Abstract
The aim of the present study was to evaluate the clinical characteristics, prognoses and predictors of mortality of patients with pulmonary tuberculosis (TB) with acute respiratory failure (ARF), and to investigate the adjunctive use of corticosteroids in such cases.
TB patients with ARF requiring mechanical ventilation (n = 90) were enrolled retrospectively during 1989–2006. The patients were divided into two groups: tuberculous pneumonia (TBP; n = 66), and miliary TB (MTB; n = 24).
The TBP patients were older than the MTB patients (mean age 68.0 versus 54.5 yrs), and the mean±sd interval from hospital admission to start of anti-TB treatment was longer for the TBP than for the MTB group (5.0±7.0 versus 2.8±2.5 days). However, there was no difference in in-hospital mortality rate between the two groups (68.2 versus 58.3%). In the TBP patients, multivariate analysis showed that advanced age and shock unrelated to sepsis were associated with poor outcomes. Even though corticosteroid use was a predictor of survival in TBP patients, it was difficult to conclusively determine the efficacy of corticosteroids in TBP with ARF because of the retrospective study design.
The present study reveals the need for randomised controlled trials to clarify the role of corticosteroids as adjunctive therapy in the management of tuberculous pneumonia with acute respiratory failure.
Following decline since the mid-1950s, the incidence of tuberculosis (TB) has, since the mid-1990s, begun to increase in many countries. The control of this disease has been impeded by co-infection with HIV 1 and the emergence of multidrug-resistant TB 2. Active pulmonary TB is a rare primary cause of acute respiratory failure (ARF) 3; however, high mortality rates have recently been reported in patients with ARF arising from TB 4–6.
Corticosteroids are the most important physiological inhibitors of inflammation. Several randomised studies have shown efficacy and safety of corticosteroid treatment in patients with severe inflammatory conditions, such as catecholamine-dependent septic shock 7, 8, severe community-acquired pneumonia 9, 10 and early acute respiratory distress syndrome (ARDS) 11, 12. With persistent unresolving ARDS, a beneficial effect of corticosteroid treatment was observed in some studies 13, but such improvements were not documented in others 14.
Corticosteroids have also been used as an adjunct in treating TB. The effective use of systemic corticosteroids in this regard is well-documented for several extrapulmonary forms of the disease, such as tuberculous meningitis 15 and tuberculous pericarditis 16. Several studies have suggested that more rapid radiological resolution of pulmonary infiltrates and closure of cavities accompany steroid use; these effects may be more pronounced in patients with severe disease 17, 18. However, corticosteroid use to modulate the harmful effects of severe inflammatory responses has not been prospectively investigated in patients with severe TB-induced ARF.
The present study was, therefore, conducted to: 1) evaluate the clinical characteristics of South Korean TB patients who develop ARF requiring mechanical ventilation; 2) determine the mortality rate and predictors of in-hospital mortality; and 3) investigate the status of adjuvant use of corticosteroids and the effect of such therapy on outcomes.
MATERIALS AND METHODS
Patients
The medical records of all relevant patients (aged >18 yrs) who had been admitted to the medical intensive care unit of the Asan Medical Center (Seoul, South Korea) between March 1989 and December 2006 were retrieved according to the following International Statistical Classification of Diseases and Related Health Problems, 10th revision, codes: A150–A153 and A160–A162 (TB of lung), and A190–A199 (miliary TB (MTB)). Medical records were reviewed and those patients selected who were bacteriologically or histologically diagnosed with active TB. Of these patients, 115 experienced ARF because of associated TB and were managed with invasive mechanical ventilation. Excluding 25 patients whose lungs had been extensively damaged by previous TB episodes, 90 patients were included in the final analysis of the present study. Of these 90 patients, 18 had already been diagnosed with TB at other hospitals and were then transferred to the Asan Medical Center, whereas the rest were diagnosed with TB after admission to the Asan Medical Center.
Chest radiography and data collection
Chest radiographs obtained during admission were reanalysed by a pulmonary radiologist and a pulmonary physician. In case of disagreement, the pulmonary physician’s view prevailed. Based on radiographic findings, the patients were divided into the following two groups: tuberculous pneumonia (TBP; n = 66), and MTB (n = 24). TBP was defined as parenchymal consolidation with or without endobronchial spread mimicking bacterial pneumonia 19. MTB was defined as the presence of bilateral diffuse millet-sized nodules.
