Introduction

During the course of adequately treated tuberculosis (TB), effective killing of Mycobacterium tuberculosis may worsen pre-existing tuberculous lesions or induce new lesions, despite an initial improvement. These events are called paradoxical reactions (PR). When PR occur in human immunodeficiency virus (HIV)-infected patients following their initiation of highly active antiretroviral therapy (HAART), the TB PR are considered to be immune reconstitution inflammatory syndrome (IRIS). This syndrome was first described at the end of the 1990s in HIV-infected patients whose condition deteriorated after being started on HAART despite a decrease of HIV-viral load and an increase of CD4 lymphocyte counts [1, 2]. It consists of an acute exaggerated restoration of cellular immunity against live or dead pathogens, leading to an uncontrolled inflammatory response to pre-existing opportunistic infection [3]. Mycobacteria are the most frequent pathogens implicated in IRIS [4]. More recently, TB PR in patients following the discontinuation of immunosuppressive tumor necrosis factor-alpha (TNFα)-inhibitor treatment have also been reported, adding further evidence supporting an immune reconstitution as the underlying pathophysiologic mechanism of PR [5, 6].

Healthcare providers have long been aware of the occurrence of PR in non-immunocompromised patients receiving anti-TB therapy, especially the development of intracranial tuberculomas in central nervous system (CNS) TB or the involvement of new lymph nodes in TB lymphadenitis [7, 8]. Nevertheless, in non-immunocompromised patients, only a few series of PR have been published to date. PR are rare (<3 %) in the course of pulmonary TB in non-HIV patients [9]. Their incidence is certainly higher in extra-pulmonary TB, but estimations differ largely depending on the published study, and the factors identified as being associated with PR occurrence are discordant. So far, only the disseminated form of TB has been consistently associated with a higher risk of PR [10–12]. Moreover, guidelines on the therapeutic management of PR in HIV-negative patients are lacking.

We have reviewed a case series with the aims of (1) to describe the incidence and the clinical and biological features of PR during the course of extrapulmonary TB in HIV-negative patients and (2) to identify predictive factors of PR occurrence.

Patients and methods

Study population

This was a single-center retrospective study in the internal medicine department of a University tertiary hospital. Patients were identified through an International Classification of Diseases-9 (ICD-9) search, and all patients hospitalized with a code of TB as a main or secondary discharge diagnosis between January 2000 and January 2010 were eligible for enrolment. Patients were included in the study if they had a definite diagnosis of TB with at least one extrapulmonary involvement, proven by mycobacteriological (direct microscopy examination or cultures) or pathological study (giant cell granulomas and caseous necrosis). Mediastinal lymph nodes were considered to be an extrapulmonary site if they were >1 cm. HIV-infected patients were excluded.

Definition of PR

The diagnosis of PR was based on the fulfillment of the four following criteria: (1) initial improvement after anti-TB treatment initiation; (2) worsening of the initial symptoms or onset of new TB-like symptoms after the initiation of anti-TB treatment; (3) absence of persistently active TB (multidrug-resistant TB, poor compliance, impaired digestive absorption or positive cultures for M. tuberculosis); (4) absence of any other explanation of clinical deterioration. Initial improvement was defined as the disappearance of fever or improvement of the initial TB clinical symptoms and diminution of inflammatory biomarkers if they were increased at diagnosis. Worsening after initial improvement was defined by clinical worsening due to either a new location of TB-like symptoms or the recurrence of initially improved TB symptoms, with or without increased levels of inflammatory biomarkers. Isolated fever was not considered to be a PR. Compliance was assessed by the treating physician(s), and plasma drug levels were not systematically measured.

Global data collection

Data were collected by two investigators (GG and BR) from the computerized database of the hospital and outpatient files. In the case of missing data, we investigated the paper file. The following data were recorded: age, gender, ethnicity, past medical history, time from symptoms to TB diagnosis, location of TB, type and duration of treatment, adverse events and outcome. TB was considered cured if the treatment was completed, there was no further evolution of clinical or radiological TB features and mycobacterial cultures were negative. Complete remission was defined as the disappearance of all clinical, biological and radiological manifestations of TB. Data specifically recorded data for PR were date of onset, site (new site or deterioration within the initial site), treatment (corticosteroids, needle-aspiration or surgical procedure) and final outcome.

