Effectiveness of Isoniazid Preventive Therapy to Reduce Tuberculosis Incidence in the Context of Antiretroviral Therapy

To the Editors: INTRODUCTION Although curable, tuberculosis (TB) is the leading cause of AIDS-related morbidity and mortality globally.1 South Africa has the world's highest burden of HIV (7 million people)2 and reports 23% of the global TB-HIV burden.3Mycobacterium tuberculosis can be latent or active in the body. Immune suppression from HIV is the strongest determinant for developing active TB4,5; however, latent TB infection can be successfully treated using isoniazid preventive therapy (IPT). Following a change in guidelines to encourage its use,5,6 IPT was offered at no charge through South Africa's public health system from 2011. Overlapping with IPT implementation, antiretroviral therapy (ART) was expanded at community level, with people living with HIV (PLWH) offered no-cost ART at progressively higher levels of immune function.7–10 ART also protects against TB due to rapidly improving immune function.11 We assessed associations between six-month IPT and ART alone and in combination to prevent TB disease across 2 years compared with no intervention in a high-burden District of KwaZulu-Natal province, South Africa. We hypothesized that IPT alone and ART alone would provide similar levels of protection against TB and greater protection in combination. METHODS Setting An estimated 70% of TB diagnoses in KwaZulu-Natal are HIV-related.12 UMgungundlovu District has a population of 1 million people, largely of Zulu descent, with high unemployment.13 HIV prevalence between ages 15–49 years is estimated at 44% and 28% among females and males, respectively.14 The district-level TB notification rate is 894 per 100,000 population.12 Multidrug resistance is estimated at 7% for all TB cases and 15% among PLWH.15 Study Design We conducted a retrospective cohort study among PLWH in a large periurban community health center catchment area of uMgungundlovu District between January 1, 2011, and December 31, 2014. The Research Ethics Boards at the Universities of Calgary and KwaZulu-Natal provided ethical approvals. The KwaZulu-Natal Department of Health and the uMgungundlovu Health District Office provided administrative approvals. We collected IPT data from paper registers and probabilistically linked them to electronic catchment-level ART and District-level TB outcome data. We defined IPT users as PLWH collecting at least 80% of pills within 9 months16 between January 1, 2011, and December 31, 2012. ART users were defined as PLWH whose first ART prescription was collected before or during study time and whose last prescription was collected after censoring or outcome of TB, with no treatment stoppage. IPT + ART users met criteria for both IPT and ART use, although may not have received both interventions at the same time. People receiving no intervention were those with no evidence of IPT use before or during the study period and who started ART after TB diagnosis or after the study period. We estimated IPT completion dates to be 30 days after the last recorded collection date. We defined study time among the exposed as the duration between IPT start and completion dates plus 540 days (18 months) of follow-up. Study time for comparison groups was considered the average start and end date of IPT users. Favoring validity over precision, we considered only bacteriologically confirmed TB events. As we used outcome data for the entire District, we assumed no loss to follow-up. Data Analysis We used Stata/IC 13.1 for analyses. At-risk time was calculated as the total disease-free time each person contributed to the cohort, with censoring occurring at the end of the study period. We calculated cumulative incidence (risk), incidence rates, and incidence rate ratios to evaluate differences between interventions. Rate differences were also calculated due to zero cell counts in contingency tables. We report rate differences as case reduction (1-rate difference) for ease of interpretation. We used stratified analysis to evaluate potential confounders and effect modifiers considering: sex,17 age at cohort start date (<35 versus ≥35 years),16,18 CD4 cell count at ART baseline (≥250 versus <250 cells/μL),19 tenofovir-containing ART regimen (yes/no), and ART duration at IPT completion (≥1 versus <1 year). RESULTS We followed 12,412 PLWH for a total of 21,473 person-years, among whom 441 completed IPT. With no intervention, TB incidence rates appeared to increase with age and decrease with rising immune function. See Table 1 for TB outcomes by intervention. Six TB-related deaths were recorded among IPT and ART-naive individuals, zero among IPT users, and 27 among ART users.TABLE 1.: Crude and Stratified TB Outcomes by InterventionIPT Alone Despite an even distribution of TB by sex at baseline (females = 48.41%), proportionally many more females initiated (71.61%) and completed IPT (73.92%) compared with men (P < 0.01 for both comparisons). Among PLWH who started IPT, 62.25% of men