Skip Navigation


American Journal of Epidemiology Advance Access originally published online on November 30, 2005
American Journal of Epidemiology 2006 163(2):160-170; doi:10.1093/aje/kwj021
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
163/2/160    most recent
kwj021v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (4)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Samore, M. H.
Right arrow Articles by Sande, M. A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Samore, M. H.
Right arrow Articles by Sande, M. A.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

American Journal of Epidemiology Copyright © 2005 by the Johns Hopkins Bloomberg School of Public Health All rights reserved; printed in U.S.A.

Original Contribution

Mechanisms by Which Antibiotics Promote Dissemination of Resistant Pneumococci in Human Populations

Matthew H. Samore1, Marc Lipsitch2,3, Stephen C. Alder1, Bassam Haddadin1, Greg Stoddard1, Jacquelyn Williamson1, Katherine Sebastian1, Karen Carroll4, Onder Ergonul2,3, Yehuda Carmeli1 and Merle A. Sande5

1 Division of Clinical Epidemiology, Department of Internal Medicine, School of Medicine, University of Utah, Salt Lake City, UT
2 Department of Epidemiology, Harvard School of Public Health, Boston, MA
3 Department of Immunology and Infectious Diseases, Harvard School of Public Health, Boston, MA
4 Department of Pathology, Johns Hopkins School of Medicine, Baltimore, MD
5 Department of Medicine, University of Washington School of Medicine, Seattle, WA

Correspondence to Dr. Matthew Samore, Department of Clinical Epidemiology, School of Medicine, University of Utah, 50 North Medical Drive, Room AC230A, Salt Lake City, UT 84132 (e-mail: matthew.samore{at}hsc.utah.edu).

Received for publication January 17, 2005. Accepted for publication August 30, 2005.


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 APPENDIX: Mathematical Models
 References
 
Mechanisms by which antimicrobials contribute to dissemination of pneumococcal resistance are incompletely characterized. A serial cross-sectional study of nasopharyngeal pneumococcal carriage in healthy, home-living children ≤6 years of age was conducted in four rural communities—two in Utah (1998–2003) and two in Idaho (2002–2003). Prevalence odds ratios for carriage of resistant pneumococci (ORres) and of susceptible pneumococci (ORsus) were estimated. Dynamic transmission models were developed to facilitate a mechanistic interpretation of ORres and ORsus and to compare the population impact of distinct antimicrobial classes. A total of 5,667 cultures were obtained; 25% of the cultures were positive, and 29% of isolates exhibited reduced susceptibility to penicillin. The adjusted ORres for recent individual and sibling cephalosporin use was 2.2 (95% confidence interval: 1.4, 3.4) and 1.8 (95% confidence interval: 1.0, 3.3), respectively. Neither individual nor sibling penicillin use was associated with increased ORres. Rather, recent use of penicillins was associated with decreased carriage of susceptible pneumococci (ORsus = 0.2, 95% confidence interval: 0.1, 0.3). In simulations, both types of effects promoted dissemination of resistant pneumococci at the population level. Findings show that oral cephalosporins enhance the risk of acquiring resistant pneumococci. Penicillins accelerate clearance of susceptible strains. The effect of penicillins in increasing resistance is shared equally by treated and untreated members of the population.

cephalosporins; drug resistance, microbial; nasopharyngeal diseases; penicillins; Streptococcus pneumoniae

Abbreviations: CI, confidence interval; ORC, odds ratio for carriage of resistant pneumococci using carriers of susceptible pneumococci as the comparison group; ORres, odds ratio for carriage of resistant pneumococci; ORsus, odds ratio for carriage of susceptible pneumococci


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 APPENDIX: Mathematical Models
 References
 
Antimicrobial resistance in Streptococcus pneumoniae became widely recognized as a significant public health threat following dissemination of beta-lactam–resistant pneumococcal clones from continent to continent during the 1980s and 1990s (1Go–11Go). Initiatives to promote judicious use of antibiotics were developed to attempt to reverse this trend (12Go, 13Go). Although it is indisputable that antibiotics select for antimicrobial-resistant pneumococci, it is still uncertain how much resistance changes when antimicrobial use increases or decreases or is altered in other ways (14Go, 15Go). The mechanisms by which antimicrobial agents contribute to the spread of resistance in human populations are incompletely understood.

We conducted a study to compare the effects of individual antimicrobial drug classes on carriage of S. pneumoniae. Two broad goals motivated this study. The first objective was to test the hypothesis that the relative risks for carriage of susceptible and resistant pneumococci would vary across drug classes as a consequence of differences in antimicrobial activities. The second was to assess the impact of distinct antimicrobial classes on dissemination of resistant S. pneumoniae at a population level.

This study was conducted in three parts. First, using data prospectively collected from a large population of healthy children in rural communities and statistical methods that accounted for household clustering, age, and other factors, we estimated the odds ratios for carriage of resistant and susceptible S. pneumoniae for three classes of antimicrobial agents. Second, using mathematical models, we related the values of the odds ratios to specific antimicrobial effects on acquisition and clearance of S. pneumoniae. Third, also using models, we examined the relation between these antimicrobial effects and the community spread of pneumococcal resistance.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 APPENDIX: Mathematical Models
 References
 
Study population
Healthy, home-living, rural community residents were recruited to participate in an annual survey of nasopharyngeal S. pneumoniae carriage in children between the ages of 6 months and 6 years. The annual surveys were performed during the period 1998–2004 in two Utah communities and from 2002 to 2004 in two Idaho communities. The results presented here encompass the period 1998–2003. The size of the community populations studied ranged from 5,000 to 35,000. Techniques to solicit household participation in the project included media coverage in local newspapers and distribution of informational flyers. Field teams performed the cultures either in the homes of participating families or in recruiting stations set up in community shopping centers and pharmacies. Cultures were collected between February and May of each year.

The University of Utah Institutional Review Board granted approval for the study. Written informed consent was obtained from all participants.

Microbiologic procedures
Nasopharyngeal cultures were collected as previously described (16Go). The field teams consisted of health-care personnel who were trained in the technique of obtaining nasopharyngeal cultures. The swab was inserted into the child's nasopharynx, rotated gently, and removed. The swabs were then placed immediately into the accompanying transport media. At the end of each collection day, the swabs were transported by courier to a central laboratory (Associated and Regional University Pathologists Laboratories, Salt Lake City, Utah) for processing.

