Low Dosage and Long Treatment Duration of β-Lactam: Title and subTitle BreakRisk Factors for Carriage of Penicillin-Resistant Streptococcus pneumoniae FREE

THE SPREAD of drug-resistant Streptococcus pneumoniae in the community poses a major challenge for clinicians and microbiologists. It is now a public health problem in both developing and developed countries.

Streptococcus pneumoniae is a major cause of community-acquired pneumonia, bacteremia, meningitis, and acute otitis media. It accounts for over 1 million deaths each year around the world in children younger than 5 years.¹ The increase in penicillin-resistant S pneumoniae (PR Sp) complicates treatment of these infections^{2 - 3} and risks to increase morbidity and mortality from pneumococcal infection.¹

For some areas in the United States, 25% to 30% of invasive pneumococcal isolates were found to have either intermediate or high-level resistance.⁴ A report from Spain, which has a high prevalence of drug-resistant pneumococci, showed that between 1979 and 1994, 57% of more than 5000 isolates were resistant to 1 or more antibiotics.⁵ For over a decade in France, the frequency of PR Sp isolates in the National Reference Center for Pneumococci increased from 0.5% in 1984 to 32% in 1994.⁶

Many authors have suggested that the dominant factor in the spread of bacterial resistance in the community is the increasing use of antibiotics^{7 - 8} ; however, the chain of causality is not clear. Ecologic correlations have been found between invasive PR Sp or erythromycin-resistant group A streptococci and antibiotic use in the community.^{9 - 10} There is also time concordance between invasive Moraxella catarrhalis and sales of cephalosporins.¹¹ Several studies have shown in vitro that β-lactams may promote PR Sp .^{12 - 13} More recently, a study has shown that antibiotic use increases the carriage rate of PR Sp in children.¹⁴ These studies do not explain how antimicrobial drug use is a risk factor for human colonization by PR Sp.

The lack of adequate community studies has prevented the development of epidemiologic models, which would predict the evolution of antimicrobial-drug resistance. As a result, policies regarding antibiotic treatment of outpatients have been hindered.

High carriage rates of PR Sp in the community have been shown to result in an increased incidence of clinical infection.¹⁵ To test the hypothesis that PR Sp carriage is related to β-lactam use, we conducted a survey for pharyngeal carriage of S pneumoniae and antimicrobial use in children.

We surveyed oropharyngeal carriage of S pneumoniae in healthy 3- to 6-year-old schoolchildren in the Loiret, an administrative department in the center of France with 600000 inhabitants, where more than 99.5% of these children are enrolled at school. This area is similar to the rest of France in terms of sociodemographic characteristics, medical activities of nonhospital practitioners, and Social Security coverage. The choice of the population (3- to 6-year-old children) was based on the possibility to recruit a randomized population sample of children. The survey was conducted over a 5-week period from April 10, 1995, to May 16, 1995.

Assuming a ratio of 50:1 non-PR Sp carriers to PR Sp carriers, and that 2% of non-PR Sp carriers were treated by low doses, our sample size enabled us to detect whether PR Sp carriage was associated with a low dosage of β-lactam with an odds ratio (OR) of 2.0 with α=.05 and power close to 80% with a 2-tailed test.¹⁶

The survey was stratified for rural/urban and socioeconomic status of the school area. Since there were not any schools with low socioeconomic status in a rural area, 3 strata were used. Twenty schools were selected at random, with equal sampling fraction from the following strata: fewer than 5000 inhabitants; at least 5000 inhabitants and low socioeconomic school level; and at least 5000 inhabitants and high socioeconomic school level. The socioeconomic school level was defined according to National Ministry of Education criteria.

Children were included in this study if they were registered for school at the beginning of the year, if the parents gave informed consent, and if the children gave oral consent for the throat screening.

Thirty days before the date planned for the throat screening, we asked the parents to complete a short questionnaire on family size, physician consultation, and diagnosis as stated by the practitioner and drug consumption (including type of drug, duration of treatment and number of daily doses).

This study was approved by the Comité Consultatif pour la Protection des Personnes dans la Recherche Biomédicale of the Kremlin-Bicêtre hospital and the Commission Nationale de l'Informatique et des Libertés.

We used an oropharyngeal bacterial screening procedure because of better acceptability by the children. A swab from the oropharynx was immediately plated onto 5% sheep blood-agar with an optochin disk. The plates were incubated at 37°C in 5% carbon dioxide for 24 hours. Children were weighed at screening.

