We thank Dr Yansouni and colleagues for their comments.

The reason that some variables which seemed significant in the regression analysis were left out of the CART model was a consequence of the CART methodology as suggested. This method begins by categorizing the strongest identifiers first, which in this case resulted in the numbers of patients in each group being too small to provide justification for categorization of these parameters.

The question of immune reconstitution inflammatory syndrome (IRIS) is an interesting one, but was not within the scope of this study and so was not reported. However, we do have some basic data on numbers of patients on ART. A total of 28 of the 112 patients with cryptococcal meningitis (CM) and 12 of the 46 patients with tuberculous meningitis (TBM) were taking ART:

 Number of patients presenting with CM and TBM after starting ART (%)
≤3 months after ART3–6 months after ART6–12 months after ART≥12 months after ARTTotal on ART
  1. CM, cryptococcal meningitis; TBM, tuberculous meningitis.

CM12 (10.7)6 (5.4)5 (4.5)5 (4.5)28 (25.0)
TBM8 (17.4)3 (6.5)01 (2.2)12 (26.1)

Following operational definitions of IRIS to cryptococcus, we could regard patients presenting with CM within the first 12 months of ART as probable cases of IRIS. Although a high proportion of our patients were on ART, we excluded most with a previous diagnosis of CM and TBM, so these patients may not be considered by some definitions to have true IRIS.

In response to the second point, we entirely agree that the diagnosis of TBM is problematic and we recognize this as the basis of the difficulty faced by clinicians caring for these patients. We acknowledge that some patients with abnormal CSF in whom no pathogen was isolated may also have had TBM. A number of definitions of TBM have been proposed but clinical, microbiological and radiological diagnostic tools all have weaknesses and none provide a gold standard which will identify all those who are found to have TBM at autopsy. However, when conducting such studies, we are obliged to choose a definition based on the information available from our living patients who enter the study. Given the lack of sensitivity in all methods of diagnosing TBM and the constraints of a cross sectional observational study protocol, we chose one definition which at least provided a high degree of specificity.

We also agree with the third point made by the authors that clinicians are not aware which patients presenting with meningitis have one of these two diagnoses and are therefore eligible to enter the algorithm. However, given the fact that the algorithm was not specific enough to answer our question when only those with TBM and CM were entered into the analysis, we did not feel it necessary to reanalyse using the whole cohort. This would only have made the tool even less specific and re-enforced our conclusion that basic clinical and laboratory features are not sufficient for the diagnosis of these two complex conditions.