How Safe Are Recently FDA-Approved Antimicrobials? A

Review of the FDA Adverse Event Reporting System

Database

Tina M. Khadem,1,2* Robbert P. van Manen,3 and Jack Brown1,2

1St. John Fisher College, Wegmans School of Pharmacy, Rochester, New York; 2University of Rochester Medical

Center, Department of Pharmacy, Rochester, New York; 3Oracle Health Sciences Global Business Unit

Kattendijke, The Netherlands

STUDY OBJECTIVE To review quantitatively and qualitatively the U.S. Food and Drug Administration(FDA) Adverse Event Reporting System (AERS) database to provide clinicians with a general under-standing of the comparative occurrence of clinically meaningful adverse events associated with 15antimicrobial new molecular entities approved by the FDA since 2006: anidulafungin, darunavir,maraviroc, raltegravir, doripenem, telavancin, ceftaroline, boceprevir, telaprevir, fidaxomicin, bedaq-uiline, dolutegravir, simeprevir, sofosbuvir, and dalbavancin.

DESIGN Retrospective analysis.DATA SOURCE FDA AERS database.MEASUREMENTS AND MAIN RESULTS Empirica Signal software was used to query the AERS database from

November 1968 to December 2012. Using disproportionality analyses, we calculated a relativereporting ratio (RRR) estimate for reports of antimicrobial adverse events. The RRR estimate com-pares the occurrence of a specific adverse event with an index drug of interest to the occurrenceof the same adverse event with similar agents or with all other FDA-approved prescription drugs.Common industry practice considers an RRR meaningful if the 5th percentile of the distributionis at least 2 (RRR05 of 2.0 or higher). Antimicrobials were compared with agents within theirrespective antimicrobial therapeutic class as well as with all agents in the AERS database. Seven-teen adverse signals with an RRR05 of 2.0 or higher were identified from the database for sixagents. Ten of the 17 signals were not included in the most up-to-date manufacturers’ packageinserts for four of the six agents: doripenem-associated hepatic dysfunction (RRR05 3.7) and hy-perchloremia (RRR05 2.6); boceprevir-associated weight loss (RRR05 2.2); darunavir-associated pre-mature labor (RRR05 3.1), sudden infant death syndrome (RRR05 2.9), ventricular hypertrophy(RRR05 2.7), acute coronary syndrome (RRR05 2.4), and congenital anomaly in offspring (RRR05

2.4); and raltegravir-associated congenital heart valve disorders (RRR05 2.5) and SIDS (RRR05

2.3).CONCLUSION Clinically meaningful adverse event signals appeared to be associated with antimicrobial

new molecular entities approved since 2006 including many not yet identified in package inserts.Although a disproportionality analysis suggests a quantitative signal for these associations, causalitycannot be inferred from the data. Due to several key limitations in this type of analysis, investigativestudies are needed to further explore these adverse event signals and the potential mechanisms bywhich they occur.

KEY WORDS antimicrobials, adverse events, spontaneous reporting, disproportionality analysis.(Pharmacotherapy 2014;**(**):**–**) doi: 10.1002/phar.1519

B R I E F R E P O R T

Since 2006 the U.S. Food and Drug Adminis-tration (FDA) has approved 15 antimicrobialnew molecular entities: anidulafungin, daruna-vir, maraviroc, raltegravir, doripenem, telavan-cin, ceftaroline, boceprevir, telaprevir,fidaxomicin, bedaquiline, dolutegravir, simepre-vir, sofosbuvir, and dalbavancin (Table 1).1

Although randomized controlled trials are con-sidered to be the gold standard methodology tostudy the safety and efficacy of a drug, these tri-als are often limited by small sample size andshort-term patient observation periods.2 As aresult, premarketing clinical trials are often onlyable to identify the most commonly occurringand acute adverse events and not able to addressall potential safety concerns with a particulardrug. In addition, strict inclusion and exclusioncriteria limit the study population to primarilyhealthier individuals. Although this may increaseinternal validity of the trial, it is at the cost ofdecreasing external validity. Databases for spon-taneous reporting of adverse reactions such asthe FDA Adverse Event Reporting System(AERS) have been useful in addressing some ofthese limitations by identifying relatively rareevents and providing safety signals, particularlywith respect to newly approved drugs.The FDA AERS database is a powerful com-

puterized drug safety tool that combines the vol-untary adverse drug reaction reports fromMedWatch and the required reports from manu-facturers.3 These reports are often the initialsource of safety signals and provide incentive foradditional studies to investigate potential seriousadverse events. The FDA defines a safety signalas a disproportionality, specifically “an excess ofadverse events compared to what would beexpected to be associated with a product’s use.”4