All swabs were processed as previously described (16Go). Briefly, swabs were inoculated onto 5 percent Columbia sheep blood agar plate with and without gentamicin. Isolates identified as S. pneumoniae were tested for susceptibility by using the E-test (AB Biodisk, Solna, Sweden). Results of testing to the following agents are described here: penicillin, ceftriaxone, and erythromycin. Categorical interpretations for the minimal inhibitory concentration values were assigned by using Clinical and Laboratory Standards Institute (formerly NCCLS) break points (17Go). Pneumococci with intermediate susceptibility to penicillin (minimal inhibitory concentration: 0.12–1.0 µg/ml) were classified as beta-lactam resistant for purposes of epidemiologic analysis; other minimal inhibitory concentration break points are given in appendix table 1.

Data collection
A questionnaire was administered to parents to collect demographic and exposure data. Information about antimicrobial prescriptions during the 6-month period prior to culture was collected directly from community retail pharmacies. Antimicrobial prescriptions were classified by time interval according to the date on which the prescription was filled and the date of culture. When the duration of the prescription encompassed the culture date, the culture was classified as being obtained during antimicrobial treatment. Child care was defined as regular attendance in a group child-care setting outside of the home or school for an average of 2 or more hours a day, Monday through Friday. Data on heptavalent conjugate vaccine use were collected from the medical records of individual children and from community health centers where vaccines were administered.

Data and statistical analysis
Measures of relative risk.
Carriage of S. pneumoniae, divided on the basis of antimicrobial susceptibility, constituted the study endpoint. Risk factors of primary interest were antimicrobial treatments. The prevalence odds ratio was calculated as the odds of carriage for those exposed (i.e., antimicrobial-treated individuals) divided by the odds of carriage for those unexposed (i.e., nontreated individuals). Three types of prevalence odds ratios were estimated according to whether children carried resistant or susceptible S. pneumoniae. The first type of prevalence odds, ORres, was computed by dividing the odds of carriage of resistant pneumococci for exposed individuals by the odds of such carriage for unexposed individuals (note that this type of odds ratio was labeled ORS in Lipsitch (18Go)). The second type, ORsus, represented the odds of carriage of susceptible pneumococci for exposed individuals divided by the odds of such carriage for nonexposed individuals. The third type, ORC, was calculated as the odds of exposure for resistant carriers divided by the odds of exposure for susceptible carriers (18Go). The ratio of ORres to ORsus approximates ORC. When ORC for an antimicrobial drug is greater than one, it means that, given pneumococcal carriage, the organism is more likely to be resistant if the individual was exposed to the antimicrobial.

Regression models.
Multilevel logistic regression was used to account for clustering by household and to estimate ORres, ORsus, and ORC adjusted for age, household size, community, and year (19Go–22Go). Variables were selected for inclusion in the final model on the basis of the Wald test statistic (p < 0.05) or a change in effect estimate of more than 15 percent for other factors or because they were judged a priori to be necessary for adjustment. First-order penalized quasi-likelihood was used as the estimation procedure. Multilevel regression models were fit by using MLwiN version 2.0 software (Centre for Multilevel Modelling, University of Bristol, Bristol, United Kingdom). Other statistical analyses were performed by using Stata version 8.0 software (Stata Corporation, College Station, Texas).

Dynamic transmission models
Deterministic model.
A compartmental model was developed to provide a framework for mechanistic interpretation of the measures of relative risk listed above (18Go, 23Go, 24Go) (Appendix). Six states were formulated according to 1) pneumococcal carriage status (carriage of resistant or susceptible pneumococci and noncarriage); and 2) antimicrobial use (whether or not using an antimicrobial agent) (25Go). The model posits that noncarriers acquire resistant or susceptible pneumococci following contact with carriers of resistant pneumococci or susceptible pneumococci, respectively. During the period of treatment, antimicrobial agents differentially affect acquisition of resistant and susceptible pneumococci. Antimicrobial treatment is presumed to accelerate loss of carriage of susceptible organisms to a greater extent than resistant organisms.

Stochastic model.
An individual-agent (microsimulation) version of this compartmental model was also constructed. This method treated transitions between states as discrete events that occurred probabilistically rather than deterministically. A benefit of the individual-agent model was that it made it easier to incorporate nonexponential distributions for parameters such as duration of antimicrobial courses. Another advantage was that it accommodated the modeling of a posttreatment effect of antimicrobial therapy.

Parameters were set according to observed or published data where possible (26Go). In the base model, the population was constituted to comprise a single pediatric age range, with prevalence of carriage and antimicrobial use comparable to that among young children. Thus, the average prevalence of antimicrobial use was set at 4 percent, and the initial prevalence of susceptible and resistant carriage was assumed to be 18 percent and 7 percent, respectively.

The population attributable fraction was calculated from simulation models by comparing the total number of resistant acquisitions generated during a 6-month period run under different conditions of antimicrobial use. The population attributable fraction for penicillins was calculated as the number of resistant infections under base conditions (antimicrobial use set at the observed ratio of three times more penicillins than cephalosporins) minus the number of infections when penicillin use was zero, divided by the number of resistant infections under base conditions. The population attributable fraction for cephalosporins was calculated similarly, by comparing the number of resistant infections under base conditions with the setting of zero cephalosporin use.

Five thousand runs of the simulation model were generated for each variation in conditions or parameters. Further details about the construction of the individual-agent model are included in the Appendix (27Go).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 APPENDIX: Mathematical Models
 References
 
Description of the population
A total of 5,667 cultures were obtained during the 6-year period from 1998 to 2003, representing 3,656 unique children and 2,123 unique households. Sixty-five percent of the children participated once, 22 percent twice, 7 percent three times, 4 percent four times, 1 percent five times, and 0.4 percent six times.

Overall, 1,392 (25 percent) of the cultures were positive for S. pneumoniae, of which 409 (29 percent) exhibited reduced susceptibility to penicillin, 263 (19 percent) resistance to erythromycin, and 54 (4 percent) resistance to ceftriaxone. Ten percent of the isolates had a minimal inhibitory concentration to penicillin of ≥2. Eighty-seven percent of erythromycin-resistant isolates were also intermediate or resistant to penicillin. In 56 percent of instances, isolates exhibiting reduced susceptibility to penicillin were also resistant to erythromycin.

