Perspectives on the use of data mining in pharmaco-vigilance

Drug-induced hepatic injury in children: a case/non-case study of suspected adverse drug reactions in VigiBase

AIM

To identify which drugs are associated with reports of suspected hepatic injury in children and adolescents.

METHODS

Using a worldwide pharmacovigilance database, VigiBase, we conducted a case/non-case study on suspected adverse drug reactions (ADRs) occurring in the population <18 years old. Cases were all the records with hepatic ADRs and non-cases were all the other ADR records. Records regarding topically administered drugs were excluded from both groups. The association between drug and suspected hepatic ADRs was calculated using the reporting odds ratio (ROR) as a measure of disproportionality while adjusting for gender, country, reporter and calendar year. Sub-analyses were performed within therapeutic class and by excluding vaccination-related reports to reduce confounding.

RESULTS

Overall, 6595 (1%) out of 624 673 ADR records in children and adolescents concerned hepatic injury. Most of the reported hepatic injuries concerned children 12-17 years of age. Drugs that were most frequently reported as suspected cause and were associated with hepatic injury comprised paracetamol, valproic acid, carbamazepine, methotrexate, minocycline, zidovudine, pemoline, ceftriaxone, bosentan, ciclosporin, atomoxetine, olanzapine, basiliximab, erythromycin and voriconazole. The association between hepatotoxicity and all these drugs, except for basiliximab, is already known.

CONCLUSIONS

Drug-induced hepatic injury is infrequently reported (only 1% of total) as a suspected ADR in children and adolescents. The drugs associated with reported hepatotoxicity (paracetamol, antiepileptic and anti-tuberculosis agents) are known to be hepatotoxic in adults as well, but age related changes in associations were observed. VigiBase is useful as a start to plan further drug safety studies in children.

Development and evaluation of a common data model enabling active drug safety surveillance using disparate healthcare databases

OBJECTIVE

Active drug safety surveillance may be enhanced by analysis of multiple observational healthcare databases, including administrative claims and electronic health records. The objective of this study was to develop and evaluate a common data model (CDM) enabling rapid, comparable, systematic analyses across disparate observational data sources to identify and evaluate the effects of medicines.

DESIGN

The CDM uses a person-centric design, with attributes for demographics, drug exposures, and condition occurrence. Drug eras, constructed to represent periods of persistent drug use, are derived from available elements from pharmacy dispensings, prescriptions written, and other medication history. Condition eras aggregate diagnoses that occur within a single episode of care. Drugs and conditions from source data are mapped to biomedical ontologies to standardize terminologies and enable analyses of higher-order effects.

MEASUREMENTS

The CDM was applied to two source types: an administrative claims and an electronic medical record database. Descriptive statistics were used to evaluate transformation rules. Two case studies demonstrate the ability of the CDM to enable standard analyses across disparate sources: analyses of persons exposed to rofecoxib and persons with an acute myocardial infarction.

RESULTS

Over 43 million persons, with nearly 1 billion drug exposures and 3.7 billion condition occurrences from both databases were successfully transformed into the CDM. An analysis routine applied to transformed data from each database produced consistent, comparable results.

CONCLUSION

A CDM can normalize the structure and content of disparate observational data, enabling standardized analyses that are meaningfully comparable when assessing the effects of medicines.

Mining pharmacovigilance data using Bayesian logistic regression with James-Stein type shrinkage estimation

Spontaneous adverse event reporting systems are widely used to identify adverse reactions to drugs following their introduction into the marketplace. In this article, a James-Stein type shrinkage estimation strategy was developed in a Bayesian logistic regression model to analyze pharmacovigilance data. This method is effective in detecting signals as it combines information and borrows strength across medically related adverse events. Computer simulation demonstrated that the shrinkage estimator is uniformly better than the maximum likelihood estimator in terms of mean squared error. This method was used to investigate the possible association of a series of diabetic drugs and the risk of cardiovascular events using data from the Canada Vigilance Online Database.

Biclustering of adverse drug events in the FDA's spontaneous reporting system

In this article, we present a new pharmacovigilance data mining technique based on the biclustering paradigm, which is designed to identify drug groups that share a common set of adverse events (AEs) in the spontaneous reporting system (SRS) of the US Food and Drug Administration (FDA). A taxonomy of biclusters is developed, revealing that a significant number of bona fide adverse drug event (ADE) biclusters have been identified. Statistical tests indicate that it is extremely unlikely that the bicluster structures thus discovered, as well as their content, could have arisen by mere chance. Some of the biclusters classified as indeterminate provide support for previously unrecognized and potentially novel ADEs. In addition, we demonstrate the potential importance of the proposed methodology in several important aspects of pharmacovigilance such as providing insight into the etiology of ADEs, facilitating the identification of novel ADEs, suggesting methods and a rationale for aggregating terminologies, highlighting areas of focus, and providing an exploratory tool for data mining.

Mining multi-item drug adverse effect associations in spontaneous reporting systems

Background

Multi-item adverse drug event (ADE) associations are associations relating multiple drugs to possibly multiple adverse events. The current standard in pharmacovigilance is bivariate association analysis, where each single drug-adverse effect combination is studied separately. The importance and difficulty in the detection of multi-item ADE associations was noted in several prominent pharmacovigilance studies. In this paper we examine the application of a well established data mining method known as association rule mining, which we tailored to the above problem, and demonstrate its value. The method was applied to the FDAs spontaneous adverse event reporting system (AERS) with minimal restrictions and expectations on its output, an experiment that has not been previously done on the scale and generality proposed in this work.

Results

Based on a set of 162,744 reports of suspected ADEs reported to AERS and published in the year 2008, our method identified 1167 multi-item ADE associations. A taxonomy that characterizes the associations was developed based on a representative sample. A significant number (67% of the total) of potential multi-item ADE associations identified were characterized and clinically validated by a domain expert as previously recognized ADE associations. Several potentially novel ADEs were also identified. A smaller proportion (4%) of associations were characterized and validated as known drug-drug interactions.

Conclusions

Our findings demonstrate that multi-item ADEs are present and can be extracted from the FDA’s adverse effect reporting system using our methodology, suggesting that our method is a valid approach for the initial identification of multi-item ADEs. The study also revealed several limitations and challenges that can be attributed to both the method and quality of data.

Determining correspondences between high-frequency MedDRA concepts and SNOMED: a case study

BACKGROUND

The Systematic Nomenclature of Medicine Clinical Terms (SNOMED CT) is being advocated as the foundation for encoding clinical documentation. While the electronic medical record is likely to play a critical role in pharmacovigilance - the detection of adverse events due to medications - classification and reporting of Adverse Events is currently based on the Medical Dictionary of Regulatory Activities (MedDRA). Complete and high-quality MedDRA-to-SNOMED CT mappings can therefore facilitate pharmacovigilance. The existing mappings, as determined through the Unified Medical Language System (UMLS), are partial, and record only one-to-one correspondences even though SNOMED CT can be used compositionally. Efforts to map previously unmapped MedDRA concepts would be most productive if focused on concepts that occur frequently in actual adverse event data. We aimed to identify aspects of MedDRA that complicate mapping to SNOMED CT, determine pattern in unmapped high-frequency MedDRA concepts, and to identify types of integration errors in the mapping of MedDRA to UMLS.

