The four-symptom screen for active tuberculosis (TB) that is recommended for all people living with HIV in lower-income settings is less likely to detect active TB in people taking antiretroviral therapy (ART) and may be improved by the additional use of a chest x-ray, according to a systematic review and meta-analysis by researchers at the Global TB Programme at the World Health Organisation (WHO).
TB remains the biggest cause of death in people living with HIV, including those already taking ART. WHO estimated that nearly 400,000 people with HIV died of TB in 2016. Preventing the development of active TB by giving isoniazid preventive treatment is a priority for reducing the incidence of TB in people living with HIV. TB prevention using isoniazid requires screening to rule out active TB, as the use of isoniazid preventive therapy in a person with active TB would lead to the development of isoniazid resistance.
To improve the uptake of isoniazid preventive treatment, WHO developed a four-symptom screening test that allows physicians to rule out the presence of active TB. In the absence of current cough, weight loss, night sweats or fever, TB preventive treatment can be started. If any of the symptoms are present, further investigation is necessary.
A meta-analysis showed that this four-symptom screen had a sensitivity of 79% and a specificity of 50% (in other words, the screen would correctly identify active TB in 79% of people who had it and correctly rule out active TB in 50% of people who do not have TB).
However, the four-symptom screen appears less sensitive in people on ART, missing more cases of active TB than in people not on ART. To determine the sensitivity and specificity of the four-symptom screen in people on ART and to establish whether the use of a chest X-ray would improve the sensitivity of screening in people on ART, WHO carried out a systematic review of studies and a meta-analysis of study data.
The systematic review looked for studies published in journals or presented at conferences after WHO recommended the four-symptom screen in 2011, in which the sensitivity and specificity of the four-symptom screen was verified with sputum or other samples, tested using culture or Xpert MTB/RIF.
The review identified 21 prospective cohort studies involving 15,247 people with HIV. Ten studies included people taking ART but three studies were excluded from the meta-analysis because they did not disaggregate data according to ART status.
The meta-analysis covered 18 studies involving 4640 people taking ART and 8664 people who were previously untreated at baseline. A median of 68% of people with HIV had at least one of the four symptoms but the median proportion of people on ART who had at least one symptom was lower, at 29.7%.
The meta-analysis found a pooled sensitivity of the four-symptom screen in people on ART of 51% (95% CI 28.4-73.2) – that is to say, the screen would identify only half of the people on ART who had active TB. The specificity was 70.7% (95% CI 47.8-86.4%). In treatment-naïve people, sensitivity was much higher in the pooled analysis (89.4%, 95% CI 83-93.4%), but specificity was lower (28.1%, 95% CI 18.6-40.1%) than in ART-treated people. However, when studies reporting no active TB in people not on ART were excluded, sensitivity improved in people on ART (62.1%, 95% CI 38.4-81.1%) but was almost unchanged in treatment-naïve people (88.5%, 95% CI 55-98).
Sensitivity of the four-symptom screen in people on ART was higher in studies that did not include pregnant women (62.1%, 95% CI 38.4-81.1%) but specificity was lower (62.9%, 33.2-85.3%).
Five studies included data on chest x-ray findings. However, only two studies provided data on 646 people that allowed the pooled sensitivity and specificity of the four-symptom screen plus chest x-ray to be calculated in people on ART.
These two studies suggested that adding the chest x-ray improved the sensitivity (84.6%, 95% CI 69.7-92.9%) but also substantially lowered the specificity (29.8%, 95% CI 26.3-33.6%).
The review authors say that the four-symptom screen may be less sensitive in people on ART because people with low CD4 counts – more likely to be initiating ART – are also more likely to present with symptoms of fever or weight loss due to HIV disease or other opportunistic infections.
Although chest x-ray significantly improved the sensitivity of screening in people on ART, the authors of an accompanying comment article, Colleen Hanrahan and David Dowdy of the Johns Hopkins Bloomberg School of Public Health, question whether the modest improvement in the probability of detecting active TB in the absence of symptoms is worth the logistical and financial challenges that national treatment programmes will face if they try to expand the use of chest x-ray.
“It is likely that a recommendation to require chest x-ray before initiation of preventive therapy will do more harm than good,” they say, pointing out that the reduction in the specificity of the four-symptom screen if chest x-ray is added would reduce the number of people eligible for isoniazid preventive treatment.
In the light of these findings, the WHO has recommended that chest x-ray may be added to the four-symptom screen in people who have already started ART.
Background: Since 2011, WHO recommends a four-symptom screening rule to exclude active tuberculosis in people living with HIV before starting tuberculosis preventive treatment (ie, absence of current cough, weight loss, night sweats, or fever). We assessed the sensitivity and specificity of the screening rule among people living with HIV based on antiretroviral therapy (ART) status and the added contribution of chest radiography.
Methods: We did a systematic review and meta-analysis. We searched PubMed, Embase, and the Cochrane Library from Jan 1, 2011, to March 12, 2018, for studies published after the WHO issued recommendations on the use of the four-symptom screening rule. We also searched abstracts from relevant international conferences. We included studies that collected sputum or any specimens (eg, urine, blood, or fine-needle aspirates from lymph nodes) from people with HIV regardless of signs or symptoms. Case-control studies were excluded because they are prone to bias. Active tuberculosis was diagnosed with bacteriological confirmation by culture or Xpert MTB/RIF of any specimens. Two investigators extracted the data, including age, sex, and ART status. We calculated sensitivity, specificity, and 95% CI. When at least four studies were available, we estimated pooled sensitivity and specificity using random and effects bivariate models; otherwise we used univariate random-effects models.
Findings: Of 4615 records identified by the search, 21 were included in the review (involving 15 427 people including 1559 with active tuberculosis). 18 eligible studies were included in the final meta-analysis. Seven studies provided data on people receiving ART. The pooled sensitivity of the four-symptom screening rule was lower for 4640 people on ART (51·0%, 95% CI 28·4–73·2) than for 8664 who were ART-naive (89·4%, 83·0–93·5). Pooled specificity for those on ART was 70·7% (95% CI 47·8–86·4) and for ART-naive people was 28·1% (18·6–40·1). On the basis of data from 646 individuals in two studies, the addition of any abnormal chest radiographic findings in people on ART improved sensitivity from 52·2% (95% CI 38·0–66·0) to 84·6% (69·7–92·9) but decreased specificity from 55·5% (95% CI 51·8–59·2) to 29·8% (26·3–33·6).
Interpretation: Our review suggested a lower sensitivity of the WHO four-symptom screening rule among people with HIV who are on ART than in those who are ART naive. The addition of chest radiography could improve the screening rule in people living with HIV who are on ART, provided it does not pose a barrier to preventive treatment.
Yohhei Hamada, Johnny Lujan, Karl Schenkel, Nathan Ford, Haileyesus Getahun