Skip to main page content

• HOME
• Current Issue
• Archive
• Alerts
• About ASM
• Contact us
• Tech Support
• Journals.ASM.org

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Reprints and Permissions Copyright Information Books from ASM Press MicrobeWorld Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Van Eldere, J. Articles by Amrine-Madsen, H. Search for Related Content PubMed PubMed Citation Articles by Van Eldere, J. Articles by Amrine-Madsen, H.

 Previous Article  |  Next Article 

Antimicrobial Agents and Chemotherapy, October 2007, p. 3491-3497, Vol. 51, No. 10
0066-4804/07/$08.00+0     doi:10.1128/AAC.01581-06
Copyright © 2007, American Society for Microbiology. All Rights Reserved.

Risk Factors for Development of Multiple-Class Resistance to Streptococcus pneumoniae Strains in Belgium over a 10-Year Period: Antimicrobial Consumption, Population Density, and Geographic Location

Johan Van Eldere,¹ Robertino M. Mera,^2* Linda A. Miller,³ James A. Poupard,⁴ and Heather Amrine-Madsen²

University Hospital Gasthuisberg, Herestraat 49, B-3000 Leuven, Belgium,¹ GlaxoSmithKline, 5 Moore Drive, Research Triangle Park, North Carolina 27709,² GlaxoSmithKline, 1250 South Collegeville Road, Collegeville, Pennsylvania 19426,³ Pharma Institute of Philadelphia, 3612 Earlham St., Philadelphia, Pennsylvania 19129⁴

Received 19 December 2006/ Returned for modification 8 May 2007/ Accepted 20 July 2007

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
We investigated the impact of the usage of antibiotics in ambulatory patients in Belgium in 147 defined geographical circumscriptions and at the individual isolate level. The study included 14,448 Streptococcus pneumoniae strains collected by the Belgium national reference lab from 1994 to 2004. Additional risk factors for resistance, such as population density/structure and day care attendance, were investigated for the same time-space window. A statistical model that included resistance to two or more antimicrobial classes offered the best fit for measuring the changes in nonsusceptibility to penicillin, macrolides, and tetracycline over time and place in Belgium. Analysis at the geographic level identified antimicrobial consumption with a 1-year lag (0.5% increase per additional defined daily dose) and population density as independent predictors of multiple resistance. Independent risk factors at the isolate level were age (odds ratio [OR], 1.55 for children aged <5 years), population density (7% increase in multiple resistance per 100 inhabitants/km²), conjugate 7-valent vaccine serotype (OR, 14.3), location (OR, 1.55 for regions bordering high-resistance France), and isolate source (OR, 1.54 for ear isolates). The expansion of multiple-resistant strains explains most of the overall twofold increase and subsequent decrease in single antimicrobial resistance between 1994 and 2004. We conclude that factors in addition to antibiotic use, such as high population density and proximity to high-resistance regions, favor multiple resistance. Regional resistance rates are not linearly related to actual antibiotic use but are linked to past antibiotic use plus a combination of demographic and geographic factors.

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Streptococcus pneumoniae continues to be a significant cause of morbidity and mortality in humans (19). Therapeutic options for pneumococcal infections are complicated by the increasing prevalence of drug-resistant strains, which in some cases may lead to treatment failure (21). Although isolates resistant to a single antibiotic class were common in the early 1990s, by the end of the century, more than 5 out of 10 isolates initially resistant only to penicillin were also resistant to macrolides (22). The first multiple-resistant isolates, demonstrating resistance to penicillin G, macrolides, tetracyclines, and trimethoprim-sulfamethoxazole (SXT), were characterized by Jacobs and coworkers in 1977 (13).

Two worldwide surveillance studies, the Alexander Project (6) and the PROTEKT study (11), which started in 1992 and 1999, respectively, have provided extensive information on multiple-resistant isolates. Penicillin and macrolide coresistance has continuously increased, reaching levels of one out of four isolates in the United States (16) and one out of three in Spain and France in 2003. Combined resistance to macrolides, tetracyclines, and SXT is now observed for more than 50% of pneumococcal isolates in Hong Kong, Taiwan, and South Korea (7). Multiple antibiotic resistance is presumed to be the direct result of antimicrobial consumption (9), and several studies of S. pneumoniae have investigated the link between the level of antibiotic consumption and the level of resistance in regions of a country (2, 3, 8, 20). However, these studies have often been rather narrow in focus, addressing either a short time window or a single geographic location. Further, although many studies have identified potential risk factors for the development of antimicrobial resistance, there is a paucity of strong evidence for factors beyond total usage and class of antibiotics.