α-Hemolytic colonies were identified as pneumococci by their typical colonial morphology and susceptibility to optochin, stored frozen at −80°C in brain-heart infusion with 15% glycerol, and then sent to the National Reference Center for Pneumococci. Streptococcus pneumoniae isolates were identified using standard methods.¹⁷ The penicillin G and cefotaxime sodium minimal inhibitory concentrations (MICs) for all S pneumoniae isolates were determined using an agar-dilution method. An inoculum of 10⁴ minus 10⁵ colony-forming units per spot was delivered with a Steers replicator onto Müller-Hinton agar plates (Sanofi Diagnostic Pasteur, Marne la Coquette, France) supplemented with 5% horse blood and appropriate antimicrobial concentration, which were incubated overnight at 37°C with carbon dioxide supplementation. The MIC was defined as the lowest concentration of antimicrobial agent that completely inhibited bacterial growth after incubation for 18 hours. Two S pneumoniae strains were used as controls (Collection Institut Pasteur, Nos. 104485 and 104471). Penicillin resistance was defined as penicillin G MIC of at least 0.1 µg/mL according to the recommendations of the National Committee for Clinical Laboratory Standards.¹⁸ Isolates were serotyped with latex agglutination.

Characteristics of S pneumoniae noncarriers, carriers, and PR Sp carriers were compared using Kolmogorov-Smirnov 2-tailed tests for quantitative variables and the χ² test or Fisher exact test for binary variables.

The daily dose in milligrams per kilogram was calculated for the last-used antimicrobial generic drug and coded as a high or low daily dose as follows. For amoxicillin, the recommended daily dose ranges between 20 and 100 mg/kg according to disease, authors, and countries.^{19 - 21} As many authors recommend an increase in daily dose to 80 to 100 mg/kg in otitis, we considered 50 mg/kg a reasonable cutoff to define low daily dose. For oral cephalosporins, we used the dosing range for children as recommended in the fourth edition of Principles and Practice of Infectious Diseases²⁰ : cefaclor, 40 mg/kg per day; cefatrizine, cefadroxil, and cefuroxime axetil, 30 mg/kg per day; and cefpodoxime proxetil and cefixime, 8 mg/kg per day. For each generic drug, the low daily dose was defined as less than or equal to the above limits. While recent studies have suggested efficacy in community respiratory infectious diseases with an antibiotic given for 5 days or less,^{22 - 25} we strictly considered more than 5 days a long duration of treatment.

Relations between PR Sp carriage and antibiotic consumption were quantified by unadjusted ORs (UORs).²⁶ To take into account a possible association between upper and lower respiratory tract infectious events (otitis, common cold, bronchitis, and pharyngitis) and PR Sp carriage, as well as possible interactions between predictive variables, these variables and their interactions were tested. An analysis was performed excluding children who had an otitis media during the previous 30 days. The Fisher exact test was used for the univariate analysis, and 95% confidence intervals (CIs) for the ORs were calculated by the logit method.²⁷ Multivariate analyses were performed to verify the lack of diminution of UORs when adjusted for factors associated with S pneumoniae carriage. Adjusted ORs (AORs) were calculated using logistic regression models²⁸ for school strata and school prevalence of the pneumococcal carriage as a quantitative variable.

Furthermore, we calculated the percentile of daily dose for each generic antibiotic and defined 2 classes of daily dosage in reference to the median (more or less than the median). The Fisher exact test was used to compare penicillin G susceptibility according to these classes. Calculations were performed using the Statistical Package for the Social Sciences program.²⁹

Of the 1168 eligible subjects, 941 participated. Nonparticipation was due to parents refusing (5.8%), children refusing (2.9%), children absent from the school because of holidays (2.4%), afternoon nap at home (2.4%), chickenpox (1.9%), other reasons (1.1%), and unknown reasons (0.8%). The median age was 4.9 years, and the median body weight was 17.5 kg.

The frequency of pneumococci carriage differed among schools, ranging from 0% to 20% of the children (Table 1). Fifty-five S pneumoniae isolates were identified. One child carried 2 different pneumococcal isolates. Among the S pneumoniae identified, the rate of penicillin resistance was 29% and the rate of cefotaxime sodium resistance was 11% (Table 2).

