Use of Penicillin MICs To Predict In Vitro Activity of Other β-Lactam Antimicrobial Agents against Streptococcus pneumoniae

Abstract

Linear regression analysis was used to compare penicillin MICs determined with 3,129 recent clinical isolates of Streptococcus pneumoniae to MICs obtained with nine other beta-lactam antimicrobial agents. A strong correlation between penicillin MICs and those obtained with other beta-lactams was demonstrated. It may be possible to test penicillin and use MICs obtained with penicillin to predict MICs of other beta-lactam antimicrobials for Streptococcus pneumoniae.

The prevalence of penicillin-resistant Streptococcus pneumoniae continues to increase. It is well documented that the mechanism of resistance to beta-lactam agents is altered penicillin-binding proteins (PBPs). Streptococcus pneumoniae contains six high-molecular-weight PBPs: 1a, 1b, 2a, 2b, 2x, and 3. Alterations in one or more of these proteins result in decreased activity in response to beta-lactam agents. For example, alterations in PBPs 1a, 2b, and 2x result in high-level penicillin resistance, and alterations in PBPs 1a and 2x confer high-level resistance to ceftriaxone and cefotaxime. PBP alterations also result in cross-resistance among the beta-lactam antimicrobial agents, as evidenced by a steady rise in resistance to amoxicillin, amoxicillin-clavulanate, and the cephalosporins as rates of penicillin resistance have increased. In infections caused by pneumococci that are resistant to penicillin, the activity of alternative therapies (beta-lactam as well as non-beta-lactam antimicrobials) is important, leading in turn to the necessity for clinical laboratories to perform susceptibility tests with a variety of antimicrobial agents in addition to penicillin.

Despite some concerns about its reliability, the Etest has become the most popular method for antimicrobial susceptibility testing of S. pneumoniae in the United States (4). Generally, laboratories are limited to testing five antimicrobials against pneumococcal isolates by this method. As the requirement for testing isolates of S. pneumoniae against multiple antimicrobial agents has grown, the utility of the Etest method has diminished.

In view of what is known about the mechanisms of beta-lactam resistance with S. pneumoniae, it was reasoned that tests with penicillin might be used to predict the activity of other beta-lactam agents, thus reducing the total number of antimicrobial agents laboratories must test against pneumococcal isolates. The intent of this investigation was to explore this possibility.

A total of 3,129 S. pneumoniae isolates from two recent multicenter U.S. surveillance studies, one in 1994 to 1995 (2) and the other in 1997 to 1998 (3), were examined in this study. Isolates were collected during the months of November to April from 30 U.S. medical centers in 1994 to 1995 and from 34 U.S. medical centers in 1997 to 1998. Isolates were sent to a central testing laboratory (University of Iowa College of Medicine, Iowa City), where they were frozen at −70°C on porous beads (ProLab Diagnostics, Austin, Tex.). Isolates were subcultured twice prior to susceptibility testing. Broth microdilution trays containing Mueller-Hinton broth plus 3% lysed horse blood were used to determine MICs. The final inoculum concentration was approximately 5 × 10⁵ CFU/ml; trays were incubated at 35°C in ambient air for 24 h before MICs were determined. Streptococcus pneumoniae ATCC 49619 was used as a quality control strain for all MIC determinations (9, 10).

MICs (log₂) obtained with penicillin were compared to those obtained with amoxicillin, amoxicillin-clavulanate, ceftriaxone, cefuroxime, cefpodoxime, cefixime, cefprozil, cefaclor, and loracarbef on a strain-by-strain basis. Simple linear regression and correlation analysis was performed with penicillin as the independent variable and all other antimicrobials as individual dependent variables. To assess the influence of outlying data points, regression analysis was performed a priori with a Student residual value of ≥5.0 and Cook's distance of 0.5 as the cutpoints. All data points above the cutpoints were reviewed for data accuracy. Statistical analyses were performed with SAS 6.12 (SAS Institute, Cary, N.C.); Sigma Plot (SPSS, Inc.) was used for graphical analyses.

Table 1 lists the results of simple linear regression and correlation analysis with penicillin as the independent variable and each comparator antimicrobial as the dependent variable. In Fig. 1, MICs obtained with penicillin are compared to MICs obtained with amoxicillin, cefuroxime, and ceftriaxone for 3,129 strains of S. pneumoniae. The two study collections (1994 to 1995 and 1997 to 1998) were examined individually as well as in combination. No significant differences were noted between studies or with either study compared to the aggregate data; therefore, the graphical data are displayed in the aggregate. Correlation coefficients of the aggregate data ranged from 0.92 to 0.97, indicating a high degree of linear association between the penicillin MICs and corresponding beta-lactam MICs. The slopes of the lines range between 0.85 and 1.10, also suggesting a nearly perfect, positive, linear relationship between penicillin and each comparator beta-lactam. The actual degree of correlation between penicillin MICs and MICs obtained with comparator beta-lactams was greatest with organisms for which the penicillin MICs were higher. This is apparent from the scattergrams (Fig. 1) and was true of all nine comparators.