Disproportionality analyses may be a useful sta-tistical approach in overcoming the most com-mon challenges associated with usingspontaneous reporting systems such as AERSincluding chronic underreporting and the lackof accurate quantitative exposure data to putsuch rare events into perspective.5 These meth-

ods focus on the relative occurrence of drug-event combinations compared with an expectedvalue derived from overall patterns of reporting,rather than attempting to determine the accuraterates of occurrence of these drug-event combina-tions.5 With the use of empirical Bayesiancorrection methodology, a disproportionalityanalysis produces a relative reporting ratio(RRR) estimate that compares the occurrence ofa specific adverse event with an index drug ofinterest to the occurrence of the same adverseevent with similar agents or with all other FDA-approved prescription drugs.5

Historically, ~3% of FDA-approved antimicro-bials are removed from the U.S. market due tosafety concerns.1, 6 Our goal was to qualitativelyand quantitatively review the FDA AERS data-base to provide clinicians with a general under-standing of the comparative occurrence ofclinically important adverse events seen in anti-microbial new molecular entities recentlyapproved by the FDA.

Methods

In this review of the FDA AERS database,Empirica Signal software (Oracle Corp., Red-wood Shores, CA) was used to query all adverseevent–drug combinations reported from Novem-ber 1968 to December 2012. With the use ofdisproportionality analysis and empirical Bayes-ian correction methodology, we calculated theRRR for all adverse events associated with anti-microbial new molecular entities approved since2006 (Figure 1). Consistent with commonindustry practice, disproportionality signals wereconsidered clinically meaningful if the 5th per-centile of the distribution was at least 2 (RRR05

of 2.0 or higher), which indicates a correctedoccurrence of the adverse event with the indexdrug that was double that of other agents.7

Wherever available, clinical information relatedto each case of an antimicrobial-reported adverseevent was reviewed.

Results

The most recently approved new molecularentities (bedaquiline, dolutegravir, simeprevir,sofosbuvir, and dalbavancin) were not repre-sented in the analysis at the time of query dueto the delay in the release of the public database.Of the remaining antimicrobials, 17 adverse sig-nals with an RRR05 of 2.0 or higher were identi-fied from the database for six agents. Ten of the

Financial support: None.Platform presentation previously presented at the 53rd

Interscience Conference on Antimicrobial Agents and Che-motherapy; September 9–13, 2013; Denver, CO. AbstractK-1544.

*Address for correspondence: Tina Khadem, Universityof Rochester Medical Center, 601 Elmwood Ave, Box 638,Rochester, NY 14642; e-mail: tina_khadem@urmc.roches-ter.edu.� 2014 Pharmacotherapy Publications, Inc.

2 PHARMACOTHERAPY Volume **, Number **, 2014

17 signals were not included in the most up-to-date manufacturer package inserts of four of thesix agents (Table 2). These include doripenem-associated hepatic dysfunction (RRR05 3.7) andhyperchloremia (RRR05 2.6); boceprevir-associ-ated weight loss (RRR05 2.2); darunavir-associ-ated premature labor (RRR05 3.1), sudden infantdeath syndrome (SIDS) (RRR05 2.9), ventricularhypertrophy (RRR05 2.7), acute coronary syn-drome (RRR05 2.4), and congenital anomaly inoffspring (RRR05 2.4); and raltegravir-associated

congenital heart valve disorders (RRR05 2.5) andSIDS (RRR05 2.3). Adverse event signals identi-fied but already listed in the most up-to-datemanufacturers’ package inserts included telavan-cin-associated acute renal failure (RRR05 2.6),doripenem-associated seizures (RRR05 3.6) andthrombocytopenia (RRR05 2.9), boceprevir-asso-ciated anemia (RRR05 2.8), maraviroc-associatedmalignancies (Hodgkin disease RRR05 7.5, rectalcancer RRR05 2.6, and squamous cell carcinomaRRR05 2.4) and myocardial infarction (RRR05

2.1), and raltegravir-associated cytolytic hepatitis(RRR05 2.4).