There was no significant difference in age, weight, sex, number of children in the family, and demographic level of the school between S pneumoniae carriers and noncarriers, or between PR Sp carriers and non-PR Sp carriers (Table 3). The percentage of PR Sp carriers was significantly higher in 2 of the schools (schools 1 and 13) compared with the others (7.7% vs 1.3%, P=.003). These schools were both urban and of high socioeconomic level, but there were no differences in β-lactam use in comparison with the other schools.

Three PR Sp isolates were nontypable and the 13 others fell into group/type 6A, 9V, 14, 15, 15C, 23A, and 24 (Table 4). Furthermore, 40% of the pneumococcal isolates and 44% of PR Sp isolates had a serotype not included in the pneumococcal vaccine.

No medical event could be identified as a risk factor for PR Sp carriage. The use of any antibiotic was not a risk factor for PR Sp carriage.

During the previous 30 days, 161 children used at least 1 β-lactam: aminopenicillin only (n=97), cephalosporin only (n=55), or aminopenicillin and cephalosporin (n=9). The β-lactams used were amoxicillin (with or without clavulanic acid), cefaclor, cefadroxil, cefatrizine, cefuroxime axetil, cefpodoxime proxetil, and cefixime. Six of the 16 children with PR Sp reported β-lactam use. Both β-lactam and, in particular, aminopenicillin use were associated with an increased risk of PR Sp carriage (UOR, 3.0; 95% CI, 1.1-8.3; and UOR, 4.4; 95% CI, 1.5-13; respectively) (Table 5).

The number of daily doses was not found to be associated significantly with a risk for PR Sp carriage. For amoxicillin, the most frequent β-lactam used during the previous 30 days, there were 2 PR Sp among 39 children who used amoxicillin with 2 daily doses or less and 14 PR Sp among 886 children with no amoxicillin or amoxicillin with 3 daily doses or more (number of missing data=16); the difference in PR Sp carriage was 3.3%. However, given the small sample sizes, it is difficult to show a statistically significant difference for the number of daily doses of amoxicillin.

Thirteen children used a β-lactam more than once a month. None of these children was a S pneumoniae carrier.

The last β-lactam used was a low daily dose in 52.2% of the children treated, which is in 8.9% of the children. As compared with no use, low daily doses of the last β-lactam were associated with an increased risk of PR Sp carriage (UOR, 5.9; 95% CI, 2.1-16.7). The association was not diminished when adjusted for strata and the school prevalence of pneumococcal carriage (Table 6) and persisted even after exclusion of children with otitis media in whom PR Sp might have been involved in β-lactam use (UOR, 5.5; 95% CI, 1.8-16.6). In contrast, PR Sp carriage was never identified with high doses of β-lactam.

Furthermore, in S pneumoniae carriers, as compared with no β-lactam treatment, low daily dose of β-lactam was associated with an increased risk of PR Sp carriage (UOR, 4.8; 95% CI, 1.1-20.5).

As compared with no use, low daily doses of the last aminopenicillin or cephalosporin were associated with an increased risk of PR Sp carriage (UOR, 6.7; 95% CI, 2.2-19.9; and UOR, 4.9; 95% CI, 1.1-22.9) (Table 6).

The median daily dose for amoxicillin was 46.2 mg/kg; for cefaclor, 31.8 mg/kg; cefadroxil, 43.5 mg/kg; cefatrizine, 41.6 mg/kg; cefuroxime axetil, 23.4 mg/kg; cefpodoxime proxetil, 8.4 mg/kg; and cefixime, 8.9 mg/kg. Seventy-seven percent of S pneumoniae isolated from children who had not taken a β-lactam had a penicillin G MIC lower than 0.1 µg/mL. All the S pneumoniae isolated from children who had taken a daily dose of β-lactam lower than the median (n=6) had a penicillin G MIC higher than 0.1 µg/mL. In contrast, all S pneumoniae isolated from children who had a daily dose of β-lactam higher than the median (n=8) had a penicillin G MIC lower than 0.1 µg/mL (P=.003). These 2 groups did not have different mean durations of treatment or different mean numbers of daily doses.

The last β-lactam used was a treatment of long duration in 85.7% of treated children, which represents 14.7% of children. A long duration of treatment with β-lactam increased the risk of PR Sp carriage (UOR, 3.5; 95% CI, 1.3-9.8, in comparison with no use) (Table 6).

Children were grouped according to their combined value for daily dose and duration of treatment for aminopenicillin and cephalosporin or the last β-lactam they had taken. The percentage of low dose with a long duration was 46.6%; high dose with a short duration, 39.1%; and missing data on dose, 14.3%. The rate of PR Sp carriage in low dose with a long duration was different than in high dose with a short duration, 8.0% and 0.0%, respectively (P=.03).

