Iressa and interstitial lung disease (ILD) in Japan - Lessons from a large nested case-control study to evaluate a safety issue Fredrik Nyberg 1, Shuji Hada 2, Kenneth J Rothman 3 and the Iressa CCS Collaborator Group. 1 AstraZeneca R&D, Mölndal, Sweden; 2 Clinical Division, R&D, AstraZeneca K.K., Osaka, Japan and 3 RTI Health Solutions, Research Triangle Park, NC, United States. Background Evaluating safety risks for pharmaceutical products can be a challenging task. Pharmacoepidemiology is one option. For events with a frequency >1%, a nested case-control approach in an ad hoc cohort is a good alternative if the required follow-up time is reasonable. After launch of Iressa TM (gefitinib) by AstraZeneca in Japan in July 2002, ILD was reported among treated non small-cell lung cancer (NSCLC) patients at a rate of several percent within the first months after treatment. This adverse event had not been noted in the clinical studies before registration and has also not been observed outside Japan either then or since. Further studies were required to elucidate this question Objectives To share study design insights from an ongoing epidemiologic investigation undertaken by AstraZeneca in Japan to estimate the relative risk of ILD among NSCLC patients treated with Iressa vs. other treatments, and elucidate mechanisms and risk factors of ILD. Methods A nested case-control design was chosen (Fig. 1). A cohort of non small cell lung cancer (NSCLC) patients with at least one previous chemotherapy treatment was recruited at 50 Japanese hospitals. Patients receiving Iressa or other treatment were followed for 12 weeks for ILD outcome (Fig. 2). For each ILD case identified, 4 controls were sampled from the cohort and detailed information was collected on these subjects. ILD diagnosis was supported by a diagnostic algorithm. The study was financed and run by clinical operations personnel of AstraZeneca, in collaboration with Clinical PIs (Fig. 3). A case review board (CRB) reviewed all identified cases and an external epidemiology advisory board EAB) supervised the scientific content of the study. Studies of genetics, proteomics and pharmacokinetics were included, in collaboration with external researchers. Results Success factors Several success factors were identified for performing a large-scale epidemiological study run by the company: - Importance of clinical buy-in for the scientific approach. Consultations with clinicians to familiarize them with epidemiologic study design and seek their advice was an essential component in the planning phase. - Close collaboration between Pharma, Academia and physicians and across specialties (epidemiology, clinical medicine, genetics, proteomics, clinical pharmacology etc). Both Astrazeneca in-house specialist input and wide outside consultation is needed to ensure a comprehensive and relevant study. Substudies of exploratory type were set up in collaboration with PI:s who are leading in their field (Fig. 3). - Internal epidemiology expertise. For an optimal design and result, a strong Epidemiology section within the company is essential. - Epidemiology support from an independent external Epidemiology Advisory Board (EAB). The Board was set up with international and local experts and provided advice to AstraZeneca throughout design, conduct and analysis. The EAB proved useful both in supporting in- house Epidemiology expertise in relation to other clinical functions that are often not extremely familiar with epidemiology, and ensuring that the study results are appropriate and acceptable to an outside audience (Fig. 3). - Good phenotyping is critical. A diagnostic algorithm was devised in consultation with clinical experts to support physicians’ diagnosis of ILD. In addition, a Case Review Board comprising oncologists, pulmonologists and radiologists reviewed all cases, blinded to study treatment and case/control status (Fig. 3). - High participation is a challenge, particularly with biomarker collection, an important component but one that adds complexity. The nested case-control design helps as patients have already consented a first time on entering the cohort. Dedicated support in-house and sincere buy-in from collaborating physicians makes a great difference. Recruitment and descriptive data The study recruitment was quite successful. Over 25 months, 4473 NSCLC patients were registered to the cohort, of whom 4423 met all study criteria. Overall, 155 cases of ILD were registered and 122 were accepted after blinded review by the CRB, and 574 controls were recruited to the case-control study, meeting the study size goals set up at the outset. Some characteristics of the study cohorts are shown in Table 1 for the data collected for all cohort members. As expected, controls were quite representative of the cohort composition. Statistical analyses are ongoing and results will be presented during the autumn of Table 1. Descriptive statistics of registrations in the cohort (per