Calculating risk is a fundamental part of pharmacovigilance: when reports of an adverse reaction are received the ideal is to accurately calculate the incidence rate. www.primevigilance.com have provided a page about the complexities of calculating risk generated as a result of spontaneous reports from doctors, pharmacists, other healthcare professionals and patients themselves. This page provides further background information to the topic, including why this area is notoriously challenging for pharmaceutical companies within the EU and the broader global community.
Spontaneous reports arise when healthcare professionals and patients around the world report a suspicion that an untoward medical event has an association with taking a particular prescribed drug. During the post marketing phase, the volume of reports can be affected by two known effects. The Webber Effect describes an increased volume of reports for new drugs within their first two years on sale. The Secular Effect describes increased reporting when a drug becomes subject to media exposure because it has been taken by a well known figure or a warning has been issued by a regulatory authority. These two considerations aside, the quality and integrity of data can still be entirely variable within spontaneous reports. Crucial areas may be incomplete, for example, scant information on the health status of the patient. There may be little or no information on whether any other medications were being taken at the same time. Busy healthcare personnel may be difficult to contact to complete the follow up. Patients may feel uncomfortable divulging sensitive information. Furthermore, pharmaceutical companies provide researchers with just the quantity of kilograms manufactured for each drug. Specialist firms then provide services attempting to quantity patient exposure to it. From this data, an approximation of demographical analysis and prescribing trends would then be made. For these reasons alone, it is not possible to draw any meaningful conclusions from spontaneous reports; yet they are still considered to be vitally important within drug safety. It would be impossible to launch clinical trials on a scale which mirrored real life prescribing for each individual drug; hence spontaneous reporting retains its function.
Epidemiological and clinical studies typically present more complete and uniformly compiled data. Since the data will be submitted to the regulatory authorities, it should be presented in standard formats. It is therefore likely to be adequately defined, including adequate descriptions of statistical methodology. Efficacy and safety analysis are also likely to be well documented and carefully controlled. The patient samples studied should be well described. Regular periodical subject or patient monitoring should be conducted specifically aimed at uncovering any Adverse Reactions by the appropriate professional teams. Using placebos or other comparators provides a clearer basis for evaluating risk using confidence intervals and significance values. As a result, the data from such studies is considered to be meaningful, with the likelihood that any problems will have been detected.
There are various methods of statistical analysis which may be employed to evaluate the results of spontaneous reporting. Collectively termed ‘data mining’, these methods are typically used to generate signals, which can be crudely characterised as indications that further study may be required. Signal screening requires analysis of large amounts of data. The aim is to identify any clear anomalies, which may then be subjected to clinical investigation. However, the results of Signal Screening may still not be sufficiently robust to be considered particularly meaningful.
Proportionality methods are sometimes also called ‘disproportionality methods’. As a rather broad summary, these methods compare the proportion of reports received for a particular Adverse Reaction to those received for other drugs using data held within the same database. The aim is a comparison with other drugs rather than the examination of exposure data for the drug under evaluation. Limitations include smaller sample sizes; compounding factors such as any issues of polypharmacy; a lack of suitable comparative data; demographic variations within the patient group; and any variations in coding. The results again may not be sufficiently robust to generate any meaningful drug safety conclusions.
Bayesian methods may also be referred to as ‘probabilistic causality’. The aim is to analyse causality in terms of conditional probability, the data from which will be routinely modified as fresh information is obtained. One of the limitations is that within the EU (and other markets) the numbers of untoward medical events which occur in real life dwarf the numbers which are actually reported. Studies since the 1980s have proved that this factor means that again, this method is not always produce the desired drug safety conclusion, i.e., a meaningful final causality prediction.
The issues involved are invariably complex. No matter which statistical methodology is employed, pharmaceutical companies and regulators find themselves faced with discrepancies between the quality and/ or quantity of reports. The challenges they face require highly senior pharmacovigilance services personnel to produce results which are likely to satisfy both the operational needs of pharmaceutical companies and fully comply with the relevant EU drug safety regulations.