METHODS

Using one years' data from the US Federal Drug Administrations Adverse Event Reporting System, we identified MedDRA preferred terms that collectively accounted for 95% of both Adverse Events and Therapeutic Indications records. After eliminating those already mapping to SNOMED CT, we attempted to map the remaining 645 Adverse-Event and 141 Therapeutic-Indications preferred terms with software assistance.

RESULTS

All but 46 Adverse-Event and 7 Therapeutic-Indications preferred terms could be composed using SNOMED CT concepts: none of these required more than 3 SNOMED CT concepts to compose. We describe the common composition patterns in the paper. About 30% of both Adverse-Event and Therapeutic-Indications Preferred Terms corresponded to single SNOMED CT concepts: the correspondence was detectable by human inspection but had been missed during the integration process, which had created duplicated concepts in UMLS.

CONCLUSIONS

Identification of composite mapping patterns, and the types of errors that occur in the MedDRA content within UMLS, can focus larger-scale efforts on improving the quality of such mappings, which may assist in the creation of an adverse-events ontology.

Data-driven approach for creating synthetic electronic medical records

BACKGROUND

New algorithms for disease outbreak detection are being developed to take advantage of full electronic medical records (EMRs) that contain a wealth of patient information. However, due to privacy concerns, even anonymized EMRs cannot be shared among researchers, resulting in great difficulty in comparing the effectiveness of these algorithms. To bridge the gap between novel bio-surveillance algorithms operating on full EMRs and the lack of non-identifiable EMR data, a method for generating complete and synthetic EMRs was developed.

METHODS

This paper describes a novel methodology for generating complete synthetic EMRs both for an outbreak illness of interest (tularemia) and for background records. The method developed has three major steps: 1) synthetic patient identity and basic information generation; 2) identification of care patterns that the synthetic patients would receive based on the information present in real EMR data for similar health problems; 3) adaptation of these care patterns to the synthetic patient population.

RESULTS

We generated EMRs, including visit records, clinical activity, laboratory orders/results and radiology orders/results for 203 synthetic tularemia outbreak patients. Validation of the records by a medical expert revealed problems in 19% of the records; these were subsequently corrected. We also generated background EMRs for over 3000 patients in the 4-11 yr age group. Validation of those records by a medical expert revealed problems in fewer than 3% of these background patient EMRs and the errors were subsequently rectified.

CONCLUSIONS

A data-driven method was developed for generating fully synthetic EMRs. The method is general and can be applied to any data set that has similar data elements (such as laboratory and radiology orders and results, clinical activity, prescription orders). The pilot synthetic outbreak records were for tularemia but our approach may be adapted to other infectious diseases. The pilot synthetic background records were in the 4-11 year old age group. The adaptations that must be made to the algorithms to produce synthetic background EMRs for other age groups are indicated.

Validation of statistical signal detection procedures in eudravigilance post-authorization data: a retrospective evaluation of the potential for earlier signalling

BACKGROUND

Screening large databases of spontaneous case reports of possible adverse drug reactions (ADRs) is an established method of identifying hitherto unknown adverse effects of medicinal products; however, there is a lack of consensus concerning the value of formal statistical screening procedures in guiding such a process. This study was performed to clarify the nature of any added benefits and additional effort required when established pharmacovigilance techniques are supplemented with statistical screening.

OBJECTIVE

To evaluate whether statistical signal detection in spontaneous reporting data can lead to earlier detection of drug safety problems and to assess the additional regulatory work entailed.

METHODS

Using the EudraVigilance post-authorization module (EVPM), a screening procedure based on the proportional reporting ratio (PRR) was applied retrospectively to examine if regulatory investigations concerning ADRs in a predefined set of products could have been initiated earlier than occurred in practice. During the same time period, between September 2003 and March 2007, the number of PRR-based signals of disproportionate reporting (SDR) that arose in the same set of products was calculated and evaluated to determine the number requiring investigation. The outcome is expressed as the ratio of the number of SDRs requiring investigation compared with the number of signals pre-empted by the statistical screening approach. In those cases where the signal was discovered earlier, the delay was calculated between identification by the PRR method and by the method that originally identified the signal.

RESULTS

In 191 chemically different products, 532 adverse reactions were added to the summary of product characteristics during the study period. Of these, 405 were designated as important medical events (IMEs) based on a comprehensive predefined list. Of the IMEs, 217 (53.6%) were identified earlier by the statistical screening technique, 79 (19.6%) were detected after the date at which they were raised by standard pharmacovigilance methods and 109 (26.9%) were not signalled during the study period. 1561 SDRs requiring further evaluation were detected during the study period, giving a ratio of 7.2 assessments for each signal pre-empted. The mean delay between the discovery of signals using the statistical methods in the EVPM and established methods in the 217 cases detected earlier was 2.45 years. A review resulted in clear explanation for why the statistical method had not pre-empted detection in all but 77 of 188 cases.

CONCLUSIONS

The form of statistical signal detection tested in this study can provide significant early warning in a large proportion of drug safety problems; however, it cannot detect all safety issues more quickly than other pharmacovigilance processes and hence it should be used in addition to, rather than as an alternative to, established methods.

Drugs associated with hepatotoxicity and their reporting frequency of liver adverse events in VigiBase: unified list based on international collaborative work

BACKGROUND

Challenges exist in the clinical diagnosis of drug-induced liver injury (DILI) and in obtaining information on hepatotoxicity in humans.

OBJECTIVE

(i) To develop a unified list that combines drugs incriminated in well vetted or adjudicated DILI cases from many recognized sources and drugs that have been subjected to serious regulatory actions due to hepatotoxicity; and (ii) to supplement the drug list with data on reporting frequencies of liver events in the WHO individual case safety report database (VigiBase).

DATA SOURCES AND EXTRACTION

(i) Drugs identified as causes of DILI at three major DILI registries; (ii) drugs identified as causes of drug-induced acute liver failure (ALF) in six different data sources, including major ALF registries and previously published ALF studies; and (iii) drugs identified as being subjected to serious governmental regulatory actions due to their hepatotoxicity in Europe or the US were collected. The reporting frequency of adverse events was determined using VigiBase, computed as Empirical Bayes Geometric Mean (EBGM) with 90% confidence interval for two customized terms, 'overall liver injury' and 'ALF'. EBGM of >or=2 was considered a disproportional increase in reporting frequency. The identified drugs were then characterized in terms of regional divergence, published case reports, serious regulatory actions, and reporting frequency of 'overall liver injury' and 'ALF' calculated from VigiBase.

DATA SYNTHESIS

After excluding herbs, supplements and alternative medicines, a total of 385 individual drugs were identified; 319 drugs were identified in the three DILI registries, 107 from the six ALF registries (or studies) and 47 drugs that were subjected to suspension or withdrawal in the US or Europe due to their hepatotoxicity. The identified drugs varied significantly between Spain, the US and Sweden. Of the 319 drugs identified in the DILI registries of adjudicated cases, 93.4% were found in published case reports, 1.9% were suspended or withdrawn due to hepatotoxicity and 25.7% were also identified in the ALF registries/studies. In VigiBase, 30.4% of the 319 drugs were associated with disproportionally higher reporting frequency of 'overall liver injury' and 83.1% were associated with at least one reported case of ALF.

