In the JournalsPerspective

Prediction score may enable prompt TB treatment in children with HIV

A tuberculosis prediction score based on diagnostic models may enable prompt treatment decisions in HIV-infected children with suspected TB and a high risk for mortality, according to research published in Pediatrics. Implementing the score would likely have significant public health benefits, the researchers noted.

“To our knowledge, this is the first study in which a diagnostic score is developed exclusively in children infected with HIV by using methods recommended for diagnostic prediction models,” Olivier Marcy, MD, PhD, a clinical epidemiologist at the University of Bordeaux in France, and colleagues wrote. “Previous pediatric TB diagnostic scores and algorithms were mostly based on expert opinion and often lacked validation.”

Marcy and colleagues noted that diagnostic challenges are greater in children infected with HIV, and that immunodeficiency reduces sensitivity of immunologic tests for TB infection. According to the researchers, of the 40,000 TB-related deaths in children infected with HIV, 90% occurred in those not receiving TB treatment.

In their study, called PAANTHER, the researchers tested scores based on clinical assessment, chest radiography, QuantiFERON Gold In-Tube (QFT) TB testing, abdominal ultrasonography and sample collection for microbiology. They enrolled HIV-infected children with suspected TB who were aged 13 years or younger in eight hospitals in Burkina Faso, Cambodia, Cameroon and Vietnam from April 2011 to December 2014.

Using the results of the tests, the researchers created four TB diagnostic models using logistic regression: 1) all predictors included; 2) QFT excluded; 3) abdominal ultrasonography excluded; and 4) QFT and abdominal ultrasonography excluded.

Among 438 children enrolled in the study, 251 (57.3%) had TB, including 55 (12.6%) with culture- or Xpert MTB/RIF assay-confirmed TB, according to the researchers. In the final four models, they included the Xpert MTB/RIF assay, fever lasting more than 2 weeks, unremitting cough, hemoptysis and weight loss in the past 4 weeks, contact with a patient who had smear-positive TB, tachycardia, military TB, alveolar opacities and lymph nodes on the chest radiograph. They combined these with data from abdominal lymph nodes on the ultrasound and QFT results.

Marcy and colleagues concluded that the score developed from the second model — based on clinical assessment of symptoms, Xpert MTB/RIF testing, chest radiography and abdominal ultrasonography — had a sensitivity of approximately 90% and a specificity of 61.2%, far exceeding the specificity of scores developed in previous studies.

“With its high sensitivity, our score should enable standardized treatment initiation in most HIV-infected children with tuberculosis,” the researchers wrote.

In a related editorial, Silvia S. Chiang, MD, assistant professor of pediatrics at Brown University, and Andrea T. Cruz, MD, MPH, research chief for pediatric emergency medicine at Baylor College of Medicine, noted the decision tool’s potential improved diagnostic sensitivity over microbiologic assays.

“When used at the primary care level by frontline providers, this tool may facilitate the diagnosis and timely treatment of TB in children coinfected with HIV and TB and, in doing so, may save lives.” – by Joe Gramigna

Disclosures: The authors report no relevant financial disclosures.

A tuberculosis prediction score based on diagnostic models may enable prompt treatment decisions in HIV-infected children with suspected TB and a high risk for mortality, according to research published in Pediatrics. Implementing the score would likely have significant public health benefits, the researchers noted.

“To our knowledge, this is the first study in which a diagnostic score is developed exclusively in children infected with HIV by using methods recommended for diagnostic prediction models,” Olivier Marcy, MD, PhD, a clinical epidemiologist at the University of Bordeaux in France, and colleagues wrote. “Previous pediatric TB diagnostic scores and algorithms were mostly based on expert opinion and often lacked validation.”

Marcy and colleagues noted that diagnostic challenges are greater in children infected with HIV, and that immunodeficiency reduces sensitivity of immunologic tests for TB infection. According to the researchers, of the 40,000 TB-related deaths in children infected with HIV, 90% occurred in those not receiving TB treatment.

In their study, called PAANTHER, the researchers tested scores based on clinical assessment, chest radiography, QuantiFERON Gold In-Tube (QFT) TB testing, abdominal ultrasonography and sample collection for microbiology. They enrolled HIV-infected children with suspected TB who were aged 13 years or younger in eight hospitals in Burkina Faso, Cambodia, Cameroon and Vietnam from April 2011 to December 2014.

