Incidence of tuberculosis and survival after its diagnosis in patients infected with HIV-1 and HIV-2

Background: In sub-Saharan Africa, tuberculosis (TB) is the most frequently diagnosed opportunistic infection and cause of death among HIV-infected patients. HIV-2 has been associated with less immune suppression, slower disease progression and longer survival.

Objective: To examine whether the incidence of TB and survival after TB are associated with CD4 cell count rather than HIV type.

Methods: Clinical and immunological data were retrospectively evaluated among an open clinic-based cohort of HIV-1- and HIV-2-infected patients to determine incidence of TB (first diagnosis > 28 days after HIV diagnosis) and subsequent mortality. Patients were grouped by CD4 cell count into those with < 200, 200-500 and > 500 × 10⁶ cells/l.

Results: Incident TB was diagnosed among 159 of 2012 patients, with 4973 person-years of observation time. In 105/159 (66.0%), the diagnosis was confirmed by direct microscopy or culture. Incidence of TB was highest in the group with < 200 × 10⁶ cells/l (9.1/100 and 8.8/100 person-years in HIV-1 and HIV-2, respectively). Adjusted for CD4 cell count, there was no significant difference in incidence or mortality following TB between HIV-1- and HIV-2-infected patients. Mortality rate was higher inthose with incident TB and HIV infection, most markedly in the group with the highest CD4 cell count (hazard ratio, 10.0; 95% confidence interval, 5.1-19.7).

Conclusion: Adjusted for CD4 cell count, incidence of TB was similar among HIV-1- and HIV-2-infected patients. Mortality rates after TB diagnosis were similar in both groups and high compared with those without TB.

Introduction

In much of West Africa, two HIV types are prevalent: HIV-1 and HIV-2. While routes of transmission and identified risk factors for both infections are similar, risk of transmission, progression to disease, rate of CD4 cell decline and excess mortality are lower among HIV-2-infected patients [¹]. Differences in pathogenicity between HIV-1 and HIV-2 can be partly explained by a lower plasma viral load in HIV-2-infected patients [²]. Clinical studies are needed to assess differences in the impact of immunosuppression following HIV-1 and HIV-2 infection.

The incidence of tuberculosis (TB) is much higher in HIV-infected patients than in HIV-uninfected patients [^3-5]. TB is the most commonly diagnosed opportunistic infection and the most frequent direct cause of death among HIV-infected patients in sub-Saharan Africa. In a postmortem study conducted in Côte d'Ivoire, TB was found to be the primary cause of death in 32% of patients with AIDS, and the disease was present in an additional 22% [⁶]. The high incidence of TB among HIV-infected subjects could be a major factor in the resurgence of TB among the general population of sub-Saharan Africa. The number of annual new cases is now estimated at 8 million, with nearly 2 million deaths [⁵]. A third of new TB cases in sub-Saharan Africa are estimated to be attributable to HIV infection [⁷].

The higher incidence of and mortality after TB diagnosis among HIV-infected subjects compared with HIV-uninfected patients has been related to their CD4 cell count [^8-10]. The immunosuppression caused by HIV leads to an increased susceptibility to (re)infection with Mycobacterium tuberculosis and to reactivation or rapid progression from infection to active TB [¹¹], and TB has been included among the AIDS-defining opportunistic infections [¹²]. Conversely, development of TB in HIV-infected patients might increase the rate of HIV disease progression and mortality, in particular for those with CD4 cell counts > 200 × 10⁶ cells/l [¹³].

Very little is known about the incidence of TB and of subsequent mortality in HIV-2- compared with HIV-1-infected subjects. The prevalence of both HIV-1 and HIV-2 is higher among patients with TB than among the general population [^14,15]. HIV infection does not progress over a prolonged period in a considerable proportion of HIV-2-infected subjects and these patients maintain high CD4 cell counts. However, once the CD4 cell count declines to < 200 × 10⁶ cells/l, clinical progression and survival are similar to those in HIV-1-infected patients [¹⁶].

The present study measured the TB incidence rate and the mortality rate after a TB diagnosis in a clinical cohort of HIV-1- and HIV-2-infected patients with long-term follow-up and related incidence and mortality to CD4 cell count to see if the differences between the two HIV types could be explained by differences in CD4 cell count.

