FDA Risk Management Workshop

Karen Goldberg reports on the FDA's development of guidance for good pharmacovigilance practices and pharmacoepidemiologic assessment

At a US FDA public workshop held in Washington, DC on 11 April 2003 the pharmaceutical industry and other stakeholders had the opportunity to express their views on the development of guidance entitled Risk Assessment of Observational Data: Good Pharmacovigilance Practices and Pharmacoepidemiologic Assessment¹.

The guidance is currently in the form of a concept paper and is being developed by the FDA as part of its obligations under the reauthorised Prescription Drug User Fee Act (PDUFA III ), which for the first time permits the agency to apply user fees to the evaluation of an applicant's risk management plan contained in its NDA or BLA (and supplements containing clinical data).

The paper provides advice on developing a pharmacovigilance plan - possibly as a component of a sponsor's comprehensive risk management plan - the sole focus of which is to detect new safety signals and/or evaluate already identified safety signals. A safety signal is defined by the FDA as an apparent excess of adverse events associated with a product's use that can be identified from post-approval observational data sources (e.g. case reports, case series and pharmacoepidemiologic [PE] studies).

Good pharmacovigilance practice

The paper emphasises that good pharmacovigilance practice usually begins by acquiring complete data from spontaneous adverse event (AE) reports, also called case reports, which are used to develop case series (when reports of two or more cases are identified in adverse event databases that associate a similar event with a product for interpretation). To this end, the FDA commissioner, Dr Mark McClellan, emphasised that the agency's strategy is to obtain better information, not more information, so that an AE can be thoroughly assessed in a timely manner.

Improving spontaneously reported case reports

The agency's concept paper outlined its views on the characteristics of a good case report, but sought comment on how the quality of these reports could be improved. The Pharmaceutical Research and Manufacturers of America (PhRMA), as represented by Linda Hostelley and Dr Wendy Stephenson, suggested simplified reporting procedures that reflect a balance between the clinical event and the intrusiveness of follow-up with the treating healthcare professional.

Other delegates agreed with PhRMA, adding that interviews should be structured to ensure that as much information as possible is collected during the first contact with the reporter. For example, a detailed checklist and/or specific questions tailored to the type of AE that has occurred should be used. The general consensus was that it is extremely difficult to obtain the missing data about an event after the initial report has been filed.

PhRMA also stressed the importance of educational initiatives and collaboration between industry, the FDA, academia, and healthcare and patient organisations. Dr Stephen Goldman, a consultant and the former medical director of the FDA safety information and adverse event reporting programme, MedWatch, noted that providing feedback to those who file a case report is a critical element in receiving high quality information.

Methods for feedback could be in the form of continuing education initiatives or notifying the healthcare professional of the outcome (e.g. labeling changes). PhRMA also brought up the possibility of an AE newsletter, although the information would have to be presented in a way that would not cause damage to companies by highlighting one product over another.

Safety signal identification through data mining

One suggestion made by the FDA in the concept paper was the development of case series through a technique called `data mining'. With this new technology, information could be extracted from large, complex AE databases in order to identify safety signals of AEs and the patterns in which they occur.

Many participants expressed concern about this technique, cautioning the FDA that its use is limited and it could result in false positive causality conclusions. Dr Robert Nelson, an FDA official-turned-consultant, stated that data mining is valuable for secondary alerts and the identification of previously unrecognised data relationships, but that its use is limited. Another consultant and former FDA official, Dr Gerald Faich, commented that false signals are inevitable but that data mining could be useful if restricted within a therapeutic class. PhRMA added that the information extracted through the use of data mining should be complementary to data taken from clinical trials and pharmacoepidemiologic studies. Finally, Dr Sidney Kahn, president of Pharmacovigilance & Risk Management, Inc. said that data mining is more appropriate for regulatory authorities than pharmaceutical companies.

Causality assessments at individual case level

According to the FDA, it is difficult to determine from an individual case report that an AE was product-induced. However, certain features in a case report may support the association between the use of a product and an AE, for example, if an event (e.g. a type 1 allergic reaction) occurred in the expected timeframe for that event. When assessing a case report as well as all other safety data available, the agency believes that it may be possible to at least determine the degree of causality (`probable', `possible' or `unlikely') between the use of a product and an adverse event.

