Amplification of Antimicrobial Resistance in Gut Flora of Patients Treated with Ceftriaxone

ABSTRACT

Although antibacterial therapy has an impact on human intestinal flora and the emergence of resistant bacteria, its role in the amplification of antimicrobial resistance and the quantitative exposure-effect relationship is not clear. An observational prospective study was conducted to determine whether and how ceftriaxone exposure is related to amplification of resistance in non-intensive care unit (non-ICU) patients. Serial stool samples from 122 extended-spectrum β-lactamase-positive (ESBL⁺) hospitalized patients were analyzed by quantitative real-time PCR to quantify the resistant gene bla_CTX-M. Drug exposure was calculated for each patient by using a population pharmacokinetic model. Multi- and univariate regression and classification regression tree (CART) analyses were used to explore relationships between measures of exposure and amplification of bla_CTX-M genes. Amplification of bla_CTX-M was observed in 0% (0/11) of patients with no treatment and 33% (20/61) of patients treated with ceftriaxone. Stepwise regression analysis showed a significant association between amplification of bla_CTX-M and the plasma area under the concentration-time curve from 0 to 24 h for the unbound fraction of the drug (fAUC_0–24), the maximum concentration of drug in serum for the unbound fraction of the drug (fC_max), and the duration of ceftriaxone therapy. Using CART analysis, amplification of bla_CTX-M was observed in 11/16 (69%) patients treated for >14 days and in 9/40 (23%) patients treated for ≤14 days (P = 0.0019). In the latter group, amplification was observed in 5/7 (71%) patients with an fAUC_0–24 of ≥222 mg · h/liter and in 4/33 (12%) patients with lower drug exposures (P = 0.0033). A similar association was found for an fC_max of ≥30 mg/liter (63% versus 13%, P = 0.0079). A significant association was found between the amplification of bla_CTX-M resistance genes and exposure to ceftriaxone. Both duration of treatment and degree of ceftriaxone exposure have a significant impact on the amplification of resistance genes. (The project described in this paper has been registered at ClinicalTrials.gov under identifier NCT01208519.)

INTRODUCTION

The likelihood of acquisition of nosocomial resistant bacteria during or shortly after therapy with different antibiotic agents is still unclear and is often confounded by scarce data on antibiotic usage at the individual patient level (1). Historically, bacteria were susceptible to a wide range of antibiotics (ABx). However, plasmid-mediated multidrug antimicrobial resistance has emerged and become widespread among the Enterobacteriaceae (2). In particular, β-lactamases have become an important mechanism of antibiotic resistance in Enterobacteriaceae, and intestinal carriage of β-lactamase-producing organisms is an important source of transmission (3). In order to understand the contributing factors of the emergence of antimicrobial resistance, investigations at the bacterium, drug, and patient levels are necessary to better elucidate the time-varying and heterogeneous role of antibiotic selection pressure on the emergence and selection of antimicrobial resistance.

While drug exposure has been associated with clinical and bacteriological outcomes in pharmacokinetic (PK)-pharmacodynamic (PD) studies (4) and is used for dose optimization and the setting of clinical breakpoints (5), its role in the development of antimicrobial resistance is under intense investigation (6). Antibiotics, particularly those excreted into the intestinal tract, may promote amplification of resistance if sufficient exposures are not attained (7). As the human colon flora is a diverse ecosystem with many and different bacteria, culture of specific species is not quantitatively related to the total numbers of individuals of each of those species. It is not practical to follow up every different colony form, if cultured at all, in a sample of feces. In addition, there is the strong potential for a shift in gene copy numbers during the culturing process. Therefore, quantitative PCR techniques are important for studying the bacterial and gene compositions and the effect of antibiotics in situ (8). We here studied the effect of antibiotic therapy and exposure on the amplification of resistance among identified carriers of enteric resistant bacteria in a prospective study. A previously developed and validated quantitative real-time PCR (qPCR) (9) was used to quantitate resistance genes in rectal samples of non-intensive care unit (non-ICU) patients colonized with bacteria carrying genes encoding extended-spectrum β-lactamase (ESBL) enzymes and β-lactamases.

