Multidrug-resistant tuberculosis has been declared a public health crisis by the World Health Organization (WHO).¹ The global burden of tuberculosis remains enormous with an incidence of 10 million new cases and a mortality of 1.3 million attributed to the disease worldwide in 2017.¹ The 2018 WHO Global Tuberculosis Report documented new tuberculosis cases, which included an additional 300 000 tuberculosis cases among HIV-positive tuberculosis patients.¹

South Africa is among the top 20 high-tuberculosis-burden countries worldwide and was previously ranked third following India and China.^1,2 Multidrug-resistant (MDR) tuberculosis is steadily increasing in South Africa; cases doubled from 7350 in 2007 to 14 000 in 2017.^1,3 A form of tuberculosis known as extensively drug-resistant (XDR) tuberculosis has been reported in 92 countries. South Africa has reported 73% of the global XDR cases following the historical outbreak of XDR tuberculosis in Tugela Ferry.⁴

The standard tuberculosis treatment regimen lasts for a minimum of 6 months. Treatments regimens for MDR tuberculosis (resistance to isoniazid and rifampicin) and XDR tuberculosis (resistance to any of the injectable drugs including fluoroquinolone, plus MDR tuberculosis),^5,6 have an extended treatment duration of 18–24 months with harmful second-line anti-tuberculosis drugs. The WHO has recently proposed a 9–12-month shortened regimen duration for MDR tuberculosis, with moxifloxacin replacing gatifloxacin (used in the original Bangladesh regimen).^7,8

The conventional, culture-based phenotypic method is recognized as the gold standard for confirmation of disease and tuberculosis drug susceptibility testing (DST). The agar proportion method (APM) compares growth of Mycobacterium colonies on drug-free and drug-containing mediums, where growth in a particular antibiotic-containing medium determines resistance. It is a low-cost method requiring no special equipment, but with a turnaround time of up to 6 weeks.^9,10 BACTEC Mycobacterium growth indicator tubes (MGIT) 960 liquid culture system (Becton Dickinson) is expensive and prone to contamination but has the advantage of a rapid turnaround time.²

The WHO has endorsed the GeneXpert MTB/RIF Ultra system, which is a real-time polymerase chain reaction-based assay for tuberculosis diagnosis and detection of rifampicin resistance, yielding results within 2 hours. The Genotype MTBDR Plus identifies common mutations for rifampicin and isoniazid.^1,11 Although molecular methods are rapid, they require costly equipment and staff expertise, and cannot detect resistant strains caused by unidentified mutations.¹¹

An innovative colorimetric assay has been designed using redox indicators to detect tuberculosis cell viability by a simple colour change. The resazurin microtitre plate assay (REMA) uses resazurin salts in a liquid culture medium.^12,13 This nontoxic compound incorporates into living cells and is reduced to the fluorescent molecule resorufin via a reduction and oxidation reaction.¹⁴ Visible colour change of the reagent from blue to pink demonstrates cell viability and therefore drug resistance. This technique has been applied for high throughput screening and determining minimum inhibitory concentrations (MIC) of anti-tuberculosis drugs.^14,15

Another advance in determining the phenotypic susceptibility of M. tuberculosis is the Sensititre^® MycoTB (MycoTB TREK Diagnostics) plate method. The Sensititre^® MycoTB plate method is a novel 96-well microtitre plate broth microdilution DST method incorporating drug concentrations of 12 anti-tuberculosis drugs for MIC determination.¹⁶ The drugs included consist of both first-line (rifampicin, isoniazid and ethambutol) as well as second-line drugs (moxifloxacin, ofloxacin, para-aminosalicylic acid, rifabutin, streptomycin, amikacin, cycloserine, ethionamide and kanamycin) existing as lyophilised forms in microtitre wells. A 7–21-day period of incubation is needed to observe culture growth noted as turbidity or cellular deposits at the base of a well.^14,16

The aim of this study was to determine the performance of the Sensititre^® MycoTB plate and REMA as potential tools for tuberculosis DST in a high tuberculosis-burden reference laboratory. This entailed comparison of the REMA and Sensititre assay using the APM as a gold standard in order to evaluate turnaround times, sensitivity, specificity, and positive and negative predictive values.

The 134 isolates utilised comprised clinically derived XDR (n = 65), MDR (n = 28), pre-XDR (n = 3), isoniazid mono-resistant (n = 15), rifampicin mono-resistant (n = 4) and susceptible (n = 19) according to the gold standard. The REMA and Sensititre^® MycoTB results for all 134 (100%) clinical isolates were tested and interpreted. The time to results for the Sensititre^® MycoTB method was as early as 7 days, with more reliable results being produced within 10 days.

Of the 100 rifampicin-resistant isolates by APM, 98 were resistant by Sensititre assay. Of these 98 isolates, 9 had an MIC of 1 µg/mL, 7 had an MIC of 2 µg/mL, 16 had an MIC of 4 µg/mL, 24 had an MIC of 8 µg/mL, 4 had an MIC of 16 µg/mL and 38 had an MIC above 16 µg/mL. Refer to Figure 1 for rifampicin MIC distributions using the Sensititre^® MycoTB.