The most commonly used antimicrobial drugs were penicillins (58 percent of antimicrobial courses), followed by cephalosporins (19 percent) and macrolides (19 percent) (refer to appendix table 2 for details). The mean duration of an antimicrobial prescription was 9 days. Eighty-seven percent of cephalosporin and penicillin prescriptions were for 10 days. During the months January–April, an average of 4 percent of children were using antimicrobials on any given day.

Factors associated with resistant and/or susceptible carriage
Cephalosporins, but not penicillins, were associated with increased beta-lactam–resistant carriage.
The prevalence of carriage of beta-lactam–resistant S. pneumoniae was 15 percent (30 of 195) among children who had used cephalosporins within 30 days, 7.5 percent (36 of 478) among children who had used penicillins within 30 days, and 6.8 percent (333 of 4,867) among children who had not used an antimicrobial within 30 days. Thus, children recently treated with cephalosporins were 2.2-fold more likely than children who had not recently used cephalosporins to be beta-lactam–resistant carriers, whereas the frequency of carriage of resistant pneumococci among children who recently used penicillins was similar to that for non-penicillin-treated children. The corresponding ORres adjusted for household clustering, age, year, and community and other factors were 2.2 and 0.9, respectively (table 1). The results were similar regardless of whether intermediately susceptible pneumococci were classified as resistant. The cephalosporin and penicillin ORres for carriage of fully resistant S. pneumoniae were 2.2 and 1.0, respectively.


View this table:
[in this window]
[in a new window]
  		TABLE 1. Adjusted prevalence odds ratios, from multilevel logistic regression, for resistant or susceptible carriage of Streptococcus pneumoniae in children in two rural communities in Utah (1998–2003) and two in Idaho (2002–2003)

 
Penicillins were strongly associated with reduced carriage of beta-lactam–susceptible pneumococci, much more so than cephalosporins.
The prevalence of carriage of beta-lactam–susceptible pneumococci was 4.2 percent (20 of 478) among recent (≤30 days) penicillin recipients, 12 percent (24 of 194) among recent cephalosporin recipients, and 19 percent (932 of 4,867) among children who had not used an antimicrobial within 30 days. The adjusted ORsus for recent penicillin and beta-lactam–susceptible carriage was 0.2 (p < 0.001); the corresponding ORsus for recent cephalosporin use was 0.7 (p = 0.26). The difference between the adjusted cephalosporin ORsus and the adjusted penicillin ORsus was statistically significant (p < 0.001).

The largest effect of antimicrobials on resistant and susceptible carriage was observed when cultures were obtained within 10 days after therapy was initiated.
Expressed another way, the values of ORres and ORsus were highest or lowest (e.g., most deviated from one) when determined on the basis of antimicrobial prescriptions given within 10 days prior to the culture date. Figure 1 depicts the ORres and ORsus for cephalosporins and penicillins measured across successive intervals between date of prescription and date of culture. The adjusted ORres for cephalosporins decreased sharply toward one when the culture was obtained more than 10 days after therapy was initiated. Thirteen of the 17 children who had used cephalosporins 1–10 days prior to culture and who were beta-lactam–resistant carriers had been treated with cefdinir or cefixime. Conversely, the adjusted ORsus for penicillins trended upward gradually, reaching one at the 61–90-day interval.



View larger version (16K):
[in this window]
[in a new window]
  		FIGURE 1. Prevalence odds ratios for beta-lactam–resistant and –susceptible carriage of pneumococci according to timing of antimicrobial exposure. For each two-sided 95% confidence interval, one direction is displayed with a T-shaped bar. Odds ratios were adjusted for age (categories: 6–23 months, 2–4 years, 5–6 years), community, year, number of children in the household cultured, prior carriage, and sibling antimicrobial use. Cultures were obtained during 1998–2003 from home-living children residing in two rural communities in Utah and two in Idaho.

 
Antimicrobial use after the culture date was not associated with pneumococcal carriage.
An analysis was also performed of antimicrobials prescribed after the culture was collected to test the proposition that the findings were spurious. If the adjusted associations were due to confounding rather than causal effects of distinct antimicrobial classes on pneumococcal carriage, antimicrobials prescribed after collection of the culture should also be associated with carriage. Use of neither penicillin nor cephalosporins 1–10 days after culture was associated with resistant carriage (adjusted ORres for penicillin = 1.5, 95 percent confidence interval (CI): 0.9, 2.6; ORres for cephalosporin = 0.8, 95 percent CI: 0.3, 2.4).

Recent macrolide therapy was strongly associated with decreased macrolide-susceptible carriage and with a nonsignificant trend toward increased macrolide-resistant carriage.
Erythromycin-susceptible S. pneumoniae were carried by 5 percent of the children recently (<30 days) treated with macrolides and 20 percent of the children not recently treated with macrolides. The adjusted ORsus for recent macrolide use was 0.2 (95 percent CI: 0.1, 0.4). Thirteen (9 percent) of 149 children recently treated with macrolides carried erythromycin-resistant S. pneumoniae compared with 5 percent of children not recently treated with macrolides. The adjusted ORres was 1.8 (95 percent CI: 0.9, 3.6).

Recent macrolide therapy was strongly associated with decreased penicillin-susceptible carriage but was not associated with increased penicillin-resistant carriage.
The adjusted ORsus for recent macrolide therapy and beta-lactam–susceptible carriage was 0.2 (95 percent CI: 0.1, 0.4); the corresponding adjusted ORres for beta-lactam–resistant carriage was 1.2 (95 percent CI: 0.7, 2.2). Penicillin use was similarly associated with decreased erythromycin-susceptible carriage but not with erythromycin-resistant carriage (adjusted ORsus and ORres were 0.3 (95 percent CI: 0.2 and 0.4) and 1.0 (95 percent CI: 0.6, 1.6), respectively).

Sibling cephalosporin use, but not sibling penicillin use, was associated with increased carriage of beta-lactam–resistant pneumococci.
Thirteen percent (21 of 163) of children with at least one sibling who had used cephalosporins within 30 days before culture carried resistant S. pneumoniae, compared with 8 percent (33 of 403) of children with at least one sibling who had used penicillins. The adjusted ORres for sibling cephalosporin use was 1.8 (95 percent CI: 1.01, 3.3; table 1) and for sibling penicillin use was 1.2 (95 percent CI: 0.8, 1.8).

