Introduction

In the beginning, there was nothing. No data were available to study the prescribing and use of drugs. Over the years more and more data have become available in the constantly growing databases on prescription drugs. This has facilitated the development of pharmacoepidemiology. During the last 50 years this research area has developed from descriptive studies counting tablets to advanced analytical models to assess the effectiveness and safety of drug therapy in clinical practice. Clinical pharmacologists have played an important role in this development and their engagement will be important for future development of the field.

Pharmacoepidemiology—the bridge between clinical pharmacology and epidemiology

Pharmacoepidemiology can be defined as the study of the use and the effects of drugs in large numbers of people [1]. The research area can be seen as the bridge between clinical pharmacology and epidemiology (Fig. 1). Clinical pharmacology, with its pharmacokinetic and pharmacodynamic principles, contributes towards an understanding of the relationship between drug exposure and response in humans. Epidemiology, on the other hand, contributes with various descriptive and analytical methods. Clinical pharmacologists have also brought the concepts of critical drug evaluation and rational use of drugs into the research field. This dual background, with pharmacoepidemiology acquiring its focus of inquiry from clinical pharmacology and its methods of inquiry from epidemiology, has created a dynamic research field undergoing a more rapid development than many other research areas.

Fig. 1
figure 1

Pharmacoepidemiology—the bridge between clinical pharmacology and epidemiology

Full size image

The birth of pharmacoepidemiology may be dated to the early 1960s [1, 2]. Growing concerns about adverse drug reactions (ADRs) emphasized the need to develop methods to study the safety of drug therapy. The thalidomide disaster in the late 1950s and early 1960s strengthened the requirements for the pharmaceutical industry to provide proof of the safety of their drugs before approval [1]. In 1960, the Federal Drug Administration (FDA) started collecting ADR reports, leading to the establishment of hospital-based drug monitoring programs. These systems were further developed, and the term pharmacoepidemiology was proposed for the new discipline [2].

At the same time in Europe, clinical pharmacologists were conducting pioneering drug utilization studies. Many of these studies focused on differences in the utilization of drugs between countries or regions and were based on existing data sources [3–11]. Other pioneering studies focused on factors influencing the prescribing patterns of physicians [12]. Originally, drug utilization research was defined as “the marketing, distribution, prescription, and use of drugs in a society, with special emphasis on the resulting medical, social, and economic consequences” [13]. The main difference between drug utilization and pharmacoepidemiology was that the latter focused to a greater extent on the quantitative benefits and risks of drug treatment in cohorts of patients, while drug utilization focused more on drug exposure and differences in the quality and quantity of drug use [14, 15]. During the last decades of the 20th century, pharmacoepidemiology shifted from being focused entirely on ADRs and risk association studies to also include other clinical outcomes and health economic aspects of drug use, thereby lessening the distinction between pharmacoepidemiology and drug utilization research.

Clinical pharmacologists discovered very early the potential of drug utilization research in promoting rational prescribing. In their contribution to “Avery’s drug treatment”, Folke Sjöqvist et al. suggested that rational drug prescribing involves: a decision on whether to use a drug, and if so, the selection of a suitable drug and regimen; consideration of compatibility between the drug and patient and any other drugs being given; a legibly written prescription; and appropriate instruction(s) to the patient about the use of the drug and the expectations of treatment and follow-up [16]. Drug utilization studies were designed to assess these aspects of rational prescribing, and the research area has provided important tools to identify areas of irrational prescribing, forming the basis for guideline development and policy decisions [17–21].

Most prophecies have been fulfilled

People have always been intrigued of what the future will look like. Pharmacoepidemiologists share these interests, and there have been many predictions about the future over the years since the discipline was formed. In 1990, Strom and Tugwell summarized the achievements during the first decades and made a number of predictions for the future [22]. At that time, the quite young discipline had already shown its ability to contribute to the rational use of drugs, and a number of possible approaches to detecting and evaluating adverse drug reactions had been developed. The first databases of medical claims information had been developed, and the first universities had started pharmacoepidemiology training programs. Strom and Tugwell predicted a rapidly increasing interest in pharmacoepidemiology and that it would have an increasing impact on clinical medicine [22].

In 1995, Alvarez and Porta reviewed the opinions of leading pharmacoepidemiologists on the prospects of pharmacoepidemiology and summarized the most important challenges the discipline would face on its way towards the next century [23]. The future was expected to bring many opportunities, including the growth of databases, the development of methods and an increased international collaboration. The introduction of computer-assisted decision algorithms to reduce the level of uncertainty in decision-making was suggested. Certain challenges were also noted, including the limited number of qualified professionals, the difficulties in using administrative databases for research purposes (including restrictions due to confidentiality) and the lack of appropriate measures to assess the benefits of drug therapy in observational studies. Still, the expectation was that pharmacoepidemiology would meet the requirements of modern society to improve medication safety. Academic researchers, regulators and the pharmaceutical industry were pointed out as the key stakeholders in this process, but the need to establish a link between pharmacoepidemiology and economics was also emphasized [23, 24].

