Recurrent tuberculosis (TB) accounts for 5% of TB cases worldwide , but more than 25% in high-incidence settings or where control programmes are weak [2,3]. HIV infection greatly increases the rate of recurrence, and in areas with generalized HIV epidemics this is due to an increase in the rate of disease due to reinfection rather than an increase in relapse [4,5]. In high TB incidence populations, the risk of reinfection disease is much higher than the incidence of first episode TB in the general population, for both HIV-positive and HIV-negative individuals [6,7], so patients who have had an episode of TB are an important group to target for prevention efforts.
Antiretroviral therapy (ART) reduces the incidence of new TB in HIV-positive individuals, although the risk remains elevated compared with HIV-negative individuals . Much less is known about the effect of ART on recurrent TB. It would be expected to reduce recurrence rates, as recurrence rates are known to be lower in HIV-positive individuals with higher CD4 cell counts . However, to date only one study has measured the direct effect of ART on recurrence rates: in a clinic-based study of an HIV-positive cohort in Brazil, ART halved the recurrence rate . Here patients who started ART during follow-up, and had not yet experienced recurrence, were analysed together with patients receiving ART before end of treatment, although the effect may be qualitatively different. Additional relevant data are essential to inform control programmes in sub-Saharan Africa,
Molecular epidemiology studies have shown that the rate of recurrence due to relapse, that is, treatment failure, is highest during the first year after the end of treatment, whereas recurrences due to reinfection with a new strain occur at a more constant rate [4,5,11]. Among HIV-positive patients in Africa, recurrences after 1 year are almost exclusively due to reinfection [4,5]. The effect of ART on these different mechanisms and timings of recurrence is unknown.
Here we present the first study to estimate rates of TB recurrence by ART status in an African cohort. We assessed TB recurrence rates by HIV and ART status, stratifying by time since the end of treatment to assess the effects of HIV and ART on the different disease mechanisms.
Patients and setting
The study was conducted as part of the Karonga Prevention Study, in rural northern Malawi. Karonga District has a population of about 300 000. The incidence of new smear-positive TB in adults is about 75 per 100 000, and nearly two-thirds of patients with TB are HIV-infected . TB cases are identified through enhanced passive surveillance: project staff are based at peripheral clinics and the district hospital to take sputum specimens from individuals with chronic cough.
All diagnosed TB cases in the district were treated following Malawi National TB Programme guidelines. For new smear-positive patients before 1997, this was 2 months streptomycin (S), isoniazid (H), rifampicin (R) and pyrazinamide (Z) (2SHRZ) and 6 months thiacetazone (T) and isonazid (6TH). Ethambutol (E) replaced thiacetazone from 1997, and streptomycin from 2001, resulting in the regimen 2EHRZ/6EH. New smear-negative patients were treated with 1SHT/11TH until 1997, then 1SHE/11EH until 2007. Since 2007, a 6-month regimen with rifampicin throughout has been used for all new patients (2HRZE/4HR), with an extended course for retreatment cases.
Three sputum specimens were taken for diagnostic purposes, and for smear-positive patients, two review specimens were sought at the completion of the intensive treatment phase and at treatment completion. Smears were read (auramine staining with fluorescence microscopy with confirmation of positives by Ziehl–Neelson staining and light microscopy) and cultures set up on Lowenstein–Jensen solid media in the project laboratory in Malawi. Species confirmation and drug resistance testing (using the ratio method) was done in the London Health Protection Agency reference laboratory, UK.
All TB patients were seen by project staff and asked for their consent to participate in the study, including interview and counselling and testing for HIV. Information on previous TB treatment was acquired through interviews at the start of the index episode, and from the project database, which holds linked information on all TB episodes recorded in Karonga District since 1986.
The first ART clinic in Karonga District opened in 2005 , and to date there are seven clinics providing ART to more than 5000 individuals. HIV-positive individuals with WHO stage 3 or 4 disease (which includes TB), or with CD4 cell counts less than 250 cells/μl (recently increased to less than 350 cells/μl) are eligible for ART. Information on when the patient first started ART was collected through interviews and linked ART clinic records.
Eligibility and follow-up
Outcomes at the end of TB treatment were ascertained for all adult patients (≥15 years) diagnosed with laboratory-confirmed TB [defined as smear-positive (excluding single scanty smears), or with positive culture or histology] between January 1996 and February 2010. They were eligible to enter the cohort if they had completed their TB treatment and were usually resident in Karonga District at the time of their TB episode. Outcomes were initially determined retrospectively. Many were already known, through our other linked studies in the district [13,14]. Home visits were conducted in 2005 for those with uncertain vital status, and relatives or other informants were asked about dates of death or migration if the TB patient had died or left the district. From August 2007, all patients were followed up annually. During the home visits, patients were asked about their HIV and ART status and screened for chronic cough. Three sputum samples were collected from those with coughs of 3 or more weeks’ duration.
Patient follow-up was started at their recorded date of completion of treatment. The two main endpoints were a recurrent TB episode (defined as starting treatment for a laboratory-confirmed TB episode) and death. If no study endpoint occurred, follow-up was censored at 30 June 2011 or the date of leaving the district if they were reported to have left permanently. In the recurrence analysis, follow-up was censored if the participant started a course of TB treatment, regardless of laboratory confirmation (as they were no longer at risk of TB).
The main analysis was restricted to new TB patients, that is, those who had not been treated for TB prior to the index episode.
Three groups were compared based on their HIV/ART status at the time of completion of TB treatment: HIV-negative; HIV-positive and not receiving ART; and receiving ART. For the main analysis, follow-up was censored if a participant moved to another group (became HIV-positive, or started ART), as rate of recurrence after completion of treatment varies with time since completion of treatment as well as with duration of HIV infection or ART. A time-updated analysis of recurrence could therefore bias the results.
Analyses were adjusted according to the period in which the index episode occurred. The first period, 1996–2000, was before the TB regimens changed. The second period was from regimen change (2001) to mid-2005, when ART became available in the district. The third period was from mid-2005 onwards.
Survival and hazard curves were generated using the Kaplan–Meier method. Incidence estimates and rate ratios were calculated using Poisson regression. Interaction was examined using the likelihood ratio test. We stratified the analysis by time since the end of treatment (looking separately at recurrences within 1 year and later) to explore the effect of HIV and ART on the different mechanisms of recurrence, as molecular epidemiological evidence has shown that relapses and reinfections tend to occur at different time points [4,15].
In order to compare our results with those from Brazil , we re-ran the analysis treating ART as a time-dependent covariate. In this analysis, HIV-positive individuals who started ART after the end of their TB episode were moved to the ‘on ART’ group instead of being censored at the point when they started their ART.
Analyses were repeated using any recurrent TB episode, that is, regardless of laboratory confirmation, and for those who had experienced a prior episode of TB at the time of their index episode.
The study was approved by the ethics committee of the London School of Hygiene and Tropical Medicine and the National Health Sciences Research Committee of Malawi.
There were 1596 eligible TB episodes during the period of the study. After excluding those occurring in patients who had had previous treatment for TB (n = 201), with missing HIV/ART status (n = 260) and two episodes for which no information beyond the end of treatment could be found, the main analysis included 1133 index patients (Table 1). Out of these, 105 participants were lost to follow-up due to leaving the district after a median of 1.5 years (range 0.03–13) of follow-up.
There were 309 deaths during follow-up: 58 in HIV-negative; 230 in HIV-positive ART-negative; and 21 in the HIV-positive ART-positive group, giving mortality rates [95% confidence interval (CI)] of 2.0 (1.6–2.6), 16.1 (14.2–18.4) and 4.9 (3.2–7.5) per 100 person-years, respectively. Figure 1a shows the overall survival by HIV/ART status at the start of the follow-up. HIV-positive ART-positive individuals had better survival than HIV-positive ART-negative individuals, but mortality rates remained elevated compared with HIV-negative index cases (rate ratio 2.42, 95% CI 1.47–3.99). This increased to 4.44 (2.27–8.66) after adjusting for period, age, sex and time since TB.
Rates of laboratory-confirmed recurrence are shown in Table 1. Overall there were 103 recurrences, with more than half occurring in the first year (Figure 1b). Follow-up for 43 participants was censored due to the occurrence of nonlaboratory-confirmed TB.
Recurrences were less common in HIV-negative individuals, and those with drug-sensitive strains isolated, and slightly less common among women and younger patients. After adjusting for age, sex, period and TB type, HIV infection increased the rate of recurrence 2.7-fold (rate ratio 2.69, 95% CI 1.69–4.26) compared with HIV-negative individuals. Those who were HIV-positive ART-positive had intermediate recurrence rates. Compared with HIV-positive ART-negative, the adjusted rate ratio for recurrence was 0.56 (95% CI 0.25–1.27).
The rate of recurrence was much higher in the first year, with some evidence of interaction between time to recurrence and HIV/ART status (likelihood ratio test for interaction P = 0.2) and drug resistance (no P-value calculated as no recurrence in drug resistant cases after 1 year, see Table 2). Table 2 shows the recurrence rates separately for the first year and beyond the first year. In the first year, HIV-positive ART-negative patients had only a slightly higher recurrence rate than HIV-negative patients, and HIV-positive ART-positive patients had similar rates to HIV-negative patients. Among HIV-positive individuals, there was no evidence that ART changed the recurrence rate within the first year (adjusted rate ratio 0.74, 95% CI 0.27–2.06).
