;The HMO Research Network; and Kaiser Permanente ’ sCenterfor Effectivenessand Safety Research. The Mini-Sentinel data partners have each organized and format-ted their data into the Mini-Sentinel Common Data Model, which is a standardized, distributed data struc-ture that enables data partners to run a single search or analysis program against data stored locally, thus creat-ing the Mini-Sentinel Distributed Database (MSDD).  11 Aggregated follow-up time and events, as will bedescribed, will be submitted to the coordinating center in 2011. Additional transfers will occur as frequentlyas quarterly, depending on the volume of new usersand frequency of data updates through mid-2013.In 2010, the Of  fi ce for Human Research Protections,Department of Health and Human Services, determinedthat the Common Rule does not apply to activitiesconducted under the Sentinel Initiative. FDA has deter-mined that the Mini-Sentinel pilot is part of the SentinelInitiative. Therefore, institutions participating in Mini-Sentinel do not need to obtain review by their institu-tional review boards to participate or provide data for Mini-Sentinel activities. Choice of comparators New users of four comparator antidiabetic agents(sitagliptin, long-acting insulins, pioglitazone, andsecond-generation sulfonylureas (glimepiride, glipizide,and glyburide)) will be identi fi ed, followed, and com-pared in separate analyses to new users of saxagliptin.Multiple comparators were speci fi ed because there aremany treatment alternatives to saxagliptin for patientswith T2DM, whether used as a second- or third-linetherapy. These agents may have differing CVD risk pro fi les, although actual risks and risk differences arecurrently not well de fi ned. Thus, the comparative safetyof saxagliptin versus each comparator is of interest, andno rationale for choosing one primary comparator wasapparent. An additional advantage of multiple compara-tors is the opportunity provided to compare risks amongthese agents. active surveillance of acute myocardial infarction  283 Copyright © 2012 John Wiley & Sons, Ltd.  Pharmacoepidemiology and Drug Safety , 2012;  21 (S1): 282 – 290DOI: 10.1002/pds   Identi  fi cation of the new-user cohort  The cohort will include all eligible patients aged18years and older during the surveillance period. A pre-vious diagnosis of diabetes or a previous prescriptionfor another antidiabetic medication will be requiredto ensure that each patient truly has diabetes. If new use is for insulin, further evidence will be re-quired of a prior prescription for an oral antidiabeticagent to reduce chances of including persons withType 1 diabetes.New dispensings for saxagliptin or a comparator will be sought beginning 1 August 2009. A 1-year period of continuous health plan enrollment andprescription drug coverage (i.e., no more than a 45-daygap in either) is required before the  fi rst new antidiabeticagent dispensing. Relatively high rates of enrolleeturnover within certain health plans preclude requiringlonger prior enrollment. This 1-year follow-back periodwill allow inclusion of some apparent new users whoare not truly new users because of past use before thefollow-back period. This may not be rare amongapparent new users of comparators that have been com-monly used for many years. We will attempt to quantifythe extent of this potential concern by looking for earlier use in persons who meet criteria for new use but havemore than 12months of prior enrollment (i.e., up to10years of data are available in the MSDD).There are few other reasons for exclusion from thenew-user cohort. In analyses of each comparator, anyprior users of saxagliptin would also be excluded.However, prior use of one comparator does not preclude inclusion as a new-user of another compara-tor; over time, patients can become new users of morethan one comparator. In comparisons of saxagliptinwith pioglitazone, patients with any history of chronicheart failure will be excluded from both groupsbecause the condition is a relative contraindicationto pioglitazone and may also predict AMI.The FDA-mandated postmarket trial of saxagliptinis excluding patients who had been discharged withinthe prior 60days for an AMI. 6 We will do the samebecause persons with such a recent history of AMIare at high risk for having a subsequent event duringthe aftermath of the  fi rst event, a period when thelikelihood of getting treatment intensi fi cation is muchgreater and the choice of agents different, with a greater tendency to initiate insulin. Thus, confoundingmay be quantitatively greater shortly after an AMI.Patients with other CVD