Discussion

Based on our analysis, 10 clinically meaning-ful adverse event signals were identified fromthe AERS database. Of note, these are statisticalsignals that have been identified, and they arenot observed adverse events. To our knowledge,none of these types of adverse events have previ-ously been reported in the most up-to-date man-ufacturer package inserts. Only one of theseevents was identified when conducting a searchof the PubMed database (doripenem and hepaticdysfunction).8 Furthermore, the type and sever-ity of many of these events is alarming. Daruna-vir signals such as premature labor, SIDS,ventricular hypertrophy, and congenital anoma-lies in offspring, and raltegravir signals such ascongenital heart valve disorders and SIDS havenot been previously reported with other agentsin their respective drug classes of protease inhib-itors and integrase inhibitors, and they warrantfurther investigation. The potential mechanismof these adverse events is unknown. Darunaviris a protease inhibitor used for the treatment ofhuman immunodeficiency virus infection. Com-mon adverse reactions include hypercholesterol-emia and gastrointestinal effects such as nausea,vomiting, and diarrhea. Less common adverseeffects include headache, fatigue, and dermato-logic and hepatic reactions. A plethora of post-marketing and case reports of adverse reactions

Equation example: Frequency of the adverse event of interest for a given agent

Frequency of all adverse events for a given agent RRR =

Frequency of the adverse event of interest for all other agents within the therapeutic drug class

Frequency of all adverse events for all other agents within the therapeutic drug class

Figure 1. Example of the relative reporting ratio (RRR) equation used in the disproportionality analysis.

Table 1. Antimicrobial New Molecular Entities Approvedby the FDA Since 2006

Antimicrobial newmolecular entity(brand name)

FDA approvaldate (month,day, year)

Therapeutic class(total no. of

agents in class)

Anidulafungin(Eraxis)

2/17/2006 Echinocandin (3)

Darunavir(Prezista)

6/23/2006 Anti-HIV proteaseinhibitor (9)

Maraviroc(Selzentry)

8/06/2007 CCR5 antagonist (1)

Doripenem(Doribax)

10/12/2007 Carbapenem (4)

Raltegravir(Isentress)

10/12/2007 Integrase inhibitor (3)

Telavancin(Vibativ)

9/11/2009 Glycopeptide (2)

Ceftaroline(Teflaro)

10/29/2010 Cephalosporin (18)

Boceprevir(Victrelis)

5/13/2011 Anti-HCV proteaseinhibitor (3)

Telaprevir(Incivek)

5/23/2011 Anti-HCV proteaseinhibitor (3)

Fidaxomicin(Dificid)

5/27/2011 Macrolide (4)

Bedaquiline(Sirturo)

12/28/2012 Antituberculosisdiarylquinoline (1)

Dolutegravir(Tivicay)

8/12/2013 Integrase inhibitor (3)

Simeprevir(Olysio)

11/22/2013 Anti-HCV proteaseinhibitor (3)

Sofosbuvir(Sovaldi)

12/06/2013 Anti-HCV polymeraseinhibitor (1)

Dalbavancin(Dalvance)

5/23/2014 Glycopeptide (3)

FDA = U.S. Food and Drug Administration; HCV = hepatitis Cvirus; HIV = human immunodeficiency virus.

SAFETY OF NEW ANTIMICROBIALS IN THE FDA AERS Khadem et al 3

are associated with darunavir, none of whichmention adverse events such as premature labor,SIDS, ventricular hypertrophy, and congenitalanomaly in offspring. Although acute coronarysyndrome is not listed as a potential adversereaction in the darunavir manufacturer’s packageinsert, protease inhibitors as a class may causedyslipidemias that include elevated cholesteroland triglyceride levels, potentially contributingto increased incidence of myocardial infarctionand unstable angina. In addition, myocardialinfarction is listed as an adverse event identifiedby postmarketing reports for protease inhibitorsincluding fosamprenavir, indinavir, lopinavir/ri-tonavir, and ritonavir. Of the 10 cases of acutecoronary syndrome reported with darunavir,medical conditions concomitantly reportedincluded myocardial infarction (five cases) andangina (two cases). However, without a moreadequate clinical history, it is difficult to deter-mine if these conditions previously existed oroccurred as a result of darunavir exposure.Raltegravir is an integrase inhibitor used for

the treatment of HIV infection, with less com-mon adverse reactions overall compared withdarunavir. Similar to darunavir, among the sev-eral postmarketing and case reports for raltegra-vir, there is no mention of reports of adversereactions such as congenital heart valve disor-ders or SIDS. Nevertheless, it is important tonote that the results of our analysis do not infera causal relationship between the drug and