Methods

Subjects

HIV patients diagnosed at or referred to the Medical Research Council clinic in Fajara, The Gambia were invited to join a prospective seroprevalent cohort study. The cohort consisted of a mixture of healthy and symptomatic patients, as described previously [¹⁶]. At recruitment, details of previous TB episodes were recorded. A Mantoux skin test, chest radiography and sputum microbiology were carried out at clinical presentation to diagnose or exclude TB. A clinical assessment was carried out at each visit, including a Karnofsky score [¹⁷] and measurement of body mass index [weight (kg)/height (m)²]. After 1 July 1999, patients with a CD4 cell count < 500 × 10⁶ cells/l were offered prophylaxis with co-trimoxazole against opportunistic infections [¹⁸]. The difficulties in excluding active TB in HIV-infected patients and the risks associated with monotherapy so far have prevented the introduction of Isoniazid prophylaxis in The Gambia.

Patients in the cohort were asked to return to the clinic once every 3 months, or sooner when indicated. Patients who did not return to the clinic were visited in their homes by a field worker to ascertain survival status and to encourage them to attend the clinic. Patients with no follow-up at all after their initial visit were not considered part of the cohort. Patients who moved out of the study area or refused further contact with the study team were classified as lost to follow-up. Patients on antiretroviral therapy, patients with dual HIV-1 and HIV-2 infection (or indeterminate HIV diagnosis) and patients under the age of 15 years were excluded from the analysis.

In order to establish incident cases of TB, clinical records were retrieved for all patients aged 15 years and above with a provisional diagnosis of TB indicated in the database between 1 August 1992 and 31 December 2001. Clinical, microbiological and radiographic data were evaluated to assign a final diagnosis. Confirmed pulmonary TB was defined by the presence of acid-fast bacilli in direct smear or culture from sputum or lavage. Smear-negative pulmonary TB was defined as the presence of strongly suggestive clinical symptoms and radiographic signs consistent with TB. Confirmed extrapulmonary TB was defined by the demonstration of acid-fast bacilli in a biopsy or aspirate of a lymph node, or from any other normally sterile site by smear or culture. Probable extrapulmonary TB was defined by strongly suggestive clinical features. All patients newly diagnosed with TB were treated with multiple drugs according to Gambia Government TB treatment protocols, provided at a Gambia Government TB treatment centre. Patients already diagnosed with TB at the time of HIV diagnosis, as well as those diagnosed with TB in the first 28 days following HIV diagnosis, were considered prevalent cases. Incident TB was thus defined as a new diagnosis of TB more than 28 days after HIV was diagnosed. Second or further episodes of TB during follow-up were excluded from the analysis.

Ethical permission for the study was obtained from the Gambia Government/Medical Research Council Ethics Committee. All patients gave informed consent prior to enrolment.

Laboratory analyses

Serum samples were screened for HIV using Wellcozyme HIV 1+2 enzyme-linked immunosorbent assays (ELISA) (Murex Diagnostics, Dartford, UK) until August 1996, and then by the ICEHIV-1.O.2 ELISA (Murex Diagnostics). All reactive samples were re-tested by type-specific ELISA. Samples that were clearly reactive in only one type-specific ELISA were assigned a serological diagnosis accordingly. Samples reactive in both type-specific ELISA were further tested by a synthetic peptide-based line immunoassay, Pepti-Lav 1-2 (Sanofi Diagnostics Pasteur, Marne la Coquette, France). The appearance of a clear (++) or very clear (+++) band was interpreted as evidence of infection with the relevant HIV type. Samples with clear or very clear lines for both virus types were considered as dually infected. A second confirmatory serum sample, usually taken 2-8 weeks later, was tested similarly. Patients with at least one positive screen result but inconclusive monospecific ELISA and Pepti-Lav results, and patients with two sample test results that were incompatible, were classified as indeterminate [¹⁶].

The CD4 cell count was estimated on site by flow cytometry using the FACScalibur with the use of simultest or Multitest staining reagents on the whole blood samples using fresh cells (Becton-Dickinson, Oxford, UK). The first CD4 cell count value obtained within 3 months of the first positive HIV test was considered as the baseline CD4 cell count. A CD4 cell count obtained within 3 months before or after diagnosis of incident TB was considered as the value at the time of TB diagnosis. Patients were considered to fall into three strata based on CD4 cell count: < 200, 200-500 and > 500 × 10⁶ cells/l.

Statistical analysis

Data were analysed using Stata v7 (Stata Corp., College Station, Texas, USA). Person-years of observation were calculated from the day of HIV diagnosis; follow-up data were censored at the date of death, the date last known to be alive or 31 October 2003, whichever came first.

Comparison of means was performed using Student's t-test if the data were approximately normally distributed; otherwise distributions were compared using the Wilcoxon rank sum test. For comparison of proportions, the chi-squared test or, where appropriate, Fisher's exact test was used.

