Between January 2018 and February 2022, 166 365 patients newly entered or reenrolled in HIV care. Of these, 34 610 (20.8%) were excluded, including 6363 (3.8%) with insufficient data to calculate IPT duration or person-time. There were 131 755 (79.2%) individuals included in the analysis (Figure 1). Overall, 40 733 (30.9%) completed IPT and 91 022 (69.1%) had no IPT. At the end of the inclusion period, 9089 (5.5%) patients were known to have died. Of the 48 573 patients who started isoniazid, 83.9% completed at least 146 days of treatment. Among those in the No IPT group, 845 (0.9%) started isoniazid but received <28 days of therapy and were categorized as having received no IPT.

Baseline and follow-up characteristics of the 131 755 patients are shown in Table 1. The median age was 38.7 years (interquartile range [IQR], 33.3–44.5 years) for the No IPT group and 40.4 years (IQR, 34.7–47.0 years) for the Complete IPT group; males accounted for about 54% of both groups. The proportion of PWH with sustained viral suppression was 73.7% (62 309) in the No IPT group and 64.7% (25 005) in the Complete IPT group.

Participants who did not receive IPT were followed for 271 039 PY, and those who completed IPT were followed for 69 604 PY (Table 1). The median follow-up time was 3.74 (IQR, 2.35–3.74) PY per person in the No IPT group and 1.53 (IQR, 0.84–2.53) PY per person in the Complete IPT group. The unadjusted TB incidence rate was 1.91 cases per 100 PY for the No IPT group (95% confidence interval [CI], 1.77–2.07) and 1.01 cases per 100 PY for the Complete IPT group (95% CI, .81–1.25) (Table 2), a rate difference of 0.9 cases per 100 PY. The adjusted TB incidence rate ratio (aIRR) was 1.99 (95% CI, 1.62–2.46) for the No IPT group when compared to the Complete IPT group (Table 3). The unadjusted mortality rate was 2.62 deaths per 100 PY for the No IPT group (95% CI, 2.44–2.81) and 2.79 deaths per 100 PY for the Complete IPT group (95% CI, 2.44–3.19). When adjusted for the same variables, the adjusted mortality rate ratio was 1.11 for the No IPT group (95% CI, .97–1.27) when compared to the group who completed IPT. Among persons who developed TB disease, MDR-TB occurred in 29.1% and 30.7% of TB cases in the Complete and No IPT groups, respectively (χ² = 0.78, P = .3780).

Patients with a history of imprisonment and self-reported alcohol misuse had 1.95 (95% CI, 1.50–2.53) and 2.07 times (95% CI, 1.56–2.75), respectively, the rate of developing active TB disease when compared to patients without these risk factors. PWH with lower CD4 counts had higher incidences of active TB disease when compared to PWH with CD4 counts ≥500 cells/μL. Patients who were not virally suppressed had 3.19 times the rate of developing active TB disease compared with patients who were virally suppressed (95% CI, 2.72–3.74) (Table 3).

We tested for interaction of TPT status by year and found it was not significant (P = .2266). To explore the impact of differential follow-up time across IPT groups, we limited the follow-up time to 18 months per person. The unadjusted TB incidence rate was 2.44 cases per 100 PY for the No IPT group (95% CI, 2.23–2.67) and 1.15 cases per 100 PY for the Complete IPT group (95% CI, .93–1.43), with an aIRR of 2.22 (95% CI, 1.77–2.77) when adjusted for age, sex, time since HIV diagnosis, CD4 count, viral suppression, history of incarceration, and alcohol misuse.

Among virally suppressed PWH, completion of IPT was associated with a difference of 0.3 cases per 100 PY in TB incidence compared to those who did not complete IPT (Figure 2). For those who were virally suppressed, adjusted TB incidence was 1.1 cases per 100 PY for the No IPT group and 0.8 cases per 100 PY for the Complete IPT group (aIRR, 1.42 [95% CI, 1.25–1.62]). Among those who were not virally suppressed, adjusted TB incidence was 3.8 cases per 100 PY for the No IPT group and 1.5 cases per 100 PY for the Complete IPT group (aIRR, 2.56 [95% CI, 2.26–2.89]; rate difference, 2.3 per 100 PY). TB incidence was lower among virally suppressed than among virally unsuppressed PWH in both IPT groups. The unadjusted MDR-TB incidence rate was 0.59 cases per 100 PY for the No IPT group and 0.29 cases per 100 PY for the Complete IPT group (IRR, 2.13 [95% CI, 1.43–3.18]; Table 2). In a supplementary analysis including patients who received a partial course of TPT, the adjusted MDR-TB IRR was 2.01 for the No IPT group when compared to the group that received any IPT.

