I would like to thank M. Sadatsafavi and M. Najafzadeh for raising an interesting concern. First, our cost minimisation analysis 1 was not a cost-effectiveness analysis between the consequences of isoniazid treatment and nontreatment over several decades; it only considered costs that accumulated over the 2 yrs following the last possible moment of exposure of the contact person to the infectious source case.
Secondly, we had to include new tuberculosis (TB) cases that were typically identified in large contact studies in our model. These generally equate to ∼1% of the contact persons investigated, and, because they do not result in further investigation costs, must be shown to drop out of the cohort under suspicion. In our model, we distributed this “dropping out” over the entire 2-yr period. The crucial point, however, is simple: previous molecular–epidemiological studies 2–4 have shown that two types of TB cases may be expected in contact investigations of the kind we studied. On the one hand, there are TB cases with no epidemiological connection to the source case in question, i.e. they must have been infected by another source case at the same time or even many years prior to their diagnosis. On the other hand, there are secondary cases (re-activations) of infections that are actually acquired from the source case. Only these cases could be directly assigned to contacts that had previously tested positive using the tuberculin skin test or QuantiFERON-gold. Depending on the risk profile of the source case and its contacts, the proportions of the two types will vary and cannot be predicted.
Thus, although we found a number of positive tuberculin skin tests 4.4 times higher than positive QuantiFERON-gold results, we currently do not know whether the positive predictive value of the QuantiFERON-gold is really higher than that of a tuberculin skin test not confounded by Bacille Calmette–Guérin vaccination or nontuberculous mycobacteria. This must be determined by ongoing prospective restriction fragment length polymorphism studies observing the clinical course of QuantiFERON-gold-positive contact persons. These use “clustering” with the source case and observe who becomes infected with the disease. As QuantiFERON-gold is only just being implemented for routine use in western industrial countries, this question is not sufficiently answered. In our view, it would be speculative to distinguish between the predictive values of the two testing procedures. Therefore, we assigned a uniform probability of re-activation to manifest tuberculosis disease to subjects testing positive to either the tuberculin skin test or the QuantiFERON-gold assay.
- © ERS Journals Ltd