Recent treatment with cephalosporins, penicillins, and macrolides was more common among resistant carriers than susceptible carriers, an association gauged by ORC.
The adjusted ORC for recent cephalosporin and penicillin use, estimated by comparing carriers of beta-lactam–resistant pneumococci with carriers of beta-lactam–susceptible pneumococci, were 3.0 (95 percent CI: 1.6, 5.9) and 4.0 (95 percent CI: 2.1, 7.5), respectively. The adjusted ORC for macrolide use, estimated by comparing carriers of erythromycin-resistant pneumococci with carriers of erythromycin-susceptible pneumococci, was 6.2 (95 percent CI: 2.2, 17.5).

The ORC for macrolide use and beta-lactam resistance was 6.2 (95 percent CI: 2.3, 19.6), and the ORC for penicillin use and erythromycin resistance was 2.9 (95 percent CI: 1.6, 5.3). The elevated ORC in these instances was explainable on the basis of the correlation between beta-lactam and macrolide resistance. Stratifying on the presence or absence of co-resistance to macrolide reduced the ORc for macrolide treatment and beta-lactam resistance to 1.2. Stratifying on the presence or absence of co-resistance to penicillin reduced the ORc for penicillin treatment and erythromycin resistance to 1.1.

Prior carriage of beta-lactam–resistant or –susceptible pneumococci, detected by previous participation in the surveillance study, was associated with susceptible carriage (ORsus for prior susceptible carriage = 1.4, ORsus for prior resistant carriage = 1.7). In contrast, neither prior carriage of resistant nor prior carriage of susceptible pneumococci was associated with resistant carriage (table 1). Younger age was more strongly associated with carriage of beta-lactam–resistant pneumococci than with carriage of beta-lactam–susceptible pneumococci (table 1). Prior heptavalent conjugate vaccine use also was not associated with decreased carriage (table 1).

Use of transmission models to interpret the observed results
The association between cephalosporin use and carriage of resistant pneumococci denoted an effect to enhance acquisition of beta-lactam–resistant pneumococci.
The transmission model was used to derive generalizations about the circumstances under which ORres exceeds one: either therapy favors acquisition of resistant organisms, compared with the non-antimicrobial-treated state, and/or therapy slows down the rate of clearance of resistant organisms, compared with the non-antimicrobial-treated state (figure 2). However, the maximum increase in ORres caused just by delayed clearance is modest.



View larger version (20K):
[in this window]
[in a new window]
  		FIGURE 2. Results of a two-way sensitivity analysis used to develop deterministic-model-derived prevalence odds ratios for carriage of resistant pneumococci (ORres) based on antimicrobial effects on acquisition and clearance of resistant organisms. m, multiplicative effect of the antimicrobial in enhancing the probability that an individual will acquire a resistant organism given contact with an infectious individual; values >1 indicate that the antimicrobial agent enhances acquisition of resistant organisms; values <1 indicate that, compared with the nontreated state, the antimicrobial reduces acquisition of resistant organisms. {delta}R, multiplicative factor by which antimicrobial treatment decelerates or accelerates clearance of resistant organisms; when values are <1, treatment delays clearance, and, when values are >1, treatment speeds up loss of carriage. Refer to the Appendix for details.

 
ORres for recent penicillin therapy was close to one, suggesting that penicillins at the doses prescribed to children in this study enhanced acquisition of beta-lactam–resistant pneumococci only minimally, if at all.

The association of penicillin use with reduced susceptible carriage indicated an effect to accelerate clearance of susceptible pneumococci during treatment.
The transmission model was also used to derive generalizations about the circumstances under which ORsus is less than one. Figure 3 demonstrates that either diminished acquisition of susceptible organisms or accelerated clearance of susceptible organisms leads to a lowering of ORsus. However, the lowering of ORsus achieved by effects on acquisition alone is modest.



View larger version (22K):
[in this window]
[in a new window]
  		FIGURE 3. Results of a two-way sensitivity analysis used to develop deterministic-model-derived prevalence odds ratios for carriage of susceptible pneumococci (ORsus) based on antimicrobial effects on acquisition and clearance of resistant organisms. ms, multiplicative effect of the antimicrobial in decreasing the probability that an individual will acquire a susceptible organism given contact with an infectious individual; values <1 indicate that, compared with the nontreated state, the antimicrobial agent inhibits acquisition of susceptible organisms. {delta}S, multiplicative factor by which antimicrobial treatment accelerates clearance of susceptible organisms. Refer to the Appendix for details.

 
Simulations of community populations demonstrated that both types of antimicrobial effects could promote dissemination of resistant S. pneumoniae in a community population.
A population of 1,000 children was simulated in which only the penicillin class of antimicrobials was used. During the period of treatment, penicillin was assumed to enhance clearance of susceptible pneumococci 10-fold and resistant pneumococci 1.2-fold during therapy (refer to appendix tables 3 and 4 for other base-case parameter values). Acquisition of resistant pneumococci was assumed to be unaffected. These parameter values were selected for illustrative purposes, not on the basis of maximum likelihood estimation or other methods of model fitting.

When 5,000 simulations of the individual-agent model were run by using these parameters, resistant carriage rose from a median of 7 percent (2.5th, 97.5th percentile: 6, 9) at baseline to a median of 12 percent (2.5th, 97.5th percentile: 5, 20) at 6 months (figure 4). The median ORsus and ORres from the simulations were 0.3 (2.5th, 97.5th percentile: 0.1, 0.6) and 1.0 (2.5th, 97.5th percentile: 0.5, 1.8), similar to the observed ORsus and ORres for penicillins.



View larger version (23K):
[in this window]
[in a new window]
  		FIGURE 4. Box plot of temporal trends in the prevalence of carriage of beta-lactam–resistant pneumococci based on running 5,000 simulations of an individual-agent model. Compared are a simulated population in which only cephalosporins were prescribed and one in which only penicillins were prescribed. The upper and lower lines of the boxes represent the 75th and 25th percentiles, respectively. The lines extending from the box are capped at the upper and lower adjacent values. Adjacent values were calculated by using the interquartile range (IQR), which is the difference between the 75th and 25th percentiles. The upper adjacent value is the largest data point that is less than or equal to the third quartile plus 1.5 x IQR; the lower adjacent value is the smallest data point that is greater than or equal to the first quartile minus 1.5 x IQR. Refer to the text and the Appendix for a detailed description of the structure of the individual-agent model.