Belgium is located among countries where S. pneumoniae clinical isolates that demonstrate very high rates of nonsusceptibility to two or more antibiotic classes are found, such as France (52%), countries with intermediate rates of nonsusceptibility, such as Luxembourg (14.8%), and countries with very low rates of multiple resistance, such as Germany (8.1%), and The Netherlands (1.3%). In 1985, an ongoing S. pneumoniae surveillance system that captured information on serotype, antibiotic susceptibility, geographic location, and patient characteristics began in Belgium. Prior to 1994, antibiotic resistance in Belgium was below 5%, and then Belgium experienced rapid changes in resistance levels. There was an overall increase in resistance from 1994 to the present, but from 2000 onward there was a stabilization and even reversal of resistance levels. This trend coincides with a decrease in antibiotic use, which has been claimed as an intervention (1). Belgium seems to be an ideal location to test the ecological hypothesis for the relationship between antimicrobial consumption and resistance. Moreover, with the inclusion of local factors like population density, consumption by postal code, and day care attendance, the country is in the unique position of facilitating a comprehensive examination of the changes in both single- and multiple-class resistance in S. pneumoniae.

Here, data from the 1994 to 2004 period of the surveillance study of pneumococcal isolates from all regions of Belgium are presented and utilized to obtain a comprehensive explanation of the risk factors that impact the development of S. pneumoniae multiple-class antimicrobial resistance over time and place.

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
The Belgian national reference laboratory has serotyped and determined the antibiotic susceptibility of pneumococci collected from patients with invasive infections from 1985 onward. Since 1994, the isolates sent to the national reference laboratory have been stored at –80°C. Patient characteristics, including age, gender, and address, were also collected. Initially, this was a nonfunded, totally voluntary surveillance system. Since 2002, partial funding has been provided by the Belgian federal authorities (Belgian Antibiotic Policy Coordinating Committee).

Geographical origin of the isolates. Isolates sent to the national reference laboratory come from laboratories all over Belgium, including hospital and non-hospital-based collecting sites. The number of isolates sent to the reference laboratory has increased every year, from 752 in 1994 to 1,746 in 2004. From 1996 onward, more than 1,000 isolates were received every year. On average, for the period 1994 to 2004, 1,324 isolates were received every year. The geographical distribution of the isolates in the collection has remained relatively constant. The contribution of the different provinces to the reference collection as a percentage of total isolates submitted varies mostly within a 3% window. Further, with the exception of one province (Liège), the contribution of each province correlates with the population size of that province.

Source of the isolates. All of the 14,488 S. pneumoniae isolates are from invasive infections, including blood, other sterile body fluids, and middle ear fluid. Over the period 1994 to 2004, between 71 and 82% of isolates were collected from blood; 11 to 21% from middle ear fluid; 4 to 6% from cerebrospinal fluid (CSF); 1 to 2% from pleural fluid, and approximately 1% from other sources, including peritoneal fluid, joints, etc. Over the time period included, there was a consistent decrease in the percentage of isolates from middle ear fluid that was mirrored by a steady increase in the percentage of blood isolates.

Microbiological analysis. Upon arrival at the reference laboratory, all isolates were reidentified and serogrouped according to the Copenhagen scheme, and susceptibility to the following antibiotics was determined via agar diffusion: erythromycin, oxacillin, tetracycline, and ofloxacin (collected from 1995 onward). For oxacillin-nonsusceptible isolates, the MIC to penicillin was determined via E-test. For each isolate, in addition to serotype and antibiotic susceptibility pattern, the following data are available: sample origin, sample collection date, and patient data (age, sex, address). For 51% of all isolates, additional patient data on therapeutic interventions (antibiotic usage during the 48 h prior to isolate collection, hospitalization, outcome of illness) and vaccination status are available. Multiple resistance is defined as nonsusceptibility to two or more antibiotic classes (beta-lactams, macrolides, tetracyclines, or quinolones). The nonsusceptibility designation was based on Clinical and Laboratory Standards Institute breakpoints, according to the most recent document available at that time (19a, 19b). Since breakpoints for the antibiotics tested and testing methods have not changed over the years, data remain current.