In Ukraine, a public health intervention that included completion of IPT was associated with a nearly 2-fold lower incidence of TB when compared to PWH who did not receive IPT. The effect was maintained among PWH who were virally suppressed: TB incidence was 27% lower among virally suppressed PWH who completed IPT. The proportion of MDR-TB among incident TB cases was similar between groups and comparable to the prevalence of MDR-TB (32.6%) in the general population in Ukraine. These findings suggest IPT is an effective intervention to prevent TB among PWH in high MDR-TB settings, regardless of viral suppression status.

While there are have been clinical trials of TPT for high-risk contacts of persons with MDR-TB [23–25], this cohort was not limited to contacts of an index MDR-TB case. In our study of a population of PWH, MDR-TB incidence also decreased with receipt of IPT despite national survey data that estimate 37.3% (95% CI: 34.6%–40.1%) isoniazid resistance among new TB patients and 65.8% (95% CI: 60.3%–71.7%) among previously treated patients [13]. There are several possible explanations for this apparent paradox and researchers have raised the possibility that the mechanism by which isoniazid reduces TB disease risk among those with latent infection may differ from its mechanism in active TB disease [26–29]. Isoniazid is known to have potent early bactericidal activity, and because mycobacterial populations in latent TB may be heteroresistant [30], isoniazid could have an effect on isoniazid-sensitive drug-resistant TB strains. Furthermore, at the early stage of infection, when MDR-TB is restricted to the granuloma and bacterial load is low, it is possible that isoniazid is more effective than when MDR-TB is released outside the granuloma and bacterial load is higher.

We found that the proportion of MDR-TB among PWH who developed TB disease was comparable among those who received IPT and those who did not. If IPT were only effective in PWH with drug-sensitive latent TB infections, the proportion of MDR-TB would have increased among PWH who received IPT and developed TB disease. Furthermore, several studies have previously demonstrated that receipt of IPT does not increase the risk of developing drug-resistant TB [31]. Overall, although one-third of new TB cases in Ukraine have MDR-TB, most TB is drug-sensitive and most PWH are likely to be exposed to a drug-sensitive index case, which may explain the lower overall TB incidence we observed in our analysis. However, for household contacts of a known MDR-TB index case, recent trials [32, 33] have supported the role of levofloxacin TPT in reduction of TB incidence. Because TB transmission is primarily driven by community exposures [34], the proportion of MDR-TB among our cohort of PWH was reflective of the prevalence of MDR-TB in the general population in Ukraine.

To our knowledge, this is the first study to investigate the effects of longitudinal viral suppression on IPT effectiveness. Studies have previously shown the effect of IPT to be independent and additive to that of ART [2]. Here, we used patient-level viral load results as an assessment of HIV control over time, an important risk factor for TB disease. As expected, TB incidence was lower among virally suppressed PWH, but even among PWH who maintained viral suppression, IPT was associated with lower TB incidence. These data suggest that there is an added benefit of treating latent TB, beyond the immune recovery that usually accompanies viral suppression.

This analysis has some limitations. First, PWH were not randomized to receive or not receive IPT, and receipt of IPT may have been correlated with other variables that are also correlated with TB disease. Additionally, we could not control for unmeasurable confounders that may have affected how vigorously newly diagnosed PWH were screened for TB disease in comparison to routinely monitored returning patients. We attempted to address the potential for bias with careful exclusion of groups of patients who did not have an opportunity to complete IPT and be followed over time, and through covariate adjustment in regression models. Further limitations include a large amount of missing data on smoking status, nutritional status, and homelessness—all known risk factors for TB that could not be adjusted for in our regression models. Last, our data source did not include phenotypic data on resistance patterns, so we were unable to determine the prevalence of isoniazid resistance in our cohort.

While we cannot infer causality in an observational analysis, our results are consistently supportive of the benefit of IPT. We conclude that countries with overlapping HIV and MDR-TB epidemics should continue scaling up IPT to all PWH as a population-level intervention. IPT should remain a critical component of care for all PWH who live in high-TB-burden settings, even among those who are on ART and virally suppressed.