 
The same simulation model was rerun by replacing penicillins with cephalosporin-class antimicrobials. The parameters representing the antimicrobial effects were altered, and other parameter values were left unchanged. During treatment, cephalosporin therapy was assumed to reduce acquisition of susceptible pneumococci following contact by 30 percent and to increase acquisition of beta-lactam–resistant pneumococci fourfold (refer to appendix table 3). Resistant carriage rose from 7 percent at baseline to 20 percent (2.5th, 97.5th percentile: 12, 28) at 6 months (figure 4). The median ORsus and ORres from the simulations were 0.7 (2.5th, 97.5th percentile: 0.4, 1.2) and 1.7 (2.5th, 97.5th percentile: 1.0, 2.7).

The model was then applied to compare the population impact of the two antimicrobial classes by simulating a population in which both drugs were used. The prevalence of penicillin use was set at 3 percent and of cephalosporin use at 1 percent to correspond to actual practice. Other parameters were the same as described above. These conditions produced a median of 715 (2.5th, 97.5th percentile: 406, 1,046) transmissions (infections) with resistant pneumococci during a 6-month follow-up period. When the same model was run by reducing the prevalence of penicillin use to zero but continuing the same level of cephalosporin use, a median of 607 (2.5th, 97.5th percentile: 343, 918) infections were observed; when the model was run by setting cephalosporin use to zero but continuing the same penicillin use, the median number of infections during 6 months was 606 (2.5th, 97.5th percentile: 342, 907). Thus, each drug class had an estimated population attributable fraction of 15 percent.

Possible occurrence of co-carriage of resistant and susceptible strains and its implication with respect to ORsus and ORres was also evaluated. The conditions of the cephalosporin-only model were changed to stipulate that 25 percent of susceptible carriers were also colonized with a subpopulation of beta-lactam–resistant S. pneumoniae. The mean duration of days from initiation of antimicrobial agents to selection of the resistant subpopulation was set at 3 for individuals who co-carried susceptible and resistant pneumococci. With other parameters set at the same values as in the cephalosporin simulation, ORres was 1.9 (2.5th and 97.5th percentiles: 1.2, 3.0). The median ORsus was 0.5 (2.5th and 97.5th percentiles: 0.2, 0.8). Resistant carriage at 6 months rose to 23 percent (2.5th and 97.5th percentiles: 15, 30).


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
APPENDIX: Mathematical Models
 References
 
The results of this study support the hypothesis that distinct antimicrobial classes promote pneumococcal resistance by different mechanisms. Individual and sibling use of cephalosporin increases the individual risk of resistant carriage by enhancing acquisition of new resistant pneumococcal strains during or shortly after treatment. In contrast, penicillins do not enhance individual risk of carriage of resistant S. pneumoniae. Rather, penicillins markedly accelerate clearance of susceptible pneumococci (5Go, 28Go). Both antimicrobial drug classes promote population dissemination of resistant pneumococci, but, in the case of penicillins, the elevated risk of resistant carriage is shared equally by treated and untreated members of the population.

This sharp contrast between cephalosporins and penicillins may, in part, be a function of the predominant oral cephalosporins prescribed in these communities, cefdinir and cefixime (29Go–31Go). Cefixime has particularly poor activity against pneumococci and also is broad spectrum. Macrolides appeared to combine effects of the penicillin and cephalosporin drug classes: they accelerated clearance of susceptible S. pneumoniae similar to penicillins while exhibiting no evidence of activity against erythromycin-resistant pneumococci (32Go, 33Go). Although not statistically significant, the trend toward an association between macrolide use and erythromycin-resistant pneumococcal carriage suggested that macrolides may also enhance acquisition of resistant organisms (34Go). Because of co-resistance, macrolides promote dissemination of beta-lactam resistance, and beta-lactams enhance spread of macrolide resistance (35Go, 36Go).

Limitations of the study design should be acknowledged. Minor discrepancies are frequent between the E-test method for susceptibility testing and microbroth dilution testing (27Go, 37Go–40Go). However, the differences rarely affect the categorical interpretation of susceptibility. Furthermore, the results of the regression models were similar whether or not isolates with intermediate susceptibility were classified as resistant. Another drawback was that the small number of communities precluded an analysis of the relation between community-level antimicrobial use and pneumococcal resistance. A strength of the study was that healthy children were recruited from home settings rather than from within health-care or day-care facilities (41Go–49Go). Lastly, the study design did not permit a detailed examination of antimicrobial drug effects at a pharmacodynamic or concentration-specific level (50Go).

Penicillin-resistant pneumococcal clones have disseminated to multiple continents (2Go, 6Go, 51Go). These clones possess mosaic penicillin-binding protein genes that contain interspersed sequences of foreign DNA (1Go). Decreased susceptibility to cephalosporins has also been associated with point mutations of specific penicillin-binding proteins (52Go). When the simulation model was modified to account for the possibility of emergence of resistant organisms through mutation or through selection of resistant strains among individuals who carry multiple strains (53Go–57Go), inferences were not materially altered. Emergence of resistance in the individually treated host is likely to be more important at a stage when resistance is rare and when antimicrobial treatment and carriage are more closely linked.

In summary, the data and models show that oral penicillins promote pneumococcal resistance primarily by enhancing clearance of susceptible pneumococci, and that this mechanism gives resistant pneumococci a competitive advantage at the population level. In contrast, the oral cephalosporins appear to directly increase acquisition of resistant S. pneumoniae. The models demonstrate that both types of effects contribute to dissemination of resistant strains. Thus, it is not necessary for antimicrobials to increase an individual's risk of resistant carriage or infection for the antimicrobial to select for resistant organisms at a population level (58Go). An appropriate message to be drawn is that narrow-spectrum antimicrobial agents that exhibit highly targeted activity still can cause plenty of damage with respect to resistance. The broader conclusions are that both susceptible and resistant carriage should be evaluated in epidemiologic studies of resistant organisms and that all antimicrobial drug classes ultimately impact resistance, albeit to varying degrees. An intervention strategy of simultaneously lowering unnecessary use of even narrow-spectrum drugs while redirecting choice away from agents with broader-spectrum activity is likely to be most effective at mitigating the public health threat of antimicrobial resistance.


