Prevention of Infection with Mycobacterium tuberculosis by H4:IC31® Vaccination or BCG Revaccination in Adolescents

Background Recent Mycobacterium tuberculosis (M.tb) infection predisposes to tuberculosis disease, the leading global infectious disease killer. We tested safety andefficacy of H4:IC31® vaccination or Bacille Calmette-Guerin (BCG) revaccination for prevention of M.tb infection. Methods QuantiFERON-TB Gold In-tube (QFT) negative, HIV-uninfected, remotely BCG-vaccinated adolescents were randomized 1:1:1 to placebo, H4:IC31® or BCG revaccination (NCT02075203). Primary outcomes were safety and acquisition of M.tb infection, defined by initial QFT conversion tested 6-monthly over two years. Secondary outcomes were immunogenicity and sustained M.tb infection, defined by sustained QFT conversion without reversion three and six months post-conversion. Statistical significance for efficacy proof-of-concept was set at 1-sided p<0.10. Results 990 participants were enrolled. Both vaccines had acceptable safety profiles and were immunogenic. QFT conversion occurred in 134 and sustained conversion in 82 participants. Neither H4:IC31® nor BCG prevented initial QFT conversion, with efficacy point estimates of 9.4% (95% confidence interval: -36.2, 39.7; one-sided p=0.32) and 20.1% (-21.0, 47.2; one-sided p=0.14), respectively. However, BCG did prevent sustained QFT conversion with an efficacy of 45.4% (6.4, 68.1; one-sided p=0.013); H4:IC31® efficacy was 30.5% (-15.8, 58.3; one-sided p=0.08). QFT reversion rate from positive to negative was 46% in BCG, 40% in H4:IC31 and 25% in placebo recipients. Conclusions This first proof-of-concept, prevention of M.tb infection trial showed that sustained infection can be prevented by vaccination in a high-transmission setting and confirmed feasibility of this strategy to inform clinical development of new vaccine candidates. Evaluation of BCG revaccination to prevent tuberculosis disease in M.tb- uninfected populations is warranted.

3. Abnormal laboratory values from the most recent blood collected prior to randomization 5 as follows: 6 • Laboratory evidence of hematologic disease (white blood cell count <3000/mm 3 or 7 >11,500/mm 3 ; hemoglobin <0.9 times the lower limit of normal of the testing 8 laboratory, by age and gender; absolute neutrophil count <1300/mm 3 ; absolute 9 lymphocyte count <1000/mm 3 ). 10 • ALT, AST, alkaline phosphatase, total bilirubin, creatinine, blood urea nitrogen (BUN) 11 >1.25 times the upper limit of normal of the testing laboratory.  20. Any current medical, psychiatric, occupational, or substance abuse problems that, in the 4 opinion of the investigator, will make it unlikely that the subject will comply with the 5 protocol. increases the immune response to peptide antigens. ODN1a is a synthetic bacterial 23 deoxyribonucleic acid (DNA) analogue that resembles a CpG pattern that will direct the 24 adaptive immune response toward a T helper cell type-1 (Th1) pattern with production of 25 interferon-g (IFNg) and interleukin-12 (IL12). The amount of adjuvant given may affect the 26 immune response. The optimal molar ratio of KLK to ODN1a in mice is 25:1 5  according to the manufacturer's instructions. After reconstitution, 1 adult dose (0.1 mL) 5 contains 2 to 8 x 10 5 CFU and was administered on D0 intradermally with a 22 gauge 6 needle. Sauton SSI diluent is composed of magnesium sulphate, dipotassium phosphate, 7 citric acid, monohydrate L-asparagine monohydrate, ferric ammonium citrate and glycerol at 8 85%, which is reconstituted with sterile water for injection. 9 QFT assay 17 The QFT assay was conducted according to the manufacturer's instructions, with 18 implementation of more stringent parameters unless otherwise stated. Briefly, 1mL of blood 19 was collected in each QFT tube and mixed by 10 manual inversions and 5 minutes on a tube 20 rotator, incubated at 37°C within 2 hours from phlebotomy for 16 to 20 hours, and plasma 21 was harvested without further delay. Supplementary quality control criteria, in addition to the 22 manufacturer's algorithm, were implemented as described 6 . 23

