Combining Xpert and LAM urine testing improves TB diagnostic sensitivity
1 April 2013. Related: Conference reports, Diagnostics, Coinfections and complications, TB coinfection, CROI 20 (Retrovirus) 2013.
Nathan Geffen, CSSR
Despite recent advances in TB diagnostics, current technologies suffer serious disadvantages.
- Mycobacterial culture tests, like the MGIT, are the closest to a gold standard. But results often take weeks, they require reference laboratories and are expensive.
- The Gene Xpert (Xpert) is also expensive and has good but not excellent sensitivity, especially in sputum-negative patients. Its place in the clinic, as opposed to laboratories, is not established. Also, with a two-hour turn-around time, it is still not a point-of-care test.
- Microscopy is affordable, but despite advances in microscope technology, its sensitivity is still poor and it has long turn-around times.
- Lipoarabinomannan (LAM) assays, which use urine to diagnose TB, are quick and reasonably affordable but have poor sensitivity, especially at higher CD4 counts. Additionally, detecting patients with extra-pulmonary TB, especially in cases where they do not have pulmonary TB, brings its own challenges.
Maunank Shah from Johns Hopkins, presented an important Ugandan study at CROI 2013 that considered whether TB diagnosis could be improved in HIV positive people using combination testing, particularly LAM assays and Xpert, and whether this has advantages in specific patient groups.
Shah pointed out that the Xpert’s advantages are that it detects smear-positive TB 95 to 98% of the time, that its specificity is 94 to 99% and that it detects rifampicin resistance with 97 to 100% sensitivity and 98 to 100% specificity. However, it detects smear-negative TB in only 55-77% of cases. Even with subsided pricing, the cheapest four-cartridge machine costs about $17,000 and each cartridge costs just under $10.
LAM is a component of mycobacterial cells walls. It is detectable in the urine of some HIV positive patients with active disease. The urine ELISA LAM has a sensitivity of 50 to 80% in HIV positive patients and is more sensitive at lower CD4 counts, especially in patients with disseminated disease, but its sensitivity is poor in most other patients. However, LAM is a point-of-care test and results are available in 20 minutes. LAM costs about 3 to 4 dollars per test.
This was a substudy of the “Feasibility of LF-LAM in HIV-infected individuals” study that included 208 patients, of whom 103 had culture-confirmed or clinically diagnosed active TB and 105 did not have active TB. Using two specimens, MGIT (the most sensitive single diagnostic test) detected 92% of cases compared to 80% for Löwenstein-Jensen. Mycobacterial blood culture detected TB in 48% of active-TB patients.
At baseline 60% of the active TB group were women versus 68% of the no-TB group. The median age was between 32 and 34 across the two groups. The CD4 median was 63 (IQR: 19-152) in the TB group and 280 (IQR: 97-486) in the TB-negative group. Most of the TB group (84%) was hospitalised versus 34% of the group without TB.
In the TB group, 54 (52%) had pulmonary TB alone. A further 43 (42%) had pulmonary TB and mycobacteremia. Six patients had mycobacteremia alone and would not be detectable using sputum, which, as Shah said, shows the value of doing blood cultures for TB in patients with HIV.
The accuracy of smear microscopy, Xpert and LAM were compared on this group of patients. Microscopy performed the worst. Using two specimens for each method, Ziehl-Neelson stains were 30% sensitive and fluorescence microscopy (FM) 42%. The lateral flow urine LAM test was 49% sensitive and was significantly better than smear microscopy. Xpert detected 76% of cases.
Specificity was 100% with microscopy, 97% with LAM and the 98% with Xpert 98%.
The urine LAM assay has graded bands on it: the higher the grade the more strongly positive the result. Reducing the grade increases sensitivity but worsens specificity. The study used grade 2 as the cut-off, which Shah said also avoided incorrectly detecting a few non-MTB mycobacterial infections. During questions, Shah said that it was worth exploring whether two LAM tests improved sensitivity.
