Risk of rifampin resistance emergence after incomplete first-line tuberculosis treatment

Dear Editor

Tuberculosis (TB) treatment is lengthy and causes side effects, making treatment completion challenging. Some patients are “lost to follow-up” (LTFU) before completing treatment. Patients sometimes subsequently return to care if symptoms motivate them, or health systems and/or personal issues that caused LTFU are resolved. Case-control studies established that drug-susceptible TB (DS-TB) treatment and incomplete adherence were risk factors for relapse with drug-resistant TB (DR-TB), but its frequency and the lengths of incomplete treatment that pose the greatest risk are unknown[1, 2]. We aimed to estimate the risk of recurrent TB, and specifically rifampin resistant TB (RIF-R), after rifampin susceptible TB (RS-TB) treatment and how these risks vary depending on previous RS-TB treatment length.

We used Ukraine’s National TB Program surveillance database, eTB Manager, which records information on all TB diagnoses. eTB Manager includes demographic information and baseline (at diagnosis) and follow-up laboratory results (smear, culture, GeneXpert, drug resistance testing, and HIV and antiretroviral [ART] status). A unique ID allows linkage of an individual’s records across multiple treatment episodes. All outcomes are recorded, including LTFU and if and when they subsequently resume care. We included everyone diagnosed 1^st January 2015 to 9th November 2018.

TB diagnosis for our analysis was based on GeneXpert, smear microscopy, or culture. RIF-R was confirmed by GeneXpert (87% of people were tested with GeneXpert), culture, or line probe assay. Outcomes were defined as per the World Health Organization, including LTFU defined as treatment interruption for ≥2 consecutive months[3]. If a patient returns to treatment after LTFU, this is recorded, with new treatment start dates, and laboratory results (e.g. new drug resistance). We defined recurrence as when someone had an initial outcome of cured, treatment completed, or LTFU and the person then returns to TB care with a positive TB diagnosis within 18 months – this return is the recurrence. We chose 18 months: to ensure equal follow-up time for all participants, and to preferentially capture likely relapses (that would occur earlier post-treatment) than reinfections.

We included all TB episodes with outcomes of cured, treatment completed (both 6+ months after diagnosis; “successful”), or LTFU before 6 months, that were recorded on or before May 8, 2017 (18 months before the November 9, 2018 download date to ensure full follow-up time). We excluded people with a drug resistant isolate during initial treatment to exclude any potential misclassification of baseline drug resistance status.

We calculated the absolute recurrence percentage (i.e. risk) among cohort members who were either LTFU or completed treatment. We also calculated the risk of RIF-R recurrence among people in our cohort who experienced a recurrence, regardless of treatment length or initial outcome.

We used log binomial regression to estimate the adjusted association between recurrence and treatment duration measured in months up to 6 months compared with one duration category for people with a successful outcome after 6+ months. To construct a multivariable model, we used backward selection, initially adjusting for HIV/ART status, self-reported alcohol and/or drug abuse, homelessness, unemployment, incarceration, migration, extra-pulmonary TB, lung cavitation, isoniazid resistance, age, and sex. We used a cutoff p-value of ≤0.15 for keeping variables and retained any confounders as well as age and sex regardless.

To estimate the association between treatment duration and RIF-R risk on recurrence, we focused on people with a recurrence after LTFU and used log binomial regression to estimate the association between RIF-R on recurrence and treatment duration before LTFU, comparing individual months of treatment with <1 month. Here, we excluded anyone without rifampin susceptibility testing after recurrence. We constructed a multivariable model as described for the recurrence model above.

We used R version 4.3.3 (R Foundation, Vienna, Austria). This study used nonidentifiable data and, therefore, was deemed exempt from institutional review board (IRB) approval by Bogomolets National Medical University (Kyiv City, Ukraine) and the Public Health Center of the Ministry of Health (Kyiv City, Ukraine) and was considered non-human subjects research and thus, ethics approval was waived by the Boston University Medical Center IRB.

Among the 29,341 people included in our cohort, the recurrence risk was 3.9% (1,143/29,341). Most recurrences (68.6%) occurred ≤6 months post-treatment (Figure). The recurrence risks among people who were LTFU versus those with a successful outcome were 654/2,331 (28.1%) and 489/27,010 (1.8%), respectively.

Figure.

Time to (a) any recurrence and (b) rifampin-resistant recurrence after stopping rifampin susceptible tuberculosis treatment. Risk of (c) any recurrence by treatment duration and (d) rifampin-resistant tuberculosis on recurrence by treatment duration. The reference (risk ratio=11) in figures (c) and (d) is shown with a red horizontal line.

