Trial of Antibiotics Useful in Predicting TB in People with HIV, Study Finds

Response to antibiotics may be an important measure for diagnosing tuberculosis in people with HIV, despite a World Health Organization move in 2007 to discourage the use of a trial of antibiotics as a diagnostic tool, a new study conducted in South Africa suggests.

The study, published in Open Forum Infectious Disease, involved 207 HIV-positive participants at a community health care clinic in Johannesburg with symptoms including cough, fever, night sweats or weight loss, between June 2018 and February 2019. The study found that 75 patients (36%) were confirmed to have tuberculosis.

It evaluated predictors of tuberculosis, and determined lack of clinical response to antibiotics to be a strong independent predictor of tuberculosis. It also explored multivariable prediction models (MPM), and determined that C-reactive protein (CRP) after antibiotics moderately improved the model, but CRP was not an independent predictor of tuberculosis.

“Using a 'trial of antibiotics' to see if a patient with HIV and cough actually has pneumonia (responds to antibiotics) or TB (doesn't respond to antibiotics) was helpful in this study,” Tom H. Boyles, MD, of the Department of Medicine at the University of Cape Town in South Africa told Contagion^®. “The WHO dropped the use of 'trial of antibiotics' a few years ago as there was limited evidence that it was a useful test but my study suggests that they should think again.”

WHO guidelines recommend antibiotics after ruling out likelihood of tuberculosis with Xpert MTB/RIF Ultra tests, chest X-rays and clinical assessments. However, trial of antibiotics is still used in settings where resources are limited, the study noted.

Among study participants with confirmed tuberculosis, 9 had positive sputum smears, 62 had positive Xpert MTB/RIF Ultra tests, 2 had rifampicin-resistant tuberculosis, and 1 had multidrug resistance.

For diagnosing tuberculosis, trial of antibiotics had a sensitivity of 0.43 and a specificity of 0.86. Sensitivity and specificity were 0.83 and 0.99 for Ultra, 0.95 and 0.26 for CRP, respectively.

“C Reactive protein (CRP) has been suggested as a good test for TB in outpatients. While that still holds for patients with limited or no symptoms, my study found that it was of no great value in sicker patients, basically because pneumonia, TB and even influenza all make CRP go up so it didn't differentiate between them,” Boyles told Contagion^®.

Boyles advice to clinicians: “Don't abandon 'trial of antibiotics' as a way to diagnose TB. Don't rely on CRP at first presentation to tell you anything useful, however if CRP is still high AFTER antibiotics (pneumonia treated), it does suggest TB.”

Boyles said next steps are: “1) Perform a randomized controlled trial of this strategy to see if it improves patient outcomes and costs; 2) Perform a similar study but focus on patients who have a negative Xpert (sputum TB test) to see what predicts TB in that group. The numbers in my study were too small to draw conclusions but the prevalence of TB in that group remained high (9%) so it is well worth looking into it.”

HIV and tuberculosis are a potentially deadly combination, according to a recent study that found a 30.7% mortality rate 56 days after diagnosis. In South Africa, tuberculosis causes most AIDS-related deaths.