Opti-Q was an international, multi-center, randomized, placebo-controlled, superiority phase II trial comparing four different levofloxacin dosing groups for the treatment of TB resistant to isoniazid and rifampicin but susceptible to fluoroquinolones. The trial was conducted as part of the US Centers for Disease Control and Prevention (CDC) Tuberculosis Trials Consortium (TBTC) with additional funding from US NIH/NIAID (DMID) and drug donation from MacLeods Pharmaceuticals, Ltd. The trial was approved by institutional review boards of Boston University, and all study sites.

Participants with smear-positive pulmonary tuberculosis with evidence of resistance to isoniazid and rifampicin and susceptibility to fluoroquinolones on rapid molecular tests were eligible for the study. Individuals with a QTcF of 450ms or more, use of a fluoroquinolone for 7 days or more within 30 days prior to randomization, poorly controlled diabetes (defined as HgB A1c >9%), aged <18 or weight <40kg were not eligible. Full details of the design have been published previously(24).

Eligible participants were randomized in the ratio 1:1:1:1 to receive one of four weight-adjusted once-daily doses of levofloxacin (250mg tablets) given for 28 weeks (168 doses): 11mg/kg (750mg), 14mg/kg (750mg if weight <60kg, 1000mg otherwise), 17 mg/kg (1000mg if <60kg, 1250mg otherwise) or 20mg/kg (1250mg if <60kg, 1500mg otherwise). Levofloxacin dosing was blinded such that all participants received 6 tablets (combination of active and placebo) once-daily for 24 weeks given as part of a background regimen that was consistent with the then current local and WHO guidelines for the treatment of MDR-TB; the background regimen was selected by local investigators(25, 26) and did not include bedaquiline, linezolid, delamanid, or pretomanid (Tables S4.6 and S4.7 in online supplement)

Randomization was achieved using pre-prepared lists with blocks of varying sizes; separate lists were prepared for each site. Randomization was stratified by site and, in South Africa, by HIV status. The randomization lists were prepared by a statistician independent of the study. Allocation was made by each site pharmacist by opening the next in a series of opaque envelopes.

Participants, clinicians, study staff, laboratory technicians, and study leadership were blinded to treatment allocation. Only the site pharmacist was aware of the allocated dose and only the DSMB reviewed aggregate safety and efficacy data by treatment arm on a 6-monthly basis, supported by the unblinded study statistician. No other individuals in the study were aware of treatment allocation.

Participants were followed for 28 weeks from randomization with sputum samples taken for Middlebrook 7H11 Selective agar and BACTEC MGIT 960 liquid culture every two weeks for the first 12 weeks, and then every four weeks to week 28. Blood samples for chemistry and hematology were taken on the same schedule. Targeted symptoms assessment, focused clinical assessment, and recording of all adverse events (irrespective of relatedness to study medications) was done weekly for the first 8 weeks and with the same schedule as sputum samples thereafter. Adverse events were graded according to the Common Terminology Criteria for Adverse Events v4.0 (CTCAE) published on 28 May 2009.

Pharmacokinetic (PK) sampling for levofloxacin exposures was performed in one 24-hour period between the 14^th and 28^th day after the start of study regimen (further details have been previously published(27)). All baseline M. tuberculosis isolates were shipped to CDC’s Division of Tuberculosis Elimination Reference Laboratory for determination of MIC for levofloxacin. MICs were measured both on 7H10 Middlebrook agar and in a broth microdilution assay; the former was considered primary for all analyses.

The primary objectives of the study were to: 1) determine the levofloxacin AUC/MIC that provides the shortest time to sputum culture conversion on solid medium; 2) determine the highest levofloxacin AUC that is both safe and associated with fewer than 25% of patients discontinuing or reducing their dose of levofloxacin; 3) develop a dosing algorithm to achieve the AUC associated with maximal efficacy and acceptable safety and tolerability. For our efficacy objective, based on murine studies, the target AUC/MIC was set as 100(28)

The primary efficacy outcome was time to negative sputum culture conversion on solid medium, which is defined as the time from initiation of treatment to the first of two consecutive negative cultures (without an intervening positive culture) that are not followed by a culture-positive specimen within 28 weeks. Time to negative sputum culture conversion in liquid medium defined similarly was an important secondary efficacy outcome. The primary safety outcome was the proportion of participants with grade 3 or higher adverse events occurring up to and including four weeks after the last dose of study drug. The primary tolerability outcome was the proportion of participants completing 168 daily doses of levofloxacin within 200 days of initiation of the study regimen (excluding patients withdrawn from treatment where the reason was not at least possibly related to study medication).

