Clinical and radiographic features of HIV-related pulmonary tuberculosis according to the level of immunosuppression

Medical charts and radiographs from 38 HIV-infected patients with positive cultures for Mycobacterium tuberculosis from sputum or bronchoalveolar lavage were reviewed in order to compare the clinical, radiographic, and sputum bacilloscopy characteristics of HIV-infected patients with pulmonary tuberculosis according to CD4+ lymphocyte count (CD4). The mean age of the patients was 32 years and 76% were male. The median CD4 was 106 cells/mm3 and 71% had CD4 < 200 cells/mm3. Sputum bacilloscopy was positive in 45% of the patients. Patients with CD4 < 200 cells/mm3 showed significantly less post-primary pattern (7% vs. 63%; p = 0.02) and more frequently reported weight loss (p = 0.04). Although not statistically significant, patients with lower CD4 showed lower positivity of sputum bacilloscopy (37% vs. 64%; p = 0.18). HIV-infected patients with culture-confirmed pulmonary tuberculosis had a high proportion of non-post-primary pattern in thoracic radiographs. Patients with CD4 lower than 200 cells/mm3 showed post-primary patterns less frequently and reported weight loss more frequently. Key-words: Pulmonary tuberculosis. Acquired immunodeficiency syndrome. Thoracic radiography. CD4 lymphocyte count.

It has been estimated that a third of the world's population is infected with Mycobacterium tuberculosis and that 9% of these cases are associated with human immunodeficiency virus (HIV) infection 6 .The proportion of this association appears to be greater in Africa, but some developed countries, like the USA, also present high numbers 6 .In developing countries, tuberculosis is the leading cause of preventable death 6 , such that it is responsible for approximately 11% of deaths in adults.Brazil is on the list of the 22 countries that concentrate 80% of the tuberculosis cases worldwide, occupying the 15th place in terms of numbers of cases (114,000 new cases per year) 25 .In Brazil, HIV infection is responsible for 3.8% of new cases of tuberculosis in the adult population 6 25 .
Atypical radiological presentations of pulmonary tuberculosis associated with HIV infection have been described before 1 21 22 , such that HIV-infected patients show a lower frequency of cavitations and higher frequency of mediastinal adenopathy, similar to primary tuberculosis.
The clinical and radiological manifestations of tuberculosis in HIV-infected patients may vary according to the degree of immunosuppression 9 19 20 , but published data on the subject is still scarce.The objective of the present study was to evaluate the clinical and radiological presentations of culture-confirmed pulmonary tuberculosis in HIV-infected patients.

MATERIAL AND METHODS
Selection of patients.Inclusion criteria: to be included, patients had to (1) be 13 years of age or older; (2) be HIV-infected according to the CDC (Centers for Disease Control and Prevention) criteria 4 ; (3) present a positive culture for Mycobacterium tuberculosis in their sputum or bronchoalveolar lavage; (4) have a record of CD4+ T lymphocyte cell counts available for a period of four months before or after the first clinical suspicion of tuberculosis; and (5) have radiographs of the thorax in either posteroanterior or anteroposterior orientation, with or without radiographs in lateral orientations, taken at most 30 days after the first clinical suspicion of tuberculosis.
Exclusion criteria: patients were excluded if (1) their culture for Mycobacterium tuberculosis was not related to a diagnostic procedure but to tuberculosis treatment control, or if (2) they had had a previous pulmonary disease that might be confused with the clinical and radiological patterns of tuberculosis.
Diagnostic tests.The tests used to diagnose HIV were Micro-ELISA AXSYM HIV1/2 (Meia) (Abbott, Abbott Labs, USA) and Western blot (Genelabs Diagnosis Pte Ltd., Singapore).Quantification of HIV RNA was done by measurement of the bDNA (Bayer Diagnostics, USA).CD4+ T-lymphocyte cell counts were assessed using a FACSCalibur flow cytometer (BD Biosciences, USA).
