Identification of β-Lactamase-Negative, Ampicillin-Resistant Strains of Haemophilus influenzae with Four Methods and Eight Media

In North America approximately 20 to 40% of all Haemophilus influenzae strains are resistant to ampicillin by virtue of their ability to produce β-lactamase enzymes that inactivate ampicillin and other β-lactams (3, 4, 7, 17,19, 21). Non-β-lactamase-producing strains of H. influenzae are generally inhibited by ≤1.0 μg of ampicillin per ml. However, over the past 20 years there have been sporadic reports of strains for which ampicillin MICs were slightly increased but which had no known β-lactamase production (10, 15, 16). Those strains have been categorized as β-lactamase negative, ampicillin resistant (BLNAR) and are defined as requiring an ampicillin MIC of ≥4.0 μg/ml (5, 14). Strains requiring an MIC of 2.0 μg/ml are defined as intermediate (indeterminate). Because repeated MICs can be expected to vary by a magnitude of plus or minus one doubling concentration, the intermediate category defies definition; i.e., such strains can be either susceptible, intermediate, or resistant if retested on another day or by another method. BLNAR strains have been shown to have diminished susceptibility because of altered penicillin binding sites and, consequently, they are also relatively resistant to other β-lactams, including some cephalosporins (11, 15, 16). There are only a few published reports of clinical failures when patients infected with BLNAR strains were treated with ampicillin (18) or a cephalosporin (12). Clinical data are sparse because BLNAR strains remain uncommon in most surveys (4, 7, 21). Doern et al. (4) reported 1.3% of all β-lactamase-negative strains to be BLNAR (MIC, ≥4.0 μg/ml) and another 2.7% to be ampicillin intermediate. Because the latter strains are not susceptible, they are occasionally included in the BLNAR category.

Standardized methods are essential if we are to create a common language for different investigators to use to communicate. The NCCLS has developed reference methods for testing H. influenzaeagainst appropriate antibiotics (13, 14). Initially, broth dilution tests were done in Mueller-Hinton broth with lysed horse blood and NAD and disk tests utilized Mueller-Hinton agar with hemoglobin and a defined supplement containing X and V factors (chocolate agar). In 1987, Jorgensen et al. (8) developed a clear medium that made it easier to read endpoints. That medium was simply Mueller-Hinton broth or agar with hemin, yeast extract, and NAD, and it was called haemophilus test medium (HTM). In 1990, the NCCLS subcommittee adopted the use of HTM as the reference method because test results have been shown to be comparable to those with media that were used previously (2, 3, 8, 19, 20). However, when commercial medium manufacturers started to provide this medium to clinical laboratories, there were many complaints that clinical isolates often failed to grow. There was significant variability in the performance of different lots of Mueller-Hinton agars when prepared as HTM agar (1). Most manufacturers have now resolved problems of mass production, but there are other inconsistencies that appear after prolonged storage of HTM agar, and strict quality control measures are essential for all users of HTM. Consequently, many microbiologists would prefer to use another, more nutritive medium. For that reason, we undertook a comparative study that evaluated eight different media that have been utilized in studies published over the past decade. This was done as part of a much larger study that was reported separately (6), but here we focus our attention on ampicillin susceptibility tests and consider criteria for separating BLNAR strains from other β-lactamase-negative H. influenzae strains.

A challenge set of 143 β-lactamase-negative H. influenzaeisolates was selected from our collection of clinical isolates that have been gathered from medical centers throughout North America over the past 5 years. About half of the strains were selected because they originally required ampicillin MICs of ≥1.0 μg/ml. For each of those strains, we selected an ampicillin-susceptible strain that was recovered at about the same time within the same geographic region. The original susceptibility test results were not considered in this analysis because of the variable time delay before this study was initiated.

