Validation of a Tuberculosis Treatment Decision Algorithm in HIV-infected Children

The ANRS 12229 PAANTHER 01 (Pediatric Asian African Network for Tuberculosis and HIV Research) study, which enrolled 438 HIV-infected children of median age 7.3 (IQR: 3.3 - 9.7) years with presumptive TB in four countries (Burkina Faso, Cameroon, Cambodia, Vietnam) from 2011 to 2014, aimed at developing a diagnostic prediction score and algorithm for TB treatment decision in HIV-infected children. This was the first study developing a TB diagnostic score exclusively in HIV-infected children, using methods recommended for diagnostic prediction models. Based on microbiological, clinical and radiological features, the best scoring system obtained had a sensitivity of 88.6% and specificity of 61.2% when Xpert MTB/RIF was included in the algorithm. The investigators showed previously that mortality is high in children with both confirmed and unconfirmed TB and that initiation of anti-TB treatment, that occurred at a median time of 7 (IQR: 5 - 11) days in the study, led to delayed ART initiation, which is associated to significantly increased mortality. The score, based on easily collected clinical features, chest radiographic features, Xpert MTB/RIF results, and abdominal ultrasonography, could enable same-day TB treatment decision. Abdominal ultrasonography has shown promise for the diagnosis of TB in both HIV-infected adults and children. In the PAANTHER study, it detected abdominal lymphadenopathy in 50% of culture confirmed TB cases and 35% of all confirmed and unconfirmed cases, with a specificity of 85%.

Developed in tertiary research-experienced health facilities, the PAANTHER score could enable standardized treatment decision in HIV-infected children with presumptive TB, and could be recommended for extensive use in secondary and primary healthcare settings where most of these children are seeking care. However, external validation studies are now needed to assess the predictive performance of the PAANTHER diagnostic score on independent datasets.

The PAANTHER TB treatment decision algorithm will be used for a TB treatment decision by site clinicians in all children enrolled in the study. Validation of the algorithm will be performed by evaluating the safety of withholding TB treatment in children not initiated on treatment as per the PAANTHER TB treatment decision algorithm. The safety of this strategy will be carefully assessed through review of safety reports every 4 to 6 months during study conduct by the safety sub-committee of the SAB.

Occurrence of algorithm failures in terms of missed TB cases (TB cases subsequently detected among untreated cases, i.e. false negatives) and cases with unlikely TB among those initiated on TB treatment as per the algorithm (cases not secondarily validated as confirmed or unconfirmed TB cases by the expert committee, i.e. false positives) will enable to estimate the negative and positive predictive values of the algorithm.

A centralized international Expert Committee will clinically review and validate TB diagnosis in children. This will enable assessment of the added value of new markers (MLR and CRP) and, if need be, to propose a new version of the score based on an optimised predicted probability.