and 65.59% of women completed IPT (P = 0.79). The association between IPT and TB appeared to be modified by sex (P = 0.05). Zero TB events occurred among females (one-sided 97.5%CIIRR = 0.00 to 0.22), while TB incidence rates decreased by 54.33% among males [incidence rate ratio (IRR) = 0.46, 95% CI: 0.15 to 1.10]. A similar trend was seen in sensitivity testing which included 237 PLWH who took less than 5 months of IPT (IRRfemale = 0.04, 95% CI: 0.001 to 0.23; IRRmale = 0.37, 95% CI: 0.13 to 0.84; P = 0.02). ART Alone The preventive association with ART appeared to vary by age group and CD4 cell count (P < 0.001). Among those receiving ART alone, those at highest risk of TB activation (ie, older age and low CD4 cell count) received the greatest protection. Among this group, ART alone appeared to perform similarly to IPT alone in females and potentially better than IPT alone in males. IPT alone among males and females appeared to perform better than ART among people with high CD4 cell counts (≥250 cells/μL), and people older than 35 years appeared to receive greater protective benefit from ART alone compared with those younger than 35 years (IRR≥35 = 0.27, 95% CI: 0.11 to 0.63 versus IRR<35 = 0.61, 95% CI: 0.34 to 1.07, P = 0.11). ART and IPT As zero cases of TB developed among IPT users who were also on ART, our data suggest that IPT increased the protective effect of ART. A similar trend was seen among those receiving <5 months of IPT (IRR = 0.18, 95% CI: 0.005 to 1.02). DISCUSSION Our analysis suggests that among clinic-going PLWH in uMgungundlovu District, six-month IPT works to decrease the incidence of active TB for 2 years with and without the use of ART, consistent with other observational studies. Golub et al reported adjusted hazard ratios (aHRs) of 0.47 and 0.36 among ART-only users, 0.32 and 0.73 among users of IPT alone, and 0.20 and 0.15 among IPT + ART users who were tuberculin skin test positive in Gauteng province, South Africa and Rio de Janeiro, respectively.20,21 Yirdaw et al22 reported an aHR of 0.36 for IPT alone (95% CI: 0.19 to 0.16), 0.32 for ART alone (95% CI: 0.24 to 0.43), and aHR of 0.18 when IPT was added to ART (95% CI: 0.10 to 0.42). Khawcharoenporn et al23 reported a four-year incidence rate ratio of 0.45 among IPT + ART users (P = 0.13). The strong effect of IPT alone demonstrated in our study, at least among females, may relate to a high HIV prevalence driving TB incidence in KwaZulu-Natal.12 Given that the one-sided confidence interval among females on IPT alone suggested a minimum 77.78% reduction in TB incidence rate compared with no intervention (97.5% CI: 0.00 to 0.22), we would expect that IPT + ART would perform even better, despite a limited sample size. Sensitivity testing suggested a similar trend by sex, despite limited follow-up time among males. Potential sex differences may relate to higher immune function; among individuals whose CD4 cell counts were known (n = 73), 39.12% of female IPT users had low CD4 cell counts compared with 59.09% of males (P < 0.001). In absolute numbers, males in our study demonstrated a similar reduction in the number of TB events compared with females, albeit with a wider margin of error (CRMale = 3176, 95% CI: 584 to 5769 versus CRFemale = 3045, 95% CI: 2478 to 3611). LIMITATIONS Although IPT guidelines recommend that all symptom-free PLWH be offered IPT, our study was limited in that the number of people initiated was much lower than ART. This may be explained by the differential push for ART expansion, an established program when IPT was introduced. Nonetheless, it is unclear whether people offered and initiating IPT differed systematically in a way that would differentially effect IPT outcomes. Given the small number of IPT users, we were unable to investigate a potential dose–response relationship for IPT. Moreover, in an attempt to maximize validity, we excluded TB events that were not bacteriologically confirmed. As bacteriologic confirmation can be limited among people who are highly immune compromised, findings among those with low CD4 counts may be less accurate. Findings are also limited to clinic-going patients for whom other health, mobility, and economic conditions did not prevent monthly IPT collection. We report on broader implications of IPT acceptability elsewhere.24 Although some studies suggest that only people with a positive tuberculin skin test benefit from IPT,25,26 guidelines no longer require the test in settings where its use would inhibit IPT initiation.5 We are therefore unable to comment on IPT given tuberculin skin test response. Furthermore, although previous TB is an independent risk factor for recurrent TB,27 comprehensive TB data were unavailable before 2010. Among ART + IPT users, the interventions may not have overlapped. CONCLUSIONS In a periurban setting with a high incidence of TB and TB-HIV in the general population, six-month IPT offered through the public system appeared to improve two-year TB incidence rates alone and in combination with ART among PLWH.