Paradoxical reaction (PR) in tuberculosis (TB) is defined as transient worsening of symptoms and signs or the appearance of new lesions after beginning appropriate anti-tuberculosis chemotherapy. Recent studies suggest that PR occurs in 10–35% of patients. It is more common and more severe in HIV co-infected individuals with disseminated disease.¹ PR is thought to be an immune mediated phenomenon but the reasons for its occurrence are unknown.¹

Infection by Mycobacterium tuberculosis (MTB) results in highly variable outcomes between individuals. The characterisation of MTB strains by molecular typing techniques suggests this may be a reflection of the infecting organism, as well as host response and environmental factors. MTB strains with distinct genotypes have been shown to evoke different immunopathological events in mouse models² and variable clinical manifestations in human population based studies.³ Furthermore, individual strain types have been linked to particular clinical outcomes; for example, a significant association was seen between the Beijing MTB lineage and transient fever unrelated to disease severity, toxicity, or drug resistance in early treatment.⁴

We sought to investigate the hypothesis that the risk of PR may be strain dependent as defined by IS6110 restriction fragment length polymorphism (RFLP) typing.

Between January 2002 and December 2003 all adult patients seen at our centre with culture positive MTB had IS6110 RFLP typing performed on one isolate. A case note review was performed retrospectively for clinical evidence of PR.¹ IS6110 RFLP typing was undertaken using a modification of the standard international protocol.⁵ All patterns were entered onto a database using Bionumerics Edition 3.0 package (Applied Maths, Kourtrai, Belgium). Comparison of DNA fingerprints and cluster analysis of profiles was performed by calculation of the Dice coefficient; optimisation was set at 1% and position tolerance at 1.2%. A cluster was defined as a series of isolates with 100% identity. A putative lineage was identified as a series of isolates with 70% or greater similarity.⁶

145 patients had isolates that were typed. 100 (69%) sets of notes were reviewed. 45 were excluded (24 were unavailable or incomplete; 21 patients were lost to follow up or care was transferred). Of the 100 patients’ notes reviewed, 52 were male, age range 16–81 years. 48% were black African, 16% Asian, and 19% from the UK. Table 1 shows the TB site and HIV status of the patients. PR occurred in 20 patients (20%) (HIV positive 10/26 (38%); HIV negative or unknown 10/74 (14%)). All patients with PR had distinct IS6110 RFLP profiles suggesting 20 separate strains.

There was only one identified lineage with a similarity of 70% within the whole patient cohort. This was not associated with increased risk of PR. The group comprised 10 patients who were all black African, only one of whom had a PR. The similarity is likely to represent an original strain of African descent (data not shown).

IS6110 typing is the recognised gold standard for MTB strain typing and has widespread application in epidemiological and outbreak investigations of TB.⁵ Using this method we found one lineage, but no association between strain type and PR. The power of our study is limited by its small sample size and retrospective nature. However, the rate of PR (20% and three times higher in HIV positive subjects: 38% v 14%) is in line with previous work.¹ IS6110 typing is not a definitive phylogenetic marker and other molecular techniques such as single nucleotide polymorphism may demonstrate an association.

The possibility exists that paradoxical TB reactions may be a consequence of specific host response genes. Particular MHC haplotypes have been linked to “immune reconstitution disease” in HIV positive patients starting antiretroviral therapy.⁷ A larger study is needed to focus on both strain type and consequent host immune response.