Beyond the first year, recurrence rates were much higher in HIV-positive ART-negative patients than HIV-negative patients (adjusted rate ratio 4.20, 95% CI 2.16–8.15). The recurrence rate after 1 year also appeared higher in those HIV-positive ART-positive than in HIV-negative patients (adjusted rate ratio 1.85, 95% CI 0.49–7.00). There was also weak evidence that those HIV-positive ART-positive individuals had a reduced risk of recurrence compared with HIV-positive ART-negative (adjusted rate ratio 0.43, 95% CI 0.11–1.73). Drug resistance increased the rate of recurrence in the first year, but not later.
The direction and size of these associations were similar when we expanded the analyses to include recurrences without laboratory confirmation. Out of 146 recurrences, 74 occurred during the first year after treatment. In the three HIV/ART groups, the rate per 100 person-years of all recurrences was 1.8 (95% CI 1.4–2.4) in the HIV-negative, 6.1 (95% CI 4.9–7.6) in the HIV-positive ART-negative and 4.1 (95% CI 2.4–6.6) in those HIV-positive ART-positive before end of treatment. Among HIV-positive, the impact of ART appeared smaller during the first year (adjusted rate ratio 0.82, 95% CI 0.33–2.05) compared with later (adjusted rate ratio 0.43, 95% CI 0.13–1.45).
For comparison with other studies, the analysis was also run including ART as a time-dependent covariate for all HIV-positive patients (Table 3). Patients who were HIV-negative during their index episode and were first known to be HIV-positive when they started on ART only contribute to person time on ART. This gave an adjusted rate ratio for recurrence of 0.61 (95% CI 0.30–1.23) comparing those on ART to those not on ART . The effect was similar in the first year after treatment and subsequently (P-value likelihood ratio test for interaction = 0.9).
Among the 201 individuals excluded from the main analysis who had already had previous TB treatment at the time of their index episode, there were 17 laboratory-confirmed episodes in 539 person-years of follow-up [rate per 100 person-years 3.2 (95% CI 2.0–5.1)]. Re-recurrence rates were high among HIV-negative patients in this group [10 in 306 person-years, rate = 3.3 (95% CI 1.8–6.1)] and similar to those of HIV-positive ART-negative patients [six in 152 person-years, rate 3.9 (95% CI 1.8–8.8), rate ratio 1.21 (95% CI 0.44–3.33)]. The numbers of HIV-positive ART-positive individual were small (one per 81 person-years).
We estimated rates of TB recurrence by ART status in a cohort in Africa with active follow-up, and showed that ART use appeared to halve the rate of recurrence in HIV-positive patients, similar to results found in the only comparable study, from Brazil .
Recurrence rates of HIV-positive ART-positive patients were also compared with those of HIV-negative patients. CIs were wide, but it appeared that the rates remained elevated. This would be expected, as it is known that ART does not reduce the incidence of first episode TB to the level of those who are HIV-uninfected. INH prophylaxis may be warranted for all HIV-positive patients who completed TB treatment, irrespective of ART status.
Drug resistance increased the recurrence rate, as expected. All recurrences in patients with drug-resistant strains had presented within 12 months after treatment, consistent with most relapses occurring early . Among those who had already had a previous episode of TB, recurrence rates were high in HIV-negative patients, with no further increase in the HIV-positive patients. High rate of recurrence – similar to that in HIV positive patients – in HIV-negative patients who have already had more than one TB episode was also found in a South African study .
In this study CD4 cell counts were not available to explore the effects of ART in detail. Another limitation is that most HIV-positive ART-negative patients were seen before 2005, and most HIV-positive patients after 2005 were put on ART during treatment. Although we adjusted for period, this may not fully account for differences between the groups, especially as other changes, addition of rifampicin to continuation phase (which should reduce relapse) and initiation of active follow-up (which might increase identified recurrence) occurred around this time point. It was not possible to compare recurrence rates in those HIV-positive ART-positive at the start of TB treatment with those who were only started on ART during TB treatment, as numbers were too low. The cohort was highly complete, for only 0.2% (two of 1135) no information on health status could be obtained. Loss to follow-up due to leaving of the district was acceptable (<10% over 13 years), and unlikely to have biased results.
The main effect of ART was on later recurrences, consistent with previous findings that HIV infection predominantly increases recurrences due to reinfection rather than relapse [16,17]. This suggests that the effect of ART will increase over time, and that it will vary between settings, depending on the risk of Mycobacterium tuberculosis infection. The benefit of getting and keeping TB patients on ART will be largest where TB incidence is highest.