Cutaneous tuberculosis tests were performed in accordance with Brazilian standards using the Mantoux technique of intradermic inoculation of two TUs (tuberculin units) of PPD (purified protein derivative) RT 23.A positive cutaneous reaction was defined as an area of induration > 5 mm in diameter.
The acid-fast bacillus (AFB) findings from microscopic examination were described in terms of crosses, according to the number of bacilli per microscopic field examined 16 : (-) negative, when AFB were not found in 100 fields examined; (+) presence of less than one AFB per field in 100 fields examined; (++) presence of one to 10 AFB per field in 50 fields examined; (+++) presence of more than 10 AFB per field in 20 fields examined.
For mycobacterial culturing, Löwenstein-Jensen solid medium was used and growth detection was performed between the 45 th and 60 th days after specimen inoculation.When growth was detected, speciation of mycobacterial isolates and susceptibility testing followed.
Study design.This was a cross-sectional study, with comparisons of clinical characteristics and radiological data for HIV-infected patients presenting pulmonary tuberculosis, who were classified according to their degree of immunosuppression, as assessed by the CD4+ T-lymphocyte count.The following information was collected from patients' charts: age, sex, race, alcohol consumption, use of steroids, presence of comorbidities (diabetes mellitus, renal failure, cancer or silicosis), risk factors for HIV transmission, signs and symptoms relating to tuberculosis (coughing, fever, expectoration, complaints of significant weight loss or breathing difficulties), tuberculin test results, bacilloscopy results, sputum and bronchoalveolar lavage cultures, previous treatment and/or chemoprophylaxis for tuberculosis infection, CD4+ T-lymphocyte count, use of antiretroviral drugs and tuberculosis as the first opportunistic infection associated with HIV.
Radiological evaluation.Radiographs of the thorax were classified according to post-primary patterns (infusion and/or cavitations in the upper lobes or upper segments of the lower lobes, fibronodular opacities or bronchogenic dissemination) and primary patterns associated with other patterns (miliary pattern, hilar and/or mediastinal lymphadenopathy, pleural effusion, consolidation or localized infiltration in mid or lower-lung fields, lobar consolidation, diffuse reticulonodular infiltration, cavitation in mid or lower-lung fields, nodules or normal radiography) 19 22 .The radiographs of the thorax were evaluated by two of the authors (GFG and CF), without previous knowledge of the CD4+ T-lymphocyte count.The level of agreement between the two evaluators for the main radiological patterns was assessed by means of kappa statistics.When evaluations were discordant, the final report was obtained by reaching a consensus between the evaluators.
Statistical analysis.Initially, a descriptive analysis was performed.Some variables did not present a significant number of patients in the groups compared and were presented only in the descriptive statistics.Next, univariate analysis was performed.The variables that presented a significance level of up to 0.25 in the univariate analysis were subjected to multivariate analysis.For the multivariate analysis, p values < 0.05 were considered statistically significant.The Epi Info statistical package, version 3.3.2,was used for the analyses.
Ethical matters.The project was approved by the Research Ethics Committee of the Eduardo de Menezes Hospital and also by the Research Ethics Committee of the Federal University of Minas Gerais.
Of the 68 patients who were HIV-positive, 38 were included in the study.The other 30 patients were excluded from the study for the following reasons: 22 did not have a CD4+ T-lymphocyte count within the period previously established for inclusion in the study; four did not have a thoracic radiograph in the hospital files; three had radiographs of the thorax which were taken more than 30 days before the first clinical suspicion of tuberculosis; and one presented a radiological image of bronchiectasis, which was considered to be a confounding factor for the radiological diagnosis of tuberculosis.