IMS data. Data on antibiotic usage for the period 1995 to 2004 were provided by IMS Health Services. These data describe antibiotic usage in ambulatory patients in Belgium at the level of 147 defined circumscriptions and at the level of the individual brand of antibiotic. The circumscriptions used to monitor antibiotic usage can be translated into postal codes and thus be linked to the patient address. Consumption data were expressed as defined daily doses (DDD) and as DDD per 1,000 inhabitants per day (DID) by postal code.

Other data. Population structure, population density, and day care membership levels were obtained from public databases (http://statbel.fgov.be/).

Statistical methods. Two main statistical analyses were performed, the first where the unit of analysis was the rate of resistance for a certain postal code and the second where the unit of analysis was the individual pneumococcal isolate (Table 1). They correspond to a single hierarchical model, but they have been separated in Results for easier presentation.

View this table: [in this window] [in a new window]

TABLE 1. Characteristics of multivariate models used for the analysis

The rate of resistance at the postal code level was analyzed using linear mixed models. Independent variables included population density, population structure, day care membership, and antimicrobial consumption by class and month. A nonlinear mixed model for repeated measures was recursively fitted in order to obtain the best possible relationship in time between consumption and resistance. This model will be referred to as the consumption-resistance model.

The odds of multiple resistance at the individual isolate level were calculated using multivariate logistic regression. The dependent variable was the presence or absence of multiple resistance. Independent factors were age, sex, population density of the location, time in years, PCV7 serotype, location (province, region, or border), and source of the isolate (blood, ear, CSF, etc.). Information on treatment during the 48 h prior to isolate collection was available for 51% of the isolates. A separate logistic model was fitted for treatment after all significant risk factors were included.

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Nonsusceptibility to antibiotics in Belgium over time. Overall, resistance to erythromycin, tetracycline, penicillin, SXT, ofloxacin, and cefotaxime increased over the time period of 1994 to 2004 (Fig. 1). The highest levels of resistance were observed for erythromycin. The prevalence of nonsusceptibility to erythromycin in Belgium reached its highest value in 2001 (36.7%) and stayed mostly stable until 2004. The second highest prevalence of nonsusceptibility was to tetracycline, which attained levels of 31.7% in the year 2000. The level of nonsusceptibility to penicillin peaked at 17.7% in the year 2000 and declined slowly to 11.6% in 2004.

View larger version (18K): [in this window] [in a new window]

FIG. 1. Single- and multiple-class resistance in Belgium.

The highest prevalence of coresistance to two antibiotics in Belgium was for erythromycin-tetracycline. It peaked during the period 2002 to 2003 at a level of 26.7% of all isolates, decreasing slightly to 25.9% in 2004. In 2004, out of all tetracycline-resistant isolates, 90.3% were also resistant to erythromycin, while out of all erythromycin-resistant isolates, 72.7% were also tetracycline resistant.

The second most common coresistant pair was penicillin-erythromycin. Resistance to erythromycin and penicillin peaked in 2001 at 12.3%, declining slowly to 9.3% in 2004. Resistance to penicillin-tetracycline peaked in 2002 at 9.9%, declining to 7.5% in 2004. The proportion of tetracycline and erythromycin resistance among penicillin-resistant isolates was 64.4% and 79.7%, respectively.

In 1994, 16.3% of all isolates were resistant to just one antibiotic class, while 14.1% were resistant to two or more antibiotic classes (multiple resistance). Ten years later, 11.7% of all isolates were resistant to just one antibiotic class, while 28.8% were resistant to two or more antibiotic classes. During that 10-year period, resistance to just one antibiotic was reduced by 36% (odds ratio [OR], 0.64; 95% confidence interval [CI], 0.53 to 0.77; P = 0.001), while multiple resistance increased 2.46 times (OR, 2.46; 95% CI, 1.95 to 3.12; P < 0.0001). Multiple resistance peaked at 30.7% in 2000, while resistance to only one antibiotic class peaked at 16.7% in 1996 (Fig. 2).

View larger version (25K): [in this window] [in a new window]

FIG. 2. Resistance to just one and two or more antibiotics in Belgium.

