3.1. Pulmonary TB
The updated systematic review of the diagnostic accuracy of commercial tests for pulmonary TB identified 67 unique studies, including 32 studies from low- and middle-income countries.6 None of these studies evaluated the tests in children. The results demonstrate that:
for all commercial tests, sensitivity (0% to 100%) and specificity (31 to 100%) from individual studies are highly variable;
using bivariate meta-analysis, for Anda-TB IgG (the most commonly evaluated test), the pooled sensitivity is 76% (95% CI 63 to 87%) in studies of smear-positive and 59% (95% CI 10 to 96%) in studies of smear-negative patients, respectively; the pooled specificity in these studies was similar: 92% (95% CI 74 to 98%) and 91% (95% CI 79 to 96%), respectively;
for Anda-TB IgG, sensitivity values in smear-positive (54% to 85%) and smear-negative (35% to 73% ) patients from individual studies are highly variable;
for Anda-TB IgG, specificity values from individual studies are variable (68% to 100%);
in the TDR evaluation of 19 rapid commercial tests, in comparison with culture plus clinical follow-up, sensitivity (1% to 60%) and specificity (53% to 99%) values are highly variable;
compared with ELISAs [60% (95% CI 6% to 65%], immuno-chromatographic assays have similar sensitivity [53%, 95% CI 42% to 64%]; and
in the single study involving HIV-infected TB patients, the sensitivity of the SDHO test is 16% (95% CI 5% to 34%).
The only commercial test (Anda-TB) that could be included in sub-analyses provided poor performance and the other commercial tests did not have enough data to analyse. None of the tests reviewed could replace smear microscopy, a finding consistent with those reported in a previous systematic review.
The sensitivity and specificity estimates in this meta-analysis are likely to be overly optimistic for at least two reasons: (1) study quality generally suffered from lack of a representative patient spectrum which could result in exaggerated estimates of test accuracy and (2) potential publication bias, where studies with poor performance were likely to be unpublished.
3.2. Extra-pulmonary TB
The updated systematic review of the diagnostic accuracy of commercial tests for extrapulmonary TB identified 25 unique studies, including 10 studies from low- and middle-income countries.6 None of these studies evaluated the tests predominantly in children. The results demonstrate that:
for all commercial tests, sensitivity (0% to 100%) and specificity (59% to 100%) values from individual studies are highly variable;
pooled sensitivity is 64% (95% CI 28% to 92%) for lymph node tuberculosis and 46% (95% CI 29% to 63%) for pleural tuberculosis;
for Anda-TB IgG, although the pooled sensitivity and specificity are 81% (95% CI 49% to 97%) and 85% (95% CI 77 to 92%) respectively, sensitivity (26% to 100%) and specificity (59% to 100%) values from individual studies are highly variable;
in the single study involving HIV-infected individuals, the sensitivity of MycoDot is 33% (95% CI 19% to 39%).
As for pulmonary TB, the only commercial test (Anda-TB) that could be included in subgroup-analyses for extrapulmonary TB provided poor performance and the other commercial tests did not have enough data to analyze. These findings are consistent with those reported in a previous systematic review.
The sensitivity and specificity estimates in this meta-analysis are likely to be overly optimistic for at least two reasons: (1) as described earlier, study quality generally suffered from lack of a representative patient spectrum which could result in exaggerated estimates of test accuracy and (2) potential publication bias, where studies with poor performance were likely to be unpublished.
3.3. Case study of economic and epidemiological impact of serologic testing for active tuberculosis in India
India is the country with the greatest burden of TB, nearly 2 million incident cases per year. Conservatively, over 10 million TB suspects need diagnostic testing for TB each year. Findings from a country survey done for the Bill & Melinda Gates Foundation showed that the market for TB serology in India exceeds that for sputum smear and TB culture; six major private lab networks (out of hundreds) perform >500,000 TB ELISA tests each year, at a cost of approximately $10 per test or $30 per patient (for three simultaneous tests).7 Overall an estimated 1.5 million TB ELISA tests are performed every year in the country, mostly in the private sector.8
The impact of serological testing was compared against that of other TB testing modalities (sputum smear and culture) with sensitivity analysis performed around the accuracy of the test and the annual number of tests performed. Results showed that replacing sputum microscopy with serological testing would result in an estimated 14,000 additional cases of TB diagnosed but also result in 121,000 additional false-positive diagnoses relative to microscopy. In addition, the results indicated that for each additional smear-negative TB case diagnosed by serology, more than six additional false-positive would be inappropriately diagnosed.7
Most serological tests on the market in developing countries have no published evidence to support their claims of sensitivity and specificity (usually in excess of 95% each, according to package inserts). These tests are often performed in an environment with no external quality assurance, and tests from different labs on specimens from the same patient often yield widely varying results. A recent survey in the 22 high TB burden countries showed that regulation of TB diagnostics is weak in most countries, allowing for poorly performing tests to enter the market. Once on the market, incentives and financial gains by stakeholders (doctors, laboratories, diagnostic companies) keep these products profitable.8
3.4. Strengths and limitations of the evidence base
Strengths of the systematic review include the use of a standard protocol and comprehensive search strategy, independent reviewers, a bivariate model for meta-analysis, and pre-specified subgroups to account for heterogeneity.
Limitations related to the evidence base include the fact that the majority of studies was not considered to have a representative patient spectrum and was not performed in a blinded manner or blinding was not explicitly stated. Also, subgroup analyses were limited by the small number of studies for a particular commercial test or type of extrapulmonary disease. Differing criteria for patient selection and greater duration and severity of illness of the study populations may have introduced variability in findings among studies. Finally, although statistical tests and graphical methods are available to detect potential publication bias in meta-analyses of randomized controlled trials, such techniques have not been adequately evaluated for diagnostic data.
Nevertheless, it was considered prudent to assume some degree of publication bias as studies showing poor performance of commercial tests may be less likely to be published. This in turn may have introduced ‘optimism bias’ in the pooled estimates of sensitivity and specificity.
Concerning the TDR evaluation,5 a few additional limitations were discussed:
Testing was done retrospectively using stored frozen sera that passed through two freeze-thaw cycles. It is possible that the use of fresh serum may increase sensitivity;
There was limited geographic diversity amongst TB and HIV-positive patients whose specimens were used for evaluating the commercial tests. It is possible that there may be variations in the anti-mycobacterial antibody responses both due to patient genetic diversity and differential antigen expression by different mycobacterial isolates that could have led to reduced sensitivity with these specimens;
The duration of illness in patients was unknown. Greater duration or severity of illness may be correlated with the likelihood of a positive diagnostic test;
It is possible that infections with nontuberculous mycobacteria or exposure to environmental mycobacteria led to cross reactivity and decreased specificity;
The systematic review focused on test accuracy (ie. sensitivity and specificity). None of the papers reviewed provided information on patient-important outcomes, ie. showing that commercial tests used in a given situation resulted in a clinically relevant improvement in patient care and/or outcomes. In addition, no information was available on the values and preferences of patients.
No studies were identified that directly assessed the value of serology over and above conventional tests such as sputum smear microscopy. The TDR study did evaluate added value of smear plus serology and reported a gain equivalent to the detection of 57% of the smear-negative, culture-positive TB cases. However, there was a corresponding unacceptable decrease in specificity (58%).
