Skip to main content
SearchLoginLogin or Signup

Review 2: "External Validation of a Treatment Decision Algorithm for Tuberculosis in Children Living with HIV - A Diagnostic Cohort Study"

The study is recognized as important work that could inform global and national policies, emphasizing the need for external validation of TDAs in different contexts.

Published onJan 21, 2025
Review 2: "External Validation of a Treatment Decision Algorithm for Tuberculosis in Children Living with HIV - A Diagnostic Cohort Study"
1 of 2
key-enterThis Pub is a Review of
External validation of a treatment decision algorithm for tuberculosis in children living with HIV - a diagnostic cohort study
External validation of a treatment decision algorithm for tuberculosis in children living with HIV - a diagnostic cohort study
Description

ABSTRACT Introduction Tuberculosis (TB) is the leading cause of death in children living with HIV (CLHIV) and is challenging to confirm the diagnosis. The PAANTHER treatment decision algorithm (TDA) was developed to improve the diagnosis of TB in CLHIV. We aimed to externally validate the PAANTHER TDA in CLHIV with presumptive TB.Methods We conducted a prospective diagnostic cohort study in seven tertiary hospitals across Côte d’Ivoire, Mozambique, Uganda, and Zambia, implementing the PAANTHER TDA in CLHIV aged between 1 month and 14 years with presumptive TB. TDA assessments included Xpert MTB/RIF Ultra (Ultra) on respiratory and stool samples, history of contact, symptoms (fever >2 weeks, unremitting cough, haemoptysis and/or weight loss in previous 4 weeks, tachycardia), chest radiography and abdominal ultrasound. A positive score (>100) prompted TB treatment initiation. Children were followed-up for 6 months, and retrospectively classified as having confirmed, unconfirmed or unlikely TB. The primary outcome was the proportion of missed TB cases (false negative) among children with negative scores; secondary outcomes included TDA diagnostic accuracy, feasibility, and time to treatment initiation. The TDA was considered validated if the negative predictive value (NPV, 1 - rate of false negative) was superior to a 75% pre-established confidence interval lower limit.Findings From 2 October 2019 to 31 December 2021, we enrolled 277 CLHIV, including 175 (63·2%) who were on antiretroviral therapy at inclusion. 272 (98·2%) children had a complete TDA evaluation; 215 (75.8%) scored >100, including 24 (8·7%) with positive Ultra. 182 (86·7%) children who scored ≥100, and 12 children who scored negative, initiated TB treatment at a median of 1 (IQR: 0-3) and 27 [8·2; 64] days after inclusion, respectively. 62/215 children (28·8%) who scored ≥100 were classified as having unlikely TB and 4/12 (33·3%) who scored negative were initiated on treatment and were classified as having unconfirmed TB. The proportion of children with TB (confirmed and unconfirmed) was 155/273 (56·8%; 95% CI: 50·9; 62·5). The NPV was 55/67 (93·3%; 95% CI: 84·1; 97·4), reaching protocol-defined validation. The TDA sensitivity was 97·4% (95% CI: 93·6; 90·0) with specificity of 47·5 (95% CI: 38·7; 56·4).Interpretation The PAANTHER TDA was validated in CLHIV. Its high sensitivity, excellent feasibility, and short turnaround time to treatment initiation, should allow rapid treatment decision-making and could reduce morbidity and mortality in CLHIV.Funding UNITAID

RR\ID Evidence Scale rating by reviewer:

  • Reliable. The main study claims are generally justified by its methods and data. The results and conclusions are likely to be similar to the hypothetical ideal study. There are some minor caveats or limitations, but they would/do not change the major claims of the study. The study provides sufficient strength of evidence on its own that its main claims should be considered actionable, with some room for future revision.

***************************************

Review: In the presented manuscript, there are some issues with the methods, singling out the sample size estimates vs the actual and the power calculations. The loss in sample size should be highlighted as a possible limitation and risk of biasing the estimates and thus the claims. We also had issues with choosing the NPV which the author termed 'unacceptable' - this affects the acceptability of the results. All these should be included in the limitations and possible ways the authors tried to counter these. Furthermore, using a prevalence of TB of 50% in children may not be plausible especially since it is not referenced. The authors may in this case have to show what the results would be at different TB prevalence for example 5%, 12.6%[referenced] and say 20%. I still note that 50% of TB in children is an overestimate therefore results may not have been interpreted correctly, which affects the conclusions. 

Some excerpts below:

  1. Calculated sample size was 550 children, but stopped at almost 50%: this would greatly affect the power calculations of this study which ultimately affects the strength of the claims.

  2. Combined with a low 'unacceptable' NPV of 75%; creates a lot of unanswered questions. Why did we choose to use this NPV?, how does this affect our results? 

  3. The hypothesised prevalence of 50% may not be plausible, of course it helps reduce the sample size, but brings our methods into question. The referenced paper, only points to 12.6% microbiologically confirmed, using the >50% may be misleading.

Comments
0
comment
No comments here
Why not start the discussion?