RESULTS

Patients.

Of the 122 patients who consented to be included in the study, 55 were from medical wards and 67 were from surgical wards. Noninterpretable or qPCR data from one swab only were obtained for 4 and 5 patients, respectively. For the remaining 113 patients, the numbers of samples [median (range)/25th to 75th percentile] with qPCR results were 5 (2 to 15)/4 to 7 per patient over a period of 15 (6 to 57) days. Only 14 patients were sampled before antibiotic treatment was initiated.

Treatment.

Of the 113 patients available for analysis, 11 did not receive any antimicrobial treatment, whereas 102 patients were treated with one (n = 31) or more (n = 71) antibiotics (Table 1). Of 31 patients who received only one drug, 20 were treated with ceftriaxone, 2 with vancomycin, 2 with ciprofloxacin, and 7 with another 7 different antibiotics. Of 71 patients treated with a combination of drugs, 36 patients received ceftriaxone, 25 patients received ciprofloxacin, 17 received metronidazole, 11 received meropenem, 11 received cefuroxime, 11 received levofloxacin, 10 received vancomycin, and another 29 received different antibiotics (<10 patients per drug). Of 46 treated patients who did not receive ceftriaxone alone or in combination with other drugs, 9 were treated with one drug (2 received vancomycin, 3 received ciprofloxacin, 1 received meropenem, 1 received cefazolin, 1 received amoxicillin, and 1 received oxacillin) and 37 were treated with a combination of drugs, with the most common drug being ciprofloxacin (12 patients), followed by vancomycin (9), levofloxacin (8), meropenem (7), metronidazole (7), and 26 other antibiotics (<5). The details of the 56 patients treated with ceftriaxone (20 as monotherapy and 36 as combination therapy) are shown in Table 1.

TABLE 1.

Details of patients included in the present study with a focus on patients treated with ceftriaxone alone or in combination with other drugs

Patient characteristic	No. of patients
No treatment	11
Treatment	102
E. coli	65
K. pneumoniae	30
Mixeda	2
Other species	5
One drug	31
Ceftriaxone	20
Drug other than ceftriaxone	11
More than one drug	71
Ceftriaxone + another drug	36
Ceftriaxone-treated patients	56
Admitted from home	47
Previous hospitalization	9
Catheters (urinary, i.v.)	12
Underlying disease
Malignancy	56
Cardiovascular disease	29
Diabetes	10
HIV	2
Renal disease	2
Primary diagnosis
Muscoskeletal system	31
Nervous system	14
Endocrine system	7
Respiratory system	2
Digestive system	1
Infectious and parasitic disease	1
Surgical ward/medical ward	42/14
Age (yr), median (range)	63 (25–88)
Wt (kg), median (range)	74 (50–158)
Ht (cm), median (range)	168 (55–187)
Male/female	30/26
ESBL colonization	56
E. coli	40
K. pneumoniae	13
Mixeda	1
Unidentified	2

^aE. coli and K. pneumoniae.

qPCR.

Of 113 patients with more than one interpretable qPCR datum, the bla_CTX-M gene was detected in 97 patients at the initial screening with bla/16S resistance gene ratios [median (range)/25th to 75th percentile] of 3% (0.01% to 206%)/0.1% to 18%; of these 97 patients, 20 were considered to have significant results, with bla/16S ratios of >20%. The bla_CTX-M gene was detected in 110 patients at any time point, with bla/16S ratios of 16% (0.01% to 1263%)/3% to 48% (46 patients had bla/16S ratios of >20%), and the bla_CTX-M gene was detected in 93 patients at the final screening with bla/16S ratios of 4% (0.01% to 1263%)/0.7% to 19% (22 patients had bla/16S ratios of >20%) (Fig. 1).

FIG 1.