Nonsusceptibility to antibiotics in Belgium across location. During the first years of surveillance (1994 to 1996), multiple-class resistance prevalence was 18% in provinces such as Hainaut and Namur, which border France, where there are high levels of antibiotic resistance; multiple-class resistance prevalence was 11% in provinces such as Antwerp and Limburg, which border The Netherlands, where there low levels of antibiotic resistance. During the same period, Brussels and its environs, which do not share borders with neighboring countries, had a multiple-class resistance level of 16%. Although multiple-class resistance increased over time in all provinces, it did so in proportion to the initial values, so that the Dutch border provinces in 2003 to 2004 had values of 25%, while the levels in Brussels and at the French border were 29% and 33%, respectively.

The distribution of multiple-class resistance by postal code and province in 2004 can be seen in Fig. 3. It can be observed that cities like Brugge, Liège, and Namur tended to have most of the resistant isolates in their respective provinces. On the other hand, there is a cluster of postal codes with a high prevalence of multiple-class-resistant isolates in the corridor from Lille (France) to Brussels.

View larger version (124K): [in this window] [in a new window]

FIG. 3. Prevalence rate of multiple-class resistance in Belgium by province and postal code.

Changes in antimicrobial consumption in Belgium over time and place. Overall antimicrobial consumption was 26.4 DID in 1995, increasing until 2000 to 29.7 DID and declining every year since then to 23.3 DID in 2004. Among antibiotic classes, the highest consumption was found for broad-spectrum penicillins, with levels around 9 DID until 2000, declining since then to 6.4 in 2004. Macrolides have followed a similar pattern, with levels of 6 and 4.5 DID in 2000 and 2004, respectively, as have cephalosporins (4.7 and 3.7 DID in 2000 and 2004, respectively). The use of fluoroquinolones has been continuously increasing but is showing some signs of stabilization in recent years. Tetracycline was the second-most-prescribed class of antibiotics in 1995 but has declined threefold to 1.9 DID in 2004 (Fig. 4).

View larger version (33K): [in this window] [in a new window]

FIG. 4. Antimicrobial consumption in Belgium by year.

Historically, total antimicrobial consumption has been the highest in provinces bordering France and lowest in those bordering The Netherlands. Although consumption overall has decreased across Belgium, its relationship with location has remained the same, so that by 2004, the highest consumption was still in provinces bordering France (e.g., Namur, 24.5 DID) and the lowest in provinces bordering The Netherlands (e.g., Antwerp, 19.0 DID) (Fig. 5).

View larger version (113K): [in this window] [in a new window]

FIG. 5. Total antimicrobial consumption by province, 2004.

Single-resistance versus multiple-resistance models. Separate single-antibiotic-class resistance models were built for penicillin, macrolides, and tetracycline and for antibiotic consumption by class. Multiple-class resistance models were built either with single-class antibiotic consumption or with total antibiotic consumption. When multiple resistance was defined as resistance to two or more antibiotic classes but not to three or more, inclusion of this variable in the models explained most of the changes over time and place of single resistance and thus was a better fit than any of the single-antibiotic-class resistance models. Furthermore, multiple-class resistance models with total antibiotic consumption fit the data significantly better than consumption by class; therefore, overall consumption was included in the final model.

Statistical analysis of the relationship between consumption and multiple-class resistance at the postal code level. Multiple-class resistance at the postal code level was analyzed using mixed effects models for repeated measures. Time, place, population density and structure, and antimicrobial consumption for all available antimicrobial classes were utilized to explain changes in multiple resistance (Table 1). Day care center membership was not found to be a significant factor and was excluded from any further modeling. The analysis found that multiple-class resistance has been increasing over time in all regions of the country at roughly the same rate. There were significant differences in the rate of multiple-class resistance among different regions in Belgium, but they were maintained over the 10-year period of observation; thus, there were significant increases over time in those regions that started with low levels, but the increases occurred at roughly the same rates as those in the regions that had high levels in 1994.

Population density was a significant predictor of multiple-class resistance independent of time and place, so that regions that have higher population density tended to have more resistance, independent of their geographical location. When antimicrobial consumption was considered in the model, after adjusting for time, place, and population density, total antimicrobial consumption with a 1-year lag was a better predictor of multiple-class resistance than the current-year total antimicrobial consumption. A statistical model that used a 1-year lag in antimicrobial consumption found that for every one additional DDD of total antimicrobial consumption, there was a 0.5% change in the prevalence of multiple resistance, so that locations with the lowest consumption (22.4 DDD) had 4% less multiple resistance on average than locations with the highest consumption (30.4 DDD). This relationship was independent of the population density, population structure, and geographic location. A cumulative consumption model did not fit any better than the lagged model, but both fit significantly better than the no-lag model.