CONCLUSIONS

This newly developed list of drugs associated with hepatotoxicity and the multifaceted analysis on hepatotoxicity will aid in causality assessment and clinical diagnosis of DILI and will provide a basis for further characterization of hepatotoxicity.

Modelling the time to onset of adverse reactions with parametric survival distributions: a potential approach to signal detection and evaluation

BACKGROUND

It has been postulated that the time to onset of adverse drug reactions is connected to the underlying pharmacological (or toxic) mechanism of adverse drug reactions whether the reaction is time dependent or not.

OBJECTIVE

We have conducted a preliminary study using the parametric modelling of the time to onset of adverse reactions as an approach to signal detection on spontaneous reporting system databases.

METHODS

We performed a parametric modelling of the reported time to onset of adverse drug reactions for which the underlying toxic mechanism is characterized. For the purpose of our study, we have used the reported liver injuries associated with bosentan, and the infections associated with the use of the tumour necrosis factor (TNF) inhibitors, adalimumab, etanercept and infliximab, which are used in Crohn's disease and rheumatoid arthritis, reported to EudraVigilance between December 2001 and September 2006.

RESULTS

The main results reflect the fact that the reported time to onset is a surrogate of the true time to onset of the reaction and combines three hazards (occurrence, diagnosis and reporting) that cannot be disentangled. Consequently, the modelling of the time to onset of reactions reported with TNF inhibitors showed differences that could reflect different pharmacological activities, indications, monitoring of the patients or different reporting patterns. These variations could also limit the interpretation of the parametric modelling.

CONCLUSIONS

Some consistency that was found between the occurrences of the infections with the TNF inhibitors suggests a causal association. There are statistical issues that are important to keep in mind when interpreting the results (the impact of the data quality on the fit of the distributions and the absence of a test of hypothesis linked to the absence of a relevant comparator). The study suggests that the modelling of the reported time to onset of adverse reactions could be a useful adjunct to other signal detection methods.

Antimicrobials and the risk of torsades de pointes: the contribution from data mining of the US FDA Adverse Event Reporting System

BACKGROUND

Drug-induced torsades de pointes (TdP) is a complex regulatory and clinical problem due to the rarity of this sometimes fatal adverse event. In this context, the US FDA Adverse Event Reporting System (AERS) is an important source of information, which can be applied to the analysis of TdP liability of marketed drugs.

OBJECTIVE

To critically evaluate the risk of antimicrobial-induced TdP by detecting alert signals in the AERS, on the basis of both quantitative and qualitative analyses.

METHODS

Reports of TdP from January 2004 through December 2008 were retrieved from the public version of the AERS. The absolute number of cases and reporting odds ratio as a measure of disproportionality were evaluated for each antimicrobial drug (quantitative approach). A list of drugs with suspected TdP liability (provided by the Arizona Centre of Education and Research on Therapeutics [CERT]) was used as a reference to define signals. In a further analysis, to refine signal detection, we identified TdP cases without co-medications listed by Arizona CERT (qualitative approach).

RESULTS

Over the 5-year period, 374 reports of TdP were retrieved: 28 antibacterials, 8 antifungals, 1 antileprosy and 26 antivirals were involved. Antimicrobials more frequently reported were levofloxacin (55) and moxifloxacin (37) among the antibacterials, fluconazole (47) and voriconazole (17) among the antifungals, and lamivudine (8) and nelfinavir (6) among the antivirals. A significant disproportionality was observed for 17 compounds, including several macrolides, fluoroquinolones, linezolid, triazole antifungals, caspofungin, indinavir and nelfinavir. With the qualitative approach, we identified the following additional drugs or fixed dose combinations, characterized by at least two TdP cases without co-medications listed by Arizona CERT: ceftriaxone, piperacillin/tazobactam, cotrimoxazole, metronidazole, ribavirin, lamivudine and lopinavir/ritonavir.

DISCUSSION

Disproportionality for macrolides, fluoroquinolones and most of the azole antifungals should be viewed as 'expected' according to Arizona CERT list. By contrast, signals were generated by linezolid, caspofungin, posaconazole, indinavir and nelfinavir. Drugs detected only by the qualitative approach should be further investigated by increasing the sensitivity of the method, e.g. by searching also for the TdP surrogate marker, prolongation of the QT interval.

CONCLUSIONS

The freely available version of the FDA AERS database represents an important source to detect signals of TdP. In particular, our analysis generated five signals among antimicrobials for which further investigations and active surveillance are warranted. These signals should be considered in evaluating the benefit-risk profile of these drugs.

Prospective data mining of six products in the US FDA Adverse Event Reporting System: disposition of events identified and impact on product safety profiles

BACKGROUND

The use of data mining has increased among regulators and pharmaceutical companies. The incremental value of data mining as an adjunct to traditional pharmacovigilance methods has yet to be demonstrated. Specifically, the utility in identifying new safety signals and the resources required to do so have not been elucidated.

OBJECTIVES

To analyse the number and types of disproportionately reported product-event combinations (DRPECs), as well as the final disposition of each, in order to understand the potential utility and resource implications of routinely conducting data mining in the US FDA Adverse Event Reporting System (AERS).

METHODS

We generated DRPECs from AERS for six of Wyeth's products, prospectively tracked their dispositions and evaluated the appropriate DRPECs in the company's safety database. We chose EB05 (the lower bound of the 90% confidence interval around the Empirical Bayes Geometric Mean) > or =2 as the appropriate metric, employing stratification based on age, sex and year of report.

RESULTS

A total of 861 DRPECs were identified - the average number of DRPECs was 144 per product. The proportion of unique preferred terms (PTs) in AERS for each drug with an EB05 > or =2 was similar across the six products (5.1-8.5%). Overall, 64.0% (551) of the DRPECs were closed after the initial screening (44.8% labelled, 14.3% indication related, 4.9% non-interpretable). An additional 9.9% (85) had been reviewed within the prior year and were not further reviewed. The remaining 26.1% (225) required full case review. After review of all pertinent reports and additional data, it was determined which of the DRPECs necessitated a formal review by the company's ongoing Safety Review Team (SRT) process. In total, 3.6% (31/861) of the DRPECs, yielding 16 medical concepts, were reviewed by the SRT, leading to seven labelling changes. These labelling changes involved 1.9% of all DRPECs generated. Four of the six compounds reviewed as part of this pilot had an identified labelling change. The workload required for this pilot, which was driven primarily by those DRPECs requiring review, was extensive, averaging 184 hours per product.