Time course of quantitative PCR expressed as a percentage of bla_CTX-M/16S rRNA ratio in patients without any treatment (A), in patients treated with ceftriaxone alone (B) and in combination with other drugs (C), and in patients treated with other drugs (D) at the time of screening (initial), the time of the maximum (max PCR), and the last sample (final PCR). The % bla_CTX-M/16S rRNA ratios are the absolute measurements calculated by dividing bla_CTX-M copy numbers by 16S rRNA copy numbers.

Amplification of resistance genes and antimicrobial therapy.

Amplification of resistance genes compared to initial screening levels in four patient groups is shown in Fig. 2. By definition, no amplification was observed in any of the 11 patients that did not receive any antibiotic (controls). Of the remaining 102 patients treated with antibiotics, amplification of resistance genes was observed in 20/56 (36%) patients treated with ceftriaxone alone or in combination (3/20 patients treated with ceftriaxone alone, 17/36 patients treated with ceftriaxone and another antibiotic) and 10/46 (22%) patients treated with other antibiotics (6/10 treated with ciprofloxacin [n = 5] or levofloxacin [n = 1] together with other antibiotics). Of 30 patients in which amplification of resistance genes was observed, 20 were colonized with ESBL-positive (ESBL⁺) Escherichia coli, 8 with ESBL⁺ Klebsiella pneumoniae, and 2 with other ESBL⁺ species. Of the 72 patients in which no amplification of resistance was found, 45 were colonized with ESBL⁺ E. coli, 21 with ESBL⁺ K. pneumoniae, 2 with both species, and 4 with other ESBL⁺ species.

FIG 2.

Amplification of resistance genes (percent increase calculated as the maximum bla_CTX-M/16S rRNA ratio at any time compared to the ratio at the initial screening) in different patient groups. At the top of the graph, the number of patents with a >20% increase in bla/16S ratios relative to the total number of patients per group is shown for each group. Error bars represent the median and interquartile range. For the CFO comb group, P < 0.05 compared to the No Abx group. The horizontal broken lines represent a relative increase of 20% of bla_CTX-M/16S rRNA ratios over the ratio at the initial screening.

To determine whether an association existed between amplification of resistance genes and antimicrobial therapy or any demographic variable, a stepwise (mixed direction with entering P value of 0.25 and leaving P value of 0.1) multivariate regression analysis was performed with the following nominal (n) and continuous (c) variables: country (n), ward type (n), gender (n), age (c), weight (c), height (c), previous hospitalization (n), previous ABx (n), ABx on admission (n), any ABx during study (n), number of ABx (c), duration of ABx (c), ABx, including ceftriaxone (n), ABx only with ceftriaxone (n), and ABx without ceftriaxone (n). Of these, the only significant associations with amplification of resistance genes found were for the variables ceftriaxone treatment alone (P = 0.010, odds ratio [OR] = 1.8) and ceftriaxone treatment alone or in combination with other drugs (P = 0.015, OR = 2.7).

Ceftriaxone exposure.

The median (range) dose in patients receiving ceftriaxone was 1,000 (1,000 to 4,000) mg given every 12 h (40 patients) or every 24 h (16 patients). Using the individual patient characteristics and the population pharmacokinetic model, the estimated ceftriaxone clearance, fAUC_0–24, fC_max, and fC_min in these patients [median (range)] were 0.89 (0.67 to 1.36) liters/h, 97 (50 to 667) mg · h/liter, 14.8 (5.8 to 134.5) mg/liter, and 3.4 (1.3 to 12.4) mg/liter, respectively. The median (range) percent free drug levels at fC_max and fC_min were 9% (7 to 34%) and 6% (5 to 9%), respectively.

Amplification of resistance genes and ceftriaxone therapy.

To assess the association between amplification of resistance genes and ceftriaxone PK parameters, duration of ceftriaxone therapy, and their interaction, a stepwise multivariate regression analysis was performed for the 56 patients treated with ceftriaxone alone or in combination. A significant association was found between amplification of resistance (used as a categorical variable based on the cutoff of a 20% bla/16 rRNA increase) and fAUC_0–24, duration of therapy, and their interaction (chi square ranging between 3.7 to 7.11 for each of three terms, P = 0.02 to 0.08).

