The application of knowledge discovery in databases to post-marketing drug safety: example of the WHO database.

After market launch, new information on adverse effects of medicinal products is almost exclusively first highlighted by spontaneous reporting. As data sets of spontaneous reports have become larger, and computational capability has increased, quantitative methods have been increasingly applied to such data sets. The screening of such data sets is an application of knowledge discovery in databases (KDD). Effective KDD is an iterative and interactive process made up of the following steps: developing an understanding of an application domain, creating a target data set, data cleaning and pre-processing, data reduction and projection, choosing the data mining task, choosing the data mining algorithm, data mining, interpretation of results and consolidating and using acquired knowledge. The process of KDD as it applies to the analysis of spontaneous reports can be exemplified by its routine use on the 3.5 million suspected adverse drug reaction (ADR) reports in the WHO ADR database. Examples of new adverse effects first highlighted by the KDD process on WHO data include topiramate glaucoma, infliximab vasculitis and the association of selective serotonin reuptake inhibitors (SSRIs) and neonatal convulsions. The KDD process has already improved our ability to highlight previously unsuspected ADRs for clinical review in spontaneous reporting, and we anticipate that such techniques will be increasingly used in the successful screening of other healthcare data sets such as patient records in the future.

Ingestion of a quadhelix appliance requiring surgical removal: a case report.

This report presents an unusual case, whereby a 13-year-old Down's syndrome boy accidentally swallowed a removable quadhelix appliance that subsequently required surgical removal. The paper discusses management strategies for patients who have accidentally swallowed components of their orthodontic appliance. It also highlights the need for orthodontists to consider limited objective treatment options for certain patient groups.

Data mining in spontaneous reports.

The increasing size of spontaneous report data sets and the increasing capability for screening such data due to increases in computational power has led to a recent increase in interest and use of data mining on such data. While data mining plays an important role in the analysis of spontaneous reports, there is general debate on how and when data mining should be best performed. While the cornerstone principles for data mining of spontaneous reports have been in place since the 1960s, several significant changes have occurred to make their use widespread. Superficially the Bayesian methods seem unnecessarily complex, particularly given the nature of the data, but in practice implementation in Bayesian framework gives clear benefits. There are difficulties evaluating the performance of the methods, but they work and save resources in managing large data sets. The use of neural networks allows more sophisticated pattern recognition to be performed.

Introducing triage logic as a new strategy for the detection of signals in the WHO Drug Monitoring Database.

PURPOSE: An important role for the WHO Programme for International Drug Monitoring is to identify signals of international drug safety problems as early as possible. The signal detection strategy, operated at the Uppsala Monitoring Centre (UMC), gave too many drug-adverse drug reaction (ADR) combinations for individual review. Therefore additional selection strategies were needed to improve the likely signal-to-noise ratio and for the UMC to complement the efforts of national centres in an efficient way. METHODS: The combinations database of the first quarter of 2001 was analysed using algorithms representing different strategies for finding relevant signals using triage logic. RESULTS: The strategies that together gave a manageable number of combinations, i.e. around 600, for further consideration in a single quarter were the algorithms for 'Rapid reporting increase', 'Serious reaction and new drug' and 'Special interests'. These filters began to be used routinely on the combinations database in late 2001. CONCLUSIONS: While stressing that human review is essential, triage strategies are useful when attempting analysis of large amounts of data. By definition, the use of triage strategies may exclude some potential signals from consideration, although the intention is to improve the chances of detection by focussing on areas of greatest importance.

Data-mining analyses of pharmacovigilance signals in relation to relevant comparison drugs.

OBJECTIVE: The aim of this paper is to demonstrate the usefulness of the Bayesian Confidence Propagation Neural Network (BCPNN) in the detection of drug-specific and drug-group effects in the database of adverse drug reactions of the World Health Organization Programme for International Drug Monitoring. METHODS: Examples of drug-adverse reaction combinations highlighted by the BCPNN as quantitative associations were selected. The anatomical therapeutic chemical (ATC) group to which the drug belonged was then identified, and the information component (IC) was calculated for this ATC group and the adverse drug reaction (ADR). The IC of the ATC group with the ADR was then compared with the IC of the drug-ADR by plotting the change in IC and its 95% confidence limit over time for both. RESULTS: The chosen examples show that the BCPNN data-mining approach can identify drug-specific as well as group effects. In the known examples that served as test cases, beta-blocking agents other than practolol are not associated with sclerosing peritonitis, but all angiotensin-converting enzyme inhibitors are associated with coughing, as are antihistamines with heart-rhythm disorders and antipsychotics with myocarditis. The recently identified association between antipsychotics and myocarditis remains even after consideration of concomitant medication. CONCLUSION: The BCPNN can be used to improve the ability of a signal detection system to highlight group and drug-specific effects.