Statistical analysis

Quantitative variables were expressed as the median and interquartile range (IQR). The incidence of PR with 95 % confidence intervals (CI) was calculated using the Kaplan–Meier method and stratified according to different TB features (each encoded as a binary variable). For quantitative variables, the first quartile was the chosen cutoff value. Because of the different follow-up durations and the absence of hazards proportionality over time, a survival analysis using the log rank test was performed to compare the risk of PR in each class of variables. A multivariate analysis was not performed because of the low number of PR events, which allowed for only two or fewer explanatory variables in the model.

Results

Characteristics of patients with extrapulmonary TB

Our ICD-9 search yielded 174 patients, of whom 76 fulfilled all inclusion criteria; the remaining patients were excluded (41 because the diagnosis of TB was not confirmed, 19 because there was only a history of TB and 38 because the TB was isolated pulmonary TB). The median age of the enrolled patients was 40 (IQR 30–55) years, and the female:male ratio was 1.2. Patient ethnicity was diverse and included European (n = 15, 20 %), Indian (n = 22, 29 %), Subtropical African (n = 20, 26 %), North African (n = 8, 11 %) and South-East Asian (n = 5, 7 %) ethnicities. Ethnicity data were not available for six patients. Twenty (26 %) patients had an underlying immuno-compromising condition (4 with severe chronic renal failure, 4 with active cancer, 3 with diabetes, 2 with alcohol abuse, 2 with ongoing TNFα-inhibitor treatments). All patients had a negative HIV serology. Extrapulmonary sites of TB were the lymph nodes (n = 55, 72 %), pleura (n = 12, 16 %), peritoneum (n = 12, 16 %), bone (n = 7, 9 %), liver (n = 6, 8 %), pericardium (n = 6, 8 %), gastrointestinal tract (n = 5, 7 %), CNS (n = 4, 5 %) and urogenital tract (n = 1, 1 %). Simultaneous pulmonary TB was present in 22 patients (29 %). Among the cases of TB lymphadenitis, the most commonly involved sites were mediastinal (51 %) and cervical (40 %) sites. Direct microscopic mycobacterial examination was positive in 14/71 (20 %) specimens: five gastric aspirations, three sputum specimens, three lymph node biopsies, one abscess and one ascites. M. tuberculosis culture was positive in 43/64 (67 %) patients. All but one mycobacterial strains were sensitive to all anti-TB antibiotics (one strain was resistant to streptomycin). Biopsy sites were the lymph nodes (n = 25), peritoneum (n = 5), gastrointestinal tract (n = 4), pleura (n = 3), bone, pericardium (n = 3 each) and liver (n = 2). Pathological examination showed granulomas in 45/53 (85 %) patients with biopsy, associated with caseous necrosis in 36/53 (68 %).

Anti-TB treatment consisted of a 2HREZ regimen [isoniazid (H), rifampicin (R), ethambutol (E) and pyazinamide (Z)], respectively, for 2 months [13]) in 63/76 patients (83 %), followed by an HR regimen that was prolonged for at least 4 months. A 3HRE/6HR scheme was used in 7/76 (9 %) patients. Another combination of anti-TB antibiotics, including a fluoroquinolone (moxifloxacin), was prescribed in five patients. Median duration of anti-TB treatment was 9 (IQR 6–11) months. Median follow-up from the beginning of treatment was 15 (IQR 8–28) months. Among the 67 patients with at least 6 months of follow-up or TB-related death, complete remission of TB was obtained in 65 (97 %) patients and TB was cured with residual damage in four patients (1 lumbar pain, 1 limp despite hip surgery, 1 paraplegia and 1 persisting minimal pleural effusion). One patient who was non-compliant with the anti-TB treatment remained stable. One patient with meningeal TB developed obstructive hydrocephalus with aqueductal stenosis (this was not considered as a PR because of progressive deterioration since TB diagnosis and treatment) and died from intracranial hypertension after 2 months of treatment. Five other deaths were not related to TB.