CONCLUSION

The number of DRPECs identified for each drug approximately correlated with the number of unique PTs in the database. Over one-half of DRPECs were either labelled as per the company's reference safety information (RSI) or were under review after identification by traditional pharmacovigilance activities, suggesting that for marketed products these methods do identify adverse events detected by traditional pharmacovigilance methods. Approximately three-quarters of the 861 DRPECs identified were closed without case review after triage. Of the approximately one-quarter of DRPECs that required formal case review, seven resulted in an addition to the RSI for the relevant products. While this pilot does not allow us to comment on the utility of routine data mining for all products, it is significant that several new safety concepts were identified through this prospective exercise.

Construction and validation of synthetic electronic medical records

There is a current and pressing need for a test bed of electronic medical records (EMRs) to insure consistent development, validation and verification of public health related algorithms that operate on EMRs. However, access to full EMRs is limited and not generally available to the academic algorithm developers who support the public health community. This paper describes a set of algorithms that produce synthetic EMRs using real EMRs as a model. The algorithms were used to generate a pilot set of over 3000 synthetic EMRs that are currently available on CDC’s Public Health grid. The properties of the synthetic EMRs were validated, both in the entire aggregate data set and for individual (synthetic) patients. We describe how the algorithms can be extended to produce records beyond the initial pilot data set.

Information technology in pharmacovigilance: Benefits, challenges, and future directions from industry perspectives

Risk assessment during clinical product development needs to be conducted in a thorough and rigorous manner. However, it is impossible to identify all safety concerns during controlled clinical trials. Once a product is marketed, there is generally a large increase in the number of patients exposed, including those with comorbid conditions and those being treated with concomitant medications. Therefore, postmarketing safety data collection and clinical risk assessment based on observational data are critical for evaluating and characterizing a product’s risk profile and for making informed decisions on risk minimization. Information science promises to deliver effective e-clinical or e-health solutions to realize several core benefits: time savings, high quality, cost reductions, and increased efficiencies with safer and more efficacious medicines. The development and use of standard-based pharmacovigilance system with integration connection to electronic medical records, electronic health records, and clinical data management system holds promise as a tool for enabling early drug safety detections, data mining, results interpretation, assisting in safety decision making, and clinical collaborations among clinical partners or different functional groups. The availability of a publicly accessible global safety database updated on a frequent basis would further enhance detection and communication about safety issues. Due to recent high-profile drug safety problems, the pharmaceutical industry is faced with greater regulatory enforcement and increased accountability demands for the protection and welfare of patients. This changing climate requires biopharmaceutical companies to take a more proactive approach in dealing with drug safety and pharmacovigilance.

An evaluation of three signal-detection algorithms using a highly inclusive reference event database

BACKGROUND

Pharmacovigilance data-mining algorithms (DMAs) are known to generate significant numbers of false-positive signals of disproportionate reporting (SDRs), using various standards to define the terms 'true positive' and 'false positive'.

OBJECTIVE

To construct a highly inclusive reference event database of reported adverse events for a limited set of drugs, and to utilize that database to evaluate three DMAs for their overall yield of scientifically supported adverse drug effects, with an emphasis on ascertaining false-positive rates as defined by matching to the database, and to assess the overlap among SDRs detected by various DMAs.

METHODS

A sample of 35 drugs approved by the US FDA between 2000 and 2004 was selected, including three drugs added to cover therapeutic categories not included in the original sample. We compiled a reference event database of adverse event information for these drugs from historical and current US prescribing information, from peer-reviewed literature covering 1999 through March 2006, from regulatory actions announced by the FDA and from adverse event listings in the British National Formulary. Every adverse event mentioned in these sources was entered into the database, even those with minimal evidence for causality. To provide some selectivity regarding causality, each entry was assigned a level of evidence based on the source of the information, using rules developed by the authors. Using the FDA adverse event reporting system data for 2002 through 2005, SDRs were identified for each drug using three DMAs: an urn-model based algorithm, the Gamma Poisson Shrinker (GPS) and proportional reporting ratio (PRR), using previously published signalling thresholds. The absolute number and fraction of SDRs matching the reference event database at each level of evidence was determined for each report source and the data-mining method. Overlap of the SDR lists among the various methods and report sources was tabulated as well.

RESULTS

The GPS algorithm had the lowest overall yield of SDRs (763), with the highest fraction of events matching the reference event database (89 SDRs, 11.7%), excluding events described in the prescribing information at the time of drug approval. The urn model yielded more SDRs (1562), with a non-significantly lower fraction matching (175 SDRs, 11.2%). PRR detected still more SDRs (3616), but with a lower fraction matching (296 SDRs, 8.2%). In terms of overlap of SDRs among algorithms, PRR uniquely detected the highest number of SDRs (2231, with 144, or 6.5%, matching), followed by the urn model (212, with 26, or 12.3%, matching) and then GPS (0 SDRs uniquely detected).

CONCLUSIONS

The three DMAs studied offer significantly different tradeoffs between the number of SDRs detected and the degree to which those SDRs are supported by external evidence. Those differences may reflect choices of detection thresholds as well as features of the algorithms themselves. For all three algorithms, there is a substantial fraction of SDRs for which no external supporting evidence can be found, even when a highly inclusive search for such evidence is conducted.

Drug-induced torsades de pointes: data mining of the public version of the FDA Adverse Event Reporting System (AERS)

AIMS

To investigate spontaneous reports of TdP present in the public version of the FDA Adverse Event Reporting System (AERS) in the light of what is already known on their TdP-liability.

METHODS

Reports of TdP from January 2004 through December 2007 were retrieved from the public version of the AERS database. All reports were selected from REACTION files and the relevant suspected and/or interacting drugs were identified from DRUG files. Qualitative analysis was performed by the case/non-case method. Cases were represented by TdP reports, whereas non-cases were all reports of adverse drug reactions other than TdP. Quantitative analysis was assessed by calculating the crude and adjusted reporting odds ratio (ROR), as a measure of disproportionality, with the 95% confidence interval.

RESULTS

Reports of TdP were 1665 over a 4-year period, involving 376 active substances. Thirty-five drugs with at least 10 reports were identified: amiodarone and methadone were associated with the highest number of cases (113 and 83 respectively) and most of the other reports were ascribable to antibacterials, antidepressants and antipsychotics; remarkable differences in number of cases and ROR were present among agents within each therapeutic class. A disproportionate reporting was also observed for other compounds such as donepezil, famotidine and mitoxantrone.

CONCLUSIONS

Large spontaneous reporting databases represent an important source for signal detection of rare adverse drug reactions (ADR), such as TdP. The number of reports associated to donepezil, famotidine and mitoxantrone could be considered unexpected on the basis of current evidence and needs further investigations on their true TdP-liability.

Drug-versus-drug adverse event rate comparisons: a pilot study based on data from the US FDA Adverse Event Reporting System

BACKGROUND

A number of published studies compare adverse event rates for drugs on the basis of reports in the US FDA Adverse Event Reporting System (AERS). While the AERS data have the advantage of timely availability and a large capture population, the database is subject to many significant biases, and lacks complete patient information that would allow for correction of those biases. The accuracy of comparative AERS-based data mining has been questioned, but has not been systematically studied.

OBJECTIVE

To determine whether AERS could be used as a data source to accurately compare the adverse event rates for pairs of drugs, using pre-defined, stringent criteria to dictate whether a given pair of drugs was considered eligible for such a comparison.