CART analysis confirmed these findings (Table 2). Amplification of bla_CTX-M genes was found in 11/16 (69%) patients treated with ceftriaxone for >14 days and in 9/40 (23%) patients treated with ceftriaxone for ≤14 days (P = 0.0019). In the latter group, a strong association was found for fAUC_0–24: amplification of resistance genes was observed in 5/7 (71%) patients with an fAUC_0–24 of ≥221.9 mg/liter and in 4/33 (12%) patients with lower fAUC_0–24s (P = 0.0033). A significant association was also found with fC_max, with amplification of bla_CTX-M genes observed in 5/8 (63%) patients with an fC_max of ≥29.3 mg/liter and in 4/32 (13%) patients with a lower fC_max (P = 0.0079). The fAUC_0–24 was highly correlated with the fC_max (Pearson correlation coefficient [r] = 0.95, P < 0.0001). No significant associations were found for fC_min or the percentage of a 24-h period that the drug concentration exceeded the MIC (%fT_>MIC).

TABLE 2.

Association between amplification of bla_CTX-M genes and duration of therapy as well as exposure of ceftriaxone therapy

Parameter	No. of patients with amplificationa		% of patients with amplificationa		Fisher exact test P value
	Yes	No	Yes	No
Treatment duration
>14 days	11	5	69	31	0.0019
≤14 days	9	31	23	77
fCmax ≥ 29.3b	5	3	63	37	0.0079
fCmax < 29.3b	4	28	13	87
fAUC0–24 ≥ 221.9b	5	2	71	29	0.0033
fAUC0–24 < 221.9b	4	29	12	88

^a“Yes” for amplification was assigned to patients with a maximum bla_CTX-M/16S rRNA ratio that was >20% higher than the initial bla_CTX-M/16rRNA ratio.

^bThese cutoff values were derived from CART analysis for patients treated with ceftriaxone for ≤14 days. Steady-state fC_max and fAUC_0–24s were estimated using the population pharmacokinetic model of ceftriaxone.

The bla/16S ratios that were plotted over time for selected patients treated with ceftriaxone are shown in Fig. 3. More than 14 days of ceftriaxone therapy was associated with amplification of resistance genes independently of fC_max and fAUC_0–24. For a shorter duration of therapy, amplification of resistance genes was observed at high drug exposures but not at lower drug exposures, for which fAUC_0–24 was more important than fC_max (Fig. 3, lower middle and right graphs).

FIG 3.

Amplification of bla_CTX-M resistance genes in rectal samples over time for six different patients treated with ceftriaxone for >14.3 days (top graphs) and <14.3 days (bottom graphs) attaining low (right graphs), intermediate (middle graphs), and high (left graphs) fC_maxs. The duration of ceftriaxone therapy and the fC_max together with the fAUC_0–24 are shown for each patient. The horizontal line represents the threshold of bla_CTX-M gene amplification observed in nontreated patients.

DISCUSSION

A significant association was found between ceftriaxone exposure and amplification of bla_CTX-M resistance genes in the gastrointestinal (GI) tract. An increased risk of resistance amplification was found when ceftriaxone therapy lasted >14 days independent of drug exposure. In addition, for shorter periods of treatment, the risk of amplification was increased in patients with an fC_max of >30 mg/liter or an fAUC_0–24 of >222 mg · h/liter.