Descriptive analysis.Of the 38 patients included in the study, 29 (76.3%) were male and 28 (73.6%) were nonwhite.The ages of the patients ranged from 22 to 58 years, with a mean age of 32.68 ± 7.65 years.The CD4+ T-lymphocyte count ranged from one to 558 cells/mm 3 , with a mean of 168.5 ± 172.6 cells/mm 3 .Eleven (28.9%) patients presented a CD4+ T-lymphocyte count greater or equal to 200 cells/mm 3 .Regarding educational levels, 7.9% were illiterate but most (65.8%) had concluded elementary school.The vast majority (91.6%) of the patients were hospitalized.Concerning risk factors for exposure to HIV, no pertinent information was reported in the charts of 42% of the patients.Exposure through heterosexual contact was described for 15 (39.4%) patients, followed by homosexual or bisexual contact for three (7.8%)patients, use of injectable drugs for three (7.8%)patients and blood transfusion for one patient.
The most common clinical manifestations were coughing and fever, found in 94.7% of the patients, expectoration in 71%, complaints of significant weight loss in 68.4% and shortness of breath in 52.6%.Coughing for a period of less than three weeks was present in 33% of the patients.Esophageal candidiasis was the most (42%) common comorbidity.Patients with higher CD4 cell counts were significantly older than were patients with lower counts (33 vs. 27 years; p = 0.03) but did not differ in relation to other demographic variables such as gender, race and educational level, nor regarding HAART use or alcohol abuse (Table 1).

Variable
Comparison of the clinical manifestations according to the CD4+ T-lymphocyte count showed that those with CD4 cell counts lower than 200 cells/mm 3 did not significantly differ regarding respiratory symptoms or fever (Table 2).
Among the radiological findings, the level of agreement between the two observers, as to the presence of primary and post-primary patterns associated with other patterns was very high (kappa = 0.80 and 0.78, respectively).Comparison of the radiological patterns showed that those with higher CD4 cell counts more frequently presented post-primary patterns (63.6 vs 7.4%; p = 0.02).The descriptions of the radiological patterns are shown in Table 3 and indicate that all the patients with CD4 cell count lower than 200 cells/mm 3 had primary pattern characteristics, especially lymphadenopathy and diffused infiltration.
A greater proportion of the patients with CD4 cell counts higher than 200 cells/mm 3 had positive findings from AFB microscopy in spontaneous sputum (63.6% vs. 37.0%), but the difference was not statistically significant (p = 0.18).The general results from bacilloscopy and mycobacterial culturing are presented in Table 4 and show that only 44% of the spontaneous sputum samples were positive on AFB microscopy.
The multivariate analysis of the relationship showed that postprimary patterns and weight loss were significantly associated with CD4 cell count after adjusting for potential confounders (Table 5).

Table 3 -Radiological findings from HIV-infected patients with pulmonary tuberculosis, stratified by CD4 + T-lymphocyte count (there may be more than one abnormality for the same patient).
CD4

DISCUSSION
The predominance of male and young patients in this study was similar to that found in other studies on tuberculosis coinfection conducted in Brazil 11 15 .The mean age of our patients was lower than what was observed in South Africa, where 86.7% of the HIVinfected patients with pulmonary tuberculosis had a mean age of more than 30 years 19 .In the present study, 73.6% of the patients were nonwhite.This finding differs from what has been found in other regions of Brazil 11 , where there is a preponderance of Caucasians among the patients with associated tuberculosis/HIV.Regarding education levels, our findings are comparable with the data from another Brazilian study 10 , in which the majority of the patients studied until the end of elementary school and a smaller proportion studied in high school or higher.
Considering the risk factors for HIV infection, the most (39.4%) reported risk was heterosexual contact, which is in accordance with Brazilian epidemiology 10 .
The most common clinical findings in the study group were coughing and fever.High frequency of these clinical manifestations was also found in Ethiopia 3 where more than 90% of the HIVinfected patients develop tuberculosis.In the present study, a third of the patients with pulmonary tuberculosis presented coughing for a period of less than three weeks, which is comparable with an African study conducted in an area of high prevalence of both tuberculosis and HIV infection 2 .