Mortality hazard ratios (HR) were assessed using Cox regression analysis, with TB diagnosis as a time-dependent variable. To satisfy the proportional hazards assumption, the data were stratified by baseline CD4 cell count. There was significant interaction between HIV type and TB disease in those with a low and a high CD4 cell count. The level for statistical significance was set at 5%.

Results

Subjects

Among 2012 patients diagnosed with HIV-1 or HIV-2 infection, records of 457 patients with a provisional diagnosis of TB were retrieved from the database. For 368 patients (80.5%), a diagnosis of TB was established. Table 1 summarizes the characteristics of the TB diagnosis among HIV-1- and HIV-2-infected patients. TB was diagnosed among 140 patients (38.0%) at the time of HIV diagnosis. Another 69 (18.8%) patients were diagnosed with TB within 28 days of the date of HIV diagnosis. The remaining 159 (43.2%) patients had incident TB; baseline CD4 cell counts were available for 156 (98.1%) of these. TB was diagnosed a median of 5.4 [interquartile range (IQR), 1.9-23.1] months after HIV diagnosis. (Table 2).

Table 1

Table 2

The majority of those with incident TB (138/159, 86.8%) had pulmonary TB. For 105/159 (66.0%) of these, the diagnosis was microbiologically confirmed.

Nearly half of the patients who developed TB had a baseline CD4 cell count < 200 × 10⁶ cells/l: 49.1% of those with HIV-1 infection and 45.1% of those with HIV-2 infection (P = 0.8). There were no differences in sex, median body mass index or Karnofsky score between the two groups, but HIV-2-infected patients were older (median age 39 versus 32 years; P = 0.01) (Table 2).

Tuberculosis incidence rate

Between 1 August 1992 and 31 December 2001, 2012 patients aged 15 years and above with a diagnosis of HIV-1 or HIV-2 were recruited into the cohort. They contributed 4973 person-years of observations. After excluding those with prevalent TB and those with less than 29 days of follow-up, 1644 patients [1134 (69.0%) HIV-1 infected, 510 (31.0%) HIV-2 infected] with 4681 person-years of observation remained to assess the incidence of TB. CD4 cell counts were available for 1561/1644 (95.0%), contributing 4497 person-years observation. At the time of TB diagnosis, HIV-1-infected patients were more likely to have been started on co-trimoxazole prophylaxis than HIV-2-infected patients, but there were no significant differences in median CD4 cell count, body mass index or Karnofsky score.

The incidence rate of TB was 3.8/100 person-years [95% confidence interval (CI), 3.1-4.5] among HIV-1-infected patients and 2.8/100 person-years (95% CI, 2.1-3.7) among HIV-2-infected patients. Incidence was highest in the group with a CD4 cell count < 200 × 10⁶ cells/l and lowest in those with > 500 × 10⁶ cells/l. There were no differences in the incidence rates in any of the CD4 cell count strata between HIV-1 and HIV-2 infection [< 200 (P = 0.5), 200-500 (P = 0.3) and > 500 × 10⁶ cells/l (P = 0.8)]. A similar pattern was observed when the analysis was restricted to those with pulmonary TB only, or to those with microbiologically confirmed TB (both pulmonary and extrapulmonary TB) only (Table 3). Significant differences in TB incidence rates between HIV-1- and HIV-2-infected patients were not observed in any of the CD4 cell count strata (all P > 0.2).

Table 3

The crude incidence rate of TB among HIV-2-infected patients was lower than that among HIV-1-infected patients [incidence rate ratio (IRR), 0.75; 95% CI, 0.54-1.00; P = 0.05]. Incidence of TB increased with age, and adjustment for age resulted in a further decreased IRR of HIV-2- versus HIV-1-infected patients (IRR, 0.69; 95% CI, 0.49-0.90; P = 0.03). After adjusting for baseline CD4 cell counts, this difference was no longer significant (IRR, 0.83; 95% CI, 0.59-1.17; P = 0.3); adjustment for all confounding variables (baseline CD4 cell count, age, sex, co-trimoxazole prophylaxis, baseline Karnofsky score), gave an overall IRR of 0.92 (95% CI, 0.65-1.30; P = 0.6).

Mortality comparing HIV-1- and HIV-2-infected patients with incident tuberculosis

Of the HIV-infected patients with incident TB, 119 (74.8%) died during follow-up and 20 (12.6%) were lost to follow-up: there was no significant difference between HIV-1 and HIV-2 infection in terms of this loss (11 and 9, respectively; P = 0.2). The follow-up time after TB diagnosis prior to the loss to follow-up was also similar among HIV-1- and HIV-2-infected patients (5 and 7 months, respectively; P = 0.7).