We utilized a quantitative PCR assay to measure bla_CTX-M resistance genes. Bacteriological techniques are not sensitive enough to detect small amounts and changes in subpopulations of resistant bacteria (10). Previous studies showed that rectal swabs are suitable for quantifying the concentration of beta-lactamase producers and that qPCR demonstrated a higher correlation between rectal swabs and stool specimens than the culture-based method. The bla_CTX-M/16S RNA ratio reflects changes in the number of resistant bacteria as a proportion of total bacteria because all the subpopulations can potentially change during therapy. Previous studies found a relationship between the qPCR ratio and surveillance cultures, indicating that this ratio reflects colonization by resistant bacteria. Antibacterial treatment may select for a wide range of cephalosporin resistance genes from different families and microorganisms (11, 12). In the present study, we studied the effect of antibiotic treatment on bla_CTX-M genes normalized by the total bacterial load by using the 16S rRNA genes with a method developed earlier (9). This method was found to give comparable results between rectal swabs and stool specimens by culture methods. Moreover, given that previous findings showed that anaerobic flora plays an important role in emergence of resistant bacteria (13), a PCR method allows quantification of all resistance genes originating from different bacteria, including nonculturable ones that would be missed by culture methods. However, since there was considerable variation in positivity, and there were some very high bla_CTX-M/16 rRNA ratios that were difficult to explain, the analysis was performed with a qualitative parameter of amplification of resistance genes using the cutoff of 20%, which represents the increase in the bla/16S ratio associated with antimicrobial therapy.

Antibiotic therapy has a severe impact on GI flora, and this effect depends on the antibiotic used, its exposure, and the duration of treatment (14). However, although the use of broad-spectrum cephalosporins has been implicated in the emergence of resistance (15, 16), few studies demonstrated a clear association between amplification of resistance and drug exposure. Animal studies in pigs found an increase in the prevalence of resistant E. coli within the first week of treatment with ceftiofur and amoxicillin (11). Similar studies in calves showed a 14% increase in fecal bacteria resistant to ceftriaxone within 3 days after treatment. This response remained stable up to 13 days, and a further increase was found at day 17 (17). Goessens et al. demonstrated a clear relationship between the duration of exposure and the percentage of time within the selection window for ceftazidime in a rat model (18).

Ceftriaxone is a highly protein-bound drug with a high volume of distribution and excellent penetration in fluids (19). Up to 67% of the drug is excreted unchanged in urine, and the remaining is excreted in bile, with average concentrations of 581 to 898 mg/liter found 1 to 3 h after dosing when the concurrent plasma concentration was 62.1 mg/liter (20). In addition, median ceftriaxone concentrations in feces of healthy volunteers treated for 7 days with 2 g of ceftriaxone intravenously (i.v.) were 2.4 mg/liter on day 4 and 161 mg/liter on day 8, indicating accumulation in the GI tract, although a wide variation among individuals was found (0 to 806 mg/liter) (21). When intestinal flora was quantitated in the latter study, the number of enterococci increased whereas the number of E. coli decreased during the ceftriaxone treatment (21). This accumulation is compatible with the strong association that we found in the present study between emergence of resistance and >14 days of ceftriaxone treatment. For a shorter period of treatment, drug exposure plays a more important role in the emergence/amplification of resistance.

As cephalosporin's activity is time dependent, the driving PK-PD index is time above the MIC, and 60 to 70% of the dosing interval is commonly thought to be associated with maximal bactericidal activity in vivo (22, 23). However, no significant association was found between amplification of resistance and the %fT_>MIC in the present study. This is in agreement with previous studies in rats where intestinal colonization by ceftazidime-resistant Enterobacter cloacae isolates was associated with the time within the mutation selection window, AUC/MIC, and fC_max/MIC rather than %fT_>MIC (18). Although concentrations in feces are not the same as those in plasma, it may well be that the location where selection occurs (close to the gut wall) is correlated with plasma concentrations.

