Patients	Patients or specimens of any age presumed to have resistance to any of the second-line TB drugs and those with confirmed MDR-TB
Prior testing	Patients who received MTBDRsl testing will first have received smear microscopy or culture (if smear-negative) or both for the detection of TB and phenotypic or genotypic DST for resistance to first-line TB drugs
Settings	Intermediate or central level laboratories
Index (new) test	MTBDRsl assay
Reference standard	Culture-based DST
Studies	Cross-sectional and case control studies in which cases and controls were sampled from the same patient population
A. MTBDRsl for second-line injectable drugs by indirect testing
Prevalence	Sensitivity (95% CI) 76.9% (61.1 to 87.6)	Specificity (95% CI) 99.5% (97.1 to 99.9)	Number of participants (studies)	Quality of the evidence (GRADE)*
15%	A diagnostic test does not always accurately detect all of the people who actually have the disease or condition in question. 150 people (out of 1000 people) have (as yet undetected) resistance. Of the 1000 people who take the MTBDRsl test, 115 people will be correctly identified as having resistance (TPs). However, 35 people with resistance will remain undetected; their "negative" test results will be incorrect (FNs).	A diagnostic test does not always accurately identify all of the people who do not have the disease or condition in question. 850 people (out of 1000 people) do not have resistance. Of the 1000 people who take the MTBDRsl test, 846 of these people will be correctly identified as not having resistance (TNs). However, 4 people will be incorrectly identified; their "positive" test results will suggest they have resistance (FPs).	1637 (14 studies)	Quality of the evidence indicates how likely it is that the accuracy of the test will be substantially different from what the research found. Moderate ⊕⊕⊕о
23%	230 people (out of 1000 people) have (as yet undetected) resistance. Of the 1000 people who take the MTBDRsl test, 177 people will be correctly identified as having resistance (TPs). However, 53 people with resistance will remain undetected; their "negative" test results will be incorrect (FNs).	770 people (out of 1000 people) do not have resistance. Of the 1000 people who take the MTBDRsl test, 766 of these people will be correctly identified as not having resistance (TNs). However, 4 people will be incorrectly identified; their "positive" test results will suggest they have resistance (FPs).
30%	300 people (out of 1000 people) have (as yet undetected) resistance. Of the 1000 people who take the MTBDRsl test, 231 people will be correctly identified as having resistance (TPs). However, 69 people with resistance will remain undetected; their "negative" test results will be incorrect (FNs).	700 people (out of 1000 people) do not have resistance. Of the 1000 people who take the MTBDRsl test. 696 of these people will be correctly identified as not having resistance (TNs). However, 4 people will be incorrectly identified; their "positive" test results will suggest they have resistance (FPs).
B. MTBDRsl for second-line injectable drugs by direct testing
Prevalence	Sensitivity (95% CI) 94.4% (25.2 to 99.9)	Specificity (95% CI) 98.2% (88.9 to 99.7)	Number of participants (studies)	Quality of the evidence (GRADE)*
15%	A diagnostic test does not always accurately detect all of the people who actually have the disease or condition in question. 150 people (out of 1000 people) have (as yet undetected) resistance. Of the 1000 people who take the MTBDRsl test, 142 people will be correctly identified as having resistance (TPs). However, 8 people with resistance will remain undetected; their "negative" test results will be incorrect (FNs). There is considerable uncertainty in these results. If the CIs are taken into account, then between 0 and 112 people might be missed (FNs).	A diagnostic test does not always accurately identify all of the people who do not have the disease or condition in question. 850 people (out of 1000 people) do not have resistance. Of the 1000 people who take the MTBDRsl test, 835 of these people will be correctly identified as not having resistance (TNs). However, 15 people will be incorrectly identified; their "positive" test results will suggest they have resistance (FPs). There is considerable uncertainty in these results. If the CIs are taken into account, then between 3 and 94 people might be misclassified as positive (FPs).	947 (6 studies)	Quality of the evidence indicates how likely it is that the accuracy of the test will be substantially different from what the research found. Very low ⊕ооо
23%	230 people (out of 1000 people) have (as yet undetected) resistance. Of the 1000 people who take the MTBDRsl test, 217 people will be correctly identified as having resistance (TPs). However, 13 people with resistance will remain undetected; their "negative" test results will be incorrect (FNs). There is considerable uncertainty in these results. If the CIs are taken into account, then between 0 and 172 people might be missed (FNs).	770 people (out of 1000 people) do not have resistance. Of the 1000 people who take the MTBDRsl test, 756 of these people will be correctly identified as not having resistance (TNs). However, 14 people will be incorrectly identified; their “positive” test results will suggest they have resistance (FPs). There is considerable uncertainty in these results. If the CIs are taken into account, then between 2 and 85 people might be misclassified as positive (FPs).
30%	300 people (out of 1000 people) have (as yet undetected) resistance. Of the 1000 people who take the MTBDRsl test, 283 people will be correctly identified as having resistance (TPs). However, 17 people with resistance will remain undetected; their "negative" test results will be incorrect (FNs). There is considerable uncertainty in these results. If the CIs are taken into account, then between 0 and 224 people might be missed (FNs).	700 people (out of 1000 people) do not have resistance. Of the 1000 people who take the MTBDRsl test. 687 of these people will be correctly identified as not having resistance (TNs). However, 13 people will be incorrectly identified; their "positive" test results will suggest they have resistance (FPs). There is considerable uncertainty in these results. If the CIs are taken into account, then between 2 and 78 people might be misclassified as positive (FPs).

*We deducted one point for limitations and, for direct testing, two additional points for imprecision (considering the very wide 95% CI for pooled sensitivity). We did not deduct points for indirectness; however we consider sensitivity and specificity to be surrogates for patient-important outcomes and high accuracy does not mean that patients will get better.

DST = drug susceptibility testing; TP = true positive; FP = false positive; TN = true negative; FN = false negative.