diagnoses in the 12-monthbaseline period also raise concerns about confounding.A broad range of CVD-related comorbidities areconsidered to be potential complications of T2DM.Their presence suggests greater severity of diseaseand can lead to confounding for the reasons mentionedearlier. There was additional concern that the relation-ships of other confounders to both treatment choicesand AMI risk may differ in the presence of CVD inways that make standard approaches to confounder adjustment inadequate. For these reasons, all analyseswill  fi rst be done separately in subgroups de fi ned bypresence versus absence of CVD. Subsequent com-bined analyses of the two subgroups will be strati fi ed. Cohort follow-up Follow-up begins on the date of the  fi rst   fi lled pre-scription for saxagliptin or comparator and continuesuntil the  fi rst of the following: occurrence of anAMI, fatal or nonfatal; death from any cause; disen-rollment, that is, a gap of greater than 45days incontinuous enrollment or pharmacy bene fi t coverage;a gap in medication possession for the drug of interest that is greater than one third of the days ’  supply at themost recent dispensing (and at least 10days); or theend of the surveillance. For patients followed assaxagliptin users, adding a comparator ends follow-upfor that comparison. For patients followed as new usersof a comparator, adding saxagliptin ends follow-up,although such patients may then qualify as new usersof saxagliptin if there are remaining comparators towhich they have not been exposed.Receipt and review of death certi fi cates in a timelymanner will not be possible in the context of activesurveillance, but any deaths identi fi ed from hospitali-zation summaries or membership/enrollment data willend follow-up. Data partners were unanimous that short enrollment gaps were typically due to adminis-trative errors rather than genuine loss of coverage but that longer gaps often re fl ected true cessation of coverage such that events or prescriptions claimscould be missed. Allowing a gap in medication posses-sion recognizes that individuals on chronic medica-tions occasionally fail to take the medication due toforgetfulness, acute illnesses, or lost medication. Thegap is expressed as a proportion of previous days ’ supply rather than a   fi xed length because typicalprescription sizes varied widely (from 30 to 100days ’ supply) across data partners. Primary analyses will not allow for   “ former use ”  exposure categories or for resumption of follow-up when medication use resumesafter a gap. Concerns that such users may differ from continuous users for unmeasured reasons as well asthe challenges of aggregating such complex exposuredata led to this decision. In secondary analysesplanned at the end of the surveillance (see below), b.  fi reman  et al. 284 Copyright © 2012 John Wiley & Sons, Ltd.  Pharmacoepidemiology and Drug Safety , 2012;  21 (S1): 282 – 290DOI: 10.1002/pds  we will examine the effects of allowing longer gaps inpossession, categories for former use, and resumptionof medication use. Outcome events The primary outcome event is an AMI occurringduring eligible follow-up time. AMI is identi fi ed as a hospital discharge with a primary or principal dischargeInternational Classi fi cation of Diseases, Ninth Revision,Clinical Modi fi cation (ICD-9-CM) code of 410.x0 or 410.x1 or, alternatively, as any record of death occurringwithin 24 h of an emergency department visit for ischemic heart disease (ICD-9-CM codes 410.x0,410.x1, 411.1, 411.8, and 413.x). No length of stayrequirement will be applied for otherwise eligiblehospitalizations. This hospitalization-based de fi nition isbeing validated within Mini-Sentinel using standardizedcriteria. 12 The emergency-department-based de fi nitionis intended primarily to capture deaths from AMI occur-ring in an emergency department that may be missed byhospital discharge claims. It is similar to that employedin a recent study of CVD endpoints in users of rosiglita-zone versus pioglitazone. 