Availability of medicines in the European Union: results from the EURO-Medicines project.

OBJECTIVE: There is at present no comprehensive directory of medicines available in European countries. Such a directory would be valuable to policy analysts, clinicians, regulatory agencies, pharmaceutical companies and consumer groups. The aim of this project was to compile such a directory of all medicines marketed in each of the European Union member countries. METHODS: Lists of medicines for each country, compiled from several national sources, classified by Anatomical-Chemical-Therapeutic (ATC) code. Census date was late 1998. RESULTS: A comprehensive directory was created using data from 14 of the 15 European Union countries. Numbers of trade names and of active ingredients varied widely, from Germany with 18,554 and 1,973, respectively, to Denmark with 1,915 and 1,016, respectively. In individual therapeutic areas, there were variations in the numbers of active ingredients available: the least variation between countries was in antineoplastic medicines (ATC code L, maximum number available in any country 101, minimum 60) and wider variation in alimentary (ATC code A, maximum 256. minimum 103) or cardiovascular (ATC code C, maximum 269, minimum 112). Only 7% of all the active ingredients were available in all the countries studied. The Scandinavian countries had the greatest proportion of active ingredients (60%) available in all other countries. Each country had a number of active ingredients available only in that country Italy had the largest number of these. CONCLUSIONS: The directory illustrates the wide variations in the availability of medicines across the European Union. The range of drugs available in each country represents differences in regulatory and market policies, as well as cultural and historic differences. This directory lends itself to many further analyses.

Antipsychotic drugs and heart muscle disorder in international pharmacovigilance: data mining study.

OBJECTIVES: To examine the relation between antipsychotic drugs and myocarditis and cardiomyopathy. DESIGN: Data mining using bayesian statistics implemented in a neural network architecture. SETTING: International database on adverse drug reactions run by the World Health Organization programme for international drug monitoring. MAIN OUTCOME MEASURES: Reports mentioning antipsychotic drugs, cardiomyopathy, or myocarditis. RESULTS: A strong signal existed for an association between clozapine and cardiomyopathy and myocarditis. An association was also seen with other antipsychotics as a group. The association was based on sufficient cases with adequate documentation and apparent lack of confounding to constitute a signal. Associations between myocarditis or cardiomyopathy and lithium, chlorpromazine, fluphenazine, haloperidol, and risperidone need further investigation. CONCLUSIONS: Some antipsychotic drugs seem to be linked to cardiomyopathy and myocarditis. The study shows the potential of bayesian neural networks in analysing data on drug safety.

The WHO Programme for International Drug Monitoring, its database, and the technical support of the Uppsala Monitoring Center.

We describe the development of an international adverse reaction database. The operational responsibility for technical aspects of international drug monitoring are run by the Uppsala Monitoring Center (UMC). The system is based on interchange of adverse reaction information between national drug monitoring centers in 60 countries. Collectively these centers provide more than 150,000 individual reports annually of reactions suspected of being drug induced. The cumulative database constructed from these reports now comprises over 2 million records. Compatibility of different data collection systems that need to communicate with each other has been achieved through harmonization rather than standardization. The design of the new system was driven by the needs of existing and prospective users in terms of data fields and functionality. The data set required in the original WHO case reports form was the lowest common denominator consistent with being useful for signal generation and evaluation. The new database has an unlimited number of data fields. The WHO system relies on information being transferred, stored, and retrieved in a timely and secure way. Through the use of sophisticated exchange server technology, the Internet can be used as a transport medium for data and document transfer with guaranteed security and client authentication.

The management of adverse drug reactions: from diagnosis to signal.

Severe adverse reactions (ADR) are uncommon and benefits of drug treatment usually outweigh the disadvantages. Definitions and guidelines for managing adverse drug reactions (ADRs) are proposed mainly with the aim of helping to keep the possibility of ADR at the front of our minds and making the right diagnosis. As spontaneous reporting is a cornerstone in post-marketing surveillance, the reasons for under-reporting are thoroughly analysed. Finally, a reminder is given that communication of information should be encouraged within the medical community and that everyone who is involved in the treatment of patients has an active part to play in the network of pharmacovigilance.

Adverse drug reactions: definitions, diagnosis, and management.

