Current Work of WHO

CURRENT WORK

·    The database of the WHO Programme is a unique reference source used in many different situations. When a national centre receives the first report of an unfamiliar drug-reaction association, on-line search facilities to Vigibase are at the disposal of national centres to find out whether a similar observation has been made elsewhere in the world. If so, the initial signal may be strengthened. For more complicated studies, the UMC staff can make customised searches in the database.

From the database, cohorts of patients affected by similar kinds of drug-associated reactions may be retrieved. By looking for common features in these reports, risk factors and hypotheses for underlying mechanisms may be revealed.

·     The methodology developed at the UMC using a BCPNN technique in analysing the database was put into routine use already in 1998. The concept of data mining, or knowledge detection as it also may be called, is now operating to support all countries in their work. It is based on artifi-cial intelligence using a Bayesian logic system. It has been fully validated and is under continuous development and has been presented in a doctoral thesis by Andrew Bate in 2003. (Bate et al., 1998a,b, 2000; Lindquist et al., 1999; Bate, 2000; Lindquist, 2000; Orre et al., 2000; Ståhl et al., 2004.)

A combination of automatic signalling devices and scanning by experienced medical personnel is consid-ered most advantageous to fulfil successfully the original aim of the programme, that is the early identification of new ADRs. This method provides a quantitative measure of the strength of association of a drug-reaction combination in the database. Combi-nations that occur more frequently than expected as compared with the generality of the database are high-lighted.

When the new data has been processed and entered into the ADR database, a BCPNN scan is run to generate statistical measurements for each drug-ADR combination. The resulting data are presented in two steps:

·     The resulting Combinations database (Combina-tion: ADR data elements occurring together in ADR reports) is made available to national centres and to pharmaceutical companies, in the latter case including only information on the company’s own patented products. The database is presented in a computerised form which facilitates searching and sorting of the information.

·     An Associations database (Association: Combina-tions selected from a database on a quantitative basis) is generated by selecting those combinations that pass a preset threshold. Based on the results of the test runs of the BCPNN, the threshold level for associations is that of the lower 95% confi-dence limit of the information component (IC) value crossing zero when a new batch of reports is added.

All associations are followed automatically for 2 years, the data being checked at 6-month inter-vals. After the final listing, an association may be reintroduced for another 2-year follow-up. The associations are also copied to a cumulative log file (history file), which will serve as a filter to exclude combinations that have in previous quarters passed the threshold level. This will prevent drug-ADR combinations with a confidence limit fluctuating around zero from being fed into the review process repetitiously.

A panel of experts has been established to analyse reactions pertaining to particular body systems. The Associations database is sent to the expert review panel for evaluation. Before distributing the database, associations are checked against standard reference sources [e.g. Physician’s Desk Reference (PDR), Martindale] and the published literature (using, e.g., Medline and Reactions Weekly). This facili-tates the review and identifies those associations that are, if not generally known, at least identified previously.

Searching and sorting of the associations data can be done not only on drug, ADR and the various statistical measurements but also on system organ class (SOC) and on therapeutic drug groups using the anatomical-therapeutic-chemical (ATC) classification. To ensure that there are at least two reviewers per SOC, we intend to extend the panel of reviewers from today’s 30 experts to around double. Recently, a special panel of experts to review reactions to herbal preparations was set up.

To the Associations stage, the process is purely quantitative, but clinical knowledge and judgement is necessary for the evaluation of associations and is provided by the national centres and expert reviewers. Short summaries of their findings are circulated to participating national centres in a memorandum called ‘Signal’. An investigation has demonstrated that the WHO Programme is successful in finding new drug-adverse reaction associations at an early stage and in providing useful information about them to national centres (Ståhl et al., 2003) Individualised sections of the Signal document are provided to companies on their patented products for their comment.

To aid the expert reviewers, and also to facili-tate interpretation of the information presented in the Signal document, a set of guidelines is being used. As with the associations, signals will be reassessed after 2 years, with a possibility of re-introduction for follow-up and also copied to a history file for easy tracking. With the new follow-up procedures that are being introduced, a mechanism by which signals can be re-evaluated following new information will be in place. This enables, for example, renewed considera-tion of associations for which there initially was not enough information to merit signalling. Signals that are later supported by new evidence can also be high-lighted. The nature of the signal will determine what measures need be taken in terms of follow-up.

