Application of QUADAS-C and GRADE in a Cochrane DTA Review: Comparison of lipoarabinomannan assays for tuberculosis in adults living with HIV

Article type
Authors
Yang B1, Bjerrum S2, Kohli M3, Schiller I3, Nathavitharana R4, Denkinger C5, Steingart K6, Leeflang M7, Zwerling A8, Shah M9
1UMC Utrecht
2University of Southern Denmark
3McGill University
4Harvard Medical School
5University Hospital Heidelberg
6Liverpool School of Tropical Medicine
7Amsterdam UMC
8University of Ottawa
9Johns Hopkins University School of Medicine
Abstract
Background: The WHO recommends the urine lipoarabinomannan assay Determineā„¢ TB LAM Ag (Determine-LAM) to help diagnose tuberculosis in people living with HIV. A new assay, SILVAMP TB LAM (SILVAMP-LAM)*, may serve as a replacement test for Determine-LAM. Two recently developed methods allow assessing certainty of evidence in comparative accuracy reviews: QUADAS-C and GRADE for comparative accuracy.

Objectives: To compare the accuracy of Determine-LAM and SILVAMP-LAM for diagnosing tuberculosis in adults living with HIV.

Methods: We searched 10 databases and registries in June 2021. We included studies comparing Determine-LAM and SILVAMP-LAM in adults living with HIV. We used two reference standards: microbiological (culture, nucleic acid amplification test, or both on any specimen type) and composite (microbiological or other tests or clinical assessment). We independently assessed study quality using QUADAS-2 and QUADAS-C. We estimated summary sensitivity, summary specificity, and their absolute differences using a Bayesian bivariate model. Using GRADE, we assessed certainty that the true difference in accuracy lies within the range of the 95% credible interval (CrI).

Results: We included five comparative accuracy studies. Using QUADAS-C, risk of bias was high in two. We present comparative accuracy estimates based on the microbiological reference standard; those based on the composite reference standard were similar. Based on four fully paired studies in inpatient settings (1021 participants, 619 cases), difference in summary sensitivity between SILVAMP-LAM and Determine-LAM was 26.5% (CrI 5.1% to 44.9%; moderate certainty) and difference in summary specificity was -4.4% (CrI -14.2% to 5.3%; moderate certainty). Based on three fully paired studies in outpatient settings (627 participants, 124 cases), difference in summary sensitivity and difference in summary specificity were 42.9% (CrI 11.4% to 66.5%; low certainty) and -4.8% (CrI -16.1% to 2.6%; low certainty).

Conclusions: In adults with HIV in inpatient settings, SILVAMP-LAM likely has higher sensitivity than Determine-LAM and specificity that likely ranges from lower to similar to that of Determine-LAM. In outpatient settings, SILVAMP-LAM may have higher sensitivity than Determine-LAM and specificity that may range from lower to similar to that of Determine-LAM. Results of an updated search will be presented at the meeting.

Patient involvement: none.
*Not commercially available.