We define an adverse drug reaction as "an appreciably harmful or unpleasant reaction, resulting from an intervention related to the use of a medicinal product, which predicts hazard from future administration and warrants prevention or specific treatment, or alteration of the dosage regimen, or withdrawal of the product." Such reactions are currently reported by use of WHO's Adverse Reaction Terminology, which will eventually become a subset of the International Classification of Diseases. Adverse drug reactions are classified into six types (with mnemonics): dose-related (Augmented), non-dose-related (Bizarre), dose-related and time-related (Chronic), time-related (Delayed), withdrawal (End of use), and failure of therapy (Failure). Timing, the pattern of illness, the results of investigations, and rechallenge can help attribute causality to a suspected adverse drug reaction. Management includes withdrawal of the drug if possible and specific treatment of its effects. Suspected adverse drug reactions should be reported. Surveillance methods can detect reactions and prove associations.

The value of reporting therapeutic ineffectiveness as an adverse drug reaction.

Therapeutic ineffectiveness is a frequent drug-related problem that can occur in a variety of different situations and be caused by different mechanisms. Examples are inappropriate use, interactions or metabolic abnormalities. Observations in patients of unexpected ineffectiveness can provide important information with regard to such situations. Therefore, ineffectiveness--especially when unexpected or unexplained--is a potentially important reportable event in pharmacovigilance. The terms regarding ineffectiveness in the WHO Adverse Reaction Terminology (WHOART) have been recently revised in order to enable optimal coding of such case reports.

Spontaneous ADR reporting and drug safety signal induction in perspective. To honour Professor Jens Schou.

Drug safety signals will continue to come mainly from the reporting of alert clinicians and every effort should be made to enhance this and to ease the process. The use of multipurpose health databases for finding signals has much potential, if they are better planned so that the appropriate data is captured and examined routinely. Consumer reports give us information about their concerns and should not be ignored. Better information is needed on poisoning, drug misuse and on herbal remedies. The analysis of signals must be improved and speeded up, if we are even to maintain our current safety standards, given the global release of 'blockbuster products'. Benefit-risk analysis of medicines needs to be better understood in relationship to actual clinical use, both from an individual and public health perspective. Such analysis should become more logical rather than just listing the benefits and risks, and then expressing an essentially unsupported opinion. This is essential if therapies are to be compared, and their costs justified. The communication of medicines safety and benefit-risk information to clinicians, other health professionals and patients is an area where there needs to be considerable improvements. We need to be better informed about the consequences of guidance and warnings given, so that we may improve the service we offer to recipients. In the future, information technology, which provides exciting possibilities with what it offers now, will help us with all the above challenges.

The accelerating need for pharmacovigilance.

The search for new drugs takes on greater complexity with increasing knowledge, allowing more sophisticated therapeutic interventions. At the same time there is increasing commercial pressure for the pharmaceutical industry to find 'blockbuster' drugs which will be marketed globally to maximise profit in the shortest possible time. Other changes in the industry--shortened times for drug development and increasing outsourcing of functions--make for an environment where some pre-marketing safety issues may go unnoticed. The increasing challenge to pharmacovigilance is not only to be able to find early signals of drug problems, but to rapidly determine the true benefits and risks. We may not have adequate systems to prevent unnecessary harm from globally marketed drugs.

A retrospective evaluation of a data mining approach to aid finding new adverse drug reaction signals in the WHO international database.

BACKGROUND: The detection of new drug safety signals is of growing importance with ever more new drugs becoming available and exposure to medicines increasing. The task of evaluating information relating to safety lies with national agencies and, for international data, with the World Health Organization Programme for International Drug Monitoring. RATIONALE: An established approach for identifying new drug safety signals from the international database of more than 2 million case reports depends upon clinical experts from around the world. With a very large amount of information to evaluate, such an approach is open to human error. To aid the clinical review, we have developed a new signalling process using Bayesian logic, applied to data mining, within a confidence propagation neural network (Bayesian Confidence Propagation Neural Network; BCPNN). Ultimately, this will also allow the evaluation of complex variables. METHODS: The first part of this study tested the predictive value of the BCPNN in new signal detection as compared with reference literature sources (Martindale's Extra Pharmacopoeia in 1993 and July 2000, and the Physicians Desk Reference in July 2000). In the second part of the study, results with the BCPNN method were compared with those of the former signalling procedure. RESULTS: In the study period (the first quarter of 1993) 107 drug-adverse reaction combinations were highlighted as new positive associations by the BCPNN, and referred to new drugs. 15 drug-adverse reaction combinations on new drugs became negative BCPNN associations in the study period. The BCPNN method detected signals with a positive predictive value of 44% and the negative predictive value was 85%. 17 as yet unconfirmed positive associations could not be dismissed with certainty as false positive signals. Of the 10 drug-adverse reaction signals produced by the former signal detection system from data sent out for review during the study period, 6 were also identified by the BCPNN. These 6 associations have all had a more than 10-fold increase of reports and 4 of them have been included in the reference sources. The remaining 4 signals that were not identified by the BCPNN had a small, or no, increase in the number of reports, and are not listed in the reference sources. CONCLUSION: Our evaluation showed that the BCPNN approach had a high and promising predictive value in identifying early signals of new adverse drug reactions.