A larger number of variables than the routine drug-ADR combinations can also be considered using the Bayesian approach, as described above. One of the advantages of a neural network, as used in the BCPNN, is that it can search for patterns of associations between fields that are not determined a priori: it can find novel complex relationships. One of the outcomes of these analyses may be to identify patient subgroups that may be at particularly high risk of getting a specific adverse reaction when they have taken a specific drug. Another possibility is to establish that a drug safety problem is related to a particular country, or region, or a certain time period. By further developing the BCPNN method-ology for the analysis of the large amount of data in the WHO database, it is expected that not yet revealed risk factors for the development of drug-related ailments may be detected. The UMC develops and uses unsuper-vised pattern recognition methods to avoid too many preconceptions influencing investigations, which are then largely driven by the data itself. The approach also picks up, and allows analysis of, data and method-ological issues such as duplications in reports which may not be obvious (Norén, Orre and Bate, 2005).

·        The UMC has an important role to play as a communication centre – a clearing house for infor-mation on drug safety at the service of drug regula-tory agencies, pharmaceutical industry, researchers and other groups in need of drug safety informa-tion. Requests for special database searches and investigations are received from these parties at a rate of around 225 per year. In addition, flexible on-line retrieval programmes are made available by which the database users may perform a variety of standardised searches by themselves. Access for non-member parties is subjected to some confiden-tiality restrictions agreed by Programme members. Use of the information released is subject to a caveat document to explain its proper use. Detailed manuals for the on-line service and the customised retrievals on request are available from the Uppsala centre.

The UMC co-operates with WHO to provide drug safety information in the WHO Pharmaceuticals Newsletter, distributed by the Health Technology and Pharmaceuticals department of WHO headquarters, leading to a wider distribution of the information to all member countries of WHO. The UMC is responsi-ble for compiling information from national pharma-covigilance centres, including their adverse reaction bulletins, on warnings, recommendations and advice provided to health professionals in relation to the safe use of medicines.

The UMC co-operates with Adis International in the journal Reactions Weekly to provide additional information in the section ‘Adverse Reaction Case Reports’ in the journal. Any claim to a first report in ‘Reactions’ is supplemented, where possible, by supporting information from the WHO adverse reac-tions database.

Uppsala Reports is the name of a bulletin which is made freely available to all interested parties by the UMC. It provides an easy-to-read account of news about pharmacovigilance, the WHO Programme, its members and services.

Communications within the WHO Programme have improved with the increasing use of electronic communications media. The UMC is maintaining an e-mail discussion group called ‘Vigimed’, which allows for rapid exchange of information around the world on drug safety matters. Membership is restricted to persons connected to national pharmacovigilance centres, which means that confidential information before a issue is fully evaluated can be circulated.

The internet home page of the WHO Programme (http://www.who-umc.org) is a dynamic tool for communications with all clients of the UMC. Recently, the Products & Services division of UMC, dealing mainly with commercial customers, set up a new website presenting all of their services (http://www.umc-products.com).

·     International comparisons of drug safety reporting have been made (Lindquist, 1990, 2003; Lindquist and Edwards, 1993). These comparisons have shown important differences in country profiles of reporting. The differences between countries may be due to a variety of factors. Some of the differ-ences may be purely technical but others may relate to differences in medical practice, the use of medi-cal terms and societal influences such as media interest (Mills and Edwards, 1999). Sometimes, the difference in indications, doses of medicines and/or the routes of administration may be significant (Lindquist et al., 1996). It is sometimes alleged that these findings are not signals, but this is to take a narrow view of a ‘signal’ as simply a previously unreported medicine/ADR association, rather than to consider that any significant new evidence on a medicine-related risk is a signal (see WHO defini-tion – Edwards, 1997).

·     Definitions for a variety of pharmacovigilance terms have been proposed and accepted widely (Edwards, 1997). Within the WHO Programme, many definitions of commonly used terms, such as adverse reaction, side effect, adverse event and signal, have been worked out. These definitions contribute to a harmonised way of communicating both inside and outside the Programme (Edwards and Biriell, 1994).

·     Guidelines for signal finding have been proposed and widely accepted (Edwards et al., 1990). It is an important concept that a medicine-related signal from spontaneous reports should be consid-ered starting with the seriousness of the apparent signal and then appraising both the quantity of reports as well as the strength/quality of the infor-mation in those reports. Because the quality of information on a report is limited does not neces-sarily mean that the observation underlying it is less valid: but it does mean that objective assess-ment may be difficult or impossible. Assessing the weight of reported evidence is a complex clinical decision, which has further been aided by defini-tions of ‘certain’, ‘probable’, ‘possible’ and so on (Edwards, 1997).