   APPENDIX: Mathematical Models
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 APPENDIX: Mathematical Models
References
 
Deterministic model
The population is assumed to belong to one of six states. The labels indicating carriage status, X, S, and R, correspond to noncarriers, carriers of susceptible pneumococci, and carriers of resistant pneumococci, respectively. Each of these groups is in turn divided into two categories depending on whether the population members are using antimicrobial(s) (subscript 1) or not (subscript 0). Thus, S0, R0, and X0 are susceptible, resistant, and noncarriers not receiving antimicrobial treatment; and S1, R1, and X1 are susceptible, resistant, and noncarriers receiving antimicrobial treatment, respectively. If a closed population is assumed (no births, deaths, or migrations), the change in size of each state is modeled by the six equations below. Homogeneous contact and instantaneous infection are assumed.






Loss of carriage occurs at rate µS for susceptible organisms and µR for resistant organisms. Antimicrobials are started at rate {alpha} and discontinued at rate {varepsilon}. Antimicrobials active against pneumococci accelerate clearance of pneumococcal carriage. Multiplying clearance rate µ by a factor {delta}S for susceptible organisms and {delta}R for resistant organisms models this effect. The {delta}R term allows for the possibility that antimicrobials increase (or decrease) clearance of resistant or intermediately susceptible strains relative to the nontreated state. However, the killing of these organisms is presumed to be slower than for susceptible organisms ({delta}R < {delta}S). The acquisition rates for nontreated subjects, resulting from contact between noncarriers and susceptible carriers and between noncarriers and resistant carriers, are proportional to the corresponding transmission rate parameters, ßS and ßR, respectively. During antimicrobial treatment, the transmission rate parameters for susceptible and resistant organisms are ßAS and ßAR, respectively. The model is formulated to assume that susceptible carriers are noninfectious during antimicrobial treatment because of reduction in organism burden and viability.

To produce the graphs in figures 2 and 3, the initial prevalence of carriage of resistant pneumococci was set at 8.4 percent, susceptible pneumococci at 12.7 percent, and antimicrobial use at 4.2 percent. The rate of starting antimicrobials ({alpha}) was fixed at 0.005 and the rate of stopping antimicrobials ({varepsilon}) at 0.1. Prevalence odds ratios were calculated after simulating 60 days of follow-up.

Stochastic model
The individual-agent (microsimulation) version of this compartmental model treated transitions between states as discrete events that occurred probabilistically rather than deterministically. An advantage of this approach was that it more easily accommodated the modeling of a posttreatment effect of antimicrobial therapy. Such an effect would be mediated by antimicrobial-induced suppression of nasopharyngeal flora that otherwise inhibits acquisition of new bacteria and presumably lasts until normal flora are restored. Following antimicrobial treatment, increased susceptibility to acquisition of new strains, if present, should extend to both susceptible and resistant pneumococci. Another benefit of the individual-agent model is that it made it easier to incorporate nonexponential distributions for parameters such as duration of antimicrobial courses.

The individual-agent model was constructed in AnyLogic 5.1 (XJ Technologies, St. Petersburg, Russia) by using Java-based code to create replicated objects to represent members of a population. Individuals transitioned between carrier and noncarrier states and between antimicrobial-treated and non-antimicrobial-treated states. An individual carrier contacted randomly selected individual noncarriers at time intervals derived from an exponential distribution; each contact event resulted in transmission with a probability determined by a log-linear equation that included terms for antibiotic status of the potential transmitter and potential recipient.

A simplifying assumption of the models used for simulation was that resistance had zero fitness costs; that is, in the absence of antimicrobial use, transmission and clearance rate parameters were equal for resistant and susceptible strains. Because of these assumptions, the rate at which resistance rose in the simulations was faster than what is observed in actual human populations. Using more complex models did not alter the interpretation of the model regarding antimicrobial effects.
APPENDIX TABLE 1. Minimum inhibitory concentration break points

Antimicrobial agent
Level of resistance
Break point (µg/ml)
Penicillin Intermediate 0.12–1.0
Resistant ≥2
Ceftriaxone Intermediate 2.0
Resistant ≥4
Erythromycin Resistant ≥1
Trimethoprim/sulfamethoxazole Intermediate 1/19–2/38

Resistant
 ≥4/76


APPENDIX TABLE 2. Categories of antimicrobial agents prescribed to the children participating in a study in two rural communities in Utah (1998–2003) and two in Idaho (2002–2003)

Antimicrobial agent
No. of courses prescribed within 6 months of culture
% of subtotal
% of total
Penicillin
    Amoxicillin 1,993 62 36
    Amoxicillin/clavulanate 1,225 38 22
    Penicillin VK 13 0.4 0.2
        Subtotal 3,231 58
Cephalosporin
    Cefdinir 378 35 7
    Cefixime 278 26 5
    Loracarbef 117 11 2
    Cephalexin 74 7 1
    Cefuroxime 63 6 1
    Ceftibuten 52 5 1
    Cefprozil 45 4 1
    Cefaclor 40 4 1
    Other 28 2 0.5
        Subtotal 1,075 19
Macrolide
    Azithromycin 911 88 16
    Clarithromycin 59 6 1
    Erythromycin 42 4 1
    Erythromycin/sulfisoxazole 26 3 0.5
        Subtotal 1,038 19
Trimethoprim/sulfamethoxazole 240 4
Fluoroquinolone 9 0.2
Other 9 0.2
Total
 5,605


APPENDIX TABLE 3. Assumed characteristics of the population in the individual-agent model

Description
Base value
Population size 1,000
Prevalence of antimicrobial use 4%
Duration of antimicrobial course Triangular distribution; mode, 10 days; upper and lower boundaries, 3 and 14 days, respectively
Contact rate 1.8 (days)
Baseline transmission parameter (ßS* = ßR*) 0.028
Rate of loss of pneumococcal carriage in the absence of antimicrobials (µS = µR)
 1/28 (days)

* S, label indicating carriage of susceptible pneumococci; R, label indication carriage of resistant pneumococci.


APPENDIX TABLE 4. Parameters for antimicrobial effects

Antimicrobial effects
Symbol*
Antimicrobial simulated
Cephalosporin
 Penicillin
During treatment
    Acquisition{dagger}
        Susceptible pneumococci ßAS/ßS 0.3 0.1
        Resistant pneumococci ßAR/ßR 4 1
    Clearance{ddagger}
        Susceptible pneumococci {delta}S 2 10
        Resistant pneumococci {delta}R 1 1.2
After treatment (within 21 days)
    Acquisition{dagger}
        Susceptible pneumococci ßPS/ßS 1.2 1
        Resistant pneumococci
 ßPR/ßR
 1.2
 1

* S, label indicating carriage of susceptible pneumococci; R, label indication carriage of resistant pneumococci.