Planned total enrollment was 120 participants giving 80 evaluable participants (allowing for up to 33% loss to follow-up). This was calculated to give 95% power to detect a correlation of −0.4 or 88% power to detect a correlation of −0.35 between AUC/MIC and time to culture conversion (log transformed) using the Pearson correlation coefficient. This sample size also gives 80% power to detect a relationship between AUC and the proportion of individuals experiencing a grade 3 or above adverse event (AE) assuming 30% AEs at the mean AUC and 54% when the AUC increases by one standard deviation. The sample size gives 80% power to detect a relationship between AUC and the proportion of participants completing treatment, assuming 85% completion at the mean AUC and 70% when the AUC increases by one standard deviation. These sample size assumptions were not based on previously published studies but based on the judgment of the study investigators as to what correlations would likely be enough to inform dosing guidelines.

The primary analysis population for the efficacy analysis included all randomized participants that took at least one dose of trial medication, analyzed in the treatment arm to which they were allocated, excluding those without a positive culture at screening or baseline on either culture media and excluding those without phenotypic evidence of resistance to isoniazid and rifampicin and those without phenotypic susceptibility to levofloxacin. The primary analysis population for safety analyses included all randomized participants who took at least one dose.

Cochrane-Armitage test for trend was used for evaluating trends in proportions across the four treatment arms or across tertiles of PK and MIC parameters.

Time to culture conversion was analyzed using stratified log-rank test and Cox proportional hazards regression. Testing for non-proportional hazards was done using Schoenfeld residuals. When building a multivariable model, covariates were included if univariable likelihood ratio test p < 0.2 and retained in the model through backward selection if p < 0.1.

This work was supported by the U.S. National Institutes of Health NIAID grant U01 AI100805, the U.S. CDC TBTC, and the Boston University/Rutgers Tuberculosis Research Unit (U19AI111276). Drugs were donated by Macleods Pharmaceuticals, Ltd., which also prepared and supplied matching placebo tablets.

Between January 2015 and December 2016, 255 participants were screened for eligibility and 111 randomized to one of four dosing groups from one site in Cape Town, South Africa and two sites in Lima, Peru. Enrollment was stopped at 111, below the target 120, since the total number of evaluable participants was achieved. The main reason for exclusion from randomization was lack of evidence of resistance to rifampicin or isoniazid, or of susceptibility to fluoroquinolones (Figure 1). Among those randomized, 75 (68%) had no prior diagnosis of TB, 83 (75%) had cavities on chest X-ray, 55 (50%) had a smear grading of 3+, 68 (61%) were of male sex, the median age was 26.0 years old and the median BMI was 20.4 kg/m². Fifty-eight (52%) had disease resistant to pyrazinamide on phenotypic DST. There were 78 (70%) participants who received 5 drugs and 28 (25%) received 6 drugs at baseline, in addition to levofloxacin (Table 1). Baseline characteristics were well balanced between randomized groups.

Three participants never started treatment and were excluded from the ITT analysis population; a further 10 participants had no baseline positive cultures, or had baseline phenotypic DST indicating ofloxacin resistance, rifampicin susceptibility, or indeterminate isoniazid susceptibility (Figure 1).

MIC results were available from baseline samples for 77 individuals (75 for primary MIC on agar plate, AP); participants characteristics of these 77 were similar to those with missing MIC (Table S4.3 in online supplement). Median (Interquartile range) levofloxacin MIC was 0.2 μg/ml (0.2, 0.2) on Agar Plate and 0.5 μg/ml (0.5, 0.5) in broth microdilution (BMD), Figure 2. Of 77, 9 participants had no levofloxacin PK data (see Figure 1). Among the remaining 68 participants, median AUC/MIC was 855.0 with the minimum being 332.1, above our pre-specified target AUC/MIC of 100 (Figure 2C.).