Age, sex, Acute Physiology, Age, and Chronic Health Evaluation (APACHE) III score, body mass index, risk factors for TB, underlying diseases, previous history of anti-TB treatment, concomitant extrapulmonary TB, and duration of symptoms before admission were recorded and compared between groups. In addition, sputum acid-fast bacillus (AFB) smear and culture data, as well as other diagnostic results, such as drug susceptibility, arterial blood gas analysis, serum albumin level and in-hospital mortality rate, were recorded and compared. Furthermore, during hospitalisation, the presence of disseminated intravascular coagulation (DIC), concomitant ARDS, shock and organ failure were recorded.
The present study also investigated the status of adjunctive corticosteroid use, including dosage, duration and interval from commencement of anti-TB treatment to steroid therapy. The use of corticosteroids in patients with severe TB was not formalised in the Asan Medical Center, being completely at the discretion of the physician in charge.
Definitions
The presence or absence of DIC was determined by platelet count, prothrombin time, activated partial thromboplastin time, fibrinogen degradation product (FDP) level, d-dimer assay and fibrinogen level. Shock was defined as a requirement for vasopressors. The causes of shock were divided into septic shock and shock unrelated to sepsis, such as heart failure or cerebrovascular accidents. The diagnosis of ARDS was based on the consensus criteria of the American–European Consensus Conference 20. Sepsis was considered present when clinical suspicion of infection was combined with evidence of systemic inflammation, based on the criteria decided upon by a recent international consensus conference 21, and septic shock was defined as sepsis combined with hypotension refractory to fluid replacement 21. Organ failure was diagnosed based on the criteria of Knaus et al. 22, and multiple organ failure (MOF) was defined as the failure of more than one organ.
Statistical analysis
Categorical values, such as sex ratio and mortality rates, were compared between groups using the Chi-squared and Fisher’s exact tests. Continuous values, such as age and APACHE III score, were compared between groups using an unpaired t-test and the Wilcoxon rank-sum test. All values are presented as mean±sd or median with range for continuous variables, and numbers or percentage of the group for categorical values. A stepwise forward multiple logistic regression model was used to evaluate independent risk factors for the prognosis of TB with ARF, including all significant or almost significant parameters (p<0.1), based on univariate analysis. A p-value of <0.05 was considered significant.
RESULTS
Patient characteristics
Baseline clinical and laboratory characteristics and risk factors for TB are listed in table 1⇓. The 44 HIV ELISAs performed all gave negative results. The median age of all patients was 61.5 (range 22–89) yrs and the sex ratio was 1.5:1.0 (male:female). The symptom duration before admission was 27±26 days. Diabetes mellitus was the most frequent underlying disease, as shown in table 1⇓. Intrapulmonary cavities on chest radiography were present in 25 (27%) patients. Diagnosis was confirmed by sputum AFB smear and/or culture (n = 80), bone marrow biopsy (n = 7) or transbronchial lung biopsy (n = 3). Drug susceptibility data were available for bacterial isolates from 24 patients; isolates from six cases showed single-drug resistance, bacteria from two patients exhibited multidrug resistance (MDR) and Mycobacterium tuberculosis from the other 16 patients showed pan-susceptibility.
Comparison of patient characteristics in the tuberculous pneumonia(TBP) and miliary tuberculosis (MTB) groups
Hospital courses and outcomes
The mean durations of hospitalisation, intensive care unit stay and mechanical ventilation were 38±47, 21±27 and 18±26 days, respectively. The mean interval from hospital admission to commencement of anti-TB treatment was 4.3±6.0 days. Systemic corticosteroids were used in 44 (49%) patients. The mean daily dosage of prednisolone equivalents was 59±6.7 mg, and the median duration of corticosteroid therapy was 20 (range 7–120) days. In 36 out of 44 (81.8%) patients, the reason for using corticosteroids was noted as treatment of ARDS caused by TB. The median interval between start of anti-TB treatment and corticosteroid therapy was 2 (0–12) days. In eight out of 44 patients, systemic corticosteroids were started immediately before the diagnosis of TB for the treatment of another presumed condition, such as exacerbation of chronic obstructive pulmonary disease (COPD; n = 6) or presumed cryptogenic organising pneumonia (n = 2); however, these patients were finally diagnosed with TB and associated ARF. The in-hospital mortality rate was 65.6% (59 out of 90). TB was the direct cause of death in 50 patients (including 34 cases of ARF and 16 cases of shock). Other causes of death were intestinal perforation (n = 1), arrhythmia (n = 4), pulmonary embolism (n = 2), massive aspiration (n = 1) and intracranial haemorrhage (n = 1).