METHODS

The Fisher's Exact test was utilized to detect differences in adverse event rates between such pairs of drugs. Concordance was determined between statistically significant AERS-based adverse event rate differences, and adverse event rate differences published in the literature from clinical trials and case-control studies. The conditions for validity included (i) data that are free of 'extreme duplication' in AERS reports; (ii) drugs used in similar patient populations; (iii) drugs used for similar indications; (iv) drugs used with the same spectrum of concomitant medications; and (v) drugs not widely disparate in time on the market.

RESULTS

For 19 drugs studied, a total of 36 evaluable adverse event rate comparisons were identified. Comparisons were classified as favouring 'drug A', favouring 'drug B' or detecting no difference. Concordance for the resulting 3x3 table (AERS vs literature) gave a kappa statistic of 0.654, indicating moderately good agreement. In only two cases was there absolute discordance, with AERS designating one drug as having a lower rate, while the published study designated the other drug as having a lower rate, with respect to a given adverse event.

CONCLUSIONS

This pilot study encourages further research regarding the use of spontaneous report databases such as AERS, under stringently defined conditions, to compare adverse event rates for drugs. While not hypothesis proving, such estimates can be used for purposes such as generating hypotheses for controlled studies, and for designing those studies.

False discovery rate estimation for frequentist pharmacovigilance signal detection methods

Pharmacovigilance systems aim at early detection of adverse effects of marketed drugs. They maintain large spontaneous reporting databases for which several automatic signaling methods have been developed. One limit of those methods is that the decision rules for the signal generation are based on arbitrary thresholds. In this article, we propose a new signal-generation procedure. The decision criterion is formulated in terms of a critical region for the P-values resulting from the reporting odds ratio method as well as from the Fisher's exact test. For the latter, we also study the use of mid-P-values. The critical region is defined by the false discovery rate, which can be estimated by adapting the P-values mixture model based procedures to one-sided tests. The methodology is mainly illustrated with the location-based estimator procedure. It is studied through a large simulation study and applied to the French pharmacovigilance database.

Time-to-signal comparison for drug safety data-mining algorithms vs. traditional signaling criteria

Data mining may improve identification of signals, but its incremental utility is in question. The objective of this study was to compare associations highlighted by data mining vs. those highlighted through the use of traditional decision rules. In the case of 29 drugs, we used US Food and Drug Administration (FDA) Adverse Event Reporting System (AERS) data to compare three data-mining algorithms (DMAs) with two traditional decision rules: (i) N >or= 3 reports for a designated medical event (DME) and (ii) any event comprising >2% of reports in relation to a drug. Data-mining methods produced 101-324 signals vs. 1,051 for the N >or= 3 rule but yielded a higher proportion of signals having publication support. For the 2% rule, the fraction of signals having publication support was similar to that associated with data mining. Data-mining signals lagged N >or= 3 signaling by 1.5-11.0 months. It may therefore be concluded that data mining identifies fewer signals than the "N >or= 3 DME" rule. The signals appear later with data mining but are more often supported by publications. In the case of the 2% rule, no such difference in publication support was observed.

Influence of the MedDRA hierarchy on pharmacovigilance data mining results

PURPOSE

To compare the results of drug safety data mining with three different algorithms, when adverse events are identified using MedDRA Preferred Terms (PT) vs. High Level Terms (HLT) vs. Standardised MedDRA Queries (SMQ).

METHODS

For a representative set of 26 drugs, data from the FDA Adverse Event Reporting System (AERS) database from 2001 through 2005 was mined for signals of disproportionate reporting (SDRs) using three different data mining algorithms (DMAs): the Gamma Poisson Shrinker (GPS), the urn-model algorithm (URN), and the proportional reporting rate (PRR) algorithm. Results were evaluated using a previously described Reference Event Database (RED) which contains documented drug-event associations for the 26 drugs. Analysis emphasized the percentage of SDRs in the "unlabeled supported" category, corresponding to those adverse events that were not described in the U.S. prescribing information for the drug at the time of its approval, but which were supported by some published evidence for an association with the drug.

RESULTS

Based on a logistic regression analysis, the percentage of unlabeled supported SDRs was smallest at the PT level, intermediate at the HLT level, and largest at the SMQ level, for all three algorithms. The GPS and URN methods detected comparable percentages of unlabeled supported SDRs while the PRR method detected a smaller percentage, at all three MedDRA levels. No evidence of a method/level interaction was seen.

CONCLUSIONS

Use of HLT and SMQ groupings can improve the percentage of unlabeled supported SDRs in data mining results. The trade-off for this gain is the medically less-specific language of HLTs and SMQs compared to PTs, and the need for the added step in data mining of examining the component PTs of each HLT or SMQ that results in a signal of disproportionate reporting.

A model for decision support in signal triage

Spontaneous reporting of suspected adverse drug reactions (ADRs) has long been a cornerstone of pharmacovigilance. With the increasingly large volume of ADRs, regulatory agencies, scientific/academic organizations and marketing authorization holders have applied statistical tools to assist in signal detection by identifying disproportionate reporting relationships in spontaneous reporting databases. These tools have generated large numbers of signals defined as drug-ADR reporting associations that meet specified statistical criteria. The challenge is to identify which signals are most likely to be medically important and therefore warrant priority for further investigation. Decisions related to signal triage are often complex and are based on a combination of clinical, epidemiological, pharmacological and regulatory criteria. There are no specific regulations, guidelines or standards that provide an objective basis for these decisions. This paper describes preliminary work to identify and quantify the specific factors that contribute to a decision to prioritize a specific drug-ADR combination for further in-depth review. We applied a tool from the discipline of decision analysis to systematically assess the important attributes of spontaneously reported ADRs. A model was created that integrates these assessments and produces rankings for the signals generated from quantitative signalling methods. Although more research is necessary to evaluate the performance of this model fully, preliminary results suggest that the use of formal decision analysis approaches to support signal triage can provide potential benefit and will help meet an important need.

An evaluation of a data mining signal for amyotrophic lateral sclerosis and statins detected in FDA's spontaneous adverse event reporting system

BACKGROUND

We detected disproportionate reporting of amyotrophic lateral sclerosis (ALS) with HMG-CoA-reductase inhibitors (statins) in the Food and Drug Administration's (FDA) spontaneous adverse event (AE) reporting system (AERS).

PURPOSE

To describe the original ALS signal and to provide additional context for interpreting the signal by conducting retrospective analyses of data from long-term, placebo-controlled clinical trials of statins.

METHODS

The ALS signal was detected using the multi-item gamma Poisson shrinker (MGPS) algorithm. All AERS cases of ALS reported in association with use of a statin were individually reviewed by two FDA neurologists. Manufacturers of lovastatin, pravastatin, simvastatin, fluvastatin, atorvastatin, cerivastatin, and rosuvastatin were requested to provide the number of cases of ALS diagnosed during all of their placebo-controlled statin trials that were at least 6 months in duration.