adverse event signal. Unfortunately, the clinicalcharacteristics provided in the AERS databasewere insufficient to evaluate external factorsappropriately that may have affected the statisti-cal association. When considering the clinicaldata available, medications most commonlyreported with darunavir and raltegravir adverseevent signals were concomitant antiretroviralagents. Diseases commonly reported were oftencardiac in nature and sometimes related to theother signals reported for the agent. As a result,it is difficult to distinguish if the adverse eventsignals were affected by exposure to the drug orif they were a result of an underlying disease ordefect. There did not appear to be a temporal ordose-response relationship between each drugand the adverse event signal, although this infor-mation was limited as well.With respect to doripenem, a total of 32

cases of hepatic dysfunction were reported. Themedian age reported was 61 years, with anapproximately even distribution among menand women. Age is a commonly known riskfactor for many diseases, including hepatic dys-function, because organ function generallydeclines with age. In addition, broad-spectrumantibiotics such as doripenem are often reservedfor patients with resistant gram-negative bacte-rial infections who are critically ill. Among the32 cases, 11 (34%) concomitantly reportedshock. Therefore it is possible that the develop-ment of hepatic dysfunction was due to a pro-gression of the disease state, rather than aprocess facilitated by doripenem. Also of note,concomitant agents with a safety profile includ-ing hepatic dysfunction were reported: fluconaz-ole (11 [34%]), micafungin (9 [28%]), anditraconazole (8 [25%]).

Limitations

As previously mentioned, chronic underre-porting and the lack of accurate quantitativeexposure data are two major limitations in spon-taneous reporting systems. This study illustratesthe deficiencies present in the data analyzedwhere variables were limited to those present inthe AERS report submitted. Information regard-ing the temporal relationship between exposureof the antimicrobial agent and the developmentof the adverse event of interest was absent fromthe reports and is a known limitation of thedata. Although our statistical analyses identifiedclinically meaningful signals according to com-mon industry standards, they were not observed

Table 2. Adverse Event Signals with RRR05 ≥ 2 Not Previ-ously Identified in Manufacturers’ Package Inserts

New molecularentity(brand name)

Adverseevent signal

No. ofsignals RRR05

Darunavir(Prezista)

Premature labor 32 3.1Sudden infantdeath syndrome

10 2.9

Ventricularhypertrophy

11 2.7

Congenital anomalyin offspring

6 2.4

Acute coronarysyndrome

10 2.4

Doripenem(Doribax)

Hepatic dysfunction 32 3.7Hyperchloremia 4 2.6

Raltegravir(Isentress)

Congenital heartvalve disorders

7 2.5

Sudden infantdeath syndrome

9 2.3

Boceprevir(Victrelis)

Weight loss 92 2.2

RRR05 = 5th percentile of the distribution of the relative reportingratio.

4 PHARMACOTHERAPY Volume **, Number **, 2014

adverse events and cannot be viewed ascause-and-effect relationships; no inferenceabout causality should be made. Disproportion-ality analysis of a spontaneous report does notgenerate an incidence that may be generalized toa population but rather a signal that quantifies apotential correlation for the adverse event–drugcombination. Nevertheless, this hypothesis-gen-erating methodology is one used by the FDA asa means of identifying clinically meaningfulsafety signals that may warrant further investiga-tion, safety communications, and/or revisedlabeling information.Challenges are associated with interpreting

disproportionality statistics. External factors,such as population demographics, may influencedisproportionality statistics. For example, theymay be overestimated if a particular adverseevent is more prevalent in a subgroup of a popu-lation that is also more likely to be exposed to acertain drug. This phenomenon is often referredto as the Simpson paradox.9, 10 Disproportionali-ty analysis may also be confounded by associa-tion with other therapeutic agents, also knownas “signal leakage” or “signal absorption.”5 Ifone drug is often administered together withanother drug already known to cause a specificadverse event, disproportionality analysis maysuggest that the first drug is also related to thespecific adverse event due to the higher thanexpected joint occurrence of both drugs in thedatabase. Similarly, interpretation of dispropor-tionality statistics can be confounded by indica-tion, where the adverse event is not actuallyassociated with a drug but rather with the dis-ease being treated or comorbidity associatedwith the disease. Confounding by association orindication may have been a factor in this analy-sis. Unfortunately, these reports provide noinformation detailing if comorbid conditionswere preexisting or developed after the report ofthe adverse event of interest. Additionally, thetiming and extent of exposure to concurrentmedications is unknown.These limitations outline the need for develop-