Of the 16 isolates that were rifampicin-resistant by the Sensititre^® MycoTB having MIC of 4 µg/mL, 75% (12/16) were XDR, 12.5% (2/16) MDR and 12.5% (2/16) mono-resistant to rifampicin by the APM. Of the 24 isolates that were resistant to rifampicin by the Sensititre^® MycoTB with MIC of 8 µg/mL, 67% (16/24) were XDR, 8% (2/24) pre-XDR and 25% (6/24) were MDR by the APM.

The REMA correctly detected rifampicin resistance in all seven isolates with an MIC of 2 µg/mL by the Sensititre^® MycoTB (43% [3/7] XDR, 43% [3/7] MDR and 14% [1/7] mono-resistant to rifampicin by APM) with an MIC of 2 µg/mL by the Sensititre^® MycoTB. The REMA confirmed rifampicin resistance in all of the 38 isolates with an MIC above 16 μg/mL by the Sensititre^® MycoTB (68% [26/38] XDR and 32% [12/38] MDR) by APM. The overall sensitivity, specificity, positive and negative predictive values of the Sensititre^® MycoTB plate for rifampicin-resistance detection were found to be 98%, 82%, 94% and 93% (Table 1). The REMA displayed a sensitivity, specificity, positive and negative predictive values of 89%, 68%, 89% and 68% for the detection of rifampicin resistance (Table 2).

Of the 111 isoniazid-resistant isolates by APM, 108 were resistant by the Sensititre assay. Of these 108 isolates, 7 had an MIC of 1 µg/mL, 9 had an MIC of 2 µg/mL, 39 had an MIC of 4 µg/mL, 52 had an MIC of greater than 4 µg/mL and 1 had an MIC above 8 µg/mL. The seven isolates that were isoniazid-resistant by the Sensititre^® MycoTB with an MIC of 1 µg/ml (72% MDR and 28% mono-resistant to isoniazid by APM) were also confirmed resistant by the REMA assay. Nine isolates resistant to isoniazid by the Sensititre^® MycoTB with an MIC of 2 µg/mL comprised clinically-derived XDR (n = 6), MDR (n = 2) and pre-XDR (n = 1). Resistance to isoniazid by the REMA method was noted in all 9 isolates. Of the 52 isoniazid-resistant isolates with an MIC of over 4 µg/mL by the Sensititre^® MycoTB (Figure 1), 25 were XDR, 23 MDR and 4 were mono-resistant to isoniazid by the APM. The REMA assay confirmed isoniazid resistance in 51 of the 52 (98%) isolates. Refer to Figure 1 for isoniazid MIC distributions using the Sensititre^® MycoTB. The Sensititre^® MycoTB assay sensitivity, specificity, positive and negative predictive values for the detection of isoniazid resistance were found to be 97%, 96%, 99% and 88% (Table 1). The REMA showed sensitivity, specificity, positive and negative predictive values of 95%, 96%, 99% and 81% for the detection of isoniazid resistance (Table 2).

Sensititre^® MycoTB testing for the detection of resistance among the MDR and XDR isolates for rifampicin, isoniazid, ofloxacin and kanamycin correlated well with the APM (Table 1). Discrepancies between APM and REMA were observed with moxifloxacin susceptible isolates (APM) where 12 isolates were falsely designated as resistant by REMA (Table 2). The overall sensitivity, specificity, positive and negative predictive values of the Sensititre^® MycoTB plate for moxifloxacin-resistance detection were found to be 91%, 98%, 91% and 96% (Table 1). The REMA assay showed sensitivity, specificity, positive and negative predictive values of 91%, 78%, 64% and 96% for the detection of moxifloxacin resistance (Table 2).

The overall sensitivity, specificity, positive and negative predictive values of the Sensititre^® MycoTB assay for ofloxacin resistance detection were found to be 100%, 97%, 97% and 100% (Table 1). The sensitivity of detection for ofloxacin resistance was not assessed by the REMA as the drug could not be procured due to limited funds. The overall sensitivity, specificity, positive and negative predictive values of the Sensititre^® MycoTB plate for kanamycin resistance detection were found to be 92%, 85%, 86% and 92% (Table 1). The REMA assay showed sensitivity, specificity, positive and negative predictive values of 83%, 94%, 93% and 85% for kanamycin resistance detection (Table 2). Refer to Figure 3 for kanamycin MIC distributions using the Sensititre^® MycoTB. Good accuracy with regard to resistance to the first-line anti-tuberculosis drugs was observed with the Sensititre method as compared to the REMA (Tables 1 and 2).

The performance of the REMA for the detection of capreomycin resistance was determined. All of the 62 capreomycin-resistant isolates by the REMA were confirmed to be XDR tuberculosis by the APM (Table 2). Capreomycin was not part of the drug panel included in the Sensititre^® MycoTB assay, therefore its performance could not be assessed.

