Evaluation of Etest Strips for Rapid Susceptibility Testing of Mycobacterium tuberculosis

Abstract

In this study, our objective was to evaluate Etest strips containing exponential gradients of isoniazid (INH), rifampin (RIF), and streptomycin (STR) for susceptibility testing of Mycobacterium tuberculosis. M. tuberculosis isolates were tested for antimicrobial susceptibilities by the standard proportion method using Löwenstein-Jensen (LJ) medium and by the Etest. The MICs determined by the Etest were obtained at 5, 7, or 10 days. In some strains with Etest-discrepant results, radiometric susceptibility testing (BACTEC) was performed to determine a consensus result. M. tuberculosis concordance between the two methods was 97% (86 of 89 isolates) for RIF, 96% for INH (84 of 87 isolates), and 80% (61 of 76 isolates) for STR. Most of the MICs determined by the Etest were easy to interpret and readable within 5 days. Results correlated well with those obtained by the LJ proportion and BACTEC methods for INH and RIF. However, a high proportion of false-sensitive and false-resistant results were observed, most often for STR. We also observed that variations in the inoculum size of M. tuberculosis isolates affected the MICs to a substantial degree. These discrepancies, along with the expense of the media, the Etest strips, and the specialized equipment required (CO₂ incubator), make this method less useful in developing countries.

In the last decade, industrialized countries have been threatened by the reemergence of tuberculosis (15). However, in developing countries where the disease has always been endemic, its severity has increased not only because of the global human immunodeficiency virus pandemic but also because of the rise in poverty and the lack of efficient tuberculosis control programs. The emergence of multidrug-resistant (MDR) Mycobacterium tuberculosis has further exacerbated the tuberculosis problem.

Mycobacterial drug susceptibility profiles are performed to give adequate information regarding the treatment of individual patients and to improve strategies applicable to large-scale populations. The proportion and radiometric (BACTEC) methods are the standards for drug susceptibility testing. Both of these methods depend upon growth of the test organism, and typically only one or a few of the drug concentrations are evaluated, depending upon the antimicrobial agent being tested (5). The BACTEC method is performed in only a few laboratories in developing countries, including Brazil, due to the cost of reagents and equipment. In these countries, the proportion method is usually performed using Löwenstein-Jensen (LJ) medium, which is less expensive than the standard Middlebrook 7H10 agar and which provides results in ≤4 weeks.

The new epsylometer test (Etest) method for mycobacterial susceptibility testing was initially evaluated by Wanger and Mills (13) as an alternative rapid susceptibility method. The potential advantages were that results were obtainable in less than 1 week, it was easy to perform, and one strain could be tested against as many as five drugs in the same 150-mm-diameter plate. These features made it attractive for use in routine mycobacterial laboratories. The Etest has not, however, been sufficiently evaluated to justify its place as an alternative to the conventional methods (3). We compared the Etest with the proportion method using LJ medium (PMLJ) by specifically testing isoniazid (INH), rifampin (RIF), and streptomycin (STR) against isolates from clinical samples obtained from Brazilian tuberculosis patients.

Eighty-nine clinical isolates of M. tuberculosis were identified and tested for susceptibilities to INH, RIF, and STR by PMLJ at the Mycobacteriology Laboratory of Helio Fraga National Reference Center and Institute Noel Nutels, Rio de Janeiro, Brazil. Thirty-six of the isolates were resistant to one drug, 16 were resistant to two or three drugs, 9 of which were MDR, and 37 were susceptible to all drugs. MDR was defined as resistance to INH and RIF.

Antimicrobial susceptibility testing by PMLJ was performed by inoculating isolates to the surface of LJ medium, into which drugs were incorporated as previously described (1). The isolates were considered resistant if the percentage of colonies on media containing drugs exceeded the growth on a drug-free LJ medium control by 1%. Drugs were tested with the following concentrations: 0.2 μg/ml of INH, 40 μg/ml of RIF, and 4.0 μg/ml of STR.