We found that a low daily dose and a long treatment duration of an oral β-lactam are risk factors for PR Sp carriage. The number of daily doses was not a risk factor for PR Sp carriage, but the power of this study was low.

The oropharyngeal bacterial sampling procedure is specific but probably weakly sensitive in identifing pneumoccoccus. Thus, the 6% prevalence of S pneumoniae carriage, the 2% prevalence of PR Sp carriers, and the calculated ORs may be underestimates. Nevertheless, this is unlikely to introduce a bias either in the 29.6% rate of penicillin-resistance among carriers or in the conclusion that a low daily dose and a long treatment duration of an oral β-lactam are risk factors for PR Sp.

The parents' information was not validated with physician records. This is unlikely to induce a bias in the association we observed for the following reasons. Questionnaires were distributed 1 month before the throat screening. When parents completed the questionnaire, they did not know whether their child was a PR Sp carrier. Furthermore, neither S pneumoniae carriage nor PR Sp carriage was associated with acute otitis media and the associations persisted even after exclusion of children with otitis media in whom PR Sp might have been involved in β-lactam use.

Associations between the recent use of oral antibiotics and drug-resistant pneumococcal pharyngeal carriage have already been described.^{14 ,30 - 32} The relation between dosage, duration of antibiotic treatment, and spread of PR Sp in the community has been suggested,³³ but has never been proven in clinical studies. Our results support the hypothesis that a low daily dose and a long treatment of β-lactam are major factors contributing to PR Sp carriage. To our knowledge, this is the first such clinical demonstration. As discussed above, the effect of the number of daily doses remains to be clarified in a study with more statistical power.

Carriage of PR Sp is the precondition for interindividual transmission. Two main causes for the spread of PR Sp in the community are antibiotic selective-pressured and interindividual transmission.^{30 ,34 - 35} Several factors could help to explain the antibiotic selective pressure: bacterial antagonism in regulating the bacterial flora of the human pharynx,^{36 - 39} horizontal gene transfer,⁴⁰ or the selection of mutants of S pneumoniae due to low tissue concentration to antibiotics.¹³ Recently, bacteriological findings have suggested that a reduced exposure to antibiotics may not necessarily lead to a significant decrease in the frequency of resistant bacteria,⁴¹ but this conclusion was only based on in vitro observations for Escherichia coli. Pharmacoepidemiologic studies are required to understand how these factors are involved in the spread of PR Sp and to assess the possibility of reversing the increase in PR Sp in the community by improving antimicrobial-drug prescription.

The brief period of this study (a 5-week period in late spring 1995) only provides a snapshot of what is happening with regard to S pneumoniae colonization. The duration of S pneumoniae carriage is likely to be an important factor in interindividual transmission. We can hypothesize that the epidemiologic advantage of PR Sp vs nonresistant S pneumoniae may be related to different durations of carriage according daily doses and durations of treatment with antimicrobials. A longer period of follow-up would be needed if durations of carriage vary with daily doses and duration of β-lactam treatment.

Nearly half of the PR Sp serotypes found in our study are not included in the pneumococcal vaccine currently used in the United States and Europe. Further, as currently used pneumococcal vaccines are effective in preventing pneumococcal infection but not pneumococcal carriage⁴² even if conjugate pneumococcal vaccines may be shown to reduce pneumococcal carriage,⁴³ it is unlikely that these vaccines would limit, in the near future, the spread of PR Sp. Furthermore, it is not possible to limit interindividual transmission of resistant bacteria in the community. To reverse the trend of the spread of PR Sp, there is no choice but to reduce the antibiotic selection pressure.

Modeling the spread of antimicrobial drug resistance is essential to anticipate its evolution in the community. Epidemiologic models of antibiotic resistance have only taken into account exposure to antibiotics through the fraction of the population using antibiotics.⁴⁴ Our results indicate that studies should also take into consideration the daily dosage, the duration of treatment, and, but this remains to be established, the number of daily doses.

In developed countries, most antimicrobials are prescribed to outpatients. A reduction in the use of antibiotics has recently been proposed to forestall the problem of antibiotic resistance.⁴⁵ To control drug-resistant S pneumoniae, community-wide education programs for clinicians and the general public are important. These programs should focus on appropriate use of antibiotics and aim to encourage higher daily doses and shorter treatment with β-lactams.