TABLE 1.

Results of simple linear regression and sample correlation analysis with penicillin as the independent variable

Antimicrobial agent	Result for samples from:
	1994–1995 (n = 1,528)			1997–1998 (n = 1,601)			Aggregate (n = 3,129)
	y Intercept	Slope	r	y Intercept	Slope	r	y Intercept	Slope	r
Amoxicillin	−0.29	0.97	0.96	0.11	0.99	0.98	−0.06	0.99	0.97
Amoxicillin-clavulanate	−0.37	0.94	0.97	0.19	0.99	0.98	−0.05	0.98	0.97
Cefuroxime	1.60	1.08	0.94	1.31	1.13	0.97	1.44	1.10	0.96
Ceftriaxone	−0.47	0.83	0.94	−0.66	0.87	0.97	−0.57	0.85	0.95
Cefpodoxime	0.85	0.94	0.93	1.07	0.99	0.96	0.98	0.97	0.95
Cefixime	3.51	0.92	0.92	0.36	0.98	0.96	3.57	0.95	0.94
Cefprozil	2.50	0.86	0.93	2.04	0.96	0.96	2.24	0.91	0.93
Cefaclor	4.65	0.91	0.91	4.96	1.07	0.94	4.84	0.99	0.92
Loracarbef	4.50	0.78	0.89	5.50	1.07	0.94	5.09	0.95	0.92

FIG. 1.

Results of simple linear regression and correlation analysis with penicillin as the independent variable and amoxicillin, cefuroxime, and ceftriaxone as the dependent variables.

The actual ratio of MICs between penicillin and the nine comparator beta-lactams examined in this study was dependent on the comparator, but in all cases, remained constant across the entire range of penicillin MICs obtained with the study isolates. Specifically, amoxicillin, amoxicillin-clavulanate, and ceftriaxone MICs were essentially the same as MICs obtained with penicillin; the ratios of MICs obtained with other agents to the penicillin MICs were as follows: cefpodoxime and cefuroxime, 2:1; cefprozil, 4:1; cefixime, 16:1; and cefaclor and loracarbef, 32:1.

In this study, we chose to examine the relationship between the activity of penicillin and that of other beta-lactams against S. pneumoniae, with the notion that perhaps the MIC of penicillin could be used to predict the MICs of other beta-lactam agents, thereby reducing the amount of susceptibility testing required. By combining the data from two recent national surveillance studies, we had a large sample of isolates and susceptibility patterns with which to evaluate this relationship. The data suggested a strong correlation between penicillin MICs and the MICs of amoxicillin, amoxicillin-clavulanate, ceftriaxone, cefuroxime, cefpodoxime, cefixime, cefprozil, cefaclor, and loracarbef. Not surprisingly, the highest degree of correlation was seen between penicillin, amoxicillin, and amoxicillin-clavulanate (r = 0.97). In general, identical MICs were obtained with all three of these agents. Only slightly less correlation was noted between penicillin and cephalosporin MICs (r = 0.92 to 0.96). The highest degree of correlation among the cephalosporins tested was between penicillin and cefuroxime (r = 0.96), and the lowest was between penicillin and both cefaclor and loracarbef (r = 0.92).

Susceptibility testing remains an important task of the clinical microbiology laboratory, both for purposes of therapeutic decision making and for continued surveillance of penicillin and beta-lactam resistance. Based on the results of this study, it may be possible for laboratories to test penicillin and use the MICs obtained with this agent to reliably predict the in vitro activity of amoxicillin, amoxicillin-clavulanate, ceftriaxone, cefuroxime, cefpodoxime, cefixime, cefprozil, cefaclor, and loracarbef. With amoxicillin, amoxicillin-clavulanate, and ceftriaxone, the MICs are essentially the same as those obtained with penicillin. Penicillin MICs would be multiplied by 2× to predict the cefpodoxime and cefuroxime MICs. The multipliers for cefprozil, cefixime, cefaclor, and loracarbef would be 4×, 16×, 32×, and 32×, respectively. Alternatively, with the linear regression equation indices presented in Table 1, the MICs of alternative beta-lactams could simply be calculated based on the penicillin MICs that had been determined.