The commentary by Connell et al1 on the data presented in our paper2 is a welcome contribution to the debate on the most appropriate method for demonstrating latent tuberculosis (TB) infection in refugee children. Although a comparison of the performance of interferon-γ release assays (IGRAs) with tuberculin skin tests (TSTs) was not the primary aim of our study, the data do allow us to make observations on this topic. We have reassessed our data on the effect of previous BCG immunisation on IGRA and TST positivity (see table 3) and suggest that the very similar ORs for IGRAs and TST might reflect the adjustment of the cut-off point for a positive TST where we had added 5 mm for children under the age of 5 years with a history of previous BCG immunisation (see Methods section). It has been shown that BCG immunisation affects TST reactivity predominantly in children of this age.3 When our data analysis was restricted to include only children aged <5 years, the OR for a positive TST in BCG-immunised children was 3.2 (p=0.1) when the adjusted cut-off was used and 5.1 (p=0.04) when a cut-off of 10 mm was used. We are grateful to Connell et al for emphasising that the antigens used in both types of IGRA are also expressed by a small number of non-tuberculous mycobacteria and therefore can only be regarded as predominantly Mycobacterium tuberculosis-specific, as indicated in the Introduction to our paper.

We do not agree that our data ‘suggest TST may have had superior sensitivity to either of the two IGRAs in household TB contacts (ie, those at highest risk of latent tuberculosis infection)’. By reference to the data in figure 3 of our paper, it can be seen that five of six children (83%) with household TB contact who had a positive TST in association with neither IGRA being positive had received BCG immunisation. Furthermore, we had adjusted the cut-off point for TST positivity by subtracting 5 mm for children with household contact; re-analysis of the data using a cut-off point of 10 mm for all children decreased the OR for a positive result to 2.1 (p=0.1). It is therefore not unlikely that the higher rate of TST positivity in this subgroup reflected previous BCG immunisation rather than false negative IGRA results. With reference to the data in table 1 of Connell et al, we note that the percentage of children without an interpretable result is particularly inflated by the large number of younger children enrolled in the study. This is evident by the breakdown according to age group in table 1 (with the single HIV case omitted). It should also be noted that the failed phlebotomy numbers are further skewed for QuantiFERON-TB gold in tube (QFT-GIT) as, for most of the study, preference was given to attempting the T-SPOT.TB in cases of limited blood volume.

Connell et al did not comment on our finding that an inconclusive test result with one IGRA was usually associated with a valid result for the other IGRA. It is for that reason that we suggested initial testing with an IGRA rather than TST, and testing with the alternative IGRA when the first IGRA gives an inconclusive result. Furthermore, our data suggest that choosing the first-line IGRA according to patient demographic/clinical considerations will minimise the need for repeat testing. This strategy will only be possible when both IGRAs are available, but we believe it has the potential to be more cost-effective and convenient for both children and their families than a primarily TST-based screening approach, particularly in children from refugee families.