The adjusted recurrence risk ratio for people who were LTFU versus people with a successful outcome was 9.2 (95% CI: 8.2, 10.4). The recurrence risk ratio was highest for people with 1 to <2 months of treatment (12.1 [95% CI: 9.8, 14.8], respectively) compared with those with a successful outcome (Figure) and decreased with increasing treatment duration (Figure).

Of the 1,005 people that experienced recurrence and had RIF-R testing on recurrence, 227 (22.6%; or 0.77% of the 29,341 people with RS-TB initially) were diagnosed with RIF-R at recurrence. Of the 571 people with recurrence after LTFU, 87 (15.2%; or 3.7% of all 2,331 people who were LTFU) were diagnosed with RIF-R at recurrence. The RIF-R risk among people with a successful outcome was 0.52% (140/27,010).

People who received 3 to <4 and 4 to <5 months of treatment had a risk of RIF-R on recurrence of 1.85 (95% CI: 0.91, 3.74) and 1.61 (95% CI: 0.79, 3.28), times that of people with <1 month of treatment (Figure). People with 1 to <2 months or 5 to <6 months of treatment had a similar RIF-R risk to those with <1 month of treatment. Baseline isoniazid resistance was associated with increased RIF-R risk (1.86 [95% CI: 1.09, 3.17]) and PLHIV had increased risk of recurrence and increased risk of RIF-R on recurrence.

We show that around 4% of people with RS-TB starting treatment in Ukraine subsequently have recurrent TB, and >20% of those have RIF-R on recurrence. Moreover, the fact that RIF-R recurrences occurred soon after treatment cessation (Figure) strongly suggests these were not re-infections. Even among people “successfully” treated for RS-TB, 0.5% will develop RIF-R, a clear explanation for the evolution of the global MDR-TB epidemic. Importantly, among those with incomplete treatment, the RIF-R risk was 3.7% (87/2,331) and further increased for treatment lengths of only 3–5 months, indicating that TB patients could benefit from treatment support throughout the full course of treatment.

Our results provide a more precise estimate of recurrence and drug resistance emergence than two previous small studies[4, 5]. To our knowledge, this is the largest study to measure RIF-R risk on recurrence following RS-TB treatment.

Baseline isoniazid resistance is likely the first step in RIF-R emergence, and, indeed, was associated with increased RIF-R risk in our study[6]. Furthermore, previous studies found that PLHIV have increased risk of recurrence and DR-TB emergence, which we also observed, although among PLHIV, recurrences may be more likely to be re-infections[7, 8]. The increased RIF-R risk on recurrence after 3 and 4 months of treatment versus <1 month of treatment, and the reduction in risk back to around 1 after 6 months of treatment is consistent with a mathematical model of drug resistance emergence on treatment[9].

Study limitations include a lack of detailed information on treatment doses taken. Some patients who were LTFU may have also had inconsistent adherence before LTFU, which may have contributed to the subsequent RIF-R; therefore, their duration of treatment prior to LTFU may have been overestimated. Furthermore, our data lacked TB genotyping, precluding definitive differentiation between reinfection with new RIF-R strains and relapse with acquired RIF-R. However, the timing of recurrences after treatment end is clustered around shorter times, suggesting that for most people, their RIF-R was more likely relapse than reinfection (Figure 1). Finally, there was missing information, including missing RIF-R testing results in 12% of participants at recurrence, mostly due to lack of bacteriological confirmation of TB at that follow-up point. The recurrence risk for people with a successful outcome of 1.8% is consistent with, or lower than, that in previous studies and demonstrates our data robustness[10].

Our study demonstrates a small but worrisome risk of acquisition of RIF-R with satisfactory completion of RS-TB treatment and a substantial RIF-R risk occurring after incomplete treatment of RS-TB. All patients should be supported consistently to ensure treatment completion through the continuation phase, and those unable to complete must be diligently searched for to identify recurrences and possible drug resistance, and help them reinstate effective treatment. A better understanding of risk factors for emergence of resistance after <6 months will be important to guide use of shortened RS-TB regimens[11].

These results are timely given the recent impact of the COVID-19 pandemic, during which TB notifications dropped globally by 18% in 2020, treatment was widely disrupted and LTFU increased[12]. The war in Ukraine further compounds these issues and increases displacement of Ukrainians[13], and these results are relevant globally at this time when funding cuts to international aid may threaten the availability of TB treatment. Thus, treatment completion failure may increase, and with it, recurrence with rifampin-resistant TB.