protocol cohort analysis set Why an epidemiological study? The experimental method – a randomized control trial (RCT) – is difficult for suspected adverse effect of an already marketed drug - For reasons of ethics, logistics, size Epidemiology – studies and quantifies associations between risk factors and disease in real life populations -A set of scientific methods to perform non-experimental studies -Appears most suited to relatively rare problem seen in Japanese patients on marketed use of Iressa – a real-life situation, suitable for an observational science Some reasons for selecting a nested case-control design - Easier to apply stringent diagnostic criteria - Easier to check for complete case ascertainment - Easier to make sure that controls are correctly sampled - Incidence (and/or absolute risk) can be calculated - In contrast to the full cohort study, detailed on formation for adjustment of analyses need only be collected for cases and controls Conclusions Large prospective epidemiologic studies can be successfully undertaken by a pharmaceutical company. The nested case-control design is useful provided a suitable cohort can be identified or established and the outcome is not too rare. Motivating all participating specialties to work together on an unfamiliar epidemiologic study design is key to success. Support by an external epidemiology advisory board is very helpful. Conflict of interest statement This study was financed in its entirety by AstraZeneca. Fredrik Nyberg is employed as Senior Epidemiological Scientist by AstraZeneca and has been responsible epidemiologist for the design and conduct of the study. He also holds a position as Adjunct Lecturer in Molecular Epidemiology at the Karolinska Institute. Shuji Hada is an employee of AstraZeneca and has acted as Study Leader for the study. Kenneth J Rothman is an employee of RTI health Solutions, an independent non-profit research organisation that does work for government agencies and pharmaceutical companies. Background: Evaluating safety risks for pharmaceutical products is challenging. For events with frequency >1%, a nested case-control approach in an ad hoc cohort is a reasonable alternative. After launch of Iressa TM (gefitinib) in Japan, ILD was reported among treated non small-cell lung cancer (NSCLC) patients at a rate of up to 6% within the first months after treatment. Objectives: To share study design insights from an ongoing epidemiologic investigation undertaken by AstraZeneca in Japan to estimate the relative risk of ILD among NSCLC patients treated with Iressa vs. other treatments, and elucidate mechanisms and risk factors of ILD. Methods: A NSCLC cohort was recruited at 50 Japanese hospitals. Patients receiving Iressa or other treatment were followed for 12 weeks for ILD outcome. For each ILD case identified, 4 controls were sampled from the cohort and detailed information was collected on these subjects. ILD diagnosis was supported by a diagnostic algorithm. A case review board reviewed all identified cases and an external epidemiology advisory board supervised the scientific content of the study. Studies of genetics, proteomics and pharmacokinetics were included, in collaboration with external researchers. Results: Setting up the study was challenging since clinicians today often are not familiar with epidemiologic study designs. Close collaboration between Pharma, Academia and physicians, and across specialties (epidemiology, clinical medicine, genetics, proteomics, clinical pharmacology etc) is required. High participation is a challenge, particularly with biomarker collection, an important component but one that adds complexity. Further recruitment data, descriptive analyses and other lessons will be presented. Conclusions: Large prospective epidemiologic studies can be successfully undertaken by a pharmaceutical company. The nested case-control design is useful provided a suitable cohort can be identified or established and the outcome is not too rare. Motivating all participating specialties to work together on an unfamiliar epidemiologic study design is key to success. Support by an external epidemiology advisory board is very helpful. AstraZeneca commitment to good science – for the good of the patient Good clinical decisions can only be based on sound science Essential to work with academics and clinicians Strive to provide the best available information about our products, based on scientifically sound study design, data collection, data analysis and interpretation Fig 2. Nested case-control study flow chart Fig 3. Structure of the nested case-control study Iressa cohort and nested case-control study Studydesign, overallscientific integrity EAB Studydesign, overallscientific integrity EAB Phenotyping CRB Phenotyping CRB PK P.I. PK P.I. AZ Genetics P.I. Genetics P.I. Proteomics P.I. Proteomics P.I. Main (”clinical”) study ClinicalPIs Main (”clinical”) study ClinicalPIs Abstract Fig 1. The nested case-control study