The overall mortality rate of patients following a diagnosis with TB was 43.5/100 person-years (95% CI, 36.4-52.1); the overall median survival time was 8.9 months (IQR, 2.1-25.0), and similar for HIV-1- and HIV-2-infected patients. The mortality HR following TB diagnosis was similar in HIV-1- and HIV-2-infected patients (crude HR, 1.0; 95% CI, 0.7-1.5), and also after adjusting for age, co-trimoxazole prophylaxis, Karnofsky score and CD4 cell count at time of TB (adjusted HR, 1.0; 95% CI, 0.7-1.6).

There were no differences between HIV-1- and HIV-2-infected patients with TB in the median CD4 cell count collected within 3 months of the date of death [50 × 10⁶ cells/l (IQR, 25-120) and 50 × 10⁶ cells/l (IQR, 10-190), respectively; P = 0.7], in the median last Karnofsky score [45 (IQR, 30-60) and 60 (IQR, 40-70), respectively; P = 0.1] or in the median last body mass index [15.1 (IQR, 13.3-17.4) and 16.4 (IQR, 14.8-18.7, respectively; P = 0.3].

Excess mortality among HIV-infected patients following incident tuberculosis

Of the 2012 HIV-1- or HIV-2-infected participants recruited into the cohort, 1136 (56.5%) died during the study period and 364 (18.1%) were lost to follow-up. Excluding the 209 patients with prevalent TB, and an additional 159 patients with less than 29 days of follow-up, patients who were diagnosed with TB had a higher mortality rate from the time of TB diagnosis. Figure 1 shows Kaplan-Meier survival probabilities of HIV-1- and HIV-2-infected patients with and without TB, stratified by CD4 cell count. As TB diagnosis was analysed as a time-dependent variable, survival time of all patients who developed TB was split into survival time with and without TB. The log-rank test for equality of survivor functions was highly significant (P < 0.001 in each CD4 strata).

Fig. 1

The HR values comparing mortality for each HIV type for those with and without TB were calculated using Cox regression and are summarized in Table 4. Time to TB infection was included as a discrete time-varying covariate. In the group with CD4 cell count > 500 × 10⁶ cells/l, there was a mild departure from the proportional hazards model, but the quoted constant HR values are meaningful summaries. In the other groups, the proportional hazards assumptions were fully satisfied. Age at recruitment, co-trimoxazole usage, HIV type and TB infection were significant effects for all CD4 cell count groups; baseline CD4 cell count was significant for the lower two CD4 cell count groups. There was significant interaction on mortality between HIV type and TB infection in those with a low and a high CD4 cell count, as reflected in Table 4 and Fig. 1. The HR values for comparing those with TB infection against those without were adjusted for age at recruitment and co-trimoxazole usage for all CD4 cell count groups and also for baseline CD4 cell count in the lower two groups. Among HIV-1-infected patients, excess mortality was most marked in those with a CD4 cell count > 500 × 10⁶ cells/l (HR, 10.0); for HIV-2-infected patients the HR values in each CD4 cell count group were similar.

Table 4

Discussion

The incidence of TB was similar among HIV-1- and HIV-2-infected patients in a clinic-based cohort in The Gambia, adjusted for CD4 cell counts. Once TB is diagnosed, survival for HIV-1- and HIV-2-infected patients was equally poor. For both HIV-1- and HIV-2-infected patients, and in each of the three CD4 cell count groups, the mortality HR of patients with TB was higher than that for patients without TB.

We are aware of one other cohort study that followed HIV-1- and HIV-2-infected patients to assess TB incidence and subsequent mortality [¹⁹]. In this study in Guinea-Bissau, TB incidence was higher in HIV-1- than in HIV-2-infected patients, and lowest among the HIV-uninfected population. Mortality following TB was increased among HIV-1-infected patients but similar in HIV-2- and HIV-uninfected patients. No adjustment was made for CD4 cell count in that study, which was also community based; therefore, one can assume that the study population in general was healthier, with higher CD4 cell counts, than our cohort.

A few other smaller studies have compared TB mortality rates in HIV-1- and HIV-2-infected patients. Studies in Guinea-Bissau [²⁰] and Burkina Faso [²¹] found similar mortality rates for the two types of HIV infection, while a study in Côte d'Ivoire reported an increased mortality rate in HIV-1- versus HIV-2-infected patients with TB [²²]. Neither of these studies adjusted for CD4 cell count. Another study in Guinea Bissau, in which adjustment was made for CD4 cell percentage, found no significant difference in mortality rate between HIV-1- and HIV-2-infected patients [²³]. These studies recruited patients in clinics at the time of TB diagnosis; therefore, it is possible that the HIV-1- and HIV-2-infected patients differed prior to diagnosis in characteristics that could not be accounted for.