When separated by antimicrobial class, both macrolide consumption and broad-spectrum penicillin consumption were related to the rate of multiple-class resistance, but when included in a model that also had total antimicrobial consumption, only macrolide consumption was statistically significant. The relationship between total antimicrobial consumption with a 1-year lag and multiple-class resistance to antibiotics was significant after adjustments were made for the effects of time, population density, population structure, and location within the country.

Risk factors for the odds of multiple-class resistance at the individual isolate level. Multivariate logistic regression analysis was used to evaluate a variety of risk factors for the development of multiple antibiotic class resistance, including age, population density, year of collection, PCV7 serotype, location (province, region, or country the province borders), and source of the isolate (Table 1). In addition, a separate model was used to evaluate prior antimicrobial treatment and the specific antibiotics used in prior treatment as independent risk factors for the acquisition of multiple-class resistance.

After adjustments were made for age, sex, time, population density, PCV7 serotype, and source of the isolate, provinces that share a border with France were found to be 1.55 times (95% CI, 1.37 to 1.77) more likely to have multiple-resistant isolates than provinces close to the border with The Netherlands.

Population density in Belgium was 338 inhabitants/km² in 2003. The areas of the country with the lowest population density are those bordering the duchy of Luxembourg (54 inhabitants/km²), while those with the highest population density are in the northeastern part of the country (Antwerp, 512 inhabitants/km²). Independent of other risk factors, including proximity to France, for every 100 inhabitants/km² there was a 7% increase in multiple resistance (95% CI, 1.018 to 1.124).

Among PCV7 and related serotypes, 40.6% of isolates were multiply resistant, while only 4.5% of non-PCV7 related isolates were multiply resistant (OR 14.3, 95% CI 12.5 to 16.4). The proportion of resistant isolates among vaccine serotypes in 1994 was 22% and increased to 44% in 2004, with the tendency toward multiple-class resistance increasing faster among vaccine serotypes, although the interaction between time and PCV7 resistance did not reach significance (P = 0.067).

Isolates from the ear were 54% more likely to be resistant than those from blood (95% CI, 1.36 to 1.75), and ear isolates were 82% more likely to be resistant than those from CSF (95% CI, 1.46 to 2.27). However, independent of the source of the isolate or the geographic origin of the sample, isolates from children less than 5 years old were 55% more likely to be multiply resistant (95% CI, 1.34 to 1.71). In other words, ear isolates were not more likely to be multiply resistant, independent of age.

Subjects that were previously treated with antibiotics were 1.33 times more likely to harbor a multiple-resistant isolate (95% CI, 1.16 to 1.51). When results for antibiotic classes were compared to those not treated with antibiotics, subjects receiving cephalosporins (OR, 1.41; 95% CI, 1.14 to 1.75) and macrolides (OR, 2.51; 95% CI, 1.61 to 2.51), but not penicillins, were more likely to harbor multiple-resistant strains.

As was demonstrated above, independent risk factors associated with multiple-class resistance to antibiotics in Belgium were age, population density, PCV7 serotype, location (province, region, or country the province borders), and source of the isolate. In addition, being treated was also an independent risk factor, and among antibiotics used, macrolides and cephalosporins were also independently associated with multiple-class resistance.

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
There is some debate but no agreement in the literature on the definition of multiple resistance, with some recent manuscripts using resistance to two or more antibiotic classes as the working definition (4, 5). Studies on mathematical models of exposure to antimicrobials (18) and surveillance studies (16) have shown that, because in the United States and Europe more than 50% of all penicillin, macrolide, or tetracycline-resistant isolates are resistant to at least one other antimicrobial class, any model fitted for only one antimicrobial class that tries to explain its variability over time does not fit well unless other antimicrobial classes are added. Resistance to two antimicrobial classes explains the changes over time in the United States and Belgium better than that to three or more antimicrobial classes and, surprisingly, it explains the changes over time and place for single antimicrobial classes. The rationale for this has been stated with great clarity by Klugman (14) and has to do with the greater selective advantage of strains resistant to more than one antibiotic. In the current study, resistance to two or more antimicrobial classes was the most important factor in understanding the changes of nonsusceptibility to penicillin, macrolides, and tetracycline over time and place in Belgium.