13 For two data partners, discharge diagnoses are not classi fi ed as principal versus secondary. However,codes are conventionally listed in order of importance,and prior research efforts from these data partners havetreated the  fi rst-listed code as primary. Dischargecodes of 410.x2 are not included as outcome eventsbecause these speci fi cally refer to prior rather thanacute events. One validation study suggested that including these codes lowers the positive predictivevalue for AMI. 14 Controlling for confounding New users of saxagliptin may differ from those whoinitiate other antidiabetic agents on demographic or clinical characteristics that are also predictive of AMIand therefore confound observed associations of saxagliptin use and AMI. Many such covariates(Table 1), including age, sex, comorbid diagnoses,and use of other medications, can be found in theMSDD. These covariates were selected  a priori  onthe basis of their availability in the MSDD and theplausibility of having an association with risk for AMI. Most are comorbidities or concomitant drugslikely to be prevalent in at least 1% of the diabetespopulation. Prevalent CVD diagnoses and procedureswill identify patients for strati fi cation as having CVD.The number of saxagliptin users and the number of outcomes will be small during the early phases of surveillance, making it dif  fi cult to adjust for such a large number of covariates. Exposure propensityscores (PS) 15 – 17 and disease risk scores (DRS) 18 – 20 have been used to reduce numerous covariates intosingle summary scores that can be used for matching Table 1. Baseline covariates to be used to adjust surveillance analyses for possible confounding*DemographicsAge at   fi rst cohort entrySexResidence or stay in nursing home, skilled nursing facility, or rehabilitationfacility during prior year Utilization measures during baseline year Any hospitalization within prior 30daysAny hospitalization during prior 31 – 365daysAny emergency department visit within prior 30daysAny emergency department visit during prior 31 – 365daysNumber of outpatient visits in prior year Number of unique medications in prior year Comorbid diagnoses during baseline year Obesity (or weight gain)HypertensionHyperlipidemia or lipid disorder Cigarette smoking — based on diagnostic codeOsteoporosisFracture † Renal insuf  fi ciencyEnd-stage renal diseaseHypoglycemia  † Dementia Cancer  { HIV/AIDSAsthma or chronic obstructive pulmonary disease † DepressionPeripheral neuropathyConcurrent medicationsOtherantidiabeticmedications(currentatbaseline,useduringthebaselineyear)Antihypertensive agents (current at baseline, use during the baseline year)Lipid-lowering agents (current at baseline, use during the baseline year)Prior cardiovascular disease diagnoses or procedures } Prior acute myocardial infarction (i.e., > 60days prior)Other ischemic heart disease diagnosis † Coronary revascularization procedures: † coronary artery bypass graft or percutaneous coronary interventionOther heart disease † Stroke † Carotid revascularization procedures: † endarterectomy or carotid bypassLower extremity amputation † Lower extremity revascularization: † endarterectomy, lower extremitybypass, or amputation*Only those diagnostic codes associated with encounters (inpatient or outpatient) will be used. † These diagnoses will be coded distinctly for inpatient stays within prior 30days versus earlier or as outpatient. { Cancer (any cancer code other than non-melanoma skin cancer) } Presence of any of these diagnoses or procedures during baseline periodwill serve to place patient in stratum with CVD. active surveillance of acute myocardial infarction  285 Copyright © 2012 John Wiley & Sons, Ltd.  Pharmacoepidemiology and Drug Safety , 2012;  21 (S1): 282 – 290DOI: 10.1002/pds  or strati fi cation to overcome this challenge. Suchscores also facilitate building and analyzing study-wide datasets from data reduced at the source so that patients are no longer identi fi able through their covariate vector, avoiding the need for transmissionof potentially identi fi able individual-level data. 21 Both PS and DRS offer potential advantages(Table 2). It will be interesting to compare their performance in the context of active surveillance in a distributed data system. PS have the intuitive advan-tage of balancing the populations under assessment in a way that mimics randomized trials. They areeasier to use when evaluating multiple outcomes inrelation to a binary exposure and are advantageouswhen there is more data or externally derived knowl-edge for modeling the exposures of interest. TheDRS is easier to use when comparing multiple typesor levels of exposure with respect to a single outcomeand when there are more data or externally derivedknowledge on outcomes of interest. In the present context, predicting the propensity to use saxagliptinwill require waiting at each data partner until suf  fi cient saxagliptin users (i.e., more than 300) accumulate at a data partner to calculate a robust PS. By contrast, theDRS can be calculated using predictors of AMI duringa baseline period before licensure of the target drug or the start of surveillance. The predictors of AMI arebetter understood