1.5.1
Sample size and interim analyses 4 Sample size was based on detection of reduction in rate of M.tb infection as determined by 5 QFT conversion. The trial was designed to distinguish QFT conversion rate reduction of 50% 6 compared to placebo for each vaccine (H4:IC31 or BCG) with 80% power and a type 1 error 7 rate of 10% (1-sided). 8 9 Design calculations were performed by computer simulation estimating the statistical 10 operating characteristic of a conditional binomial test, which, due to relative rarity of the 11 endpoint, provides a reliable approximation to the operating characteristic of the log-rank 12 test. The following assumptions, based on preliminary data from the Worcester study site 8  Analysis Plan (SAP), which was finalized prior to database lock. 32 The final analyses associated with the primary and secondary endpoints remained entirely 33 unchanged from the primary analysis. Some of the exploratory endpoints that had not been 34 included within the primary analysis, including the evaluation of alternative QFT thresholds 35 and the evaluation of sustained QFT conversions and late reversions at EoS, were proposed by blinded members of the study team prior to database lock. While members of the JSC 1 approved the final SAP, they were not involved in the selection of alternative QFT thresholds 2 or the determination of methodology to evaluate trends in QFT over time. 3 4 The study was not powered to assess vaccine efficacy for prevention of TB disease. Based 5 on studies in this population [8][9][10] we would expect to observe between 1 and 13 TB incident 6 cases 95% of the time. 7 8 Unless otherwise indicated, SAS 9.4 (SAS Institute, Cary, NC) was used for data analysis, 9 and the analyses were completed by Quintiles (now called IQVIA). 10 11

1.5.2
Efficacy 12 Prior to database lock and trial unblinding, we chose to prioritize the modified intent-to-treat 13 analyses over the per-protocol analyses to be more reflective of a potential vaccine effect 14 from real-world use rather than in a more idealistic clinical trial setting. 15 The trial's primary efficacy endpoint, time to first QFT conversion after D84 using the 16 manufacturer's recommended threshold of IFNg ≥ 0.35 IU/mL, was evaluated using a one-17 sided log-rank statistic to compare H4:IC31 versus placebo and BCG versus placebo without 18 adjustment for multiplicity. The rationale for not adjusting for multiplicity to control the Type 1 19 error rate over the two tests is that the tests pertain to efficacy assessments of two unrelated 20 vaccines. If these two evaluations were performed in two separate trials then no adjustment 21 for multiplicity would be made. Thus, one should not be compelled to adjust for this 22 multiplicity simply because the two evaluations are performed administratively in a single 23 trial. 24 Estimates of vaccine efficacy were based on the hazard ratio estimate from a Cox 25 regression model. Participants who did not experience a primary endpoint were censored at 26 the date of their last available, non-positive QFT result; indeterminate results were treated as 27 negative results. The log-rank statistic was evaluated using a one-sided test to reflect the 28 way in which the trial was originally designed and powered, and two-sided 80% and 95% CI 29 of vaccine efficacy are reported. Sensitivity analyses associated with the primary efficacy 30 endpoint, including time to first QFT conversion after D0 (i.e., including conversions detected 31 on D84) and time to event using interval-censoring, were also evaluated. 32

33
The secondary efficacy endpoint was defined as sustained QFT conversion without a ≥ 0.35 IU/mL maintained at 3 and 6 months post-initial conversion). This endpoint was 1 evaluated in the same manner as described above for the primary efficacy endpoint. 2 3 Several exploratory efficacy analyses further evaluated the vaccine ability to prevent or 4 modulate infection: 5 • Median QFT levels at the time of initial conversion were compared between 6 treatment groups using Mann-Whitney-Wilcoxon tests. 7 • Time to QFT reversion among those participants who converted was evaluated using 8 a log-rank test. 9 • Time to first QFT conversion among those who ever reverted was evaluated using a 10 log-rank test. 11 • Time to end-of-study sustained conversion (as opposed to sustained conversion for 6 12 months) was evaluated in the same manner as the secondary efficacy endpoint 13 (described above). 14 15 The proportion of participants classified as converters according to alternative QFT 16 thresholds (i.e., ≥0.2IU/mL; <0.2IU/mL at any time point prior to conversion and >0.7IU/mL 17 at any time point post-D84; >0.7IU/mL; >4.0IU/mL) were compared across study arms. 18 Vaccine efficacy was calculated based on the conditional binomial distribution, and 80% and 19 95% CI were calculated using the Clopper-Pearson method with mid-p correction. 20

21
Following an IDMC request at the time of the primary analysis to obtain Month 24 QFT 22 results for all participants, including those who experienced conversions earlier in the study 23 (at Month 6 or Month 12), additional exploratory analyses included an assessment of the 24 QFT conversion and reversion rates at the end of the 24-month study follow-up period. 25 Trends in QFT status were analysed similarly to the alternative QFT thresholds (i.e., based 26 on a conditional binomial distribution and using the Clopper-Pearson method with mid-p 27 correction for 80% and 95% CI) and were compared between treatment groups based on 28 Pearson's Chi-square for the following endpoints of interest: 29 • Participants who were QFT negative at Month 24 or the Month 24 call-back visit. 30 • Participants who never converted or who converted without sustained conversion. 31 • Participants who never converted or who converted with reversion and had non-32 positive 6-month post-initial-conversion QFT values.
Note: Some analyses were based on subsets of participants determined by a post-1 randomization event such as reversion post-conversion, and some involved participants 2 who experienced differential follow-up. These types of analyses have the potential to be 3 biased, therefore they were not prioritized.