Report on the 4th EpiCom Workshop at ECCO'18
Naila Arebi, EpiCom Member
Naila Arebi © ECCO
An introduction to pharmaco-epidemiology: How to interpret real world data for clinical practice
The Epidemiological Committee (EpiCom) organised their 4th workshop at the ECCO’18 Congress in Vienna. Delegates were welcomed by Ebbe Langholz, who introduced the sessions with a short video on the UR-CARE database to illustrate the importance of databases and their potential across Europe.
The first session covered the principles of pharmaco-epidemiology (PE), with a succinct introduction by Laurent Peyrin-Biroulet. He defined PE as the study of the use and effects of drugs on a large scale and reminded us how, prior to PE, the medical community relied on spontaneous reports of random adverse events; this approach entailed a declaration bias and data were very difficult to interpret. Laurent outlined the limitations of clinical trials aimed at gaining regulatory approval to market a drug: Highly selected patients, short duration of follow-up and limited sample size. He discussed several study designs used in research, highlighting that the choice of design is dependent on the purpose of the study. Among the types of epidemiology study design outlined were case reports and case series, ecological studies, and cohort and case-control studies. Examples were provided to show in which circumstances specific study designs may be useful and to illustrate their differences. Laurent concluded by discussing situations in which PE databases are of limited value.
Jonas Halfvarson focussed on real world data, examining the dissimilarities with randomised controlled trials (RCTs) and comparing the questions that these two designs are intended to answer. He emphasised that while RCTs can be said to address efficacy, i.e. whether a treatment can work, real world data address effectiveness, i.e. whether a treatment works in real life. Jonas acknowledged that real world data often do not align with guidelines and treatment algorithms. He discussed the various definitions of real world data (ranging from everything that goes beyond data usually collected in a Phase III trial to data relating to patient health status) and real world data sources. He also addressed both the limitations of real world data and the potential of such data to improve standards and practice, citing as an example the Swedish SWIBREG project, which has the aim of driving consistent care through a national algorithm on patient pathways in IBD.
|4th EpiCom Workshop at ECCO'18, Vienna © ECCO
Valerie Strassmann described the role of PE in regulatory agencies with respect to benefit-risk assessment throughout the drug cycle, from post-authorisation safety studies to post-efficacy studies. She outlined the key EMA activities and gave examples of ‘soft’ epidemiology studies. The most important regulatory tasks rely on supporting (requested or voluntary) streams of evidence, covering, for example, signal detection and risk management throughout the product life cycle and characterisation of the impact of regulatory action. Such streams of evidence may derive from post-authorisation safety studies (PASS), post-authorisation efficacy studies (PAES) and also non-clinical RCT observational data and case series.
Marieke Pierik presented the central threats to the validity of data from PE. The talk started by considering the scenario of an 18-year-old with terminal ileal inflammation and examined how problem-based learning for medical students channelled the discussion towards which studies are best suited to answer clinical questions. The choice of study design depends on several factors, with the commonest study designs in PE being case-control and cohort studies that compare for events in exposed vs non-exposed subjects. The types of bias and additional measures of reliability and validity were discussed.
Julien Kirchgesner then presented the role of propensity scores and other approaches in overcoming potential bias. He started by discussing the balance of risks and benefits in IBD treatment and addressing the nature of data available to assess this balance: Observational studies remain invaluable in assessing the long-term impact of treatment but are subject to major potential biases. One approach to ensuring validity is to simulate randomisation, whereby a population is reweighted. That weight is the propensity score (PS). The advantage of this approach is the inclusion of large pools of data from observational studies with a balance for potential confounders. The PS is generated from the data and then re-applied to data so that all confounding factors are included, as well as factors associated with outcome and treatment. The second half of the work focussed on clinical applications of PE data.
Naila Arebi presented differences unique to the over 60s and ethnic populations. She referred to the ECCO Topical Review on IBD in the elderly and how precise definition of elderly-onset IBD will allow future comparative studies. Several epidemiological studies have consistently shown under-prescribing of biologic and immunomodulatory drugs in patients with elderly-onset IBD compared to those with adult-onset IBD, although the former show similar use of steroid treatment and a higher surgical rate. This difference may be due to awareness that in elderly-onset IBD, there is a higher risk of both infection when using anti-TNF and immunomodulatory therapy and malignancies (non-melanoma skin cancer and lymphoproliferative disorders) when using thiopurines. Clinicians accordingly need to be vigilant when prescribing biologics and thiopurines to patients with elderly-onset IBD. Naila concluded by discussing challenges of PE in ethnic populations, including definitional issues and heterogeneity of populations.
The workshop was concluded by Laurent Beaugerie, who gave a seminal talk on PE studies in IBD from various data sources. He discussed the challenges involved in deciding on a clinical question when different data sources show different risk. He focussed in particular on the drug-associated risk of lymphoma and compared the risk from three large datasets – CESAME, a Danish database and a French database – identifying key strengths and limitations of the respective study designs. Data on the risk of lymphoma due to thiopurines are consistent but various data sources show conflicting risks in association with anti-TNF drugs. Both the French and the Danish database are limited by the retrospective nature of the data. The ongoing I-CARE (Ibd CAncer and seRious infections in Europe) project aims to collect prospective data from 16,000 IBD patients and may settle the uncertainty around the benefit-risk balance of treatment with anti-TNF drugs.