There were 108 participants who started treatment and are included in the safety analysis population. Of these, 25 (23.1%) experienced a grade 3–5 adverse event (AE), 10 (9.3%) experienced a serious AE (SAE) and one participant died as a result of an intracranial IRIS reaction in the setting of unrecognized CNS TB (on the 17mg/kg treatment arm, see supplement table S9.8 for more details). There was evidence of an increasing trend (Cochrane-Armitage test for trend) in the proportion of participants experiencing a grade 3–5 AE (p=0.042) and by tertile of AUC (p=0.022) and in the proportion of participants experiencing any SAE by tertile of AUC (p=0.011, Table 3). The participants with the three highest exposures all experienced grade 3–5 AEs and one of these was the participant that died (see Online Supplement, Figure S9.1). The only grade 3–5 AE that occurred in more than two participants was blood uric acid increase (7 participants). The most common grade 3–5 adverse events (shown in Table 3) are more likely related to companion drugs than to levofloxacin.

A single participant had a QTcF exceeding 500ms (Grade 3); this participant was in the 20mg/kg treatment arm and had the highest recorded AUC, although the QT prolongation was considered likely attributable to concomitant use of azithromycin, a protocol-prohibited drug with known QT prolongation potential. The AE resolved after discontinuation of azithromycin; levofloxacin was not discontinued. One further participant experienced a QTcF of between 470 and 500ms after randomization in the lowest tertile of AUC. Both participants with QT interval of 470 ms or greater subsequently completed study treatment. A further 8 participants experienced a QTcF of greater than between 450 and 470 ms after randomization. These events occurred more frequently among participants in the highest tertile of AUC, 7 (21.2%), compared to the lowest tertile, 1 (3.0%), p=0.014, and p=0.029 excluding the participant who received azithromycin. Furthermore, excluding this participant, there was evidence of a linear relationship between maximum QTcF from visits at or after randomization and AUC (p < 0.001, Figure 4) with a mean increase of 7.3ms, 95% CI (3.1ms, 11.5ms), per increase in 100 h*μg/ml. The largest difference in QTcF between lower and upper tertile of AUC was seen at 16 weeks (see supplement Figure S12.4) with a mean change from baseline of QTcF of 8.0ms, 95% CI (1.6s, 14.4ms), in the lower tertile and 14.9ms, 95% CI (6.0ms, 23.8ms) in the upper tertile (excluding the participant who received azithromycin), a ΔQTcF of 6.9ms (95% CI −4.0ms, 17.9ms).

Since the actual dose of levofloxacin received depended both on allocated regimen and participant weight, we present safety outcomes by both in the online supplement (Table S9.7). Considering analyses within each weight band which allows us to compare doses directly, 750mg and 1000mg had similar safety outcomes (11–17mg/kg among those <60kg and 11–14mg/kg among those ≥60mg/kg) however an increase was observed at the higher doses (1250mg in the lower weight band and 1500mg in the highest weight band).

In addition to the three that never started treatment, six participants were excluded from the tolerability analysis population for withdrawing from study medications for reasons not related to treatment (see Table S11.1 in supplement and Tables S11.2–S11.8 for sensitivity analyses). There were 71 (70%) participants who completed 168 doses within 200 days. There was no change in this tolerability outcome by dose (p = 0.442, supplementary table S11.1, Cochran-Armitage test for trend) but tolerability increased slightly by tertile of AUC (p = 0.025, supplementary table S11.5, 27 (84%) in highest tertile compared to 25 (81%) in lowest tertile). Of those that did not complete treatment, only 1 participant had a treatment failure outcome (20mg/kg arm, online supplement Table S8.2).

There was no evidence of a difference in time to culture conversion on solid media by treatment arm (stratified log-rank p = 0.282), by tertile of AUC/MIC (p =0.350, Table 2, Figure 3), or by actual mg dose received (p=0.723). There was also no difference when considering MGIT liquid culture instead (p=0.313 by treatment arm, p=0.121 by tertile of AUC/MIC, p=0.171 by actual mg dose received), or when considering PK parameters as continuous in a PK-PD analysis rather than as tertiles (supplement S14.23 and S14.30). There were 73.4% participants who achieved culture conversion after week 8 (by week 9) on solid culture (7H11S) and 64.1% in MGIT liquid culture; culture conversion was 93.3% and 87.5% after week 12 (by week 13) on solid culture and in liquid MGIT culture respectively. In multivariable Cox regression models, these findings did not change even after adjusting for important predictors of time to culture conversion. Important predictors were smear grade (likelihood ratio test p < 0.001) and study site (p = 0.07) for solid culture outcome; and baseline MGIT days to positivity (DTP) on the log scale (likelihood ratio test p < 0.001) and age (p = 0.005) for MGIT (see online supplement, Tables S15.1–S15.4). Culture results at 6 months are shown in Tables S16.1 and 16.2, 94% of participants had negative MGIT cultures.