There are several limitations in our study. In this study, we explored only the association between ceftriaxone exposure and amplification of bla_CTX-M genes, and therefore the effect on ceftriaxone therapy for other resistant genes is unknown. It may therefore well be that the effect of ceftriaxone therapy on resistance amplification is even stronger than demonstrated. A second limitation is the fact that pharmacokinetic parameters were estimated by using a population pharmacokinetic model based on creatinine clearance calculated by the Cockcroft-Gault equation using patient demographic data rather than by actual measurement, as this was not feasible (24). Although this model well described the serum concentrations of ceftriaxone, it may not have captured the entire variation within our patient population and it does not predict ceftriaxone concentrations in the gut. However, concerning the latter, preclinical studies also showed correlations with plasma levels rather than concentrations in the gut (18). Therefore, even though the quantitative association may be different in the gut, the association does exist, either directly or indirectly. Finally, the study was a prospective observational cohort study comparing the effects of various treatments of ceftriaxone, rather than a randomized controlled trial; this may have resulted in both bias and confounding. The analysis of specifically ceftriaxone exposure presented here rather than that of a specific antibiotic was not defined in advance, since the study ran in several centers and it was not known beforehand for which antibiotics the analysis could be performed. Nevertheless, the data to be collected were defined in advance, and there was an anticipation that only few drugs would be taken by a sufficient number of patients to perform the analysis here. Ceftriaxone was the only drug that was taken by a sufficient number of patients to allow for the present evaluation. Even with the relatively low number of patients included in this analysis, we were able to show a significant result, indicating that the effects may be quite strong. Nevertheless, the association should be confirmed in a prospective randomized trial.

Amplification of a particular known resistance gene by ceftriaxone therapy may have important clinical implications. Amplification of this gene may indicate amplification of the resistant bacteria over susceptible ones or transmission of this gene across different species. Heavily colonized patients are at risk of developing difficult-to-treat bloodstream infections. Thus, identifying patients with high ceftriaxone exposure and therefore at risk of developing such infections may promote rigorous surveillance studies and high awareness on the part of the clinicians. Interventional strategies could be adopted by optimizing ceftriaxone dosing regimens in order to minimize the risk of amplification of bla_CTX-M genes. Thus, in addition to optimizing therapy in terms of efficacy, exposure could or should be optimized with regard to resistance gene amplification.

In conclusion, amplification of bla_CTX-M resistance genes was observed during ceftriaxone therapy, with a strong association found with the duration of therapy and drug exposure. Long treatment and high drug exposures increased the risk of amplification.

MATERIALS AND METHODS

Study.

An observational prospective study was conducted in three hospitals, one each in Italy, Serbia, and Romania, with a high prevalence of ESBL at hospital admission, as part of the FP7 European Union project SATURN (NO241796; ClinicalTrials.gov identifier NCT01208519). The project aimed to identify the prevalence and molecular background of ESBLs, the rate of acquisition, the duration of colonization, the emergence of resistance by way of appearance of β-lactamase genes not present at admission, the amplification of resistance, and whether antimicrobial exposure could be related to emergence of resistance. All adult patients admitted to medical and surgical wards were screened for ESBLs using standard culture-based surveillance techniques (25). Overall, 10,035 patients were screened at hospital admission and up to four times within the first month of hospitalization for ESBL-producing Enterobacteriaceae carriage in rectal swabs. Of 1,102 patients screened positive at admission, 122 patients consented to be included in the longitudinal study in which serial (every 2 to 3 days) rectal samples were collected and analyzed using qPCR for the detection of resistance genes. Before the start of the study, specific forms were developed for the collection of demographic and other variables to allow population pharmacokinetic models to be used to predict patient-specific exposures. More details of the study can be found in reference 25.

Sampling method.

At the sampling point, the swab was taken and put into ESwab medium and transported to the local laboratory. After arrival at the local laboratory, the tubes were vortexed for 1 min. The ESwab medium suspension (∼600 μl) was transferred into new tubes and mixed with an equal volume of 50% glycerol solution. The samples were stored at −80°C until analysis. Prior to the DNA extraction, the total volume of each sample was centrifuged (16,000 × g, 10 min), and the supernatant was discarded. The remaining pellet was subjected to total DNA extraction with a QIAamp DNA stool minikit (Qiagen). Four microliters of such DNA was used as a template in qPCR analysis in duplicates.

qPCR assay.