Although DID can be an imperfect measure of exposure, it allows for comparisons of different areas of a country at different times. Unless there are large differences in the way prescriptions are issued and in patient compliance in different parts of the country, or differences in the structure of the population, exposure to antibiotics measured using DDDs should be comparable in time and place (9). Belgium is roughly homogenous in the way prescriptions are issued, patient compliance, and population structure, and so DID is the appropriate measure of exposure.

Overall antimicrobial consumption in Belgium peaked in the year 2000 and then showed levels in 2004 that were 21% lower than at the peak. Broad-spectrum penicillins and cephalosporins followed roughly the same pattern. Consumption of macrolides has not declined as steeply as that of beta-lactams, but that of tetracyclines declined considerably, from being the second most popular class in 1995 to the least consumed in 2004. Fluoroquinolones showed a tendency toward a steady increase in consumption until 2003, with stabilization afterwards. By use of reimbursement data, similar trends in antibiotic usage were documented in Belgium (10). The relationship between consumption and resistance is not direct and linear (i.e., it is not a simple relationship of more consumption equaling more resistance) but rather, as mathematical models have suggested (15), indirect and nonlinear. To understand this indirect relationship, it is better to characterize antimicrobial consumption in terms of cumulative consumption instead of year-to-year changes. As the cumulative consumption of antibiotics has slowed down in Belgium since the year 2000, the upward trend of multiple-class resistance has also abated.

An improved explanation for the changes over time in multiple resistance allows us to understand better the changes by geographic region in the 10 years of surveillance. In 2004, multiple-class resistance was more prevalent in the French border region, in areas with higher population density, and in places that had more overall antibiotic consumption. These three factors are independent and additive, so the multiple resistance rate is the highest where all three factors converge (West Flanders), intermediate in areas of intermediate consumption and low population density that border France (Luxembourg province), and lowest in areas with low consumption and high population density bordering The Netherlands (Antwerp). Since the changes in population density and antimicrobial consumption in the 10 years of surveillance have been similar in different regions, the corresponding changes in multiple-class resistance over time have mirrored them. In other words, the maps showing the difference in multiple resistance prevalence in 1994, 2000, and 2004 look very similar; just the magnitude of the numbers has changed.

The analysis of multiple resistance rates at the postal code level had the added benefit of contrasting changes over time in the rates of nonsusceptibility to two or more antibiotics with consumption of antimicrobials at the population level in the same postal codes. Population structure (number of children as a percentage of the population) and population density are also available for this analysis. Previous studies (17) have shown that the relationship between consumption and resistance at the population level is better explained when a 1-year lag in consumption is used. A statistical model that took into account repeated measures showed that multiple-class resistance was significantly related to total antimicrobial consumption in Belgium. The stabilization of the rate of multiple resistance in the country was explained mostly by the decline in the overall consumption of antibiotics.

Turning from the analyses at the population level to the analyses at the individual isolate level, the present study showed that significant risk factors for the development of multiple resistance are age, population density, year of collection, PCV7 serotype, location, and source of the isolate. Interestingly, the analysis demonstrates that there is a border effect in Belgium, which is independent of the population density, source of the isolate, serotype, changes over time, and age of the individual. Isolates coming from the provinces that border France are more likely to be multiply resistant than those coming from provinces that border The Netherlands. Interestingly, population density adds to the explanation of the differences between locations, because there are regions that have higher rates of multiple-class resistance which do not depend on the border effect but rather on the higher population density; these are mostly cities in the middle of the country (Liège, Brussels). Both factors are tied to the place where the isolate was obtained and converge in places such as West Flanders, which has a high population density and borders France, but diverge in other places, like Luxembourg province, which has a low population density and a small share of the French border. Moreover, this more complex explanation elucidates why in the Antwerp region, which has a high population density but low antibiotic use and borders The Netherlands, the prevalence of multiple-class resistance is low.