and are expected to be more consis-tent across data partners (and practice settings withindata partners) and over time. Nevertheless, it is possiblethat the availability and quality of data on some risk factors (e.g., smoking, hypertension, and hyperlipid-emia) could change over time. Given these considera-tions, we will use and evaluate both PS and DRSapproaches, and neither will be designated as primary.We considered two distinct ways to use the PS andDRS: matching and strati fi cation. Matching (1:1) mini-mizes confounding by restricting comparisons to thesingle best match for each saxagliptin user and facili-tates simple transparent bivariate analyses such as isoften done with data from a randomized trial. However,1:1 matching reduces power by discarding informationfrom users not matched. When there are more compara-tors than saxagliptin users, the matching ratio could intheorybedesignedtobe1:  N  (or   M  :  N  )toincreasepower.However, the advantage of simplicity would be lost unless  N   remains  fi xed across matched sets. This wouldlikely mean that   N   would remain small (1 or 2) becauseat least some saxagliptin users would have only haveone or two acceptable matches and we would bereluctant to discard many saxagliptin users. Strati fi -cation is less burdensome to implement in multisite,sequential surveillance. These considerations led theworking group to propose using both matching (for the PS) and strati fi cation (by deciles) for the DRS. Propensity score estimation and matching.  Program-ming for summarizing the data, deriving the four PS,matching, and evaluating the balance achieved by PSmatching will be written centrally and distributed to data partners.EachsitewillestimatePSlocallyandseparatelywithin subgroups de fi ned by the presence or absence of  Table 2. Comparative advantages of the exposure PS and the DRS for active surveillance of medical productsAdvantages of the PS(1) A single PS readily accommodates analyses of multiple outcomes, whereas separate DRS would be needed unless the two outcomes share predictors completely.(2) Fixed-ratio PS matching facilitates examination of the achieved balance in covariates between comparison groups in the actual population under assessment.(3) PS matching effectively  “ trims ”  away patients in areas of nonoverlap, which has been shown to reduce confounding. 29 (4) If the outcome events are rare, and the target and comparator drugs are commonly used, then the PS strategy is more feasible than is the DRS strategy; it ismore feasible to model — and adjust for  — the confounders ’  associations with the exposure than their associations with the outcome.(5) PS matching (especially if 1:1) permits transparent intuitive analyses as are frequently done in randomized controlled trials.Advantages of the DRS(1) A single DRS readily allows comparisons of event rates between any two or more exposures, whereas PS is typically calculated as pairwise scores for various comparators.(2) A stable DRS can in theory be calculated before surveillance of a new agent begins, assuming that there is a suf  fi cient period of data availability before thesurveillance period and assuming further that key predictors of the outcome do not change much over time in the data, which would require monitoring. Bycontrast, the PS will require accumulation of suf  fi cient numbers of users at a source after use of the new agent begins before stable models can be estimated.(3) Predictors of drug choice are more likely than predictors of the outcome to change over time, requiring periodic recalculation of PS, although associationsof covariates with outcomes will also require monitoring during the surveillance period.(4) The major predictors of certain outcomes (e.g., acute myocardial infarction) are well established and should also be relatively similar across data sources,increasing comfort with pooling data across sources, whereas PS models are likely to be very different across data sources.(5) Because the meaning of the DRS is clearer than that of the PS, comparisons of DRS distributions between drugs of interest, across data sources, or acrosstime may be more readily understood than similar comparisons of PS distributions.(6) Because the DRS re fl ects risk status for the outcome, subgroup analyses of drug safety by DRS level are also of interest and readily interpretable.PS, propensity score; DRS, disease risk score. b.  fi reman  et al. 286 Copyright © 2012 John Wiley & Sons, Ltd.  Pharmacoepidemiology and Drug Safety , 2012;  21 (S1): 282 – 290DOI: 10.1002/pds
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks