Unrecognised tuberculosis at antiretroviral therapy initiation is associated with lower CD4+ T cell recovery

Authors

  • Sabine M. Hermans,

    1.  Infectious Diseases Institute, Makerere University College of Health Sciences, Kampala, Uganda
    2.  Department of Global Health, Academic Medical Center, Amsterdam Institute for Global Health and Development, University of Amsterdam, Amsterdam, The Netherlands
    3.  Department of Internal Medicine and Infectious Diseases, University Medical Center Utrecht, Utrecht, The Netherlands
    Search for more papers by this author
  • Frank van Leth,

    1.  Department of Global Health, Academic Medical Center, Amsterdam Institute for Global Health and Development, University of Amsterdam, Amsterdam, The Netherlands
    Search for more papers by this author
  • Agnes N. Kiragga,

    1.  Infectious Diseases Institute, Makerere University College of Health Sciences, Kampala, Uganda
    Search for more papers by this author
  • Andy I. M. Hoepelman,

    1.  Department of Internal Medicine and Infectious Diseases, University Medical Center Utrecht, Utrecht, The Netherlands
    Search for more papers by this author
  • Joep M. A. Lange,

    1.  Department of Global Health, Academic Medical Center, Amsterdam Institute for Global Health and Development, University of Amsterdam, Amsterdam, The Netherlands
    Search for more papers by this author
  • Yukari C. Manabe

    1.  Infectious Diseases Institute, Makerere University College of Health Sciences, Kampala, Uganda
    2.  Division of Infectious Diseases, Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, MD, USA
    Search for more papers by this author

Corresponding Author Sabine M. Hermans, Infectious Diseases Institute, Makerere University College of Health Sciences, PO Box 22418, Kampala, Uganda. Tel.: +256 414 307000; Fax: +256 414 307290; E-mail: shermans@idi.co.ug or s.hermans@aighd.org

Abstract

Objectives  To investigate whether an unrecognised diagnosis of tuberculosis (TB) at the start of antiretroviral therapy (ART) influences subsequent CD4+ T cell (CD4) count recovery in an urban HIV clinic in Uganda.

Methods  In a retrospective cohort study, a multivariable polynomial mixed effects model was used to estimate CD4 recovery in the first 96 weeks of ART in two groups of patients: prevalent TB (started ART while on TB treatment), unrecognised TB (developed TB within 6 months after start ART).

Results  Included were 511 patients with a median baseline CD4 count of 57 cells/mm3 (interquartile range: 22–130), of whom 368 (72%) had prevalent TB and 143 (28%) had unrecognised TB. Compared with prevalent TB, unrecognised TB was associated with lower CD4 count recovery at 96 weeks: −22.3 cells/mm3 (95% confidence interval −43.2 to −1.5, = 0.036). These estimates were adjusted for gender, age, baseline CD4 count and the use of zidovudine-based regimen.

Conclusions  Unrecognised TB at the time of ART initiation resulted in impaired CD4 recovery compared with TB treated before ART initiation. More vigilant screening with more sensitive and rapid TB diagnostics prior to ART initiation is needed to decrease the risk of ART-associated TB and sub-optimal immune reconstitution.

Abstract

Objectifs:  Déterminer si un diagnostic de la tuberculose méconnue au début de l’ART influence le taux de récupération ultérieure des lymphocytes T CD4+ (CD4) dans une clinique urbaine du VIH en Ouganda.

Méthodes:  Dans une étude de cohorte rétrospective, un modèle polynôme multivariéà effets mixtes a été utilisé pour estimer la récupération des CD4 au cours des 96 premières semaines de l’ART dans deux groupes de patients: TB prévalente (qui ont commencé l’ART durant le traitement de la TB), TB méconnue (qui ont développé la TB 6 mois après le début de l’ART).

Résultats:  511 patients ont été inclus avec un taux médian de CD4 de base de 57 cellules/mm3 (intervalle interquartile 22, 130), dont 368 (72.0%) avaient une TB prévalente et 143 (28.0%) avaient une TB méconnue. Par rapport à la TB prévalente, la TB méconnue était associée à une récupération inférieure des CD4 à 96 semaines: -22,3 cellules/mm3 (intervalle de confiance 95% de -43.2 à -1.5; P = 0.036). Ces estimations ont été ajustées pour le sexe, l’âge, le taux de CD4 de base et l’utilisation d’un traitement à base d’AZT.

Conclusions:  La TB méconnue au moment de l’initiation de l’ART mène à une récupération altérée des CD4 par rapport à la TB traitée avant le début de l’ART. Un dépistage plus vigilant avec un diagnostic plus sensible et rapide de la TB avant le début de l’ART est nécessaire pour diminuer le risque de TB associée à l’ART et une reconstitution immunitaire sous-optimale.

Abstract

Objetivos:  Investigar si un diagnóstico no reconocido de TB al comienzo del TAR tiene posteriormente influencia sobre la recuperación del conteo de linfocitos T CD4+ en una clínica urbana para VIH en Uganda.

Métodos:  En un estudio de cohortes retrospectivo, se utilizó un modelo polinomial multivariable de efectos mixtos para calcular la recuperación de CD4 durante las primeras 96 semanas de TAR en dos grupos de pacientes: TB prevalente (comenzó TAR mientras recibía tratamiento para TB), TB no reconocida (desarrolló TB dentro de los 6 meses siguientes al comienzo del TAR).

Resultados:  Se incluyeron 511 pacientes con una mediana en el conteo de base de CD4 de 57 células/mm3 (rango intercuartil 22, 130), de los cuales 368 (72.0%) tenían una TB prevalente y 143 (28.0%) tenían una TB no reconocida. Comparada con la TB prevalente, la TB no reconocida estaba asociada con una menor recuperación en el conteo de CD4 a las 96 semanas: -22.3 células/mm3 (IC 95% -43.2 a -1.5, = 0.036). Estas estimaciones se ajustaron para género, edad, conteo de base de CD4 y el uso de un régimen basado en AZT.

Conclusiones:  La TB no reconocida en el momento de iniciar el TAR resultó en un deterioro en la recuperación de CD4 comparada con la TB tratada antes de la iniciación del TAR. Se requiere una mayor vigilancia antes de la iniciación del TAR, con pruebas diagnósticas para TB más sensibles y rápidas, para disminuir el riesgo de la TB asociada al TAR y una reconstitución inmune subóptima.

Introduction

Tuberculosis (TB) is a major cause of morbidity and mortality in HIV-infected patients (Corbett et al. 2003), in particular in sub-Saharan Africa which accounts for 82% of the burden of co-infected patients (World Health Organization, 2011). Antiretroviral therapy (ART) reduces the incidence of HIV-associated TB by 67%, but TB rates never reach those of the non-HIV-infected population (Lawn et al. 2005, 2010b).

Early after ART initiation reactivation of TB-specific immune responses causes an increase in TB incidence, a phenomenon called ‘unmasking’ of TB (Lawn et al. 2008b; Manabe et al. 2009). This high incidence of TB early after the start of ART contributes to the early mortality among ART initiators in resource-limited settings (Lawn et al. 2008a; Castelnuovo et al. 2009). The majority of these cases have subclinical TB at the start of ART, undiagnosed due to lack of sensitive diagnostics in resource-limited settings. In sub-Saharan Africa, estimates of subclinical TB in ART-naïve patients range from 19% in patients about to start ART to 29% in ambulatory patients with high CD4+ T cell (CD4) counts (Mtei et al. 2005; Bassett et al. 2010). It is clear that there is an urgent need for rapid, affordable and sensitive diagnostics to improve TB screening before ART initiation.

Active TB disease is known to decrease CD4 counts, which usually normalise within a month of starting treatment in HIV-negative patients (Jones et al. 1997). In HIV-infected, ART-naïve TB patients, treatment of TB also leads to CD4 count increases; this effect, however, was not seen in patients with TB at higher CD4 counts (Martin et al. 1995; Kalou et al. 2005; Mahan et al. 2010). It is unknown whether the timing of TB treatment as a result of correct or missed diagnosis impacts CD4 recovery on ART, and whether it could be associated with differences in HIV treatment outcomes. In this analysis, we sought to investigate whether an unrecognised diagnosis of TB at the start of ART influences subsequent CD4 recovery.

Methods

Setting

The Infectious Diseases Institute (IDI) at the Makerere University College of Health Sciences in Kampala, Uganda, houses the Adult Infectious Diseases Clinic (AIDC). Since its inception in 2002, this large urban HIV clinic has provided free outpatient care to over 25 000 registered patients. By mid 2011, more than 10 000 patients were in active follow-up, over 3500 of whom had been initiated on ART. Treatment, including ART and TB treatment, followed the national guidelines of the Ugandan Ministry of Health, and has been described in detail previously (Hermans et al. 2010, 2012a,b). In brief, all patients received daily co-trimoxazole prophylaxis irrespective of their CD4 count, and ART was initiated in patients with WHO stage IV disease or a CD4 count <250 cells/mm3. Screening for TB took place prior to ART initiation. Available investigations for TB included sputum microscopy using the Ziehl-Neelsen stain, chest radiology, abdominal ultrasonography and fine-needle aspiration of lymph nodes for acid fast bacilli microscopy and cytology. No routine mycobacterial culture facilities were available. Diagnosis of TB was made on the basis of these investigations, but very often on the presentation of symptoms only. Patients diagnosed with active TB were treated with a standard 8-month regimen comprising an 8-week, four drug (isoniazid, rifampicin, ethambutol and pyrazinamide) intensive phase and a subsequent 6-month, two drug (isoniazid, ethambutol) continuation phase. Isoniazid preventive therapy was not prescribed.

CD4+ T cell counts were performed every 6 months from the date of registration in the clinic by FACS calibur (Becton Dickinson, Franklin Lakes, NJ, USA) in the College of American Pathologists-accredited Makerere University–Johns Hopkins University core laboratory. Viral load monitoring was only available for patients suspected of virological failure on clinical and immunological grounds.

Data collection

Scheduled clinic appointments took place every 4 weeks and data on clinical parameters, ART, adherence, WHO stage, toxicities and opportunistic infections were routinely collected into a database. Laboratory data was directly downloaded electronically. The pharmacy database on TB drug prescriptions was used to validate the TB diagnosis of all patients in this database, as described previously (Hermans et al. 2010). The TB status of cases identified in only one of the two databases was ascertained by chart review, confirming a TB diagnosis by a clinician and the starting date of ART and TB treatment. If the charts of such cases were unavailable for review, they were excluded from the analysis. Cases treated for TB whose chart was available but for whom a date of TB diagnosis could not be ascertained were also excluded. A team of trained nurses and medical officers performed data verification using the patient’s medical notes and by auditing the database as part of regular clinic monitoring and evaluation.

Ethical approval

The collection and use of this clinical data was approved by the Institutional Review Boards of IDI, Makerere University College of Health Sciences and the Uganda National Council for Science and Technology. De-identified data for our study were extracted from this database and analysed.

Study design, selection criteria and study outcome

In this retrospective cohort study, we identified two mutually exclusive groups of patients: those with a TB diagnosis at start of ART (within 8 months prior to ART initiation, from now on referred to as prevalent cases) and those with a missed TB diagnosis at start of ART (referred to as unrecognised cases). A TB diagnosis was assumed unrecognised at screening if it occurred within 6 months after start of ART. All patients initiated on first-line ART at IDI from January 2003 to January 2009 with at least 96 weeks of follow-up after ART initiation were eligible for inclusion. Patients who had initiated ART elsewhere were excluded, as were patients initiated on second-line ART. The primary study outcome was defined as the CD4 change between the start of ART and 96 weeks of ART use.

Statistical methods

The baseline CD4 counts were defined as those that were recorded closest to the ART starting date, with a maximum of 6 months pre- and 15 days post-ART initiation. Follow-up CD4 measurements were grouped per 4 weeks after ART initiation. In the case of multiple CD4 counts per grouped 4 weeks, the count closest to the date defining the number of weeks after ART initiation was used. For each group, we ascertained and plotted CD4 change in the first 96 weeks of ART; this change being defined as the difference between the baseline CD4 count and each follow-up CD4 measurement. We fitted a multivariable mixed effects model with random slopes and random intercepts to incorporate repeated measurements of CD4 count. The covariance structure was kept independent. The model included fractional polynomial functions of time to account for non-linearity of the data. Visual inspection of plots of the crude and the modelled data determined the model fit and the number of polynomial variables included. From the final model, we estimated CD4 changes over time. Covariates included baseline CD4 count (per 50 cells/mm3 increase), sex, age (per increment of 10 years), zidovudine (AZT) use and protease inhibitor use. All were explored using univariable regression analysis and were retained in the multivariable model based on forward selection; hypothesised risk factors were included and confounders were retained based on a univariable < 0.2. All statistical tests were two-sided at an α-value of 0.05 and were conducted using stata version SE 11.1 (StataCorp, College Station, TX, USA).

Results

Patient characteristics

A total of 7139 HIV-infected adults were initiated on ART. Of those, 3549 (50%) completed 2 years of follow-up; 963 (14%) were lost to follow-up, 429 (6%) died and 731 (10%) were transferred out. The remaining 1467 (20%) had not completed 2 years on ART at the time of censoring of our data, but were still in active follow-up. Of the 3549 patients who had a minimum of 96 weeks of follow-up after ART initiation, 511 were diagnosed with TB between 8 months pre-ART and 6 months post-ART initiation and were included in our analysis. The majority of the patients were women (53%) with a mean age of 36.9 years [standard deviation (SD), 8.0] and a median baseline CD4 count of 57 cells/mm3 [interquartile range (IQR): 22–130]. Data on baseline CD4 counts were not available for nine patients (1.8%). However, these patients still contributed information, as the mixed effects model estimated the missing values from the available data. Three hundred and sixty-eight (72%) patients started ART during TB treatment and 143 (28%) developed TB in the first 6 months of ART. The median time between TB treatment initiation and ART initiation was 96 days (IQR: 150–68) for the prevalent TB cases, and 50 days between ART initiation and TB diagnosis (IQR: 21–99) for the unrecognised TB cases. Eighty-one per cent (n = 298) of prevalent TB cases were on continuation phase of TB treatment and 19% (n = 70) on intensive phase at the time of ART initiation. The patient characteristics at ART initiation are shown in Table 1. Because of their TB diagnosis, prevalent TB cases were categorised as WHO stage III or IV more often than unrecognised cases. Because of the interaction between nevirapine and rifampicin, more prevalent TB cases were initiated on efavirenz) than unrecognised cases.

Table 1. Patient characteristics at ART initiation
CategorySubcategoryPrevalent TB cases [N = 368 (72%)]Unrecognised TB cases [N = 143 (28%)]
  1. TB, tuberculosis; ART, antiretroviral therapy; d4T, stavudine; 3TC, lamivudine; NVP, nevirapine; AZT, zidovudine; EFV, efavirenz.

  2. *Data on baseline CD4 count were not available for nine patients (five with prevalent TB and four with incident TB); CD4 counts were closest recorded values to the baseline start date, maximum 6 months pre-ART and 15 days post-ART initiation. Data on WHO stage were not available for 33 patients.

  3. †Other first-line triple ART regimens.

Sex [female, n (%)] 195 (53)75 (53)
Age [years, mean (SD)] 37 (7.6)37 (8.8)
WHO stage,*n (%)I and II8 (2)43 (32)
III127 (37)53 (40)
IV209 (61)38 (28)
Baseline CD4 count* [cells/mm3, median (IQR)] 51 (22–131)67 (25–124)
Baseline CD4 count* [cells/mm3, n (%)]≥20020 (6)11 (8)
50–199168 (46)70 (50)
<50175 (48)58 (42)
ART regimen, n (%)d4T + 3TC + NVP215 (58)102 (71)
AZT + 3TC + EFV145 (39)36 (25)
Other 1st line†8 (3)5 (4)

CD4+ T cell recovery after antiretroviral therapy

The best model fit of the mean change of CD4 count from start to 96 weeks of ART was achieved with a model containing five polynomial levels of the time variable (Figure 1). In this model, CD4 counts were estimated to increase by 18.0 cells/mm3 [95% confidence interval (CI): 6.3 to 29.7, = 0.003] every 4 weeks (Table 2). Unrecognised TB was associated with less CD4 recovery during the period of follow-up compared with prevalent TB [−22.3 cells/mm3 (95% CI: −43.2 to −1.5), = 0.036], as was male sex [−32.1 cells/mm3 (95% CI: −50.6 to −13.6), = 0.001]. Baseline CD4 count, older age and AZT use were not significantly associated with lower CD4 change. No patients were switched to a protease inhibitor (PI) based regimen during the 96 weeks of follow-up. A sensitivity analysis including suspected TB cases whose charts were unavailable for review and therefore excluded from the primary analysis (32 with prevalent TB, 18 with unrecognised TB) did not significantly change any of the estimates (results not shown). The fitted CD4 change trajectories by TB group are shown in Figure 2.

Figure 1.

 Model fit by TB group. Mean CD4 counts over time in both groups: actual (solid line) and fitted (dashed line) CD4 counts calculated from the final multivariable mixed effects model using five polynomial levels to evaluate the fit of the model. ART, antiretroviral therapy; TB, tuberculosis.

Table 2. Multivariable mixed effects model of CD4 change in the first 96 weeks of ART
 SubcategoryCD4 count increase (95% CI) P-value
  1. ART, antiretroviral treatment; TB, tuberculosis; AZT, zidovudine.

  2. *Compared to prevalent TB, female sex and non-AZT-based regimens, respectively.

TB statusUnrecognised TB*−22.33 (−43.19 to −1.46)0.036
Time (per 4 weeks’ increase) 17.98 (6.27 to 29.69)0.003
Baseline CD4 count (cells/mm3, per 50 cells increase) −2.79 (−9.51 to 3.92)0.415
SexMale*−32.07 (−50.55 to −13.58)0.001
Age (years, per increment of 10 years) −0.37 (−1.57 to 0.82)0.537
RegimenAZT-based*−6.44 (−21.09 to 8.20)0.388
Figure 2.

 Trajectories of estimated mean CD4 counts in the first 2 years of ART by TB status. Trajectories of the predicted mean CD4 counts in the first 96 weeks of ART in the two groups included in the analysis. ART, antiretroviral treatment; TB, tuberculosis.

Discussion

In our study of ART initiators in Uganda, patients with unrecognised TB at the time of ART initiation had a significantly lower CD4 recovery during 96 weeks of ART compared with patients who were being treated for TB at ART initiation. Risk factors for suboptimal immune recovery and for TB seem to be overlapping: low CD4 nadir, low body mass index and male sex (Lawn et al. 2006a; Tuboi et al. 2007). Even though the potential for TB to diminish CD4 recovery has been discussed widely, published data are conflicting. We and others have previously reported the association of ART-associated TB with lower CD4 recovery (Lawn et al. 2005; Hermans et al. 2010). TB diagnosed before ART initiation was associated with impaired immune reconstitution in South Africa (Bassett et al. 2012), but not in Asia (Patel et al. 2004; Breen et al. 2006; Zhou et al. 2010). When analysing patients with prevalent and ART-associated TB together, a South African cohort reported no difference in immunological outcomes (Lawn et al. 2006b), whereas an Italian cohort did (Cingolani et al. 2011).

Impaired immune recovery on ART despite a suppressed viral load has been associated with higher levels of T cell activation leading to apoptosis (Hunt et al. 2003; Negredo et al. 2010; Nakanjako et al. 2011). It is intriguing to consider the role of inflammatory programmed cell death, pyroptosis, which can be triggered by bacterial infections (Bergsbaken et al. 2009; Doitsh et al. 2010). There are some reports that immune reconstitution on ART might more often be impaired in resource-limited settings (Egger et al. 2009; Geng et al. 2010). Higher levels of T cell activation in Africans have been attributed to frequent infections by various endemic pathogens, including Mycobacterium tuberculosis, and might limit the capacity for CD4 recovery (Lawn 2004; Eggena et al. 2005).

Explanations for impaired immune reconstitution due to ART-associated TB also include higher rates of virological failure or poor adherence to ART. Viral loads could be higher in patients with TB because of their associated lower CD4 count or because of the effect of active TB disease on HIV replication (Toossi 2003). Our data did not allow analysis of viral suppression rates, but an Italian cohort found a longer time to achieve viral suppression on ART in patients who had TB within 6 months before or after ART initiation (Cingolani et al. 2011).

It is unclear whether the conflicting evidence on the effect of active TB on CD4 recovery is attributable to a differing categorisation of patients (by timing of TB diagnosis before or after ART initiation), or to new infections rather than unrecognised TB disease. Further research is needed to elucidate this, preferably in a large collection of ART cohorts with routine viral load monitoring in a high TB prevalent setting.

Our results imply that CD4 recovery over the ensuing 2 years is compromised if TB is not recognised before ART initiation. Our findings underscore the importance of TB screening and diagnosis in HIV-infected patients. The lack of affordable and adequately sensitive diagnostic tools for TB leads to potential underdiagnosis, especially in an HIV-infected population in which TB is often smear-negative (Mtei et al. 2005; Bassett et al. 2010). Intensive pre-ART TB screening strategies have been shown to reduce incident TB (Lawn et al. 2010a). We hypothesie that the immunological insult from active TB might be more profound after ART, resulting in a long-lasting impairment of CD4 recovery capacity. From the curve in Figure 2, it seems that this effect occurs early, during the initial phase of active TB. Initiating TB treatment before ART initiation might lead to a rapid decrease in T cell activation (and therefore T cell apoptosis) due to lowering of mycobacterial load prior to immune reconstitution. The duration of the effect on CD4 recovery is unclear, it could represent a temporary delay in immune recovery followed by the same recovery rate as non-TB patients as suggested by the similarity in slope in Figure 2.

The clinical relevance of the decreased CD4 recovery in patients with ART-associated TB is unclear. Patients with unrecognised TB after ART may remain longer at lower CD4 counts compared with patients who initiated TB treatment before ART initiation. This may put them at higher risk for other opportunistic infections and death (Battegay et al. 2006).

Our study had several limitations. The study design limited us to an analysis of a survivor’s population, possibly leading to a biased estimation of CD4 recovery. Patients not included due to loss of follow-up or death may have had lower CD4 recovery. Therefore, this analysis represents a conservative estimate of the association of unrecognised TB and impaired immune reconstitution in our clinic.

Our analysis was based on the assumption that all TB cases diagnosed within 6 months after ART initiation were unrecognised prevalent TB cases at baseline. Others have also suggested that the higher incidence of TB early after the start of ART is due to unmasking of subclinical, unrecognised TB at ART initiation, especially in patients with CD4 counts <200 cells/mm3 (Lawn et al. 2009). However, it is possible that some of these cases may have been true incident TB cases, rather than missed cases at baseline. Furthermore, this analysis was based on routinely collected data with known issues of missing data and outcome ascertainment. However, the mixed effects model used was capable of incorporating information from patients with missing data on outcome and covariates. The data presented were collected at only one clinic, potentially affecting the generalizability of our results.

Because of data collection limitations, we were not able to assess the proportion of TB-immune reconstitution inflammatory syndrome in our patient population. However, we believe that this would only be a minority of the cases presented, and that almost none of these would have been prescribed corticosteroids. Finally, we were not able to exclude higher rates of poor adherence or virological failure as an alternative explanation for slower immune recovery. Insensitive measurements of adherence in our clinic (self-report, pill counts) and infrequent viral load testing made it difficult to exclude this possibility. However, marked non-adherence to ART or virological failure would have severely impacted survival, making it unlikely that these patients could have been included in this survivors’ population.

In conclusion, unrecognised, untreated TB at the time of ART initiation resulted in impaired CD4 recovery over 96 weeks compared with TB treated before ART initiation. As CD4 levels remain suppressed longer, the risk for other opportunistic infections may be increased, although prospective study is needed. In addition to sustained efforts to initiate ART at a higher CD4 count, more vigilant screening with more sensitive and rapid TB diagnostics prior to ART initiation is needed to decrease the risk of ART-associated TB and suboptimal immune reconstitution.

Acknowledgements

We thank the Infectious Diseases Institute (IDI) TB working group for their support and endless motivation to improve TB care in our clinic, and we gratefully acknowledge the IDI data management and validation team for their efforts in improving the quality of our data. This work was carried out as part of the Infectious Diseases Network for Treatment and Research in Africa (INTERACT) programme, financially supported by the Netherlands Organization for Scientific Research–WOTRO Science for Global Development: NACCAP and the European Union. Data were presented previously at the 18th Conference on Retrovirology and Opportunistic Infections (27 February to 2 March 2011, Boston, Massachusetts, United States) and published as abstract in the conference book.

Ancillary