A quantitative real-time PCR was performed as a measure for resistance as described previously (9). Briefly, two singleplex assays were performed, one for assessing the total bacteria using the 16S rRNA gene primer set 16S_E939F (GAATTGACGGGGGCCCGCACAAG) and 16S_1492R (TACGGYTACCTTGTTACGACTT) (length, 597 bp) and the other for assessing colonic CTX-M producers with bla_CTX-M primers CTX-M-A6 (TGGTRAYRTGGMTBAARGGCA) and CTX-M-A8 (TGGGTRAARTARGTSACCAGAA) (length, 175 bp) (26–28). Samples were included in the analysis only if (i) the bla_CTX-M copy numbers were ≥1 and the cycle threshold (CT) repeat1 minus CTrepeat2 values were ≤1, (ii) the 16S rRNA gene copy numbers were ≥1,000 and the CTrepeat1 minus CTrepeat2 values were ≤1, and (iii) the bla_CTX-M copy numbers were ≤1 and the counterpart number of 16S rRNA gene copies was ≥1,000. Samples were considered noninterpretable if the 16S rRNA gene copy numbers were ≤1,000 and bla_CTX-M and/or 16S rRNA gene samples had only one repeat. The results were expressed as the percentage of the number of copies of the bla_CTX-M gene relative to the number of copies of 16S rRNA genes (100% × bla_CTX-M/16S rRNA genes, bla/16S ratio), as previously described (9). A linear correlation between bla/16S ratios and the ratio of CFU/ml of resistant bacteria to total culturable aerobic bacteria was previously found in stool samples and rectal swabs (9). The bla/16S ratio was used in order to overcome the problem of unknown quantities in fecal material among different samples (9).

Exposure parameters.

Since plasma sampling was not feasible in this study, a population pharmacokinetic model was used to estimate exposure parameters for individual patients, using their demographic characteristics and other relevant parameters such as creatinine clearance. The plasma PK parameters fAUC_0–24, fC_max, and fC_min based on notional patient-specific protein binding were calculated for each patient using KinFun1.07 software (Maastricht, Netherlands) and a previously published population two-compartment pharmacokinetic model (24). The input parameters for the KinFun software were V_c of 10.1 liters, k₁₂ of 0.51 h⁻¹, and k₂₁ of 0.71 h⁻¹, whereas the k₁₀ was individualized based on the estimated ceftriaxone clearance of each patient determined as CL = 0.56 + [0.32 × (CL_CR/4.26)], where CL_CR was the creatinine clearance calculated with the Cockcroft-Gault equation based on gender, age, weight, and serum creatinine (29). In order to calculate the free fraction, the following equation was used: fC = 0.5 × {−(nP + 1/K_aff − tC) + √[(517 + 1/K_aff − tC)^2 + 4tC/K_aff]}, where fC and tC are the free and total concentrations in micromolar, nP is the total concentration of protein binding sites of 517 μM, and K_aff is the binding affinity constant of 0.0367 μM.

Significant amplification and analysis.

To determine the threshold for significant amplification of resistance, a CART analysis was performed on the amplification results for treated patients and controls. None of the controls had a >20% absolute increase in bla/16S ratios, indicating that this level represents an increase in bla/16S ratio not associated with antimicrobial therapy. In subsequent analyses, patients were categorized as those with an increase in bla/16S ratios of ≤20% or >20% at any time point in comparison to that at the initial screening. A univariate analysis (Fisher's exact test) was used to test whether amplification of resistance genes was associated with antibiotic therapy as such, and a multivariate stepwise logistic regression analysis was used to associate different variables with amplification of resistance (JMP10; SAS Institute, Inc., Cary, NC). In order to assess the significance of time above the MIC in emergence of resistance, %fT_>MIC was calculated by using the resistance breakpoint of Enterobacteriaceae of 4 mg/liter. CART analysis was performed to determine relationships between the highest increase in resistance genes and the PK parameters or duration of therapy until the time that the highest increase in resistance genes was observed (JMP10; SAS Institute, Inc., Cary, NC). The significance between groups with PK parameters lower or higher that the cutoff PK parameter determined with CART analysis was analyzed with Fisher's exact test (GraphPad Software 4.0; GraphPad, San Diego, CA).