Main features of PR

The main features of PR are shown in Table 1. PR occurred in 19 (25 %) patients during or after completion of anti-TB therapy, with a mean delay of 86 (IQR 36–125, range 7–307) days after its initiation. Fever was observed in 12/19 (63 %) patients. PR involved the same organ as the initial one at TB presentation in 16 patients (84 %) and consisted of deterioration of the initial lesions in 13 (68 %) of these patients. Enlargement of lymph nodes occurred in these 13 patients (68 %): seven times in the initial site (54 %), three times in a new site only (23 %) and three times in both (23 %). Other locations of PR included the lung (n = 3, 16 %), pericardium, pleura and spine (n = 2, 11 % each), brain, liver, ovary, psoas muscle and paravertebral muscles (n = 1, 5 %, each). PR that did not involve peripheral lymph nodes were mainly diagnosed because of inflammatory pain, with the exception of CNS PR that was revealed by neurological deficit. For example, one patient with bone PR was initially investigated for the occurrence of lumbar inflammatory pain during the TB treatment course whereas no back pain was noted at TB diagnosis. Spondylitis of the 12th dorsal and the first lumbar vertebral with an epidural abscess was observed on the lumbar magnetic resonance imaging (MRI) scan. Similarly, the patient with liver PR reported right upper quadrant abdominal pain, and ultrasound exploration revealed isolated hepatomegaly; the patient with ovarian PR reported abdominal pain and an increase of the her right ovarian abscess was observed on the computed tomography (CT) scan, while her initial ascites had completely resolved. Prednisone was prescribed in 9/19 (47 %) patients at a median dose of 0.5 (IQR 0.05–1) mg/kg/day. Anti-TB treatment was extended in four (21 %) patients for a median duration of 3 (IQR 0.5–4.5) months. Patients who were prescribed prednisone had lymph node (n = 8), pulmonary (n = 2), pericardium (n = 1), spine (n = 2), CNS (n = 1) and gastro-intestinal tract (n = 1) PR involvement. Additional treatment was performed in 11 (58 %) patients including five needle aspirations of lymphadenitis, four surgical drainages (3 lymphadenitis and 1 psoas abscess) and two surgical excisions of lymphadenitis. Five patients received no specific PR treatment.

Table 1 Main characteristics of the 19 patients with paradoxical reactions complicating tuberculosis with extrapulmonary manifestation
Full size table

PR recurred at least once in four patients and involved the same location as the first PR episode in most cases: lymph node (4 and 1 recurrences in 2 patients), ovarian abscess (1 recurrence in one patient). Interestingly, one patient with miliary, lymph nodes and spinal locations at TB diagnosis presented a first PR involving the liver, a second one involving a lymph node and a third one involving the lumbar spine. Median interval between PR was 2 (range 1–11) months. Two of these patients had surgical treatment (1 lymph node excision and 1 surgical subcutaneous abscess drainage) and three received oral steroids. Anti-TB treatment was prolonged until PR resolution in two of the latter.

Complete recovery was noted in 18/19 (95 %) patients after a median of 2 (1–7) months. The longest disease duration was almost 4 years (with 5 successive lymphadenitis and/or subcutaneous PR) before complete resolution. One patient with T8 level extradural and intramedullary tuberculoma enlargement developed paraplegia that persisted after TB recovery.

Manifestations at TB presentation associated with the occurrence of PR

No association was found between PR occurrence and age, gender, ethnicity, past medical history including immunosuppressive treatment, mycobacterial documentation and initial TB locations with the exception of peripheral lymph nodes (p = 0.009) (Table 2). Patients with low lymphocyte counts (<1,000/mm3, p = 0.03) and anemia (hemoglobin level <10.5 g/dL, p = 0.002) developed PR more frequently. The final outcome was similar in patients with or without PR.

Table 2 Predictors of paradoxical reaction according to the characteristics of patients and manifestations of tuberculosis at presentation
Full size table

Discussion

The results of our study highlight pertinent clinical messages: (1) PR are quite common in extrapulmonary TB in HIV-negative patients, (2) peripheral lymph node location of TB, lymphopenia and anemia may be predictive factors of PR, (3) outcome of PR patients is excellent, except in the presence of CNS involvement.

