Evaluation of the Auxacolor System for Biochemical Identification of Medically Important Yeasts

Abstract

We compared the Auxacolor yeast identification system (Sanofi Diagnostics Pasteur) with the API 20C Aux (bioMerieux-Vitek) using 105 isolates of medically important yeasts. The Auxacolor system provided more rapid, accurate results and displayed less interobserver variability than the API 20C Aux.

There has been a marked increase in the incidence of systemic fungal disease caused by pathogenic yeasts (5). Further, the increasing role of non-Candida albicans species, some of which are intrinsically or potentially resistant to antifungal agents, has made rapid species-level identification important for optimizing therapy of acutely ill patients (3, 6). Conventional assimilation techniques for the identification of pathogenic yeasts are slow and impractical for routine use and have led to a reliance on commercial micromethod systems. The API 20C Aux (API 20C), the most widely used of these systems, has been shown to provide reliable, accurate identification of most yeast species (4). Although the API 20C represents a significant improvement over traditional identification techniques, it requires 48 to 72 h of incubation before reading, and it can be difficult to interpret (1, 4).

The Auxacolor system (AUX) is a commercial yeast identification kit which uses colorimetric tests for conventional assimilation substrates, actidione resistance, and phenoloxidase production. Preliminary studies have demonstrated accuracy comparable with that of the API 20C (1, 2) but have not evaluated the speed of identification or interobserver reliability. Because these factors are crucial in the evaluation of a new identification strategy for use in the clinical microbiology laboratory, we undertook a comparison of the AUX and the API 20C with respect to these criteria.

Methods.

One hundred five strains comprising 16 species of yeast were examined (Table 1). Sixty-eight specimens were sterile-site isolates obtained from inpatients at the Royal Victoria Hospital, and the remaining 37 were clinical isolates obtained from a reference collection at the Laboratoire de Santé Publique du Québec. All isolates had been identified by traditional assimilation, biochemical, and morphology tests. Isolates were subcultured to Sabouraud dextrose plates, incubated at 30°C for 48 h, and inoculated into both test systems. API 20C strips were inoculated according to the manufacturer’s instructions, incubated at 30°C, and scored by two independent blinded observers at 24, 48, and 72 h. Final identification was made with reference to the printed and telephone databases. The AUX microwell plate was inoculated according to the manufacturer’s instructions by using 2 drops of the test strain suspended in the medium supplied. This was then incubated at 30°C and read in the same manner as the API 20C. Color changes were noted, and a code was developed; this was compared to the manufacturer’s database in order to generate an identification. Since both the AUX and the API 20C require morphological examination to complete coding, for either system we considered any identification that yielded the correct organism and a group of others differentiated only by morphological criteria to be a correct result. Rates of correct identification and of interobserver agreement were calculated and compared by the chi-square test. A kappa statistic was developed for each system to evaluate the reliability of interpretation between the primary and secondary observers.

TABLE 1.

Number of isolates tested and correctly identified at 24, 48, and 72 h by the AUX and the API 20C

Species	No. of strains	No. of strains correctly identified at:
		24 h		48 h		72 h
		AUX	APIa	AUX	API	AUX	API
Candida albicans	21	21	0	21	14	21	21
Candida glabrata	14	14	0	14	10	14	14
Candida krusei	11	11	5	11	11	11	11
Candida lusitaniae	10	1	0	7	0	10	3
Candida parapsilosis	11	4	0	10	6	11	10
Cryptococcus neoformans	11	2	0	10	4	11	10
Candida tropicalis	10	8	1	8	6	8	7
Cryptococcus albidus	4	2	1	3	4	3	4
Saccharomyces cerevisiae	4	1	0	2	2	2	2
Candida guilliermondii	2	0	1	2	2	2	2
Candida famata	2	0	0	0	0	0	0
Candida inconspicua	1	1	0	1	0	1	0
Candida utilis	1	0	0	0	0	0	0
Rhodotorula rubra	1	1	0	1	0	1	0
Candida lipolytica	1	1	0	1	1	1	1
Trichosporon cutaneum	1	0	0	0	0	0	0
Total	105	67	8	91	59	96	84

^aAPI, API 20C. 

Results.

Significantly more strains were identified by the AUX than by the API 20C at 24, 48, and 72 h (Table 1). This effect was consistent across species with the exception of Candida guilliermondii, for which the API 20C identified one of two strains at 24 h compared with neither of two strains by the AUX. At 48 h each system identified both isolates correctly. There were nine errors in identification with the AUX (Table 2). Of these nine errors, four were misidentifications and five strains were not identifiable by using the manufacturer’s database. While the AUX incorrectly identified two isolates each of Candida tropicalis, Candida famata, and Saccharomyces cerevisiae, only one of these six was correctly identified by the API 20C. We found no single reaction that was consistently inaccurate in these specimens.

TABLE 2.

Isolates not identified or incorrectly identified by the AUX

Organism	Identification by:		Species in AUX database
	AUX	API 20C
C. albidus	T. cutaneum	C. albidus	Yes
S. cerevisiae	None	None	Yes
S. cerevisiae	None	None	Yes
C. tropicalis	None	C. tropicalis	Yes
C. famata	C. tropicalis	T. cutaneum	Yes
T. cutaneum	C. tropicalis	C. tropicalis	Yes
C. famata	None	None	Yes
C. utilis	C. albidus	None	No
C. tropicalis	None	None	Yes

Both observers noted that interpretation of the individual reactions was more easily performed with the AUX. The rate of interobserver agreement was significantly higher with the AUX (294 of 315 interpretations) than with the API 20C (265 of 315 interpretations) (kappa statistic, 0.82 versus 0.70; P < 0.01).

Discussion.

The ideal commercial identification system must be cheap, be easy to use, provide rapid accurate results, and have high interobserver reliability. The cost of the AUX is comparable to that of the API 20C, growth requirements prior to inoculation are identical, and the inoculum preparation for the AUX is less complicated, with only a single dilution.

The overall rate of correct identification of 91.4% found in this study for the AUX was in line with previously documented rates of 98.3 and 85.7% (1, 2) and was superior to that of the API 20C. This superiority was preserved in the species-by-species analysis with the exception of Cryptococcus albidus (four of four isolates correctly identified by the API 20C versus three of four with the AUX). In particular, a marked superiority in the identification of Candida lusitaniae, a species often misidentified by the API 20C and other systems (4), was noted.

Rapidity of identification was significantly higher with the AUX. This was most true for C. albicans, Candida glabrata, and Candida krusei; all strains of these three species were identified by 24 h. This is of particular clinical relevance because it allows for detection of the two most common azole-resistant organisms within 24 h. A reduction of 24 to 48 h in turnaround time for these organisms can be translated directly into a more rapid transition from empirical to directed antifungal therapy.

There was a significant reduction in interobserver variability of interpretation with the AUX, likely due to the advantage of reading color changes rather than turbidity. In clinical practice, where multiple technicians are often responsible for a given subspecialty bench on a rotating basis, this is useful in ensuring consistent interpretation and reporting of results.

While the traditional rapid-screening tests, such as germ tube and rapid trehalose, remain the initial procedures of choice for identification of clinical yeast specimens, many species are unidentifiable by these tests. Our findings suggest that the AUX may be a useful tool for the biochemical identification of such isolates.