{dagger} If >1, acquisition increased and if <1, acquisition decreased compared with non-antimicrobial-treated individuals.

{ddagger} If >1, clearance increased and if <1, clearance decreased compared with non-antimicrobial-treated individuals.


   ACKNOWLEDGMENTS
 
Funding was provided by Centers for Disease Control and Prevention grant RS1 CCR820631 and the Thrasher Research Fund. M. L.'s work was supported by the Ellison Medical Foundation and National Institutes of Health grant 5R01AI048935.

Dr. Carmeli, his laboratory, and studies that he has conducted during the past 4 years received grants, honoraria, travel support, and other forms of financial support from the following companies: Bayer Corp., West Haven, Connecticut; Biomedicum Ltd., Jerusalem, Israel; Bristol-Myers Squibb, Wallingford, Connecticut; Merck & Co., Inc., Whitehouse Station, New Jersey; Neopharm Ltd., Petach Tikva, Israel; Pfizer Pharmaceuticals, New York, New York; Teva Ltd., Cumbria, United Kingdom; Vicuron Pharmaceuticals, King of Prussia, Pennsylvania; Wyeth, Madison, New Jersey; and XTL Pharmaceuticals Ltd., Rehovot, Israel.


   References
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 APPENDIX: Mathematical Models
 References
 