RESULTS

There were 91 US and foreign reports of ALS with statins in AERS. The data mining signal scores for ALS and statins ranged from 8.5 to 1.6. Data were obtained from 41 statin clinical trials ranging in duration from 6 months to 5 years and representing approximately 200,000 patient-years of exposure to statin and approximately 200,000 patient-years of exposure to placebo. Nine cases of ALS were reported in statin-treated patients and 10 cases in placebo-treated patients.

CONCLUSIONS

Although we observed a data mining signal for ALS with statins in FDA's AERS, retrospective analyses of 41 statin clinical trials did not reveal an increased incidence of ALS in subjects treated with a statin compared with placebo.

Risk of Hepatotoxicity Associated with the Use of Telithromycin: A Signal Detection Using Data Mining Algorithms

Background:

With the exception of case reports, limited data are available regarding the risk of hepatotoxicity associated with the use of telithromycin.

Objective:

To detect the safety signal regarding the reporting of hepatotoxicity associated with the use of telithromycin using 4 commonly employed data mining algorithms (DMAs).

Methods:

Based on the Adverse Events Reporting System (AERS) database of the Food and Drug Administration, 4 DMAs, including the reporting odds ratio (ROR), the proportional reporting ratio (PRR), the information component (IC), and the Gamma Poisson Shrinker (GPS), were applied to examine the association between the reporting of hepatotoxicity and the use of telithromycin. The study period was from the first quarter of 2004 to the second quarter of 2006. The reporting of hepatotoxicity was identified using the preferred terms indexed in the Medical Dictionary for Regulatory Activities. The drug name was used to identify reports regarding the use of telithromycin.

Results:

A total of 226 reports describing hepatotoxicity associated with the use of telithromycin were recorded in the AERS. A safety problem of telithromycin associated with increased reporting of hepatotoxicity was clearly detected by 4 algorithms as early as 2005, signaling the problem in the first quarter by the ROR and the IC, in the second quarter by the PRR, and in the fourth quarter by the GPS.

Conclusions:

A safety signal was indicated by the 4 DMAs suggesting an association between the reporting of hepatotoxicity and the use of telithromycin. Given the wide use of telithromycin and serious consequences of hepatotoxicity, clinicians should be cautious when selecting telithromycin for treatment of an infection. In addition, further observational studies are required to evaluate the utility of signal detection systems for early recognition of serious, life-threatening, low-frequency drug-induced adverse events.

Heterogeneous but "standard" coding systems for adverse events: Issues in achieving interoperability between apples and oranges

Monitoring adverse events (AEs) is an important part of clinical research and a crucial target for data standards. The representation of adverse events themselves requires the use of controlled vocabularies with thousands of needed clinical concepts. Several data standards for adverse events currently exist, each with a strong user base. The structure and features of these current adverse event data standards (including terminologies and classifications) are different, so comparisons and evaluations are not straightforward, nor are strategies for their harmonization. Three different data standards - the Medical Dictionary for Regulatory Activities (MedDRA) and the Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT) terminologies, and Common Terminology Criteria for Adverse Events (CTCAE) classification - are explored as candidate representations for AEs. This paper describes the structural features of each coding system, their content and relationship to the Unified Medical Language System (UMLS), and unsettled issues for future interoperability of these standards.

The safety profiles of orlistat and sibutramine: results of prescription-event monitoring studies in England

INTRODUCTION

Observational cohort studies were conducted using prescription-event monitoring (PEM) to examine the safety profiles of the anti-obesity agents orlistat and sibutramine. Adverse events reported as case reports were also evaluated to determine whether these events were also identified by PEM.

RESEARCH METHODS AND PROCEDURES

Patients were identified from dispensed prescriptions written by general practitioners (GPs) in England for orlistat or sibutramine. Patient demographic and clinical event information, including reasons for stopping and adverse drug reactions, were requested on questionnaires posted to GPs at least 6 months after the first prescription for individual patients. Event incidence densities (IDs) (number of first reports of event/1000 patient-months treatment) were calculated for month 1 (ID(1)) and months 2-3 (ID(2-3)). Published case reports were identified by searching Medline and Embase.

RESULTS

The cohorts comprised 16,021 and 12,336 patients prescribed orlistat and sibutramine, respectively. Both cohorts had a median age of 45 years, and approximately 80% were female. The most common reason for stopping orlistat within 3 months was diarrhea (332 patients; 2.1% cohort), and for stopping sibutramine it was hypertension (203 patients; 1.6%). Clinical events significantly associated with taking orlistat were mainly gastrointestinal and those for sibutramine included central nervous system effects, nausea/vomiting, palpitation, and sweating. We identified 8 published case reports for orlistat and 10 for sibutramine that had equivalent or similar events assessed as causally related in the PEM studies.

CONCLUSIONS

The PEM studies highlighted different adverse event profiles for orlistat and sibutramine that were consistent with their distinct pharmacological mechanisms and other published information.

Criteria revision and performance comparison of three methods of signal detection applied to the spontaneous reporting database of a pharmaceutical manufacturer

BACKGROUND AND OBJECTIVE

Several statistical methods exist for detecting signals of potential adverse drug reactions in spontaneous reporting databases. However, these signal-detection methods were developed using regulatory databases, which contain a far larger number of adverse event reports than the databases maintained by individual pharmaceutical manufacturers. Furthermore, the composition and quality of the spontaneous reporting databases differ between regulatory agencies and pharmaceutical companies. Thus, the signal-detection criteria proposed for regulatory use are considered to be inappropriate for pharmaceutical industry use without modification. The objective of this study was to revise the criteria for signal detection to make them suitable for use by pharmaceutical manufacturers.

METHODS

A model comprising 40 drugs and 1000 adverse events was constructed based on a spontaneous reporting database provided by a pharmaceutical company and used in a simulation to investigate appropriate criteria for signal detection. In total, 1000 pseudo datasets were generated with this model, and three statistical methods (proportional reporting ratio [PRR], Bayesian Confidence Propagation Neural Network [BCPNN] and multi-item gamma Poisson shrinker [MGPS]) for signal detection were applied to each dataset. The sensitivity and specificity of each method were evaluated using these pseudo datasets. The optimum critical value for signal detection (i.e. the value that achieved the highest sensitivity with 95% specificity) was identified for each method. The optimum values were also examined with the adverse events classified into two categories according to frequency. The three original detection methods and their revised versions were applied to a real pharmaceutical company database to detect 173 known adverse reactions of four drugs.

RESULTS

The 1000 pseudo datasets consisted of an average of 81 862 reports and 11,407 drug-event pairs, including 1192 adverse drug reactions. The sensitivities of PRR, BCPNN and MGPS methods were 49%, 45% and 26%, respectively, whereas their specificities were 95%, 99.6% and 99.99%, respectively; these sensitivities were unacceptably low for pharmaceutical manufacturers, whereas the specificities were acceptable. The highest sensitivity for each method, obtained by changing critical values and maintaining specificity at 95%, was 44%, 62% and 62%, respectively. When adverse events were classified into two categories, sensitivities as high as 75% for regular events and 39% for rare events were achieved with the revised BCPNN method. The critical values of the information component minus two standard deviations (IC - 2SD) index of the revised BCPNN method were greater than -0.7 for regular events and greater than -0.6 for rare events. The revised BCPNN method yielded 51% sensitivity and 89% specificity for the real dataset.