·    The idea of the possibility of an exhaustive dataset being stored was initiated, and has become the ICH E2B project. A new WHO database, called Vigibase, containing all the fields was completed in 2002 (see paragraph below) (Lindquist, 1998). First with the Council of International Organisa-tions of Medical Sciences (CIOMS, 1995) and then with ICH, the UMC has developed a comprehen-sive set of data fields, which have been included in the new database, which is fully operational. In this data model, much more detailed information on each case may be stored.

The new UMC database has great complexity, and it seems unlikely that many of the available fields will be completed until a ‘paperless’ system comes into operation in several countries. The new database is fully compatible with the old one, so that reports both in the old WHO format and the new E2B format can still be accepted. To provide flexibility for users with varying requirements and sophistication is a great challenge, but we are hopeful that the new database will pave the way for the international availability of much more useful case data, without the need to go back to the original provider for more details.

Along with the provision of the new database, the UMC gives more active support to national centres over their IT development by offering VigiFlow, a web-based reporting software solution. The system accesses Vigibase over the internet, so no local instal-lations are required. Reports can be entered and accessed through password-protected, secure internet connection by the reporting doctor, regional centres or the national centre, and will automatically be trans-ferred to the international database. Many delays in the transmission of reports to the WHO are secondary to a variety of technical issues, which can now be minimised.

·        The UMC organises training courses to foster education and communication in pharmacovigi-lance with the main aim of supporting the devel-opment of national programmes for spontaneous ADR reporting. Since 1993, the UMC offers every second year a 2-week training course in adverse reactions and adverse reaction monitoring to which 25 healthcare profes-sionals are accepted from all over the world. The course is for 2 weeks and is divided into three consec-utive modules. The first is focused on spontaneous monitoring and the practicalities of managing a drug monitoring centre. This section also offers hands-on experience in using the database of the WHO Programme. The second module is an introduction to wider issues in pharmacoepidemiology. As it is more and more recognised that being able to communicate, often difficult issue in drug safety, is important, a third module on effective communication in pharmacovig-ilance has been added to the course.

There is an increasing trend towards local and regional meetings and courses in pharmacovigilance. The WHO Programme often takes part in such meet-ings, particularly those organised in developing coun-tries, to provide support and technical advice. UMC’s expertise is sought in the important WHO Public Health Programmes against HIV/AIDS and malaria, where new drugs causing new problems are used. UMC staff are commonly invited all over the world to speak at professional meetings.

·    Every year, representatives of national centres are invited to a meeting arranged jointly by WHO and one of the participating countries. At these meetings, technical issues are discussed, both in relation to how to improve global drug moni-toring in general and concerning individual drug safety problems. Because the meetings have very high attendance rates, they are important for the establishment and maintenance of personal relationships subsequently contributing to good communications.

The WHO Programme has developed a standard-ised adverse reaction terminology (WHOART) and a comprehensive index of reported drugs (WHO-DD), both of which have a utility beyond their importance to the monitoring system. These tools are used in the pre-marketing safety area, as well as for post-marketing studies by many pharmaceutical compa-nies. The WHO Drug Dictionary is unique in its coverage of drugs marketed throughout the world. It is available as computer files for inclusion in users’own software. The UMC has in conjunction with the introduction of the new database developed it further to incorporate more detailed information and make it compatible with the pre-standard proposed by the European Committee for Standardisation (CEN). A cooperation with IMS Health has started which will further improve the dictionary by making it more comprehensive and more up to date.

More recently, WHO has invited the UMC to be a partner in the WHO Family of Classifications (WHO-FIC). This includes all of the WHO classi-fications, notably the International Classification of Diseases. The plan is to link all of the WHO clas-sifications, which will include the WHO-DD and WHOART being linked to ICD to harmonise drug-induced disease with other illness classification.

·    There is a general need to quantify adverse reaction information. The WHO centre is working jointly with IMS International to analyse adverse reaction reports together with drug use data from different countries, and results of pilot studies have been published (Lindquist and Edwards, 1997; Lindquist et al., 1994, 1996, 1997). These analyses allow national differences in reporting rates to be further analysed for reasons that may be due to differ-ences in indications for use, medical practice and demographics. It is hoped that this type of analysis of international data will serve as a guide to the need for more precise pharmaco-epidemiological investigations and will be taken into regular use.