Most of the prophecies above have been fulfilled. In less than 50 years, pharmacoepidemiology has undergone a rapid development, similar to development of epidemiology (Table 1) [25].

Table 1 Threads in the fabric of the development of pharmacoepidemiology
Full size table

The predicted data explosion has become a reality. Progress in computer technology has allowed rapid access to descriptive data that in the early days were time consuming to collect and compile. Databases based on administrative claims data or medical records were established in North America and Europe [26–32]. The first Nordic population-based pharmacoepidemiological databases were established in Sweden in the 1970s [15]. Today, all Nordic countries have established population-based registries covering all dispensed drugs to their entire respective population [33]. An increasing number of healthcare organizations in North America, Europe and Asia have established large datasets on prescribed or dispensed drugs—in many cases with the possibility of linkage to clinical data in other registries [34–37].

In many countries, disease-based registers have been established for the specific reporting of clinical information and management of certain diseases and procedures [38]. The potential for analyzing the effects and safety of drugs has further developed with the introduction of electronic health records containing not only prescription drug data but also other clinical parameters, such as diagnosis, vital signs, laboratory data and more or less structured clinical notes [39]. In some countries, biobanks have also been established enabling linking drug utilization to genetic data [33, 40].

Bias and confounding are still great challenges

The increasing amount of data does not imply that studies have become easier to conduct. In many databases, registration is incomplete and, therefore, there is the potential for bias [41–43]. Furthermore, healthcare data are difficult to analyze due to the large number of variables and the interrelated nature of these variables. These challenges have been addressed in some reviews on how to use administrative data for pharmacoepidemiological research [41–43].

Pharmacoepidemiological studies are generally subject to three sources of bias; information bias, selection bias and confounding (Table 2) [44, 45].

Table 2 Bias and confounding in pharmacoepidemiological studiesa
Full size table

A particular problem with pharmacoepidemiological studies is the potential for confounding by indication, due to the fact that the drug may have been prescribed to patients having, a priori, a higher or lower risk of presenting the event studied [46]. Potential factors influencing the decision to prescribe—thus potentially leading to confounding-by-indication—may vary by physician and over time and involve a mix of clinical, functional and behavioral patient characteristics [39]. The Channeling of drug prescribing to certain patients may also occur as a result of guidelines or reimbursement restrictions favoring certain drugs.

While randomization balances the groups with respect to both known and unknown covariates, observational studies are limited with respect to the latter, making confounding an important issue. The development of new techniques, such as propensity scores and other matching algorithms, has enabled studies handling large numbers of potential confounders [39, 47, 48]. In theory, many confounding factors unknown today could—with the discovery of the human genome—be identified, taken into account and adjusted for [47, 49]. This was suggested by Concato and Horwitz who set up the ambitious goal “to conquer confounding by 2015” and noted that it might be within our reach, since “the number of pertinent factors is likely to be smaller than the 3 billion base-pairs in the human genome” [49]. However, this goal was suggested 9 years ago, and 2015 is now rapidly approaching, with confounding still far from being conquered. Most stakeholders still consider observational studies to be less reliable than clinical trials, and a recent review found no evidence of any shift in recognition of observational studies by the Cochrane Collaboration [50].

Drug utilization research has developed beyond descriptive studies

The area of drug utilization research has also developed beyond the descriptive studies that dominated the field during the first years. Folke Sjöqvist and Donald Birkett suggested that drug utilization research could be divided into descriptive and analytical studies [21]. The emphasis of the former would be to describe patterns of drug utilization and to identify problems deserving more detailed studies. Analytical studies, on the other hand, would link drug utilization data to figures on morbidity, outcome of treatment and quality of care, with the ultimate goal being to assess whether drug therapy is rational or not. Analytical drug utilization studies could also be designed to assess potential explanatory factors behind the observed utilization patterns [20]. An increasing number of drug utilization studies published today are just such analytical studies, and many of these apply record-linkage with other data. Another positive development is the increased number of studies focusing on the patient’s perspective, taking advantage of existing databases to follow dispensing histories in patients over time [51].

Today, drug utilization data are routinely used to promote rational prescribing in many hospitals and to assess the clinical appropriateness, cost effectiveness and, in some cases, outcome of therapy. Drug utilization data are also increasingly used in ambulatory care as a benchmarking tool and in educational interventions to prescribing physicians [52]. In recent years, regulations have been strengthened in many countries linking data to financial incentives and paying physicians for reaching identified targets [52]. The wider role of drug utilization research in health policy corresponds well to the more recent proposed definition of the research area: “Drug Utilization Research is an eclectic collection of descriptive and analytical methods for the quantification, the understanding and the evaluation of the processes of prescribing, dispensing and consumption of medicines, and for the testing of interventions to enhance the quality of these processes”. [20].

