Phase IIb (8-week) studies

1. Phase IIb (8-week) studies D A Mitchison St George’s, University of London

2. Comparison of the bactericidal activities of the Fluoroquinolones: Gatifloxacin, Moxifloxacin and Ofloxacin, substituted for Ethambutol in the 2-Month initial phase of standard treatment Oflotub phase 2 surrogate marker study South African Medical Research Council, Durban Dr R Rustomjee; Dr F Sirgel; Prof W Sturm; Dr B Fourie and staff Prof D Mitchison; Dr C Lienhardt (Co-ordinator); Dr C Merle (Trial Monitor) Supported by European Commission WHO TDR

3. Design • Open label Phase II Randomised Controlled Trial Aims • Compare three fluoroquinolones substituted for ethambutol in 2HRZE initial phases using serial sputum colony counting (SSCC) • Compare the use of different surrogate endpoints in Phase II studies

4. Screening Newly diagnosed smear +ve 412 Willing to collaborate 226 Excluded 11 HRZE Control 54 HRZO Oflo 55 HRZG Gati 54 HRZM Moxi 54 8-weeks RH continuation 4 months Summary of recruitment

5. Comparative bactericidal assessments 14 hour sputum collection Sputum colony counts on selective 7H11 medium without decontamination at 10 time points during initial 8-week phase 0 2 7 14 21 28 35 42 49 56 Standard 7H11 culture + indirect susceptibility tests Standard 7H11 culture Standard 7H11 culture + Liquid culture (MGIT)

6. Jonathan Levin (Durban): Proportions +ve at 2 months Survival analysis using SSCC Polynomial / spline mixed effects modelling of SSCC Geraint Davies (Bangkok): Non-linear mixed effects modelling of SSCC Denny Mitchison (London): Summary regression estimates on SSCC Analysis

7. Summary of SSCC results Limit of detection

8. Analytical approaches to Phase II surrogate endpoint studies 10 8 Summary statistics & regression models of colony counts Proportion positive at 2 months 6 Log10 CFU /ml 4 2 Survival analysis & time to conversion Relapse Limit of detection 0 0 10 20 30 40 50 Days on therapy

9. Proportion positive at two months Chi-square test 2 (3)=2.63 p= 0.451 58.5 49.0 46.9 43.4

10. Survival analysis Log rank test 2 (3)=10.69 p= 0.0136

11. Survival analysis * vs. HRZE

12. Regression modelling of serial sputum colony counts Limit of detection

13. Model fitting sequence

14. Regression estimatesSigns of coefficients reversed) * <0.05 M vs.E **<0.01 M vs.E * <0.05 M&G vs. E&O **<0.01 M&G vs. E&O

15. Limit of detection Estimated treatment effects

16. -17.8 9.2 -7.0 23.8 1.4 35.2 9.0 103.1 -100 -50 0 50 100 Forecasting duration of therapy by effect size Approximate 95% confidence intervals derived from NLME model HRZO -5.3 HRZG 6.5 HRZM 17.3 48.7 (SHRZ) % Change in versus control

17. Forecasting duration of therapy by extrapolation ~100 days ? 138 days Triexponential model Biexponential model

18. Cost assessment (US$) Duration of study: 12 months Number of sputum specimens: 2100 Technicians employed: 3

19. Conclusions 1 • When substituted for Ethambutol, both Moxifloxacin and Gatifloxacin killed significantly faster in the early and late phases than control • Ofloxacin substitution had no significant effect • The observed increases in late phase killing with these regimens support a probable reduction in the duration of therapy of at least one and possibly two months

20. Conclusions 2 • Several different methods of analysis detected a treatment effect but culture conversion at 8 weeks did not • It is important that methods discriminate between the early and late phases of killing and avoid overestimating “sterilizing activity” • Non-linear mixed effects modelling appears to be both an informative and sensitive approach

21. Conclusions 3 • Modelling using SSCC data has significant advantages over methods based only on culture conversion • SSCC is also cheaper as it requires a smaller sample size • SSCC modelling should be the basis of future phase II studies aiming to evaluate new regimens suggested by mouse studies.