The pathogenesis of PR remains unclear in HIV-negative patients. Biological studies have demonstrated that M. tuberculosis develops several mechanisms to escape the human immune system. Only live M. tuberculosis can incite this specific host immunodeficiency through alterations of cell-mediated immune responses, such as inhibition of the phagosome–lysosome fusion in macrophages or repression of MHC-II expression on the dendritic cell membrane [14]. In the presence of bactericidal therapy, an immune rebound may occur, possibly enhanced by the release of M. tuberculosis antigens during the destruction of infected macrophages, which may explain the worsening of clinical symptoms. A similar phenomenon also occurs in other human mycobacterial diseases, such as leprosy (in which case PR are called “reversal reactions” [15]) and Buruli ulcer, due to M. ulcerans infection [16].

A recent large observational study of pulmonary TB reported a low occurrence of PR (2.4 %) in HIV-negative patients [9]. The prevalence of PR in studies involving patients with extrapulmonary TB is much higher and ranges from 16 to 50 % [10, 17–20]. The two largest series investigating PR in extrapulmonary TB both reported a PR prevalence of 23 %. In these latter two studies, 25 of the 109 patients with lymph node TB [10] and 32 of the 139 patients with pleural TB [20] developed PR. Among the patients enrolled in our study, who presented with various types of extrapulmonary TB, but mainly lymphadenitis, we found a remarkably similar prevalence of 25 %. In a review of TB PR case-reports including 122 patients, CNS was the primary location of TB in 52 (43 %) cases, the respiratory system in 19 (16 %) cases and lymphadenitis in only five (4 %) cases [10], but this distribution is likely to have been distorted by a publication bias. In our study, peripheral lymph node TB was associated with a higher incidence of PR as compared to other primary TB locations. However, this association could have resulted from a diagnostic bias since peripheral lymph node enlargement can be detected at clinical examination unlike when PR affects visceral lymph nodes or other internal organs. Too few CNS TB were included in our series to draw any conclusions regarding this specific location.

We observed a higher incidence of PR in patients with baseline anemia, as did Cheng and colleagues [9] in their prospective study of PR occurring in pulmonary TB. Anemia can have several causes in TB, including systemic inflammation and bone marrow TB involvement. In our case series study this association with anemia was not attributable to an inflammatory syndrome based on C-reactive protein levels, but it could reflect TB dissemination. The incidence of PR was also higher in patients with low baseline lymphocyte counts as reported by two previous studies [18, 20], whereas these laboratory features were not predictive of PR in three other case series studies [9, 11, 12]. However, our study has some limitations, including its retrospective design, small sample size that precludes the detection of predictors having a small effect and a possible selection bias because of the single-center tertiary recruitment. Therefore, we are currently conducting a large prospective study to assess the validity of these possible PR predictors.

The vast majority of the patients with PR enrolled in our study had a favorable outcome, as has been previously reported [10, 11], including the few ones who experienced several PR episodes for months (up to 3 years). However, some CNS PR with enlargement of intracranial tuberculomas may lead to hydrocephaly, increased intracranial pressure and death [7, 21] or severe neurological damage, as observed in one of our patients.

The treatment of PR remains a matter of debate. Although frequently reported, the effectiveness of corticosteroids cannot be inferred from observational studies because of spontaneous remission and selection bias, since physicians are more likely to prescribe corticosteroids for the most severe PR [10, 11, 22]. Nevertheless, corticosteroids seem to be safe and are strongly recommended by several authors for the treatment of severe PR [23]. Likewise, additional procedures, such as surgical drainage, aspiration of lymph nodes or drainage of pleural effusion, have been reported to be very effective, but only in case-reports or small observational case series [11, 12, 24]. Randomized controlled prospective studies are needed to assess the indications and benefit of any therapy in PR.

Conclusions

Paradoxical reaction is not uncommon during the treatment of extrapulmonary TB in HIV-negative patients. Peripheral lymph node location, lymphopenia and anemia at presentation are potential predictors of PR occurrence. Corticosteroids and drainage or aspiration are commonly used but should be evaluated by randomized controlled trials. With the exception of a CNS location, the outcome of PR patients is usually excellent.