1. 27 Nov 2009 Dar es Salaam 1 CEM analyses

2. 27 Nov 2009 2Dar es Salaam

3. What are we trying to measure? First, the risks associated with each ARV regimen First, the risks associated with each ARV regimen Each regimen will be treated like a single drug Each regimen will be treated like a single drug Second, the risks associated with individual drugs Second, the risks associated with individual drugs 27 Nov 2009 3Dar es Salaam

4. 27 Nov 2009 4Dar es Salaam Approach Keep it simple Keep it simple most of the important results can be revealed by simple measures most of the important results can be revealed by simple measures Keep it accurate Keep it accurate clean, quality data clean, quality data controlled at input controlled at input Example: variation in spelling Example: variation in spelling

5. 27 Nov 2009 5Dar es Salaam Small things 1 Rates Rates What to do with drop-outs? What to do with drop-outs? Choosing the denominator Choosing the denominator total cohort includes drop-outs total cohort includes drop-outs demographics demographics total follow-up questionnaires returned total follow-up questionnaires returned general rates eg rates of events; reactions & incidents general rates eg rates of events; reactions & incidents Sites / regions Sites / regions Total per question answered Total per question answered Rate per data element eg traditional meds, TB Rate per data element eg traditional meds, TB

6. Small things 2 Examples of rates Overall reporting rate (response rate) Total cohort enrolled = 8432 Total follow-up forms received = 6998 Reporting rate = 6998/8432 = 83% Rates of events Total follow-up forms received = 6998 Total patients with oral candidiasis = 843 Rate of oral candidiasis = 843/6998 = 12% 27 Nov 2009 6Dar es Salaam

7. Small things 3 Examples of rates Rates per total question answered for use of traditional medicines Rates per total question answered for use of traditional medicines Total follow-up forms received = 6998 Total follow-up forms received = 6998 Total TM questions answered = 5432 Total TM questions answered = 5432 Total answered ‘Yes’ = 3954 Total answered ‘Yes’ = 3954 Rate of use of traditional medicines = 3954/5432 = 73% Rate of use of traditional medicines = 3954/5432 = 73% 27 Nov 2009 7Dar es Salaam

8. Small things 4 Calculation of risk The rate of occurrence of an event in the exposed cohort is a measure of absolute risk The rate of occurrence of an event in the exposed cohort is a measure of absolute risk Attributable risk: this is a measure of the increased risk associated with the medicine. It is calculated as follows: Attributable risk: this is a measure of the increased risk associated with the medicine. It is calculated as follows: rate in the exposed cohort minus rate in the exposed cohort minus rate before exposure (in the control period) rate before exposure (in the control period) 27 Nov 2009 8Dar es Salaam

9. 27 Nov 2009 9Dar es Salaam Small things 4 Doses Doses Analyses are done on the total daily dose Analyses are done on the total daily dose Total daily dose x weekly e.g. Total daily dose x weekly e.g. 300mg 3 x weekly 300mg 3 x weekly Or 100mg 7 x weekly Or 100mg 7 x weekly Duration Duration Duration in days Duration in days 1 m = 30 days 1 m = 30 days 1 y = 365 days 1 y = 365 days Between x & y –take mid-point Between x & y –take mid-point

10. 27 Nov 2009 10Dar es Salaam Data manipulation Collation summary of reporting rates for males, females and totals; summary of reporting rates for males, females and totals; age/sex profiles of the cohort; age/sex profiles of the cohort; patient numbers by region or site; patient numbers by region or site; event profiles by clinical category; event profiles by clinical category; within a clinical category within a clinical category reactions & incidents reactions & incidents

11. 27 Nov 2009 11Dar es Salaam

12. 27 Nov 2009 12Dar es Salaam IMMP example –COX-2 Age & sex CelecoxibRofecoxib Mean6358 Mode5953 <40 years 6.9%12.6% Highly significant 70+ years 32.7%25.7% Highly significant Women61.6%60.5%

13. 27 Nov 2009 13Dar es Salaam Celecoxib dose mg/no./%

14. 27 Nov 2009 14Dar es Salaam

15. 27 Nov 2009 15Dar es Salaam Event profiles CelecoxibRofecoxib Odds Ratio Circulatory 315 (17%) 214 (20%) NS Alimentary 302 (17%) 218 (20%) 0.8 (0.66,0.97) Died 293 (16%) 179 (16%) Skin 160 (9%) 59 (5%) 1.7 (1.24-2.30) Respiratory 108 (6%) 54 (5%) Psychiatric 88 (5%) 49 (5%) Renal 72 (4%) 46 (4%)

16. 27 Nov 2009 Dar es Salaam 16 Events collation

17. 27 Nov 2009 17Dar es Salaam Event collations Look Carefully

18. 27 Nov 2009 Dar es Salaam 18 Understanding the dictionary

19. 27 Nov 2009 19Dar es Salaam Examples Annex 8 events collation Annex 8 events collation Annex 12 eye events with signal Annex 12 eye events with signal Annex 11 deaths Annex 11 deaths Annex 13 concomitant medicines Annex 13 concomitant medicines

20. 27 Nov 2009 20Dar es Salaam Risk Risk calculation Risk calculation Relative risk (rate ratio) Relative risk (rate ratio) Rate of A / rate of B = relative risk Rate of A / rate of B = relative risk Confidence intervals Confidence intervals Risk factors eg age Risk factors eg age RR RR Multiple logistic regression Multiple logistic regression

21. Risks of individual drugs Compare events before and after drug substitutions with the patients acting as their own controls eg Compare events before and after drug substitutions with the patients acting as their own controls eg d4t(30)-3TC-NVP to d4t(30)-3TC-EFV d4t(30)-3TC-NVP to d4t(30)-3TC-EFV Compare events between regimens where substitutions have taken place Compare events between regimens where substitutions have taken place Use multiple logistic regression to test each drug as a risk factor for events of interest Use multiple logistic regression to test each drug as a risk factor for events of interest Large database needed for these analyses (pooled international data) Large database needed for these analyses (pooled international data) 27 Nov 2009 21Dar es Salaam

22. 27 Nov 2009 22Dar es Salaam Life table (survival) analysis Helps to characterise a reaction Helps to characterise a reaction time to onset time to onset spread of onset times spread of onset times Testing a possible signal Testing a possible signal

23. Statistical programmes CemFlow CemFlow automated outputs automated outputs MedCalc MedCalc free for 25 sessions free for 25 sessions 27 Nov 2009 23Dar es Salaam

24. 27 Nov 2009 24Dar es Salaam Don’t be afraid! of statistics