The Etest was performed as previously described (14). Briefly, isolates were grown for 3 weeks at 37°C on LJ medium, the fresh growth was removed and placed into a tube containing 3-mm-diameter glass beads, and the beads were vortexed for 1 min. Following the addition of 5% phosphate-buffered saline-Tween 80, the large particles were left to settle down and the supernatant turbidity was adjusted to approximate a no. 4 McFarland standard by the addition of 5% phosphate-buffered saline-Tween 80. Fifty milliliters of Middlebrook 7H11 medium supplemented with 0.2% tyloxapol (Sigma, St. Louis, Mo.) and 10% oleic acid-albumin-dextrose-catalase (BBL) was added to petri dishes (diameter, 150 mm) and inoculated by swabbing mycobacterial suspensions onto the agar surface. The plates were sealed with adhesive tape and incubated at 37°C in an atmosphere of 5% CO₂. Etest strips (kindly provided by AB Biodisk, Solna, Sweden) impregnated with gradients of INH, RIF, and STR ranging in concentration from 0.016 to 256 μg/ml were placed on the agar surface, and the plates were resealed and incubated as described above. The MIC was determined by the point where the elliptical zone of growth inhibition intersected the MIC scale on the Etest strip. Organisms were considered resistant if growth was observed at the following concentrations on the Etest strip: 0.2 μg/ml for INH, 1.0 μg/ml for RIF, and 5.0 μg/ml for STR (9). Susceptibility test results were read 5, 7, and 10 days after application of the Etest strip. Isolates for which Etest and PMLJ susceptibility results were discordant were tested further by the BACTEC 460 tuberculosis radiometric system (Becton Dickinson, Sparks, Md.). The critical concentrations of the drugs used for the BACTEC assays were 0.1 μg/ml for INH and 2.0 μg/ml for RIF and STR. The positive and negative predictive values of Etest results were obtained by using the PMLJ results as the standard method (6). Statistical analyses were done using the EpInfo5.0 software for DOS. The level of significance for distribution of concordance between the test results was assessed by McNemar's chi-square test, with a P of ≤0.05 considered significant.

Etest MICs were read 5 days after 5% CO₂ incubation, and the results did not change after additional incubations for 7 or 10 days. However, retesting was required for some isolates for which growth inhibition was difficult to interpret due to insufficient growth. To ensure the reproducibility of the Etest, seven strains were tested two to four times and at different periods. For five out of seven strains, susceptibility profiles were confirmed; in the remaining strains, the MIC variation was higher than that for a twofold dilution.

Although overall concordance between the PMLJ and Etest results ranged from 80 to 97%, STR resistance was correctly identified by the Etest in only 8 out of 23 STR isolates (Table 1). Overall discordance ranged from 3% (for RIF) to 20% (for STR). Among susceptible isolates, there were 173 concordant results compared to three discordant results. Among resistant isolates, there were 58 concordant and 18 discordant results. The Etest showed positive predictive values of 97, 88, and 35%, respectively, for INH, RIF, and STR. The negative predictive value was higher for RIF (97%), followed by INH (96%) and STR (78%).

TABLE 1.

Comparison of PMLJ and Etest results for M. tuberculosis susceptibility testing

Determination by PMLJ	Etest results
	No. of concordant isolates/total no. (%)			No. of discordant isolates/total no. (%)
	INH	RIF	STR	INH	RIF	STR
Susceptible	50/51 (96)	70/72 (97)	53/53 (100)	1/51 (2.0)	2/72 (3.0)	0/53 (0)
Resistant	34/36 (94)	16/17 (94)	8/23 (35)	2/36 (5.5)	1/17 (6.0)	15/23 (65)
Total	84/87 (96)	86/89 (97)	61/76 (80)	3/87 (3.4)	3/89 (3.0)	15/76 (20)

The 21 discordant isolates were evaluated further by the BACTEC system (Table 2). Initial PMLJ results were confirmed among 18 of the 21 discordant isolates compared to three confirmations of Etest results (P = 0.0002). Concordance with PMLJ results, BACTEC results, or both improved, increasing to 8 of 21 isolates when the isolates were retested by the Etest assay. The overall consensus agreement for INH and RIF increased to 100 and 99%, respectively, and to 84% for STR upon retesting by the Etest method.

TABLE 2.