Diagnosis of TB in the absence of positive sputum or culture results is difficult, as many clinical symptoms and even radiological signs are non-specific. This is even more difficult in severely immunocompromised HIV-infected patients. Because of the non-specific symptoms and difficulty in confirming a tentative clinical diagnosis, over- and underdiagnosis are common [²⁴]. In this analysis, we only included patients for whom a diagnosis could be ascertained by the available records. The percentage of cases confirmed by sputum or culture is, therefore, high. Other studies have concluded that underdiagnosis of smear-positive pulmonary tuberculosis might contribute to the relatively high proportion of smear-negative tuberculosis among HIV-infected patients [²⁵]. Therefore, the incidence rate we report here is most likely an underestimate.

The immunosuppression caused by HIV may lead to an increased susceptibility to (re)infection or reactivation of TB, in particular where TB treatment is less accessible. Indeed, TB is an AIDS-defining disease [¹²]. The poor prognosis following a diagnosis of TB in this cohort suggests that, even with high background population prevalence, TB in HIV-infected patients could be indicative of severe immunosuppression, with associated poor survival. There is a debate whether a diagnosis of TB among HIV-infected patients should be an AIDS-defining condition in areas where there is such a high background prevalence of TB [²⁶]. Our data show a higher mortality rate in those with incident TB and HIV infection, in particular among HIV-1-infected patients with high CD4 cell counts, which provides justification for TB as an AIDS-defining condition in sub-Saharan Africa. As ours was a retrospective study based on routine clinical records, there were some limitations to the analysis. In particular, we have not been able to account here for the effect of rate of CD4 cell decline on incidence and mortality. A community-based study comparing incidence and mortality of TB between HIV-infected and HIV- uninfected participants would be able to confirm prospectively the value of using a TB diagnosis as an AIDS-defining criterion.

We have previously shown that HIV disease progression for patients with CD4 cell counts < 200 × 10⁶ cells/l was similar among HIV-1- and HIV-2-infected patients [¹⁶]. We now show that the incidence of TB is similar among HIV-1- and HIV-2-infected patients stratified by CD4 cell count. Once a diagnosis of TB is made, mortality rates are similar among HIV-1- and HIV-2-infected patients. This supports the hypothesis that HIV-2, like HIV-1, causes ill health and premature morbidity through its effect on CD4 cells. Although generally HIV-2 less often and less rapidly leads to decline of CD4 cells, once the CD4 cell count has decreased, it is immaterial whether this immunosuppression is caused by HIV-1 or HIV-2.

The much increased mortality in HIV-1-infected patients with a high CD4 cell count suggests that these patients may have immune defects specific to TB or to intracellular pathogens prior to the onset of severe immunosuppression, which are revealed early in the course of HIV-1 infection. The limited excess mortality among HIV-1-infected patients with TB with lower CD4 cell counts could reflect the high mortality from other infections among this immunosuppressed group. Patients in this group may die before their TB can be diagnosed, as has been shown previously in a postmortem study [⁶]. Among HIV-2-infected patients, excess mortality with TB was more similar in each CD4 cell group.

The high incidence of TB among HIV-infected subjects might be a major factor in the resurgence of TB among the general population of sub-Saharan Africa, while the large burden of direct and indirect HIV-related TB puts a major strain on national programmes to achieve or maintain high detection and cure rates. The high mortality in HIV-infected patients with TB suggests that treatment is often initiated too late to prevent death. Highly active antiretroviral therapy (HAART) could reduce the incidence of HIV-1-associated TB by more than 80%, in particular among those with poor clinical condition and with a low CD4 cell count [²⁷], but widespread introduction of HAART in sub-Saharan Africa is hampered by financial, logistic, human resource and political restraints [²⁸]. Furthermore, drug interactions between antituberculosis medications and antiretroviral drugs could complicate this strategy [²⁹]. There is an urgent need for a strategy combining improved prevention, early diagnosis and treatment of TB disease in HIV-infected patients. This should be linked to the introduction of HAART to enable simultaneous treatment of TB and HIV [³⁰]. This will not only reduce morbidity and mortality among HIV-infected patients but might also reduce the increased transmission in the general population.

Acknowledgements

We thank the field and laboratory staff for their dedicated work, and thank Shabbar Jaffar, Saihou Ceesay, Philip Hill, Kebba Manneh and Christian Lienhardt for support and input. We are grateful to the patients who joined our cohort.

Sponsorship: This study was funded by the Medical Research Council UK.