1. 26 Nov 2009 Dar es Salaam 1 Selecting the cohort David Coulter

2. 26 Nov 2009 2Dar es Salaam Simple Record all patients consecutively until you have 10,000 – 15,000 Record all patients consecutively until you have 10,000 – 15,000 But there’s more to think about But there’s more to think about

3. 26 Nov 2009 3Dar es Salaam Numbers In general, 10,000 patients gives a 95% chance of identifying an event with an incidence of 1:3000 (uncommon-rare) In general, 10,000 patients gives a 95% chance of identifying an event with an incidence of 1:3000 (uncommon-rare) Smaller numbers can still be very useful (eg IMMP & nifedipine) Smaller numbers can still be very useful (eg IMMP & nifedipine) Greater numbers may be needed to test for any significant increase in incidence: Greater numbers may be needed to test for any significant increase in incidence: in sub-groups of interest eg tuberculosis in sub-groups of interest eg tuberculosis if the background incidence of an event of special interest is common if the background incidence of an event of special interest is common when comparing comparator drugs when comparing comparator drugs

4. 26 Nov 2009 4Dar es Salaam Numbers At the same level of seriousness, common events are more important because they affect more people At the same level of seriousness, common events are more important because they affect more people A non-serious common reaction could be more important than a rare serious reaction A non-serious common reaction could be more important than a rare serious reaction So, a 95% certainty of identifying an event at 1:1000 frequency (a smaller cohort) may be OK So, a 95% certainty of identifying an event at 1:1000 frequency (a smaller cohort) may be OK Be aware of the limitations / strengths of your numbers Be aware of the limitations / strengths of your numbers

5. 26 Nov 2009 5Dar es Salaam Specific age groups eg children The whole population of users will still need to be monitored to enable comparison of children with adults and determine any specific risks or risk factors for children The whole population of users will still need to be monitored to enable comparison of children with adults and determine any specific risks or risk factors for children It may be necessary to continue monitoring children longer in order to get more numbers for higher statistical power It may be necessary to continue monitoring children longer in order to get more numbers for higher statistical power

6. 26 Nov 2009 6Dar es Salaam Other sub-group analyses eg malnutrition, concomitant drugs Larger numbers of these patients might be needed to detect differences with sufficient statistical power Larger numbers of these patients might be needed to detect differences with sufficient statistical power It all depends on the incidence It all depends on the incidence Common events no problem Common events no problem Rare events need larger numbers Rare events need larger numbers

7. 26 Nov 2009 7Dar es Salaam Approach Aim for 15,000 Aim for 15,000 Review data during monitoring (‘real time’ –interim analyses) Review data during monitoring (‘real time’ –interim analyses) This will provide a better idea of the problems needing evaluation and their incidence This will provide a better idea of the problems needing evaluation and their incidence Reassess needs at intervals (every 3-6 months) Reassess needs at intervals (every 3-6 months)

8. 26 Nov 2009 8Dar es Salaam Choice of patients Principles –aim for: Principles –aim for: Unselected Unselected Consecutive or random Consecutive or random Typical of HIV/AIDS management in the country Typical of HIV/AIDS management in the country Determinants of choice Determinants of choice Expected numbers of cases Expected numbers of cases Time frame Time frame Logistics of involvement in the clinics Logistics of involvement in the clinics

9. 26 Nov 2009 9Dar es Salaam Choice of patients Timing / logistics Every day, all day would avoid biases Every day, all day would avoid biases If workload would be too great for every day, then: If workload would be too great for every day, then: All patients seen on specific days All patients seen on specific days Choose typical days Choose typical days Maybe choose mornings (or afternoons) Maybe choose mornings (or afternoons) Discuss with clinic staff Discuss with clinic staff Run a pilot Run a pilot Try and be consistent across sites Try and be consistent across sites

10. 26 Nov 2009 10Dar es Salaam Choice of patients Health facility Which health facilities? Which health facilities? Selection needs to be representative Selection needs to be representative Urban / rural Urban / rural Hospitals / clinics Hospitals / clinics Geographically Geographically Language areas (cultural differences) Language areas (cultural differences) Selection needs to be in proportion to numbers of HIV/AIDS patients normally seen Selection needs to be in proportion to numbers of HIV/AIDS patients normally seen eg if 40% of HIV/AIDS patients are urban, have 40% of cohort from urban sites eg if 40% of HIV/AIDS patients are urban, have 40% of cohort from urban sites Choose (sentinel) health facilities Choose (sentinel) health facilities Set target numbers for each Set target numbers for each

11. 26 Nov 2009 11Dar es Salaam Choice of patients The study will be dynamic –new patients are added as you go along The study will be dynamic –new patients are added as you go along The study must be inceptional –all patients must be monitored from the commencement of their treatment to exclude ‘drop-out’ bias The study must be inceptional –all patients must be monitored from the commencement of their treatment to exclude ‘drop-out’ bias

12. 26 Nov 2009 12Dar es Salaam ‘Controls’ Increase cost Increase cost Control cohorts create an artificial situation Control cohorts create an artificial situation The aim is a non-interventional study in normal clinical practice The aim is a non-interventional study in normal clinical practice Comparison of different regimens Comparison of different regimens not always possible if numbers low not always possible if numbers low pooled international data increases power pooled international data increases power A good study of a single regimen A good study of a single regimen provides valuable data provides valuable data has benchmark value has benchmark value ‘Self-controls’ are a helpful option ‘Self-controls’ are a helpful option

13. 26 Nov 2009 13Dar es Salaam Planning Select health facilities Select health facilities Hospitals Hospitals Clinics Clinics Make them representative as far as practical Make them representative as far as practical urban / rural urban / rural Set target numbers for each Set target numbers for each Determine what times monitoring will take place Determine what times monitoring will take place

14. 26 Nov 2009 14Dar es Salaam ____ ___ ____ ________ ____ ___ ____ ________ ____ ___ ____ ________ ____ ___ ____ ________ Involve everyone