The reported weight loss was significantly lower in the group with a CD4+ T lymphocyte count > 200 cells/mm 3 .This had been expected, taking into account that this symptom is frequently found in patients with an advanced stage of HIV infection, regardless of tuberculosis coinfection.These results are comparable with those found in another African study where weight loss greater than 10% was strongly correlated with CD4+ T-lymphocyte counts < 200 cells/mm 3 for patients with both tuberculosis and HIV 18 .
The efficacy of sputum bacilloscopy for comparing HIVpositive and negative patients with tuberculosis is still a matter of debate.Some studies found a greater proportion of smear-negative tuberculosis in HIV-positive patients 3 15 22 , whereas several authors did not find any difference in the efficacy of bacilloscopy between HIV-positive and negative patients 15 19 21 24 .According to Smith et al 23 , the level of CD4+ T lymphocytes and the absence of cavitation or any other abnormality in the parenchyma did not alter the possibility of positive bacilloscopy in HIV-positive patients.In our study, the proportion of positive AFB microscopy on spontaneous sputum of patients with a CD4+ > 200 cells/mm 3 was almost twice that of patients with lower CD4+ cell counts, although this difference was not statistically significant.
The radiological findings for tuberculosis in association with HIV infection, in our study, were similar to those presented in other studies 14 20 , in which diffused or localized infiltration was more frequent, as well as hilar or mediastinal lymphadenopathy.Pleural infiltration, cavitation and normal radiography were the least common findings.However, studies conducted in Africa 5 19 demonstrated greater frequency of cavitations in HIV-positive patients with tuberculosis.One explanation for this could be the high incidence of tuberculosis in that region 6 , where tuberculosis infection appears at earlier stages of HIV infection, when the immune system is relatively preserved 8 17 .
Pleural infiltration as a manifestation of tuberculosis in HIVpositive patients is more commonly reported in patients with a Tlymphocyte count of CD4+ + > 200 cells/mm 3 , which reflects a strong immune reaction in the pleura 13 .In contrast, we observed that the vast majority of cases with pleural infiltration in our study were found in patients with a T-lymphocyte count of CD4+ < 200 cells/mm 3 .
Normal radiography of the thorax may be present in 5% to 10% of HIV-positive patients 8 and in up to 20% of the patients with an advanced level of immunodepression 9 .It remains uncertain whether the absence of findings represents early stages of either primary disease or reactivation, or disease caused by intrathoracic adenopathy that has not been detected by simple radiographic examination 20 .Upon comparing different radiological patterns with the CD4+ T-lymphocyte count, an association between the radiological pattern of post-primary tuberculosis and a CD4+ Tlymphocyte count of > 200 cells/mm 3 was found.All patients with CD4+ T-lymphocyte counts < 200 cells/mm 3 presented radiological patterns that were different from the post-primary ones.These results are in accordance with other studies, where cavitation 19 20 , post-primary or typical tuberculin radiological patterns 9 were correlated with CD4+ T-lymphocyte counts > 200 cells/mm 3 and lymphadenopathy and atypical findings with lower counts 9 19 20 .
Studies on the relationship between radiological patterns in tuberculosis cases and the time of infection, using molecular epidemiology, have suggested that the radiological appearance of tuberculosis is more related to host immunity than to whether the infection occurred recently or remotely 7 12 .
The limitations to the present study include the small specimen bank, the predominance of hospitalized patients and the exclusion of a considerable number of patients due to the rigid inclusion criteria.In addition, retrospective collection of data limits the availability and accuracy of the information.
In conclusion, patients with a CD4+ T-lymphocyte count > 200 cells/mm 3 demonstrated a greater likelihood of presenting radiological patterns of post-primary tuberculosis and a smaller likelihood of reporting weight loss, in comparison with patients with a CD4+ T-lymphocyte count < 200 cells/mm 3 .