Adverse reactions may not present only immediately after the release of a new drug onto the market; on the contrary, they are entirely possible even decades after the drug was first prescribed. Information on fundamental aspects of pharmacovigilance for complex untoward medical events can be found on a page at www.primevigilance.com This page provides three examples of complex reactions where establishing the true drug safety profile proved more challenging to detect.
It is tempting to reason that through the combination of premarketing clinical trials and post marketing spontaneous reporting, safety concerns could be expected to be resolved within several years of a new drug being placed onto the market. The time period elapsed after taking the drug and the reaction is termed the ‘latency period’. It is important to note that there have been many well established drugs which have provoked reactions following a latency period of years or even decades. The time elapse and confounding issues means that such reactions are more difficult to detect, whereas those with a short latency period are likely to arouse a higher level of suspicion.
There may be any number of reasons for this phenomenon:
However, this does not mean that detecting those reactions with a longer latency period is impossible. There have been some notable cases where vital drug safety conclusions have been reached even though the adverse reactions affected the patient’s children and grandchildren rather than the patients themselves. Epidemiologic studies may prove vital in detecting drug safety problems long after the product launch.
Just as it is a fallacy to assume that problems will be detected within a short time of launch, the clinical severity of the disease being treated is no indicator of whether there can be late onset problems. The following three examples illustrate these concepts.
Nitrofurantoin is an example of a drug which has been known to cause mild to extremely serious problems within weeks of taking it – and after at least 6 months therapy. It is an antibiotic used for prophylaxis in cases of recurrent urinary tract infections or the treatment of such current infections. There have been reports of early onset reactions, within 3 weeks of starting to take the drug. Problems have been documented to range from a multisystem hypersensitivity reaction to fatal cases of suspected pulmonary fibrosis. Late onset reactions have also been documented at least six months after starting to take the drug, featuring compromised lung function, sometimes due to pulmonary fibrosis and infiltration, and a small number of fatalities.
Minocycline is a medication to combat acne vulgaris which has been documented as causing late onset effects as late as two years after commencing therapy in at least one case. Reactions have been reported when the drug was administered in combination with other typical acne medications. When an Adverse Reaction is relatively rare, this factor which can contribute to the complexity of detecting problems. It is of note that this medication had been prescribed for at least 30 years when a doctor initiated literature review identified a series of rarer reactions characterised by such polypharmacy.
Diethylstilbestrol (DES) is a drug was widely prescribed in the USA, Europe and further afield aiming to prevent miscarriage. A period of some thirty years elapsed before suspicions were raised. The issue lay with the fact that the drug was found to cause serious adverse reactions not within the original patient group, but within their children (and even their grandchildren). Cases studied in the early 1970s included structural abnormalities of reproductive organs; rare vaginal cancers usually prevalent in patients at least 56 to 48 years older; and higher instances of ectopic pregnancies, spontaneous abortions and preterm births.
This brief list of products which have been associated with late onset reactions is by no means exhaustive. Instead it is provided as a simple illustration of how even well established drugs have been found to provoke mild to serious reactions. Drug safety problems may be more difficult to detect not only due to the longer latency period, but due to a rare occurrence rate or a highly unconventional manifestation within clinical practice. The need for robust pharmacovigilance services remains of paramount importance at every stage of a products life cycle, no matter how widely prescribed the product, nor how long lived its life cycle.
A safety database is both a core element of any pharmacovigilance system and a legal pre-requisite for pharmaceutical companies wishing to place a medicine or medicinal product onto any EU/EEA market . To enter those markets, each product must first be granted a Marketing Authorisation (MA) by the regulatory authorities . The safety database must already be in place and described within the Detailed Description of Pharmacovigilance System (DDPS) accompanying the MA application . This article explains key functions of safety databases, to create a better understanding of their overall purpose.