Finally, it is important to emphasize that drug utilization research also includes qualitative studies to provide a deeper understanding of the subjective aspects of prescribing, dispensing and utilization of drugs and the interaction between healthcare providers and patients [20]. Such studies are also extremely important for the deeper understanding of medication use in the society, but these will not be covered in this paper.

The intriguing future

Pharmacoepidemiology has undergone a rapid development that will continue during the 21th century. About 2500 years ago the Chinese philosopher Confucius stated that “If a man take no thought about what is distant, he will find sorrow near at hand.” Consequently, it may be wise to make some predictions about the future and the important challenges that will have to be met.

In the future—all drugs will be monitored post-launch

The availability of electronic healthcare databases will enable researchers to identify a growing number of cases in which effectiveness does not match efficacy [53]. This will challenge the actions of all concerned—industry, regulators, payers and healthcare providers. Risk management plans (RMPs) or risk evaluation and mitigation strategies are already required by European Medicines Agency and FDA as part of the medicine approval process to help ensure that the benefits of a particular medicine outweigh its risks [54, 55]. Observational studies will also be increasingly requested by reimbursement agencies and other payers to assess the value of medicines. It is also likely that patients will demand better systems to monitor effectiveness and safety. In the future, it is therefore likely that all new drugs will be monitored post-launch. This monitoring may range from descriptive drug statistics to sophisticated comparative effectiveness studies, depending on the budget impact and level of uncertainty about the benefit–risk of the medicine (Fig. 2). The difficulties conducting studies also indicate that the nature of the monitoring activities will be limited by the availability of data and competent pharmacoepidemiologists.

Fig. 2
figure 2

Different types of monitoring of new medicines with increasing levels of complexity

Full size image

Descriptive statistics are already being used in most countries to monitor the introduction of new medicines. Diffusion patterns of new drugs may be compared with existing alternatives, often stratified by age, gender and geography. Without patient identity data, the number of defined daily doses (DDD)/1000 inhabitants/day or DDD/100 bed-days can be used as rough estimate of the proportion of the population exposed [14]. The opportunities increase dramatically with the availability of encrypted patient identifiers, facilitating studies of the prevalence and incidence of drug use [33, 56]. Patient identity drug data, with or without record-linkage, may also be used to identify if the utilization is appropriate, i.e., to what extent patients prescribed the drug correspond to the labeled indication, the reimbursed indication or the patient groups for which there are evidence of efficacy in clinical trials. In these studies, relevant patient characteristics (e.g. age, gender, co-morbidity, disease status and concomitant drug therapy) may be analyzed for those initiated on the drug and potentially compared with those initiated on other drugs. Although crude, such analyses may be a valuable tool to estimate the likely risk–benefit of the drug when used in clinical practice.

Payers will also increasingly require health economic studies to assess the cost-effectiveness of the new premium priced therapies compared to the existing standards. Increasing difficulties for national health services financing new expensive therapies will lead to the development of various risk-sharing models. This has been defined as ‘agreements concluded by payers and pharmaceutical companies to diminish the impact on payers’ budgets for new and existing schemes brought about by uncertainty and/ or the need to work within finite budgets’ [57]. Such agreements can include financially based schemes, such as price–volume agreements or performance-based/ outcomes-based schemes. The latter includes schemes whereby pharmaceutical companies refund monies where the new drug does not achieve the desired outcomes [57].

It is essential that the increasing amounts of data collected to monitor the utilization, effectiveness and safety of new medicines will be used to improve clinical decision-making. Avorn and Fisher reviewed current barriers to the dissemination of recommendations and proposed a number of solutions, including more effective educational outreach programs, requirements for practitioners to master important findings and alignment of incentives to encourage evidence-based practice [58].

New therapeutic areas needs to be addressed

The majority of pharmacoepidemiological studies performed today are based on data generated in ambulatory care [59]. This is also the case for drug utilization studies. A recent literature study found that one-third of all European cross-national comparative studies of drug utilization conducted between 2001 and 2011 examined antibiotic use [60]. Other common areas were cardiovascular drugs and those drugs acting on the nervous system. A similar pattern was observed in a review of pharmacoepidemiological studies conducted in the Nordic countries between 2005 and 2010, with the most commonly studied drugs being those acting on the nervous system, including psychotropic drugs, analgesics, antiepileptics and drugs for dementia and Parkinson’s disease (Wettermark et al, The Nordic Prescription Databases as a resource 690 for pharmacoepidemiological research—a literature review, Under Review). Many studies were also conducted on cardiovascular drugs, such as antihypertensives and lipid-lowering agents, as well as on drugs acting on the alimentary tract, including antidiabetic agents and proton pump inhibitors (Wettermark et al, The Nordic Prescription Databases as a resource 690 for pharmacoepidemiological research—a literature review, Under Review). This dominance may partly be attributable to data availability, with most databases being restricted to prescription drugs dispensed and/or reimbursed to outpatients [59]. However, the drug market is rapidly changing, with more and more new drugs being introduced onto the market for the treatment of cancer and autoimmune diseases [61, 62]. Consequently, the future of pharmacoepidemiology must include more studies conducted in the hospital setting.

