New approaches to fill the methods gap in pharmacovigilance

The pharmacovigilance ‘tool kit’5Wise et al Nature Drug Disc 2009

European Medicines Agency priorities foradverse drug reaction research2009-4th Call FP72010-5th Call FP7Long-term effects in children and inyoung adults of methylphenidateInsulin/insulin analogues anddevelopment of cancerLong-term adverse effects ofimmunomodulators (monoclonalantibodies)Anti-diabetes drugs and cardio/cerebrovascular adverse effectsand pancreatitis/pancreatic cancerLong-term adverse skeletal effectsof bisphosphonatesMedicine use in pregnancySuicidal behaviour in relation tocertain drug use(antidepressants,antipsychotics, varenicline,montelukast)Safety aspects of antipsychoticsin demented patientsEpoetins: Risk of tumour growthprogression and thromboembolicevents in cancer patients andcardiovascular and cancer risk inchronic kidney diseaseAsthma treatments: risk ofmyocardial ischaemia and longtermsafety, especially in childrenGadolinium-containing contrastagents: Nephrogenic systemicfibrosis (NSF) and long-term effectson the bone7

European Medicines Agency priorities foradverse drug reaction research2009-4th Call FP72010-5th Call FP7Long-term effects in children andin young adults of methylphenidateInsulin/insulin analogues anddevelopment of cancerLong-term adverse effects ofimmunomodulators (monoclonalantibodies)Anti-diabetes drugs and cardio/cerebrovascular adverse effectsand pancreatitis/pancreatic cancerLong-term adverse skeletaleffects of bisphosphonatesMedicine use in pregnancySuicidal behaviour in relation tocertain drug use (antidepressants,antipsychotics, varenicline,montelukast)Safety aspects of antipsychotics indemented patientsEpoetins: Risk of tumour growthprogression and thromboembolicevents in cancer patients andcardiovascular and cancer risk inchronic kidney diseaseAsthma treatments: risk ofmyocardial ischaemia and longtermsafety, especially in childrenGadolinium-containing contrastagents: Nephrogenic systemicfibrosis (NSF) and long-termeffects on the bone8

9Struggle 1: long-term adverse effects

10Struggle 1: long-term adverse effects

Long-term treatment: >6 monthsbuilding the safety database for approval The number of patients treated for 6 months at dosagelevels intended for clinical use, should be adequate tocharacterise the pattern of ADEs over time. (n=300-600) ADEs that increase in frequency or severity with time orthat may occur only after > 6 months of treatment. (n=100treated for 1 year) total number of individuals treated with the investigationaldrug, including short-term exposure, will be about 1500.11

Benefit-risk balance at time of approval--> Many long-term effects are assessed post-approval12

Efalizumab & PML riskArch Dermatol 2009Safety database at time of approval (2003): 2764 patients treated 218 >1 yearFDA post–approval commitment: 5 year surveillance study, n=500013

How to perform long-term follow-up?build a registry Clinicaltrials.gov (July 16, 2010) Observational registry n=553 Obvservational registry long-term n=92• Drug n=24• Drug+comparator n=6• Disease n=23• Device n=21• Cell therapy n=2• Procedure n=16Disease registriesD:A:D Data collection on Adverse events of anti-HIV DrugsTNF inhibitor users (British Society of Rheumatology)UK epilepsy and Pregnancy RegisterThe Hunter outcomes Survey (set up by idursulphase company)14

From single-product to disease-basedregistries?15Askling J. ENCePP Scientific Convention ‘EU health care databases for pan-EUpharmacoepidemiological research’, London, 25 November 2008.

Limitations of registries Comparator group Limited active surveillance time Adverse event of interest may be diagnosed by physiciansoutside the disease registry16

Limitations of registries Comparator group Limited active surveillance time Adverse event of interest may be diagnosed by physiciansoutside the disease registryHypothetical solution? Patients included in Phase III trials should give informedconsent to follow them for life through passive surveillance17

Studying long-term effects in databasesmethodological challenges Selection bias Who are long-term users? Health-seeking behavior Well-controlled Co-medicationIf selection to remain in the cohort is based on factorsrelated to outcome, then results will be biased18

Studying long-term effects in databasesmethodological challenges Selection bias Who are long-term users? Health-seeking behavior Well-controlled Co-medicationIf selection to remain in the cohort is based on factorsrelated to outcome, then results will be biased Some analytic options Comparison of initiators vs. initiators and model time-varyinghazards Comparison of switchers vs. switchers considering cumulativemedication history from initiation until switch (dose, duration,etc).19

20Struggle 2: drug-induced malignancies

21Struggle 2: drug-induced malignancies

22Struggle 2: drug-induced malignancies

Insuline glargine and cancer risk 4 observational studies Hemkens et al submitted to ‘Diabetologia’ The journal wanted confirmation and asked 3 other groupsto perform replication studies23

Insuline glargine and cancer risk 4 observational studies Hemkens et al submitted to ‘Diabetologia’ The journal wanted confirmation and asked 3 other groupsto perform replication studies Conclusions Hemkens: the cancer incidence with glargine was higherthan expected compared with human insulin. Jonasson: women using insulin glargine alone had anincreased incidence rate of breast cancer as compared withwomen using types of insulin other than insulin glargine. Currie: Use of insulin analogues was not associated withincreased cancer risk as compared with human insulin. Colhoun: insulin glargine use was not associated with anincreased risk of all cancers or site-specific cancers and then there was panic......24

25Insuline glargine and cancer risk

Investors are reacting promptly torumours of safety publications26Gale EAM. Diabetologia 2009; 52: 1975-82.

Let’s have a closer lookstudy design issues that could have been tackled Exposure misclassification Whole follow-up period was looked at: continuous users vsswitchers vs non-use Exposure measured during a 4 -6 month run-in period Solution: time-varying exposue measument Selection bias Incident users mixed with prevalent users, while hazardchange over time Proportion of prevalent users differs per drug and per study Type 1 and type 2 diabetes, different proportions Solution: subgroup analyses Different comparator groups Human insulin Insulin other that glargine27

28European Medicines Agengy conclusions

Struggles remainin observational drug-induced malignany research Hazard patterns: tumour initiation (5 yrs) , tumourpromotion (1 yr) Comparing users of drug A with users of drug B given thattreatment allocation is never random. (Longterm) exposure measurement and classification giventhe dynamics of diabetes treatment. The underlying disease (diabetes) is associated with theoutcome; insulin itself has also mitogenic characteristics. How to factor in relevant data on BMI, alcohol, smoking,etc?29

30Take home messages

Take home messages Active risk management needs more time to ‘infiltrate’ thewhole ‘drug landscape’31

Take home messages Active risk management needs more time to ‘infiltrate’ thewhole ‘drug landscape’ Every new safety issue has its own specific (drug, patient ordisease related) methodological problems and may need tobe studied in detail, I prefer common sense over guidelines32

Take home messages Active risk management needs more time to ‘infiltrate’ thewhole ‘drug landscape’ Every new safety issue has its own specific (drug, patient ordisease related) methodological problems and may need tobe studied in detail, I prefer common sense over guidelines Observational (post-marketing) studies, including registriesand database studies can be very useful, but stay alert andask for help if methods are challenging33

34Thank you for your attention