Comparison between the tuberculous pneumonia and miliary tuberculosis groups
Comparative data for the two groups are summarised in table 1⇑. The TBP patients were older than the MTB patients (median ages 68.0 and 54.5 yrs, respectively; p = 0.009). The mean interval from hospital admission to start of anti-TB treatment was longer in the TBP group than in the MTB group (5.0±7.0 and 2.8±2.5 days, respectively; p = 0.048); in addition, four patients for whom TB treatment was delayed for >14 days (15, 25, 28 and 30 days) were in the TBP group. Concomitant extrapulmonary TB, ARDS or DIC were more common in the MTB group than in the TBP group (all p<0.05). However, there were no significant differences in in-hospital mortality rates between the two groups (68.2 and 58.3%, respectively; p = 0.385).
Factors predicting in-hospital mortality according to groups
In the TBP group, univariate analysis showed that advanced age, longer duration of symptoms before hospital admission, the presence of shock unrelated to sepsis and non-use of steroids were factors influencing patient survival (table 2⇓). There were no significant differences in septic shock frequency, MOF, ARDS, number of lobes involved (based on chest radiography, which indicated the extent of disease; p = 0.448; data not shown) or presence of cavitation (70 and 67%, respectively; p = 0.860) between the survivor and nonsurvivor groups. Multivariate analysis revealed that advanced age and presence of shock unrelated to sepsis were independently associated with poor outcomes; however, the use of corticosteroids was a favourable prognostic factor for patients with TBP (table 3⇓).
Comparisons between survivor and nonsurvivor groups
Multivariate analysis of predictors of mortality in the tuberculous pneumonia(TBP) group
In the MTB group, unlike the TBP group, there were no factors predictive of nonsurvival (table 2⇑).
Comparison between steroid-use and non-use groups in patients with tuberculous pneumonia
Among TBP patients, those receiving corticosteroid therapy showed a lower mortality rate (56.7%; 17 out of 30) than those not receiving corticosteroid therapy (77.8%; 28 out of 36; p = 0.046; table 4⇓). There were no differences in clinical characteristics, such as age, duration of symptoms and risk factors for TB, including diabetes mellitus (20.0 versus 14.0%, respectively; p = 0.387; data not shown), between the steroid-use and non-use groups. There were no significant differences in severity indices (such as the oxygenation ratio), shock unrelated to sepsis, septic shock and MOF between the two groups. Corticosteroids did not affect either the duration of mechanical ventilation (p = 0.603) or oxygenation ratio (arterial oxygen tension/inspiratory oxygen fraction), measured on the seventh day of steroid therapy (p = 0.182; data not shown).
Comparisons between steroid-use and non-use groups
DISCUSSION
The present study was performed in a country with an intermediate burden of TB (73.0 cases per 100,000 general population in 2005) 23 and a low prevalence of HIV infection (the cumulative incidence of HIV infection during 1985–2004 was 3,153 individuals in a population of 47 million people) 24, and, to the best of the present authors’ knowledge, is the largest study investigating cases of TB with ARF.
In the current period following the development of anti-TB drugs, ARF arising because of TB has become relatively rare. However, TB remains a major cause of severe community-acquired pneumonia 4–6. In the present study, the mean interval from hospital admission to start of anti-TB treatment was 4.3±6.0 days. This indicates a somewhat shorter treatment initiation interval than in other studies (14.9 days 6 and 7.2 days 25). The probable reason is that, when Korean physicians examine patients with any atypical manifestations, TB is always suspected because there remains a relatively high incidence of TB in Korea, and medical professionals are aware that TB can have various manifestations. The overall mortality rate in the present study was 65.6% (59 out of 90), which is similar to those previously reported (66–81%) 4–6. When it is considered that most of the patients were prescribed anti-TB treatment immediately following admission, and that most cases had non-MDR-TB without concomitant HIV, the mortality rate was high.
In many previous reports, MTB was identified as the main cause of ARF or ARDS 22–28. Patients with MTB were more prone to the development of ARF that required mechanical ventilation, and some MTB patients initially presented with interstitial infiltration rather than a miliary pattern on chest radiography, resulting in delayed diagnosis of TB 5. In the present study, ARDS was reported in 83.3% of MTB patients, but this result may have been biased. In a previous report, despite clinical and radiological features compatible with ARDS in some patients, histology revealed confluent tuberculous bronchopneumonia with no evidence of ARDS 29.
In contrast to MTB, TBP has rarely been identified as a cause of ARF 19, 30. It is difficult to differentiate radiologically between TBP and severe bacterial pneumonia as causes of ARF, meaning accurate diagnosis can be delayed. In the present study, the mean interval from hospital admission to commencement of anti-TB treatment was longer in the TBP group than in the MTB group (5.0±7.0 and 2.8±2.5 days, respectively). The most important factor differentiating TB from other infectious causes was the longer duration of symptoms before admission 19. In the present study, the mean duration of symptoms before admission (29±28 days) was similar to those reported in previous studies 19, 31. Therefore, AFB sputum examination should be performed routinely in patients at risk of TB with severe pneumonia, particularly in endemic areas.