Evaluation of strains for which Etest and PMLJ results were discordant by BACTEC and retesting by Etest

Strain	Drug	Results of test method(s)a
		PMLJ	BACTEC	Etest/MIC (μg/ml)		PMLJ vs BACTEC	Etest vs PMLJ-BACTEC
				Initial	Retested		Initial	Retested
1	INH	S	R	R/1.5	R/1.0	D	C	C
2	INH	R	R	S/0.016	R/1.0	C	D	C
3	INH	R	R	S/0.016	R/4.0	C	D	C
4	RIF	S	S/S	R/256	R/256	C	D	D
5	RIF	R	R	S/0.016	R/256	C	D	C
6	RIF	R	R	S/0.047	R/4	C	D	C
7	STR	R	R	S/0.75	R/256	C	D	C
8	STR	R	R	S/1.5	S/1.5	C	D	D
9	STR	R	S	S/1.5	S/1.5	D	C	C
10	STR	R	R	S/1.5	S/1.5	C	D	D
11	STR	R	R	S/1.0	S/1.0	C	D	D
12	STR	R	R	S/1.5	S/1.5	C	D	D
13	STR	R	R	S/2.0	S/2.0	C	D	D
14	STR	R	R	S/4.0	S/2.0	C	D	D
15	STR	R	R	S/4.0	S/4.0	C	D	D
16	STR	R	R	S/4.0	S/4.0	C	D	D
17	STR	R	R	S/1.5	S/1.5	C	D	D
18	STR	R	R	S/1.0	S/1.0	C	D	D
19	STR	R	S	S/2.0	S/4.0	D	C	C
20	STR	R	R	S/1.0	S/1.0	C	D	D
21	STR	R	R	S/1.0	S/1.0	C	D	D

^aMICs at which organisms were considered resistant on the Etest strip: >0.2 μg/ml for INH, >1.0 μg/ml for RIF, and >5.0 μg/ml for STR. C, concordant; D, discordant; S, susceptible; R, resistant.

The use of the stable-gradient antimycobacterial drug strips has been reported to be useful for rapid M. tuberculosis susceptibility testing mainly by Wanger and colleagues (4, 13, 14) but recently by other authors as well (7, 8, 11, 12). All these reports suggest that the Etest could be a rapid, reproducible, and reliable method to replace the conventional proportion methods, primarily in developing countries. However, Hausdorfer et al. (3) noted that the Etest gave a significant number of false-sensitive and false-resistant results when compared with PMLJ. Our Etest results were obtained under laboratory conditions found typically in mycobacteriology laboratories throughout South America and by using isolates of M. tuberculosis from Brazilian patients, which were submitted for susceptibility testing to INH, RIF, and STR. We nonetheless found good correlations (≥96%) between the Etest and PMLJ for susceptible isolates, which supports the previous findings. It appears likely that the conversely low concordance among resistant isolates is related to difficulties in achieving uniform inocula in terms of the true numbers of bacilli plated prior to the placement of Etest strips and the subsequent misinterpretation of the susceptibility profile. This inability to achieve uniform inoculum sizes may have affected the proportion of resistant bacilli. Growth phase status may have also affected these results, e.g., the proportions of resistant bacilli may differ in log-phase versus stationary-phase cultures used for Etest inoculations. Hazbón et al. (4) also suggest that this event may relate to in vitro selection of a resistant subpopulation during subculturing; however, they did not mention if the subculturing was done in the presence of antibiotic that could promote the selection. Additional unexplained technical problems may have played roles in the two isolates (no. 9 and 19) (Table 2) that changed from resistant to susceptible upon retesting.

Our findings of false-sensitive and false-resistant Etest results do not support the usefulness of this method for routine susceptibility testing of M. tuberculosis isolates in mycobacteriology laboratories, particularly in developing countries faced with much larger workloads of M. tuberculosis isolates and where quality control and standardization of procedures may be more difficult to achieve. Hausdorfer et al. (3) agreed with this conclusion, suggesting that the Etest for M. tuberculosis susceptibility testing could not be recommended for practical use in clinical laboratories because it yielded a substantial rate of false-resistant strains, mainly for ethambutol. In our study, most of the Etest-discordant strains showed false-susceptible profiles, mainly for STR. The availability of reliable CO₂ incubators and 7H11 medium supplement are foremost among the deficiencies in laboratories in developing countries that may adversely affect the heterogeneity and overall health of test inocula.

The Etest may nonetheless prove useful for some studies, particularly those in which mutations and MICs must be correlated, for the evaluation of new drug breakpoints or for the rapid screening of MDR M. tuberculosis isolates.

Further multicenter evaluations of the Etest involving mycobacteriology laboratories in both developing and developed countries may be warranted to clearly establish the usefulness of this test. This type of collaborative study should perhaps also compare the Etest with other novel susceptibility methods such as the MABA, a colorimetric, microplate-based Alamar Blue assay method (2, 10). It has been reported to be simple and easy to perform, and the results may be obtained in 8 days.