·        The UMC has been active in refining the concept of benefit–harm analysis for drug safety (Edwards, Wiholm and Martinez, 1996). The previous common pairing of benefit and risk does not provide a logical or helpful contrast: we need to know what are the benefits and their chance of occurring (benefit and effectiveness); and what is the harm and its chance of occurring (harm and risk). Effectiveness–risk analysis, often referred to as ‘benefit–risk assessment’ and also ‘risk manage-ment’ should be more than the subjective opin-ion of a group of experts and is in its infancy an objective way of considering drug therapy. The needs of managed care and the adoption of guide-lines for therapy in all therapeutic areas mean that there needs to be satisfactory methods for measuring effectiveness–risk in clinical practice for all major therapeutic interventions, so that those interventions may be compared. Safety must be seen as relative: there is no absolute safety. There is relativity in the risk or harm that one drug causes compared with another and in the risk or harm caused by a drug in relation to its effective-ness or benefits. Risk, above a certain incidence 1/1000 , is measurable in clinical trials, but we have much less information on safety than we need, because clinical trials are not well designed to elicit information about adverse effects. Addi-tional information on lower incidences comes from individual case reports of varying quality and quan-tity and from observational studies, which are not consistently performed with all drugs and all ADRs. The observational material, reports or stud-ies, is susceptible to various biases to different degrees. Most studies actually measure effective-ness (the frequency of a defined beneficial effect) and risk (the frequency of various defined aspects of harm). They do not measure benefit (the degree to which an individual feels improved by a ther-apy) or harm (the degree to which a person feels damaged by a therapy)

·     The concept and needs for benefit–risk commu-nication have been explored and developed. One of the widely quoted outcomes is the ‘Erice Declaration’, which proposes principles for such communication (Bowdler, 1997; Edwards, 1999; Edwards and Hugman, 1997; Edwards et al., 2000). With the aim of improving communications in pharmacovigilance, initiatives have been taken to call together representatives of all major groups involved in the provision of drug safety infor-mation. The Erice report on communicating drug safety information sets out the basis for further development in this area (UMC, 1998).

It is important that everyone should have a basic understanding of how science and medicine affect their lives and of the basis on which they should make decisions which will influence their health and welfare. Drug safety issues are high on the list of priorities in everyday life. The UMC has been actively committed for some years to the development of open, ethical communication and effective, modern commu-nications practice.

The publication of the two parts of ‘Viewpoint’ (see http://www.who-umc.org – Publications – View-points 1 and 2) is an example of this commitment. ‘Viewpoint’ provides a comprehensive picture of the complex and vital issues and questions surround-ing drug safety and the part played by the UMC, the WHO Programme and its members in improving public health and reducing the potential hazards to patients.

Viewpoint has been written and designed to be accessible to the widest possible audience: among the first of its kind in the world. Part 1 explains the basic concepts and issues in drug safety and risk management while Part 2 offers a more detailed and technical account of the science of international phar-macovigilance, but still in relatively simple language and concepts. Both are in full colour and extensively illustrated with pictures, graphs and diagrams.

A recognition of the importance of good communi-cation skills has led to many initiatives, UMC person-nel have been involved in the training of journalists in drug safety in various countries, and a new module on ‘Effective communications in pharmacovigilance’ has been added to the UMC training course on ADRs

·    A great number of publications are produced annually from the UMC, both technical which are intended for national centres in the WHO programme directly working with drug safety issues and publications for a wider audience with an interest in the field.

Some of the publications, as The Importance of Phar-macovigilance (WHO and UMC, 2002) and Safety Monitoring of Medicinal Products: Guidelines for Setting up and Running a Pharmacovigilance Centre (WHO and UMC, 2000) have been published jointly with WHO. The scientific work at UMC has led to the publication of two doctoral theses [Bate 2003; The use of Bayesian confidence propagation neural network in pharmacovigilance (Bate, 2003), and Lindquist, 2003, Seeing and observing in international pharmacovigi-lance – achievements and prospects in worldwide drug safety (Lindquist, 2003)].

Over the last 10 years, there have been 66 publications in scientific journals actively involving UMC staff.

·        It has become increasingly clear that adverse reaction monitoring must be extended to herbal remedies, not least because of the cultural change in developed countries where more and more people are turning to natural products. In response to this challenge, the UMC has taken initiatives to improve the classification systems for such medicines. In a joint project with institutions in the United Kingdom and the Netherlands, a system compatible with the ATC system used for modern, synthetic medicines has been developed. Input from experts from all parts of the world, represent-ing different therapeutic traditions, is indispensable for this project. The work has been done in collab-oration with experts in South Africa, the United States and Germany.

·    The UMC database is far enough advanced to be finding some herbal signals and an expert panel to analyse these herbals signals has recently been set up (Farah, 1998, 2000a, Farah et al., 2000b; Lindquist, Farah and Edwards, 2000).

·    The need for new pharmacovigilance approaches to deal with the aggressive global marketing of drugs has been identified (Edwards, 2000).