  1. Tomasz A. Antibiotic resistance in Streptococcus pneumoniae. Clin Infect Dis 1997;24(suppl 1):S85–S88.[Medline]
  2. Tomasz A. New faces of an old pathogen: emergence and spread of multidrug-resistant Streptococcus pneumoniae. Am J Med 1999;107:55S–62S.[Medline]
  3. Whitney CG, Farley MM, Adler J, et al. Increasing prevalence of multidrug-resistant Streptococcus pneumoniae in the United States. N Engl J Med 2000;343:1917–24.[Abstract/Free Full Text]
  4. Baquero F. Pneumococcal resistance to beta-lactam antibiotics: a global geographic overview. Microb Drug Resist 1995;1:115–20.[ISI][Medline]
  5. Dagan R, Klugman KP, Craig WA, et al. Evidence to support the rationale that bacterial eradication in respiratory tract infection is an important aim of antimicrobial therapy. J Antimicrob Chemother 2001;47:129–40.[Abstract/Free Full Text]
  6. Klugman KP. The successful clone: the vector of dissemination of resistance in Streptococcus pneumoniae. J Antimicrob Chemother 2002;50(suppl S2):1–5.[Abstract]
  7. Forward KR. The epidemiology of penicillin resistance in Streptococcus pneumoniae. Semin Respir Infect 1999;14:243–54.[Medline]
  8. Finkelstein JA, Huang SS, Daniel J, et al. Antibiotic-resistant Streptococcus pneumoniae in the heptavalent pneumococcal conjugate vaccine era: predictors of carriage in a multicommunity sample. Pediatrics 2003;112:862–9.[Abstract/Free Full Text]
  9. Turnidge JD, Bell JM, Collignon PJ. Rapidly emerging antimicrobial resistances in Streptococcus pneumoniae in Australia. Pneumococcal Study Group. Med J Aust 1999;170:152–5.[ISI][Medline]
  10. Schuchat A, Hilger T, Zell E, et al. Active bacterial core surveillance of the emerging infections program network. Emerg Infect Dis 2001;7:92–9.[ISI][Medline]
  11. Karlowsky JA, Thornsberry C, Jones ME, et al. Factors associated with relative rates of antimicrobial resistance among Streptococcus pneumoniae in the United States: results from the TRUST Surveillance Program (1998–2002). Clin Infect Dis 2003;36:963–70.[CrossRef][ISI][Medline]
  12. Gonzales R, Steiner JF, Lum A, et al. Decreasing antibiotic use in ambulatory practice: impact of a multidimensional intervention on the treatment of uncomplicated acute bronchitis in adults. JAMA 1999;281:1512–19.[Abstract/Free Full Text]
  13. Schwartz B, Bell DM, Hughes JM. Preventing the emergence of antimicrobial resistance. A call for action by clinicians, public health officials, and patients. JAMA 1997;278:944–5.[CrossRef][ISI][Medline]
  14. Lipsitch M, Samore MH. Antimicrobial use and antimicrobial resistance: a population perspective. Emerg Infect Dis 2002;8:347–54.[ISI][Medline]
  15. Baquero F, Baquero-Artigao G, Canton R, et al. Antibiotic consumption and resistance selection in Streptococcus pneumoniae. J Antimicrob Chemother 2002;50(suppl S2):27–37.[Abstract]
  16. Samore MH, Magill MK, Alder SC, et al. High rates of multiple antibiotic resistance in Streptococcus pneumoniae from healthy children living in isolated rural communities: association with cephalosporin use and intrafamilial transmission. Pediatrics 2001;108:856–65.[Abstract/Free Full Text]
  17. Clinical and Laboratory Standards Institute (CLSI). Performance standards for antimicrobial susceptibility testing—15th informational supplement. Wayne, PA: CLSI, 2005. (Approved standard, CLSI document M100-S15, vol 25).
  18. Lipsitch M. Measuring and interpreting associations between antibiotic use and penicillin resistance in Streptococcus pneumoniae. Clin Infect Dis 2001;32:1044–54.[CrossRef][ISI][Medline]
  19. Bingenheimer JB, Raudenbush SW. Statistical and substantive inferences in public health: issues in the application of multilevel models. Annu Rev Public Health 2004;25:53–77.[CrossRef][ISI]
  20. Snijders TA, Bosker RJ. Multilevel analysis: an introduction to basic and advanced multilevel modeling. London, England: SAGE Publications, 1999:266.
  21. Burton P, Gurrin L, Sly P. Extending the simple linear regression model to account for correlated responses: an introduction to generalized estimating equations and multi-level mixed modelling. Stat Med 1998;17:1261–91.[CrossRef][ISI][Medline]
  22. Ekholm A, Jokinen J, Kilpi T. Combining regression and association modelling for longitudinal data on bacterial carriage. Stat Med 2002;21:773–91.[CrossRef][ISI][Medline]
  23. Temime L, Guillemot D, Boelle PY. Short- and long-term effects of pneumococcal conjugate vaccination of children on penicillin resistance. Antimicrob Agents Chemother 2004;48:2206–13.[Abstract/Free Full Text]
  24. Melegaro A, Gay NJ, Medley GF. Estimating the transmission parameters of pneumococcal carriage in households. Epidemiol Infect 2004;132:433–41.[CrossRef][Medline]
  25. Lipsitch M. The rise and fall of antimicrobial resistance. Trends Microbiol 2001;9:438–44.[CrossRef][ISI][Medline]
  26. Ekdahl K, Ahlinder I, Hansson HB, et al. Duration of nasopharyngeal carriage of penicillin-resistant Streptococcus pneumoniae: experiences from the South Swedish Pneumococcal Intervention Project. Clin Infect Dis 1997;25:1113–17.[ISI][Medline]
  27. Van Dyke Parunak H, Savit R, Riolo R. Agent-based modeling vs. equation-based modeling: a case study and users' guide. Paris, France, Proceedings of Workshop on Modeling Agent Based Systems (MABS), 1998.
  28. Schrag SJ, Pena C, Fernandez J, et al. Effect of short-course, high-dose amoxicillin therapy on resistant pneumococcal carriage: a randomized trial. JAMA 2001;286:49–56.[Abstract/Free Full Text]
  29. Dabernat H, Geslin P, Megraud F, et al. Effects of cefixime or co-amoxiclav treatment on nasopharyngeal carriage of Streptococcus pneumoniae and Haemophilus influenzae in children with acute otitis media. J Antimicrob Chemother 1998;41:253–8.[Abstract/Free Full Text]
  30. Dubois J, St-Pierre C. In vitro study of the post-antibiotic effect and the bactericidal activity of Cefditoren and ten other oral antimicrobial agents against upper and lower respiratory tract pathogens. Diagn Microbiol Infect Dis 2000;37:187–93.[CrossRef][ISI][Medline]
  31. Perry CM, Scott LJ. Cefdinir: a review of its use in the management of mild-to-moderate bacterial infections. Drugs 2004;64:1433–64.[CrossRef][ISI][Medline]
  32. Dagan R, Johnson CE, McLinn S, et al. Bacteriologic and clinical efficacy of amoxicillin/clavulanate vs. azithromycin in acute otitis media. Pediatr Infect Dis J 2000;19:95–104.[CrossRef][ISI][Medline]
  33. Jacobs MR, Bajaksouzian S, Windau A, et al. Susceptibility of Streptococcus pneumoniae, Haemophilus influenzae, and Moraxella catarrhalis to 17 oral antimicrobial agents based on pharmacodynamic parameters: 1998–2001 U S Surveillance Study. Clin Lab Med 2004;24:503–30.[CrossRef][ISI][Medline]
  34. Dagan R, Leibovitz E, Greenberg D, et al. Dynamics of pneumococcal nasopharyngeal colonization during the first days of antibiotic treatment in pediatric patients. Pediatr Infect Dis J 1998;17:880–5.[CrossRef][ISI][Medline]
  35. Garcia-Rey C, Aguilar L, Baquero F, et al. Pharmacoepidemiological analysis of provincial differences between consumption of macrolides and rates of erythromycin resistance among Streptococcus pyogenes isolates in Spain. J Clin Microbiol 2002;40:2959–63.[Abstract/Free Full Text]
  36. McCormick AW, Whitney CG, Farley MM, et al. Geographic diversity and temporal trends of antimicrobial resistance in Streptococcus pneumoniae in the United States. Nat Med 2003;9:424–30.[CrossRef][ISI][Medline]
  37. Assessment of susceptibility testing practices for Streptococcus pneumoniae—United States, February 2000. MMWR Morb Mortal Wkly Rep 2002;51:392–4.[Medline]
  38. Fasola EL, Bajaksouzian S, Appelbaum PC, et al. Variation in erythromycin and clindamycin susceptibilities of Streptococcus pneumoniae by four test methods. Antimicrob Agents Chemother 1997;41:129–34.[Abstract]
  39. Davies TA, Kelly LM, Jacobs MR, et al. Antipneumococcal activity of telithromycin by agar dilution, microdilution, E test, and disk diffusion methodologies. J Clin Microbiol 2000;38:1444–8.[Abstract/Free Full Text]
  40. Jorgensen JH, Ferraro MJ, McElmeel ML, et al. Detection of penicillin and extended-spectrum cephalosporin resistance among Streptococcus pneumoniae clinical isolates by use of the E test. J Clin Microbiol 1994;32:159–63.[Abstract/Free Full Text]
  41. Huang SS, Finkelstein JA, Rifas-Shiman, SL, et al. Community-level predictors of pneumococcal carriage and resistance in young children. Am J Epidemiol 2004;159:645–54.[Abstract/Free Full Text]
  42. Sa-Leao R, Tomasz A, Santos Sanches I, et al. Pilot study of the genetic diversity of the pneumococcal nasopharyngeal flora among children attending day care centers. J Clin Microbiol 2002;40:3577–85.[Abstract/Free Full Text]
  43. Belongia EA, Sullivan BJ, Chyou PH, et al. A community intervention trial to promote judicious antibiotic use and reduce penicillin-resistant Streptococcus pneumoniae carriage in children. Pediatrics 2001;108:575–83.[Abstract/Free Full Text]
  44. Stratchounski LS, Kretchikova OI, Kozlov RS, et al. Antimicrobial resistance of Streptococcus pneumoniae isolated from healthy children in day-care centers: results of a multicenter study in Russia. Pediatr Infect Dis J 2000;19:196–200.[CrossRef][ISI][Medline]
  45. Huebner RE, Wasas AD, Klugman KP. Prevalence of nasopharyngeal antibiotic-resistant pneumococcal carriage in children attending private paediatric practices in Johannesburg. S Afr Med J 2000;90:1116–21.[ISI][Medline]
  46. Barnes DM, Whittier S, Gilligan PH, et al. Transmission of multidrug-resistant serotype 23F Streptococcus pneumoniae in group day care: evidence suggesting capsular transformation of the resistant strain in vivo. J Infect Dis 1995;171:890–6.[ISI][Medline]
  47. Henderson FW, Gilligan PH, Wait K, et al. Nasopharyngeal carriage of antibiotic-resistant pneumococci by children in a group day care. J Infect Dis 1988;157:256–63.