The authors would like to thank the Malawi National Tuberculosis Programme for their collaboration on this work. The contributions of the field, laboratory and data entry staff at the Karonga Prevention Study are deeply appreciated, as is the cooperation of the participants in the cohort.
J.R.G., A.C.C. and R.M.G.J.H. designed the study; R.M.G.J.H. managed and analysed the data. A.C.C., R.M.G.J.H., S.M. and T.M. coordinated the study and N.J.M., L.M. and M.M. contributed to the study logistics. J.R.G. and R.M.G.J.H. wrote the first draft of the report; all authors contributed to writing the report and have seen and approved the final draft.
The study was funded by the Wellcome Trust.
Conflicts of interest
N.F. holds a grant from Novartis for research on group B streptococcus. Novartis had no involvement in the design, analysis or reporting of this study. The other authors declare they have no conflicts of interest.
1. WHO. Global tuberculosis control: WHO report 2011
. Geneva: WHO; 2011.
2. Chaisson RE, Churchyard GJ. Recurrent tuberculosis: relapse, reinfection, and HIV
. J Infect Dis
3. Wood R, Lawn SD, Caldwell J, Kaplan R, Middelkoop K, Bekker LG. Burden of new and recurrent tuberculosis in a major South African city stratified by age and HIV-status
. PLoS One
4. Crampin AC, Mwaungulu JN, Mwaungulu FD, Mwafulirwa DT, Munthali K, Floyd S, et al. Recurrent TB: relapse or reinfection? The effect of HIV in a general population cohort in Malawi
5. Sonnenberg P, Murray J, Glynn JR, Shearer S, Kambashi B, Godfrey-Faussett P. HIV-1 and recurrence, relapse, and reinfection of tuberculosis after cure: a cohort study in South African mineworkers
6. Verver S, Warren RM, Beyers N, Richardson M, van der Spuy GD, Borgdorff MW, et al. Rate of reinfection tuberculosis after successful treatment is higher than rate of new tuberculosis
. Am J Respir Crit Care Med
7. Glynn JR, Murray J, Bester A, Nelson G, Shearer S, Sonnenberg P. High rates of recurrence in HIV-infected and HIV-uninfected patients with tuberculosis
. J Infect Dis
8. Lawn SD, Harries AD, Williams BG, Chaisson RE, Losina E, De Cock KM, et al. Antiretroviral therapy and the control of HIV-associated tuberculosis. Will ART do it?
. Int J Tuberc Lung Dis
9. Charalambous S, Grant AD, Moloi V, Warren R, Day JH, van Helden P, et al. Contribution of reinfection to recurrent tuberculosis in South African gold miners
. Int J Tuberc Lung Dis
10. Golub JE, Durovni B, King BS, Cavalacante SC, Pacheco AG, Moulton LH, et al. Recurrent tuberculosis in HIV-infected patients in Rio de Janeiro, Brazil
11. Swaminathan S, Narendran G, Venkatesan P, Iliayas S, Santhanakrishnan R, Menon PA, et al. Efficacy of a 6-month versus 9-month intermittent treatment regimen in HIV-infected patients with tuberculosis: a randomized clinical trial
. Am J Respir Crit Care Med
12. Crampin AC, Glynn JR, Fine PE. What has Karonga taught us? Tuberculosis studied over three decades
. Int J Tuberc Lung Dis
13. Jahn A, Floyd S, Crampin AC, Mwaungulu F, Mvula H, Munthali F, et al. Population-level effect of HIV an adult mortality and early evidence of reversal after introduction of antiretroviral therapy in Malawi
14. Crampin AC, Glynn JR, Fine PE. What has Karonga taught us? Tuberculosis studied over three decades
. Int J Tuberc Lung Dis
15. Narayanan S, Swaminathan S, Supply P, Shanmugam S, Narendran G, Hari L, et al. Impact of HIV infection on the recurrence of tuberculosis in South India
. J Infect Dis
16. Houben RMGJ, Crampin AC, Ndlovu R, Sonnenberg P, Godfrey-Faussett P, Haas WH, et al. HIV increases the risk of TB due to recent infection more than that due to reactivation of latent infection
. Int J Tuberc Lung Dis
17. Houben RMGJ, Crampin AC, Ndlovu R, Sonnenberg P, Godfrey-Faussett P, Haas WH, et al. HIV increases the risk of TB due to recent infection, but the effect varies with age.
In: Research in progress
. London, UK: Royal Society of Tropical Medicine and Hygiene; 2008.