Safety databases should allow pharmaceutical companies to rapidly gather relevant information about the medicine or medicinal products in question. There will be information which needs to be supplied to regulatory authorities via statutory electronic reporting and the database must support this. The database must therefore be validated and acceptable to those regulatory authorities. In essence, the database must be entirely suitable to the task, in compliance with Part III of the relevant legislation, Volume 9A of The Rules Governing Medicinal Products in the European Union – Guidelines on Pharmacovigilance for Medicinal Products for Human Use . Details of the database are assessed during the application stage for each MA, as they will form part of the DDPS . To paint perhaps a more rapidly accessible picture here, the safety database is the tool used to help staff collate and analyse key drug safety data. One example could be any Adverse or Suspected Adverse Event Reports for one particular product. The database will be electronic but not ‘autonomous’: it will not perform tasks without key interaction by highly skilled staff, no matter which process is underway.
A compliant, suitable Safety Database is able to process data related to signal detection. Signal detection essentially uses data to detect any new patterns or findings. These can be identified by analysing data tables according to principles of statistical disproportion. Upon detection the subsequent course will be variable but can involve action by the regulatory authorities.
Any medicine or medicinal product granted an (EU) MA is legally required to become the subject of post-marketing Periodic Safety Update Reporting. PSURs are always more than an in-house assessment. Presented directly to regulatory authorities, as you might expect whenever drug safety is concerned, they will of course be subjected to an extremely thorough inspection of their contents. A comprehensive database will be able to process the required data in a manner which facilitates the production of PSURs in the appropriate format as far as those authorities are concerned. For example, modern databases integrate regionally-tailored support on regulations.
Certain safety databases are able to incorporate functionality to file expedited and aggregate reports to regional regulatory authorities. Again, this would always take place with core staff input, rather than becoming any type of scenario where a computer simply somehow ‘spits out’ a report. Electronic reporting will be required within strict time deadlines and differing formats by differing authorities throughout the world. In the EU for example, expedited reporting is obligatory within 15 days of a spontaneous adverse drug reaction case report.
These are just some of the core functions of a suitable safety database, which make it easy to see why it’s one of the pillars of good quality drug safety monitoring. It may be web-based, but whatever the platform it will be electronic. It is not however necessary for pharmaceutical companies to purchase and install a dedicated ‘in-house’ safety database. Instead it is common to find them using commercially available fully compliant products introduced by external pharmacovigilance services firms, who simply provide them with the necessary ongoing support.
Signal detection based on spontaneous reporting involves looking for any new patterns or seemingly significant new findings in the safety database. It may be one or more reports showing particularly strong evidence of a previously unknown adverse reaction for that drug, or involving adverse events that are usually caused by drugs, such as aplastic anaemia or toxic epidermal necrolysis. More often, it is a matter of looking for patterns or clusters of reports that stand out from the background. These clusters may be identified by looking at data tables, or – for large databases, such as those held by regulatory authorities or by a company with product with many cases reported – using computerised methods involving statistical disproportion. Examples include the Proportional Reporting Ratio, Bayesian Combination Propagation Neural Network used by the WHO Uppsala Monitoring Centre and the Modified Gamma Poisson Shrinker method used by FDA and MHRA.
Having identified potential tentative signals it is necessary to evaluate them and to consider whether they are real or not. Often, an apparent signal can result just from the disease that the drug is treating – so if a drug is used in patients for the treatment of high blood pressure, it would not be surprising to find reports of high blood pressure itself, and also kidney disease, stroke and heart failure, because these are either contributory or complications of high blood pressure. This is often a difficult process: evaluating signals involves looking at the individual case reports and assessing the strength of causation of the drug in each case. Other sources of safety data are usually also checked at this time – for example, the data from clinical trials and toxicology studies.
It may then be appropriate and necessary to set up a study specifically investigating a particular signal or safety concern. Other possible actions could be to institute a review of the benefits and risks of the product; to suspend or revoke the marketing authorisation; to propose a change to the product information (SPC, labelling); to issue a warning to health professionals in the form of a Dear Health Professional communication; or perhaps just to keep the issue under review. At some point, it may become apparent that the signal was not real, and the issue can be shelved.