Where MIC-based tuberculosis susceptibility testing can be trusted

Design

systematic review and diagnostic accuracy meta-analysis, PROSPERO-registered, studies published 2010-2025

Population

14 studies comprising 1,728 Mycobacterium tuberculosis complex isolates compared against reference drug susceptibility testing

Primary outcome

pooled sensitivity and specificity of Sensititre MYCOTB by drug

Effect

rifampicin sensitivity 0.976 (0.94-0.99), specificity 0.958; isoniazid 0.977 (0.95-0.99) and 0.957; cycloserine sensitivity 0.436 (0.190-0.725); moxifloxacin 0.801 (0.585-0.924)

Regimen design in multidrug-resistant tuberculosis increasingly depends on knowing minimum inhibitory concentrations rather than a binary susceptible-or-resistant call. The Sensititre MYCOTB plate gives MICs for many drugs at once, and this meta-analysis pooled 14 studies and 1,728 isolates comparing it against reference susceptibility testing.

For the two drugs that define the diagnosis, performance was excellent and consistent: rifampicin sensitivity 0.976 (95% CI 0.94-0.99) and specificity 0.958, isoniazid sensitivity 0.977 (0.95-0.99) and specificity 0.957, both with low heterogeneity. Amikacin, kanamycin and ofloxacin had good accuracy with specificity above 0.98. Ethambutol, streptomycin, ethionamide and rifabutin were moderate. Cycloserine, moxifloxacin and para-aminosalicylic acid were inconsistent despite high specificity; after sensitivity analysis moxifloxacin reached 0.801 (0.585-0.924) and para-aminosalicylic acid 0.76 (0.518-0.894), while cycloserine stayed at 0.436 (0.190-0.725).

High specificity with low sensitivity means the same thing for each of these drugs: a resistant result can be believed, a susceptible result cannot. For cycloserine that is a serious limitation, because it is a drug clinicians reach for when options are exhausted and the consequence of building a regimen on a false susceptible call is a failing regimen discovered months later. Read a MYCOTB report drug by drug rather than as a single document, and let phenotypic and genotypic results argue with each other rather than accepting whichever arrives first.

• Trust rifampicin and isoniazid results from this platform; they matched reference testing closely
• Treat a susceptible cycloserine result with scepticism — pooled sensitivity was 0.436
• High specificity with low sensitivity means resistant results are reliable and susceptible ones are not
• Cross-check phenotypic MICs against molecular results rather than acting on one alone
• Relevant directly to Indian programme practice, where MIC-based testing is expanding under the national programme

The statistics, in plain English

Sensitivity is the proportion of truly resistant isolates the test calls resistant; specificity is the proportion of truly susceptible isolates it calls susceptible. Cycloserine's sensitivity of 0.436 with an interval from 0.190 to 0.725 means the test misses more resistant isolates than it finds, and the width of that interval shows the studies disagreed badly. Low heterogeneity for rifampicin and isoniazid is what makes those pooled estimates usable; persistent heterogeneity elsewhere means the pooled number describes no particular laboratory well. All of this is measured against reference phenotypic testing, which is itself imperfect for several of these drugs.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.