Spontaneous reporting of suspected adverse reactions to antihistamines: a national and international perspective.

Twenty-one sedative and five non-sedating antihistamines are presently available on the UK market. Analysis of spontaneous reports of suspected adverse drug reactions was performed for those drugs with more than 90 reported reactions on the UK ADROIT database to the period ending 31 December 1997. Thus nine antihistamines, (four sedative antihistamines and five non-sedating antihistamines) were included. For each of the four sedative agents (azatidine, chlorpheniramine, diphenhydramine and trimeprazine), reactions associated with the cardiovascular, gastrointestinal, central nervous systems, the skin, general and psychiatric reactions made up more than 70% of total reported reactions. For all four agents, fatal reactions constituted less than 2.5% of total reactions. For each of the five non-sedating agents (acrivastine, astemizole, cetirizine, loratidine and terfenadine), reactions associated with the cardiovascular, gastrointestinal central nervous systems, the skin, general and psychiatric disorders together made up more than 75% of total reported reactions. For all five non-sedating agents, fatal reactions constituted less than 1% of total reactions. However, there were 21 reports of fatality in association with terfenadine, 11 (52%) of which were either sudden deaths or those associated with a cardiac rate of rhythm reaction. Analysis of the WHO database for non-sedating drugs showed a similar pattern, with terfenadine being associated with the highest frequency of reports of potentially serious arrhythmias and of sudden death and death related to disturbances of cardiac rate and rhythm combined. Despite the limitations of spontaneous reporting systems, comparison of the benefit-risk profiles of drugs using this data within a class of drugs can provide valuable information, and pharmacovigilance of antihistamines (and all other agents) using this and other means should continue for the lifetime of their use in humans.

From association to alert--a revised approach to international signal analysis.

From the inception of the WHO international drug monitoring programme, the main aim has been to detect signals of adverse reaction problems as early as possible. The Uppsala Monitoring Centre (UMC), is now in a better position to fulfil this mission. Using the latest technology, new tools have been developed which allow for rapid, robust and comprehensive data mining of the WHO database. Based on retrospective time scans made during the pilot phase the current threshold used is the 97.5% confidence level of difference from the generality of the database. To maximize the capacity for picking up signals, we intend to extend today's panel of expert consultants, as well as doing our own review. The new system includes an enhanced follow-up list of signals, a 're-signalling' procedure and a cumulative historical file of all drug-ADR associations. Already we produce some 50 signals per year, cisapride and tachycardia being an example of a controversial signal only recently accepted. With the addition of new tools for follow-up of important signals such as complex variable data mining techniques, and the combination of WHO ADR data with sales and prescription figures from the IMS, we will be able to provide more information that should benefit regulators, producers, prescribers, and most importantly, the users of medicines.

A Bayesian neural network method for adverse drug reaction signal generation.

OBJECTIVE: The database of adverse drug reactions (ADRs) held by the Uppsala Monitoring Centre on behalf of the 47 countries of the World Health Organization (WHO) Collaborating Programme for International Drug Monitoring contains nearly two million reports. It is the largest database of this sort in the world, and about 35,000 new reports are added quarterly. The task of trying to find new drug-ADR signals has been carried out by an expert panel, but with such a large volume of material the task is daunting. We have developed a flexible, automated procedure to find new signals with known probability difference from the background data. METHOD: Data mining, using various computational approaches, has been applied in a variety of disciplines. A Bayesian confidence propagation neural network (BCPNN) has been developed which can manage large data sets, is robust in handling incomplete data, and may be used with complex variables. Using information theory, such a tool is ideal for finding drug-ADR combinations with other variables, which are highly associated compared to the generality of the stored data, or a section of the stored data. The method is transparent for easy checking and flexible for different kinds of search. RESULTS: Using the BCPNN, some time scan examples are given which show the power of the technique to find signals early (captopril-coughing) and to avoid false positives where a common drug and ADRs occur in the database (digoxin-acne; digoxin-rash). A routine application of the BCPNN to a quarterly update is also tested, showing that 1004 suspected drug-ADR combinations reached the 97.5% confidence level of difference from the generality. Of these, 307 were potentially serious ADRs, and of these 53 related to new drugs. Twelve of the latter were not recorded in the CD editions of The physician's Desk Reference or Martindale's Extra Pharmacopoea and did not appear in Reactions Weekly online. CONCLUSION: The results indicate that the BCPNN can be used in the detection of significant signals from the data set of the WHO Programme on International Drug Monitoring. The BCPNN will be an extremely useful adjunct to the expert assessment of very large numbers of spontaneously reported ADRs.