ment of new tools to detect potential safety signalsin longitudinal health care databases that support,for example, the analysis of temporal relation-ships between drugs and adverse events.11 Inaddition, some of these limitations of dispropor-tionality analysis can be addressed by using statis-tical techniques. For example, the effects of theSimpson paradox, signal leakage, and signalmasking or cloaking can be alleviated by usingstratification and/or logistic regression. In this

respect, a new methodology, the Regression-adjusted Gamma Poisson Shrinker (RGPS) algo-rithm, may also prove beneficial.12 Nevertheless abalance needs to be maintained between the sensi-tivity and specificity of the type of signal detectionmethodology used. Furthermore, it is importantto consider carefully the limitations of quantita-tive signal detection when generating hypotheses.Therefore, these disproportionality statisticsshould be interpreted with caution.

Conclusion

In our review of the FDA AERS database, weidentified clinically meaningful adverse eventsignals associated with antimicrobial new molec-ular entities approved since 2006 includingmany not yet identified in package inserts. Thepotential mechanism of these adverse events isunknown. Although disproportionality analysissuggests a quantitative signal for these associa-tions, causality cannot be inferred from the data.Due to several key limitations in this type ofanalysis, investigative studies are needed to fur-ther explore these adverse event signals and thepotential mechanisms by which they occur.

References

1. U.S. Food and Drug Administration. Drug Approval Reports.Available from http://www.accessdata.fda.gov/scripts/cder/drug-satfda/index.cfm?fuseaction=Reports.ReportsMenu. AccessedJune 12, 2014.

2. Ahmad SR, Marks NS, Goetsch RA. Spontaneous Reporting inthe United States. In: Strom BL, Kimmel SE, eds. Textbook ofpharmacoepidemiology. West Sussex, UK: John Wiley & SonsLtd, 2006: 89–116.

3. U.S. Food and Drug Administration. FDA Adverse EventReporting System. Available from http://www.fda.gov/Drugs/GuidanceComplianceRegulatoryInformation/Surveillance/AdverseDrugEffects/default.htm. Accessed June 28, 2013.

4. U.S. Department of Health and Human Services. Food andDrug Administration. Center for Drug Evaluation and Research(CDER). Center for Biologics Evaluation and Research (CBER).Guidance for Industry: Good Pharmacovigilance Practices andPharmacoepidemiologic Assessment. Available from http://www.fda.gov/downloads/regulatoryinformation/guidances/ucm126834.pdf. Accessed June 28, 2013.

5. van Manen RP, Fram D, DuMouchel W. Signal detectionmethodologies to support effective safety management. ExpertOpin Drug Saf 2007;6:451–64.

6. U.S. Food and Drug Administration. CDER 2005 Report tothe Nation: Improving Public Health Through Human Drugs.Rockville, Maryland. 2005. Available from http://www.fda.gov/downloads/AboutFDA/CentersOffices/CDER/WhatWeDo/UCM078935.pdf. Accessed March 5, 2013.

7. Szarfman A, Machado SG, O’Neill RT. Use of screening algo-rithms and computer systems to efficiently signal higher-than-expected combinations of drugs and events in the USFDA’s spontaneous reports database. Drug Saf 2002;25:381–92.

8. Akiyama N, Kanamaru A, Tamura K, et al. Efficacy and safetyof doripenem for sepsis with neutropenia in Japanese patientswith hematologic diseases. Jpn J Antibiot 2012;65:251–62.

SAFETY OF NEW ANTIMICROBIALS IN THE FDA AERS Khadem et al 5

9. Simpson EH. The interpretation of interaction in contin-gency tables. J R Stat Soc Series B Stat Methodol1951;13:238–41.

10. Harvey JT, Turville C, Barty SM. Data mining of the Austra-lian adverse drug reactions database: a comparison of Bayesianand other statistical indicators. Int Trans Oper Res2004;11:419–33.

11. Oracle Health Sciences Empirica Signal. Available from http://www.oracle.com/us/industries/life-sciences/empirica-signal-ds-396068.pdf. Accessed October 29, 2013.

12. DuMouchel W, Harpaz R. Regression-adjusted GPS algorithm(RGPS). Available from http://www.oracle.com/us/industries/health-sciences/hs-regression-adjusted-gps-wp-1949689.pdf. Acc-essed October 29, 2013.

6 PHARMACOTHERAPY Volume **, Number **, 2014