The methods of the NCCLS (13, 14) were utilized to perform broth microdilution, agar dilution, and disk diffusion tests, except that media other than HTM were also evaluated. The agar and broth media that were compared are described in detail in Table1. E-test strips were applied to plates that were inoculated for disk tests, and MICs were read according to the manufacturer's instructions. For the purposes of this study, E-test MICs were rounded up to the next even log₂concentration. All agar plates were incubated in 5 to 7% CO₂, whereas broth microdilution trays were incubated in ambient air at 35°C. Controls demonstrated that the increased CO₂ did not significantly affect the MICs of ampicillin. Two quality control strains of H. influenzae (ATCC 49247 and ATCC 49766) were included throughout this study. The NCCLS broth microdilution method was performed on 41 separate days. For the BLNAR control strain (ATCC 49247), ampicillin MICs ranged from 2.0 μg/ml (intermediate) to 8.0 μg/ml (resistant): 36 of 41 values were 4.0 μg/ml. For the ampicillin-susceptible strain (ATCC 49766) MICs ranged from 0.12 to 0.5 μg/ml (all susceptible).

Table 2 describes the distribution of MICs around the breakpoints for tests of ampicillin on HTM agar or broth. According to microdilution tests with two broth media, 31 to 33% of our selected strains were BLNAR and another 9 to 10% were ampicillin intermediate (40 to 43% were nonsusceptible). However, when agar dilution tests were performed with a more nutritive medium, the number of BLNAR strains increased to include as many as 44% (chocolate agar). As reported by others (9), MICs with the E-test tended to be somewhat lower than those with the agar dilution method. MICs with the E-test were rarely ≥8.0 μg/ml, but such elevated MICs were commonly seen with the antibiotic dilution procedures (Table 2). Interpretive criteria for E-tests of ampicillin against H. influenzae on HTM agar would have to be adjusted by one doubling concentration (MIC, ≤0.5 μg/ml for susceptible and ≥2.0 μg/ml for resistant) in order to achieve parity with NCCLS methods. To the best of our knowledge, the manufacturers of E-tests have not yet addressed that issue.

Disk diffusion susceptibility test results are summarized in Table3. E-tests and disk diffusion tests on HTM agar declared very few strains to be BLNAR but a disproportionate number to be ampicillin intermediate. Others have reported “false-susceptible” ampicillin disk tests when performed on HTM agar (3, 5). On agar media other than HTM, disk tests identified only 20 to 23% of our selected strains to be BLNAR strains. For E-tests on the same agar plates, 24 to 29% were BLNAR. This is in contrast to the 31 to 32% values obtained by broth or agar dilution tests on HTM or the 35 to 44% values obtained by agar dilution tests on media other than HTM.

Within this highly selected challenge set of isolates, BLNAR should be much more common than would be expected among fresh clinical isolates. If only one test was performed on one of the eight media, the percentage of BLNAR strains in our set of isolates would range from 5% (disk tests on HTM agar) to 44% (agar dilution tests on chocolate agar). By including ampicillin-intermediate strains in the BLNAR category, those percentages would range from 32% (disk and E-tests on HTM agar) to 50% (agar dilution on chocolate agar). Media such as chocolate agar support better growth than HTM agar, and that is thought to be an advantage. However, the improved nutritive quality results in higher MICs and smaller zones of inhibition. We are not able to make a valid decision about which method or medium gives the “correct” answer; we can only conclude that the results can be markedly different. It would be interesting to know how well qualitative changes in penicillin binding proteins can be predicted by gradually increasing ampicillin MICs as determined by standard methods. Such determinations may or may not reflect clinically relevant resistance.

The clinical relevance of BLNAR strains can be debated since there are few data to support or refute the assumption that meningeal or nonmeningeal infections will fail to respond to therapy with ampicillin or other β-lactams (12, 18). However, clinical data cannot be generated until there is a standardized reference method and a universal definition of MIC criteria for the BLNAR category. HTM is not without its problems, and its value as a reference medium might be questioned. However, adoption of a richer medium can result in a sudden unwarranted increase in the number of BLNAR strains in clinical specimens. Changes in standardized methods must be made with caution because there are often unexpected consequences. Until further information is available, the current NCCLS methods and interpretive breakpoints should be used when testing ampicillin against H. influenzae.

• Copyright © 2001 American Society for Microbiology