Using the results of the tests, the researchers created four TB diagnostic models using logistic regression: 1) all predictors included; 2) QFT excluded; 3) abdominal ultrasonography excluded; and 4) QFT and abdominal ultrasonography excluded.

Among 438 children enrolled in the study, 251 (57.3%) had TB, including 55 (12.6%) with culture- or Xpert MTB/RIF assay-confirmed TB, according to the researchers. In the final four models, they included the Xpert MTB/RIF assay, fever lasting more than 2 weeks, unremitting cough, hemoptysis and weight loss in the past 4 weeks, contact with a patient who had smear-positive TB, tachycardia, military TB, alveolar opacities and lymph nodes on the chest radiograph. They combined these with data from abdominal lymph nodes on the ultrasound and QFT results.

Marcy and colleagues concluded that the score developed from the second model — based on clinical assessment of symptoms, Xpert MTB/RIF testing, chest radiography and abdominal ultrasonography — had a sensitivity of approximately 90% and a specificity of 61.2%, far exceeding the specificity of scores developed in previous studies.

“With its high sensitivity, our score should enable standardized treatment initiation in most HIV-infected children with tuberculosis,” the researchers wrote.

PAGE BREAK

In a related editorial, Silvia S. Chiang, MD, assistant professor of pediatrics at Brown University, and Andrea T. Cruz, MD, MPH, research chief for pediatric emergency medicine at Baylor College of Medicine, noted the decision tool’s potential improved diagnostic sensitivity over microbiologic assays.

“When used at the primary care level by frontline providers, this tool may facilitate the diagnosis and timely treatment of TB in children coinfected with HIV and TB and, in doing so, may save lives.” – by Joe Gramigna

Disclosures: The authors report no relevant financial disclosures.

    Perspective
    Jeffrey R. Starke

    Jeffrey R. Starke

    Although the incidence of TB disease in children living with HIV is many times higher than that in the general population in both high and low TB burden areas, the diagnosis can be very difficult to establish accurately. Microbiologic confirmation can be achieved in only approximately 30% of clinically suspected cases. For many decades, investigators have attempted to establish diagnostic scoring systems and algorithms for cases of childhood TB that cannot be microbiologically confirmed. Most of these methods have been inadequately validated and perform poorly in both sensitivity and specificity. None of them were designed for or tested in children also living with HIV infection.

    Diagnosing TB in children living with HIV is particularly difficult. The rates of microbiological confirmation are no higher in general than among children who are not infected with HIV. Clinical features such as symptoms and radiographic findings are less specific because other infections and conditions often mimic TB in these patients. The sensitivity of immunologically based tests, such as the tuberculin skin test and interferon-gamma release assays, is lower.

    This excellent study by Marcy and colleagues developed and tested a clinical model for diagnosing TB based on data from 335 children living with HIV in four countries. They looked at four different models using several clinical and radiographic features, along with a positive QFT result and visualizing abdominal lymph nodes on ultrasound. They evaluated several combinations of the clinical features and other studies, and determined that their four final versions of the model, with and without inclusion of the QFT result and abdominal ultrasound, had sensitivities from 84% to 89%. The model with the highest sensitivity — the one that excluded the QFT result — had a specificity of 61%. Interestingly, the degree of immunodeficiency and whether the child also was receiving ART did not improve model predictions. Although the score sensitivity did not differ among the four countries, specificities were lower in Cambodia and Cameroon than in Vietnam and Burkina Faso, possibly because of higher rates of nontuberculous mycobacterial infection, which is difficult to distinguish from TB, in the former two countries.

    When considering the diagnosis of TB in children living with HIV in high TB burden countries, sensitivity and positive predictive value are more important than specificity and negative predictive value because the morbidity and mortality of undiagnosed/untreated TB in these children is so high. In other words, it is preferable to “overtreat” rather than “undertreat,” especially initially, because the medications are generally well tolerated and the consequences of missing the diagnosis are high. This scoring system is designed for use in the field with limited resources to enable rapid treatment decisions for children living with HIV who have presumptive TB. Although further validation of the system is needed, especially for younger children with severe clinical status, this is an excellent step forward until better microbiological modalities for diagnosis in children are developed.

    • Jeffrey R. Starke, MD
    • Infectious Diseases in Children Editorial Board member
      Professor of pediatrics
      Baylor College of Medicine