CONCLUSION

A lower critical value may be needed when signal-detection methodology is applied to the spontaneous reporting databases of pharmaceutical manufacturers. For example, it is recommended that pharmaceutical manufacturers use the BCPNN method with IC - 2SD criteria of greater than -0.7 for regular events and greater than -0.6 for rare events.

Signal detection methodologies to support effective safety management

The increased focus on the safety of medical products, as well as the growing volume of available safety information, has created a need for objective quantitative approaches to supplement the medical review of individual case safety reports. Statistical algorithms can be used to identify trends and relationships in both clinical and postmarketing safety databases in support of safety signal detection. Powerful data visualization tools facilitate the medical review of the complex information generated by these methods. In addition, all these approaches need to be integrated into the daily practice of clinical safety and postmarketing pharmacovigilance.

Serious adverse events with infliximab: analysis of spontaneously reported adverse events

BACKGROUND & AIMS

Serious adverse events such as bowel obstruction, heart failure, infection, lymphoma, and neuropathy have been reported with infliximab. The aims of this study were to explore adverse event signals with infliximab by using a long period of post-marketing experience, stratifying by indication.

METHODS

The relative reporting of infliximab adverse events to the U.S. Food and Drug Administration (FDA) was assessed with the public release version of the adverse event reporting system (AERS) database from 1968 to third quarter 2005. On the basis of a systematic review of adverse events, Medical Dictionary for Regulatory Activities (MedDRA) terms were mapped to predefined categories of adverse events, including death, heart failure, hepatitis, infection, infusion reaction, lymphoma, myelosuppression, neuropathy, and obstruction. Disproportionality analysis was used to calculate the empiric Bayes geometric mean (EBGM) and corresponding 90% confidence intervals (EB05, EB95) for adverse event categories.

RESULTS

Infliximab was identified as the suspect medication in 18,220 reports in the FDA AERS database. We identified a signal for lymphoma (EB05 = 6.9), neuropathy (EB05 = 3.8), infection (EB05 = 2.9), and bowel obstruction (EB05 = 2.8). The signal for granulomatous infections was stronger than the signal for non-granulomatous infections (EB05 = 12.6 and 2.4, respectively). The signals for bowel obstruction and infusion reaction were specific to patients with IBD; this suggests potential confounding by indication, especially for bowel obstruction.

CONCLUSIONS

In light of this additional evidence of risk of lymphoma, neuropathy, and granulomatous infections, clinicians should stress this risk in the shared decision-making process.

Novel statistical tools for monitoring the safety of marketed drugs

Robust tools for monitoring the safety of marketed therapeutic products are of paramount importance to public health. In recent years, innovative statistical approaches have been developed to screen large post-marketing safety databases for adverse events (AEs) that occur with disproportionate frequency. These methods, known variously as quantitative signal detection, disproportionality analysis, or safety data mining, facilitate the identification of new safety issues or possible harmful effects of a product. In this article, we describe the statistical concepts behind these methods, as well as their practical application to monitoring the safety of pharmaceutical products using spontaneous AE reports. We also provide examples of how these tools can be used to identify novel drug interactions and demographic risk factors for adverse drug reactions. Challenges, controversies, and frontiers for future research are discussed.

An evaluation of computer-aided disproportionality analysis for post-marketing signal detection

To understand the value of computer-aided disproportionality analysis (DA) in relation to current pharmacovigilance signal detection methods, four products were retrospectively evaluated by applying an empirical Bayes method to Merck's post-marketing safety database. Findings were compared with the prior detection of labeled post-marketing adverse events. Disproportionality ratios (empirical Bayes geometric mean lower 95% bounds for the posterior distribution (EBGM05)) were generated for product-event pairs. Overall (1993-2004 data, EBGM05> or =2, individual terms) results of signal detection using DA compared to standard methods were sensitivity, 31.1%; specificity, 95.3%; and positive predictive value, 19.9%. Using groupings of synonymous labeled terms, sensitivity improved (40.9%). More of the adverse events detected by both methods were detected earlier using DA and grouped (versus individual) terms. With 1939-2004 data, diagnostic properties were similar to those from 1993 to 2004. DA methods using Merck's safety database demonstrate sufficient sensitivity and specificity to be considered for use as an adjunct to conventional signal detection methods.

Accounting for multiplicity in the evaluation of "signals" obtained by data mining from spontaneous report adverse event databases

Surveillance of drug products in the marketplace continues after approval, to identify rare potential toxicities that are unlikely to have been observed in the clinical trials carried out before approval. This surveillance accumulates large numbers of spontaneous reports of adverse events along with other information in spontaneous report databases. Recently developed empirical Bayes and Bayes methods provide a way to summarize the data in these databases, including a quantitative measure of the strength of the reporting association between the drugs and the events. Determining which of the particular drug-event associations, of which there may be many tens of thousands, are real reporting associations and which random noise presents a substantial problem of multiplicity because the resources available for medical and epidemiologic followup are limited. The issues are similar to those encountered with the evaluation of microarrays, but there are important differences. This report compares the application of a standard empirical Bayes approach with micorarray-inspired methods for controlling the False Discovery Rate, and a new Bayesian method for the resolution of the multiplicity problem to a relatively small database containing about 48,000 reports. The Bayesian approach appears to have attractive diagnostic properties in addition to being easy to interpret and implement computationally.

Fluvastatin and hepatic reactions: a signal from spontaneous reporting in Italy

BACKGROUND

Signal detection is a crucial element in recognising new adverse drug reactions (ADRs) as soon as possible. HMG-CoA reductase inhibitors ('statins'), the most potent cholesterol-lowering drugs, are generally well tolerated but can occasionally lead to liver toxicity. Pre- and postmarketing studies on statins revealed an incidence of 0.1-3% elevation in hepatic transaminase levels. However, these elevations are asymptomatic, reversible, dose related or probably due to other causes. Postmarketing studies clearly showed the lack of evidence of hepatotoxicity from statins, apart from some isolated case reports of serious hepatic damage described in the literature. It is still unclear whether serious hepatic reactions are dose related and more frequent than the expected rate in the general population.

OBJECTIVE

In this study, the hypothesis that fluvastatin could cause serious liver injuries more than the other statins is investigated, in the light of a quantitative and qualitative signal analysis, drug consumption data and evidence from the literature.

METHODS

The Italian Interregional Group of Pharmacovigilance (Gruppo Interregionale di Farmacovigilanza; GIF) is an example of signal detection within the Italian spontaneous ADR reporting system. The GIF database holds reports of suspected ADRs submitted by five Italian pharmacovigilance regional centres. In the GIF database, all reports of suspected ADRs are classified according to the WHO criteria for causality assessment. The reactions are coded according to the WHO Adverse Reaction Terminology and classified as serious or non-serious events on the basis of the WHO Critical Term List. Every 6 months the GIF database is analysed to extract potential signals through a qualitative case-by-case analysis and using a quantitative methodology called proportional reporting ratio (PRR). This methodology permitted us to identify the potential signal 'fluvastatin and hepatic reactions'.