It may also be wise to look at the World Health Organization’s report entitled “Priority Medicines for Europe and the World,” which identifies a number of gaps for diseases for which treatments do not exist, are inadequate or are not reaching patients [63] (Table 3). The authors recommend that pharmaceutical innovation be encouraged by a shorter medicine development process, a reviewed reimbursement procedure and a more attractive research environment [63]. Pharmacoepidemiology could contribute to better treatment in these areas, particularly in addressing obstacles where effective medicines could be better delivered to the patient. The report also highlights the special needs of three population groups (elderly, women and children) which should be addressed. Pharmacoepidemiology will be an important tool here, and efforts are now underway to use data from automated databases to conduct observational studies on drug use in children [64]. Similar strategies are urgently needed for the elderly, especially in light of demographic trends in Europe [65].

Table 3 Priority disease areas for which treatments do not exist, are inadequate or are not reaching patientsa
Full size table

The future will also bring an increased focus on the impact of drugs on our environment. Since the 1990s water contamination by pharmaceuticals has been an issue of concern [66]. Most pharmaceuticals are deposited in the environment through human consumption and excretion. The term “pharmacoenvironmentology” has been proposed for the science dealing with the environmental impact of drugs given to humans and animals at therapeutic doses, and “ecopharmacovigilance” as the science and activities associated with the detection, evaluation, understanding and prevention of the adverse effects of pharmaceuticals in the environment [67]. Pharmacoepidemiology could help these young disciplines by contributing important methods and concepts.

Further battles on confidentiality may arise

We have been fortunate to live in an era where large amounts of data have become available for research. It is important to recognize, however, that the increasing amounts of sensitive data available in registries may add fuel to the debate on confidentiality. Based on the Declarations of Helsinki from the World Medical Association, it is a basic right of the patient to be assured that all of his/her medical and personal data are confidential [68]. Researchers active in the field have recognized these ethical issues and adopted methods to ensure that the confidentiality of individually identifiable data is maintained [69]. However, the way forward has not always been easy, with years of discussions on the usefulness and integrity of collecting and storing personal data in some countries. The European Commission has recently proposed a stronger and more coherent data protection framework for the European Union [70]. Hopefully, this will not prevent/hinder the opportunities for conducting research. A too strict application of privacy rules might hinder the further development of pharmacoepidemiology.

The golden days are here to stay

There are several reasons why pharmacoepidemiology will remain to be recognized as one of the most dynamic and challenging research areas of clinical pharmacology. The data explosion in modern society will continue. By the end of 2010, one-half a zettabyte (unit of information equal to 1021 bytes) of data traveled across the Internet, which is equivalent to the information contained on a bookshelf 36 billion miles long (10-fold the distance from Earth to Pluto) [71]. And every 5 min, digital data are created equivalent to all of the information stored in the U.S. Library of Congress [72]. Considerable amounts of healthcare data will also be generated, including data used in various electronic decision-support systems [73]. Some data will be unstructured and tricky to analyze, but new techniques and methods will be developed to confront those challenges. The advanced biostatistics methods have no doubt been important for the scientific rigor in the field, but it is also important to acknowledge that they have added to the complexity, sometimes making it difficult for clinicians to understand and critically evaluate the methods applied.

Pharmacoepidemiology is also likely to continue to be an arena for collaboration between different stakeholders, including physicians, regulators, payers, manufacturers and the general public. Clinical pharmacologists played an important role in the birth and development of pharmacoepidemiology, and clinical pharmacologists have all the prerequisites to be the most important player in the future as well. It is important to recognize the multidisciplinary and multiprofessional nature of the research area, and it is likely that the future success will require alliances to other professions, including specialists within specific clinical areas but also those involved in health economy and health policy. Twenty years ago, Brian Strom stated that pharmacoepidemiology needed better science, better scientists, more science and more scientists [74]. This statement is still very relevant, although it is important to acknowledge the progress that has been made during the first 50 years of pharmacoepidemiology.

A final wish for the future is that pharmacoepidemiology and drug utilization will take further steps beyond the description and identification of problems to providing tools for improvement in drug utilization. No one knows what the future will bring, but it is obvious that those of us active in the field may look forward to interesting times. It is strange to imagine that one day we will call this period for “the good old days.”