Xpert detected 98% of the smear-positive FM cases but only 60% of smear-negative FM cases. LAM detected 56% of smear-positive FM and 45% of smear-negative FM cases.
Table 1 shows the LAM and Xpert sensitivities by CD4 count and that, as in other studies, LAM sensitivity is dramatically higher in patients with lower CD4 counts.
|CD4 count (cells/mm3)||LAM sensitivity||Xpert sensitivity|
|51 to 100||71%||76%|
|101 to 200||21%||84%|
In patients with pulmonary TB alone the LAM and Xpert sensitivities were 22% and 74% respectively. For patients with pulmonary TB and mycobacteremia, these were 79% and 86% respectively. And for patients with mycobacteremia alone, the LAM actually performed better than Xpert, finding TB in four of the six patients versus the one found by Xpert.
The main purpose of the study was to compare combinations of tests. FM combined with MGIT and blood culture yielded close to perfect sensitivity, but obviously this is not a practical combination in the vast majority of settings or even for most patients in very well resourced settings.
By combining LAM with smear microscopy, sensitivity rose to 67% (95%CI: 57-76), significantly better than either test alone (p<0.001 in both cases) and non-significantly different from Xpert alone (p=0.15). Specificity was 97% (95%CI: 92-99).
Using the LAM with Xpert produced even better results. This combination was 85% sensitive (95%CI: 77-92), significantly better than either alone and the combination of LAM and microscopy. It was non-significantly different from MGIT (p=0.167). Specificity was 95% (95%CI: 89-98). Also, combining LAM and Xpert yielded a sensitivity to smear-negative TB of about 80%. Shah explained that the difference in sensitivity by CD4 count seen with LAM alone disappeared using both tests in combination, that there is incremental gain using them together.
Shah also presented a slide summarising cost-effectiveness analyses of diagnostic algorithms, although it was beyond the scope of his talk to present how these estimates were derived (all monetary amounts rounded to nearest dollar):
- Smear alone costs $4 per person. The total cost of treating a TB patient using smear is $48.
- Smear plus LAM costs $8 per person. The total cost of treating a TB patient using smear plus LAM is $63. Per 10,000 people who might have TB, this adds an additional 4,046 disability adjusted life years (DALYs) over smear alone and costs $37 per DALY averted.
- Xpert alone costs $18 per person (presumably this incorporates the cost of the machine). The total cost of treating a TB patient using Xpert is $75. Per 10,000 people who might have TB, this adds an additional 4,706 disability adjusted life years (DALYs) over smear alone and costs $57 per DALY averted.
- Xpert plus LAM costs $22 per person. The total cost of treating a TB patient using Xpert plus LAM is $80. Per 10,000 people who might have TB, this adds an additional 6,284 disability adjusted life years (DALYs) over smear alone and costs $50 per DALY averted.
Shah concluded that all the incremental benefit that can be obtained from combining tests gains sufficient DALYs to be cost-effective.
This excellent study provides guidance to TB clinicians that can help improve diagnosis cost-effectively.
However, clinicians need to consider the data carefully in relation to the profile of patients at their site before implementing combinations of diagnostics. In this setting, patients with TB had very low CD4 counts and this benefited the relative sensitivity of the LAM. In sites with a different patient profile, the advantages of combining LAM and Xpert might not be as clear.
We still do not have anything close to an acceptable point-of-care TB diagnostic. While Xpert is an important advance, results still take too long and it requires greater expertise than many health facilities have available. LAM has important applications as a point-of-care test, but it is not good enough to be used alone.
Shah M et al. 2013. Comparative performance of rapid urinary lipoarabinomannan assays and Xpert MTB/RIF in HIV positive TB suspects: Uganda. 20th Conference on Retroviruses and Opportunistic Infections, 3-6 March 2013, Atlanta. Oral abstract 146.