RESULTS

At 31 December 2004, the GIF database contained 35 757 reports with an annual reporting rate of 170 reports per million inhabitants. We found a total of 1260 reports of ADRs related to statins, including 178 of hepatic reactions. Sixty-nine reports were attributed to fluvastatin, which showed the highest PRR in comparison with the other statins. Fluvastatin was associated with 33 serious reactions, mainly hepatitis and cholestatic hepatitis. The number of reports of severe hepatotoxicity associated with fluvastatin started to increase from 2002. About half of them did not report other suspected or concomitant drugs and in one third the hepatotoxicity occurred after <1 month of therapy. Twenty-seven out of 33 patients were female, and fluvastatin was administered at 80 mg/day in 81% of cases reporting complete data on drug dosage.

CONCLUSION

In the literature, serious hepatic reactions are rarely described in patients taking statins; however, data gathered by GIF suggest that cases of hepatotoxicity are reported more often than expected. In addition, GIF data seem to reveal that fluvastatin is more likely to cause hepatic reactions than the other statins. However, this is a preliminary signal and future evaluations are certainly needed to confirm it and to quantify this possible risk.

Use and safety profile of antiepileptic drugs in Italy

OBJECTIVE

To analyse and discuss the use and the safety profile of individual antiepileptic drugs (AEDs) in Italy.

METHODS

The AED safety data referred to the period January 1988-June 2005 and were obtained from the database of the Italian Interregional Group of Pharmacovigilance (GIF). This database collects all spontaneous reports of suspected adverse drug reactions (ADRs) from six Italian regions which are the main contributors to the Italian spontaneous reporting system. Individual AED consumption data (defined daily dose/1,000 inhabitants per day) in the GIF area and in the whole of Italy referred to the period January 2003-June 2005 and were derived from drug sales data (Institute for Medical Statistics Health).

RESULTS

Phenobarbital was the most frequently used AED in the GIF area (4.26 DDD/1,000 inhabitants per day) followed by carbamazepine (1.97), valproic acid (1.33) and gabapentin (1.10). AED consumption in the whole of Italy showed a similar pattern. Gabapentin was the most frequently used AED among newer AEDs. In the GIF database 37,906 reports (up to June 2005) were present; 666 of them (1.76%) were associated with at least one AED (Anatomical Therapeutic Chemical code N03A). The AED with the highest number of reports was carbamazepine (208 reports) followed by phenobarbital (98), gabapentin (80), phenytoin (56), valproic acid (55), lamotrigine (51), oxcarbazepine (43) and vigabatrin (35). Use and toxicity profile were evaluated only for AEDs associated with at least 30 reports. Skin reactions were the most frequently reported ADRs, followed by haematological, general condition, hepatic, neurological and gastrointestinal adverse reactions. Phenobarbital, lamotrigine, carbamazepine and phenytoin had the highest percentage of skin reactions (69, 67, 60 and 54%, respectively). Many haematological reactions were reported for each AED; the highest percentage was related to valproic acid (25%). Vigabatrin was associated with the highest percentage of reactions related to hearing, vision and other senses (97%). Phenytoin and valproic acid had the highest percentage of hepatic reactions (30 and 20%), whereas gabapentin of nervous system, psychiatric, gastrointestinal and urinary reactions (26, 21, 21 and 14%, respectively) and phenobarbital of musculoskeletal reactions (13%).

CONCLUSIONS

In Italy antiepileptic drug therapy appears to be still dominated by traditional drugs. Our analysis showed a different safety profile related to each AED. Some of the drug-adverse reaction associations discussed are not included in the Italian drug leaflets or have not been reported before in the literature.

Innovations for the Future of Pharmacovigilance

Post-marketing pharmacovigilance involves the review and management of safety information from many sources. Among these sources, spontaneous adverse event reporting systems are among the most challenging and resource-intensive to manage. Traditionally, efforts to monitor spontaneous adverse event reporting systems have focused on review of individual case reports. The science of pharmacovigilance could be enhanced with the availability of systems-based tools that facilitate analysis of aggregate data for purposes of signal detection, signal evaluation and knowledge management. GlaxoSmithKline (GSK) recently implemented Online Signal Management (OSM) as a data-driven framework for managing the pharmacovigilance of marketed products. This pioneering work builds upon the strong history GSK has of innovation in this area. OSM is a software application co-developed by GSK and Lincoln Technologies that integrates traditional pharmacovigilance methods with modern quantitative statistical methods and data visualisation tools. OSM enables the rapid identification of trends from the individual adverse event reports received by GSK. OSM also provides knowledge-management tools to ensure the successful tracking of emerging safety issues. GSK has developed standard procedures and 'best practices' around the use of OSM to ensure the systematic evaluation of complex safety datasets. In summary, the implementation of OSM provides new tools and efficient processes to advance the science of pharmacovigilance.

Atypical Antipsychotics and Pituitary Tumors: A Pharmacovigilance Study

STUDY OBJECTIVE

To analyze the disproportionality of reporting of hyperprolactinemia, galactorrhea, and pituitary tumors with seven widely used antipsychotic drugs.

DESIGN

Retrospective pharmacovigilance study.

DATA SOURCE

United States Food and Drug Administration's Adverse Event Reporting System (AERS) database.

INTERVENTION

We initially identified higher-than-expected postmarketing reports of pituitary tumors associated with risperidone, a potent dopamine D2-receptor antagonist antipsychotic, by analyzing reporting patterns of these tumors in the AERS database. To further examine this association, we analyzed disproportionate reporting patterns of pituitary tumor reports for seven antipsychotics with different affinities for blocking D2 receptors: aripiprazole, clozapine, olanzapine, quetiapine, risperidone, ziprasidone, and haloperidol.

MEASUREMENTS AND MAIN RESULTS

To conduct both of these analyses, we used the Multi-item Gamma Poisson Shrinker (MGPS) data mining algorithm applied to the AERS database. The MGPS uses a Bayesian model to calculate adjusted observed:expected ratios of drug-adverse event associations (Empiric Bayes Geometric Mean [EBGM] values) in huge drug safety databases. The higher the adjusted reporting ratio, or EBGM value, the greater the strength of the association between a drug and an adverse event. Risperidone had the highest adjusted reporting ratios for hyperprolactinemia (EBGM 34.9, 90% confidence interval [CI] 32.8-37.1]), galactorrhea (EBGM 19.9, 90% CI 18.6-21.4), and pituitary tumor (EBGM 18.7, 90% CI 14.9-23.3) among the seven antipsychotics, and one of the highest scores for all drugs in the AERS database. Some tumors were associated with visual field defects, hemorrhage, convulsions, surgery, and severe (>10-fold) prolactin elevations. The EBGM values for risperidone for these adverse events were higher in women, but high EBGM values for these events were also seen in men and children. Moreover, the rank order of the EBGM values for pituitary tumors corresponded to the affinities of these seven drugs for D2 receptors.

CONCLUSION

Treatment with potent D2-receptor antagonists, such as risperidone, may be associated with pituitary tumors. These findings are consistent with animal (mice) studies and raise the need for clinical awareness and longitudinal studies.