The use of corticosteroids to modulate the harmful effects of severe inflammatory responses has not been prospectively investigated in patients with severe TB-induced respiratory failure. Nevertheless, some clinicians prescribe corticosteroids when TB lesions are severe and progressive. Erbes et al. 30 reported that ARDS developed in seven out of 58 TB-with-ARF patients; steroids were used in all seven cases. Lee et al. 6 also reported that ARDS developed in 25 out of 41 TB-with-ARF patients; 13 out of 25 patients received >2 mg·kg−1·day−1 methylprednisolone on the seventh day following the onset of ARDS for the treatment of the fibroproliferative stage. The beneficial effects of corticosteroids in the management of TBP with ARF are suggested by several reports. First, mycobacterial antigen can induce release of pyrogens from monocytes, lymphokines from specifically sensitised lymphocytes and cytokines, such as tumour necrosis factor, from macrophages and peripheral blood mononuclear cells, which may be responsible for constitutional symptoms and tissue damage 32. Corticosteroids can inhibit the release and activities of lymphokines and cytokines. Secondly, the granulomatous host response to TB may paradoxically protect sequestered M. tuberculosis from anti-TB therapy. The adjuvant corticosteroids may be beneficial in permitting anti-TB drugs to penetrate into granulomas, by disrupting granuloma formation 33.
Among TBP patients, those receiving corticosteroid therapy showed a lower mean mortality rate. However, this may be an epiphenomenon; the data are insufficiently strong to permit the conclusion that steroids may be useful in TBP patients. In the present study, systemic corticosteroid use was based entirely on the attending physician’s decision and/or the patient’s underlying conditions. The use of corticosteroids was not formalised. The presence of ARDS was more frequent in steroid-use patients than in non-use patients (80 and 44%, respectively; p = 0.003), suggesting that the presence of ARDS may be part of the decision-making process when corticosteroids are prescribed. A history of previous TB may have influenced a decision not to prescribe steroids even though the frequencies of TB history did not differ significantly between the steroid-use and non-use groups (13.3 versus 33.3%, respectively; p = 0.059). A number of patients were already immunosuppressed or had other chronic diseases, and these findings may have prevented attending doctors from prescribing corticosteroids. Such patients, denied prescribing corticosteroids because of additional underlying diagnoses, may have had worse outcomes. When a positive bias for prescribing steroids was sought, it was observed, for example, that six patients were given steroids in order to relieve exacerbations of COPD but not to ameliorate the severity of the TB per se. It is conceivable that these patients were more likely to develop respiratory failure because of underlying lung disease, even though they had less severe TB than others, and were indeed more likely to survive.
In the MTB group, corticosteroid therapy did not improve survival rate; however, this result was inconclusive because of the small sample size. Any benefit of adjuvant corticosteroids in patients with MTB was not clear, since only limited evidence with conflicting results was available. A beneficial response was observed in one study 34, but such benefit was not documented in another 35.
The present study has some limitations inherent to retrospective studies; some unrecorded bias may be present. Even though large numbers of patients were included over a long period of time and multivariate analysis was performed, the data do not permit final conclusions to be drawn regarding the best treatment of TB with ARF. In addition, the small number of subjects in the MTB group reduced the statistical power of analysis of these patients. In cases of corticosteroid use in TB with ARF, possible positive or negative biases for steroid use (as described earlier) prevent the definitive conclusion being reached that corticosteroids may be useful. This is the case even though the survival rate was higher in the steroid-use group than in the non-use group, and steroid use was a favourable predictor of survival on multivariate analysis.
In conclusion, although South Korean clinicians usually consider tuberculosis to be a cause of severe community-acquired pneumonia and commence anti-tuberculosis treatment promptly, the mortality rate remains high. Factors independently associated with mortality in the tuberculous pneumonia group were advanced age and the presence of shock unrelated to sepsis. Even though multivariate analysis showed that corticosteroid use in tuberculous pneumonia patients was a favourable prognostic factor in lowering mortality, the retrospective study design prevents conclusion that corticosteroids may be useful in the treatment of cases of tuberculous pneumonia with acute respiratory failure. Further randomised controlled trials are necessary to clarify the role of corticosteroids in the management of tuberculous pneumonia with acute respiratory failure.
Statement of interest
None declared.
- Received July 1, 2007.
- Accepted June 14, 2008.
- © ERS Journals Ltd
References