The levels of resistance among the MDR and XDR tuberculosis isolates to the additional first- and second-line antibiotics were further assessed using the Sensititre^® MycoTB assay using established critical concentrations (Figures 4 and 5). Resistance to rifabutin was observed in 93.5% (58/62) of XDR tuberculosis isolates with the majority 27% (17/62) showing an MIC of 16 µg/mL (Figure 4) at a critical concentration of 0.5 µg/mL. In contrast to this, 80% (20/25) of MDR isolates were confirmed to be susceptible to rifabutin with an MIC of 0.5 µg/mL (Figure 5).

Most (58/61, 95%) of the XDR isolates displayed resistance to para-aminosalicyclic acid (Figure 4). At an MIC of 1 µg/mL, 76% (19/25) of MDR isolates displayed susceptibility to para-aminosalicyclic acid (Figure 5). Resistance to amikacin was observed in 46 of the 62 (74%) XDR isolates (Figure 4). Most (59/62, 95%) XDR isolates were resistant to ethambutol. Conversely, 91% (21/23) MDR isolates were susceptible to ethambutol (Figure 5).

Resistance to amikacin was observed in 46 of the 62 (74%) XDR isolates; the majority (n = 23) of amikacin-resistant isolates had an MIC of greater than 16 µg/mL (Figure 4). In contrast to this, all of the 25 MDR isolates tested were susceptible to amikacin with a large proportion 76% (19/25) showing an MIC of 0.25 µg/mL (Figure 5). More than half (37/62, 60%) of XDR isolates were resistant to ethionamide with an MIC of 40 µg/mL or more (Figure 4). Among the ethionamide susceptible isolates, 92% (23/25) were MDR with 83% (19/23) showing an MIC of 0.6 µg/mL (Figure 5).

Resistance to streptomycin was observed in 90% (54/60) of XDR isolates with a large proportion of the streptomycin-resistant isolates, 40% (22/54), showing an MIC of more than 32 µg/mL (Figure 4). A susceptible streptomycin MIC of 0.5 µg/mL was observed in 19 of the 25 MDR isolates (Figure 5). Most XDR isolates (54/61, 80%) and 12% (3/25) of MDR isolates had MICs for cycloserine of 32 µg/mL or more (critical concentration of 30 µg/mL; resistant by APM) (Figures 4 and 5).

The objective of tuberculosis DST is to determine resistant strains that are prognostic of treatment failure and relapse. The Sensititre^® MycoTB assay provides a faster method of testing first- and second-line tuberculosis drugs and displayed 99.3% concordance with the agar proportion method in prior studies.¹⁹

In our study, the percentage agreements between Sensititre plate and APM for resistant isoniazid (97%), rifampicin (98%), moxifloxacin (91%) and ofloxacin (100%) were found in similar studies; however, kanamycin had the lowest categorical agreement with the APM (92%).¹⁹ Discrepant moxifloxacin (APM susceptible) isolates could be due to the subjectivity of the microtitre plate reading. A discrepant analysis would be useful to clarify discordant results, for example, repeat testing of isolates and ruling out technical errors. According to the WHO, current critical concentrations of all anti-tuberculosis drugs will be reviewed.¹ As phenotypic DST is required for the determination of moxifloxacin susceptibility (WHO recommendation due to poor concordance of Genotype MTBDRsl with APM), this might prove problematic if the Sensititre^® MycoTB assay is used.¹ The distribution of MIC with regard to second-line tuberculosis antimicrobials among MDR and XDR isolates was clearly evident as rising MIC values with the increasingly MDR strains. Resistance to capreomycin has been documented in KwaZulu-Natal in approximately 90% of XDR isolates of newly diagnosed individuals.^17,19 Both ethambutol and pyrazinamide resistance has been reported as exceeding 60%, the first-line drugs composing the Bangladesh regimen, as described in previous studies.⁸

The Sensititre^® MycoTB plate rapidly produced results in comparison to the APM, however, it produced 44% (15/34) false rifampicin-resistant isolates (sensitive by APM). The treatment implication would include patients who were mismanaged with prolonged, toxic second-line therapy in an MDR facility.

The REMA assay had a relatively short turnaround time of 9 days.²¹ Delays in results retrieval and therefore suitable treatment options for resistant tuberculosis may result in the further selection of resistant M. tuberculosis, morbidity and mortality in patients afflicted with the disease. Advantages of the REMA format are that it is faster, low cost, easy to interpret and does not need special equipment.²¹ A disadvantage of this method is the potential for aerosolisation since the plates utilise a liquid medium resulting in a biosafety hazard.¹⁵ In our study, REMA proved to be labour-intensive, requiring individual drug and dye preparation, as well as tedious microtitre plate inoculation. The current study showed the sensitivity and specificity of the REMA are comparable to that of the APM for isoniazid, rifampicin, capreomycin, moxifloxacin and kanamycin to be comparable to previous studies.²¹ Agreement between REMA and the APM was over 90% for resistance testing among MDR and XDR isolates (isoniazid 95%, capreomycin 93%, moxifloxacin 91%).