Most commenters expressed the view that causality assessments on a routine basis are of little value because information is often incomplete and there is no standardised method to ensure consistency. Dr Faich pointed out that routine assessments by the FDA were terminated in 1983 as they were not deemed valuable and because they contributed to backlogging. However, participants also noted that under certain circumstances (e.g. for serious trial ADR reports) causality assessments can be useful.

Pharmacoepidemiologic assessment

In the paper the FDA stated that when safety signals are identified pre- or post-approval, or when `at risk' populations have not been adequately studied, it may request that the sponsor develop a pharmacovigilance plan above and beyond the routine post-marketing spontaneous reporting. This may include sponsor commitments to use registries as surveillance tools, perform additional PE studies and implement active surveillance strategies to identify as yet unreported AEs.

Usefulness of registries as surveillance tools

For the purpose of this concept paper, the FDA defined registry as `a systematic collection of defined events or product exposures in a defined patient population for a defined period of time'. However, participants commented that this definition needs further clarification and that it is too narrow in scope. For example, Dr Annette Stemhagen of the International Society for Pharmacoepidemiology and Dr Faich suggested that the definition be expanded to include large simple safety trials. Dr Nelson added that registries are best used to assess risk factors for known reactions in order to find the root cause of an AE.

Although participants recognised that registries could be useful to obtain information not available elsewhere and for certain types of data, the overall message was that they must be carefully considered especially in terms of methods of data collection and analysis and the trial size. In the words of Dr Nelson, `don't just assemble a registry to say you have one'.

Additional pharmacoepidemiologic studies

PE studies are typically started at the time of initial marketing based on questions that remain after review of the pre-market data or when a safety signal has been identified from spontaneous case reports, literature reports or other sources. The FDA believes that carefully designed PE studies, in addition to registries and surveys, may allow the sponsor to further characterise one or more safety signals associated with a product as it is used in the `real world'.

When asked whether additional PE studies would be useful, PhRMA remarked that no single study will ever be definitive, but that additional PE studies could be beneficial for, amongst other things, hypothesis testing, characterisation of patients at high risk of specific AEs (risk factors) and medicines utilisation, such as physician prescribing and monitoring, or characteristics of patients who are treated. In these studies, observed PhRMA, false positives or negatives may result from bias and confounding factors (i.e. the presence of other disease, drugs or chemicals).

In addition, Dr Stemhagen noted that the use of PE studies should be expanded so that they are carried out during the developmental phases as well as the post-marketing phase.

Active surveillance to identify unreported AEs

Active surveillance activities recommended by the FDA concentrate on events associated with the use of certain products, events occurring at selected healthcare settings such as hospitals or emergency departments, and events that are often product-related, such as acute liver failure. The agency also suggested that AE collection mechanisms include the use of health information systems and/or databases maintained by the Department of Health and Human Services.

Views on active surveillance were mixed. For example, PhRMA felt that it should be reserved for situations where information is not otherwise obtainable. Dr Nelson, on the other hand, expressed the opinion that active case-control surveillance methodology should be re-explored. He also pointed to poison control and drug information centres as a valuable resource for data acquisition. A further suggestion was that the strategy used in MedSun, a programme to educate hospital staff in the identification and reporting of injuries and death associated with medical devices, should be extended to pharmaceuticals. Dr McClellan also mentioned that the agency hopes to expand MedSun to cover pharmaceuticals.

Conclusions

Despite some constructive criticism from participants, the FDA's concept paper on pharmacovigilance and PE assessment was generally well received. The stakeholders presenting their views were in agreement that none of the components of a comprehensive pharmacovigilance plan could be used as a stand-alone measure and that each element should be developed with extreme caution.

References

1. Concept Paper: Risk Assessment of Observational Data: Good Pharmacovigilance Practices and Pharmacoepidemiologic Assessment, 3 March 2003, www.fda.gov/cder/meeting/riskManageIII.htm