The interplay of all the factors included in the individual as well as in the population analyses not surprisingly provides a consistent explanation for most of the changes in the rate at which multiple resistance (and, consequently, all other individual resistance rates) in S. pneumoniae has increased and reached a plateau over time and geographic region. Due to the impact of cumulative consumption of antibiotics, which is more important in the same regions bordering France, and aided by population movement and density, multiple resistance increased to levels around 30%, with less tendency toward a continued increase mostly due to stabilization in the cumulative consumption of antibiotics. The differences between the regions have been maintained over time, so that regions bordering France continue to have the highest levels of resistance.

The identification of the PCV7 serotype as a significant risk factor for resistance development suggests that the pneumococcal vaccine undoubtedly would have a favorable short-term impact on resistance in Belgium. Using results from the United States as a guide, if coverage is more than 70%, a relative switch in the prevalence of the vaccine by nonvaccine serotypes should be expected, with an initial drop in single resistance (penicillin and macrolides) but no noticeable long-term effect on multiple resistance (12).

Overall antimicrobial consumption has shown a trend toward a slow decline in Belgium in the period 2000 to 2004. Since resistance, once acquired, is not readily lost (15), this small drop in consumption has slowed the continued increase in multiple-class resistance. Our nonlinear consumption-resistance model therefore would predict that multiple-class resistance would remain stable in the next few years in Belgium unless other unforeseen changes in consumption occur.

Nonetheless, it is expected that as the overall consumption of antibiotics diminishes in Belgium, the relative rise in consumption of other antibiotic classes may become more important. We have shown that the changes of prevalence in resistance to each class of antibiotics separately follow the increase in multiple resistance except, thus far, for fluoroquinolones. We speculate that if multiple-resistance strains acquire fluoroquinolone resistance, it would help maintain fluoroquinolone transmission cycles and presumably increase the prevalence of fluoroquinolone resistance.

Our consumption-resistance model applied to other countries with multiple-resistance rates above 20% (United States, France) indicates that sustained increases in the use of one antimicrobial or one antimicrobial class replacing currently used antimicrobials would continue to exert pressure on multiple resistance, even in an environment of slowly declining overall consumption (unpublished data). We speculate that the introduction of two new antibiotics with different mechanisms of action, and the slow replacement of current failing antimicrobials, is a scenario under which resistance might stay stable, as long as no single antimicrobial dominates usage.

This study has taken advantage of the unique geographical position occupied by Belgium among countries with different pneumococcal resistance rates to apply statistical methods designed to isolate and characterize the various factors that impact drug resistance in S. pneumoniae. The analyses were made powerful by the temporal, ecological, and demographic information and patient characteristics present in the surveillance data. Although many risk factors were significant, multiple antibiotic class resistance is the key factor in understanding the evolution of resistance over time and place in Belgium.

 
We thank Jan Verhaegen from the University Hospital Gasthuisberg for critical reading of the manuscript, Veerle Saelaert from Kind en Gezin and Anne-Françoise Bouvy from the Office de la Naissance et de l'Enfance for the childcare attendance data, and IMS Health Services for access to the Belgian antibiotic consumption data.

Since 2002, partial funding has been provided by the Belgian federal authorities (Belgian Antibiotic Policy Coordinating Committee).

GlaxoSmithKline provided expertise for the statistical analysis.

 
* Corresponding author. Mailing address: 5 Moore Drive, MAI-B252.2E, Research Triangle Park, NC 27709. Phone: (919) 583-6084. Fax: (919) 315-4081. E-mail: robertino.mera{at}gsk.com

Published ahead of print on 6 August 2007.

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 

1
1. Bauraind, I., J. Lopez-Lozano, A. Beyaert, J. Marchal, B. Seys, F. Yane, E. Hendrickx, H. Goossens, P. Tulkens, and L. Verbist. 2004. Association between antibiotic sales and public campaigns for their appropriate use. JAMA 292:2468-2470.[Free Full Text]
2
2. Bronzwaer, S. L., O. Cars, U. Buchholz, S. Molstad, W. Goettsch, I. K. Veldhuijzen, J. L. Kool, M. J. W. Sprenger, and J. E. Degener. 2002. A European study on the relationship between antibacterial use and antibacterial resistance. Emerg. Infect. Dis. 8:278-282.[Medline]
3
3. Diekema, D., A. Brueggemann, and G. Doern. 2000., Antibacterial drug use and changes in resistance in Streptococcus pneumoniae. Emerg. Infect. Dis. 6:552-556.[Medline]
4
4. Draghi, D. C., M. E. Jones, D. F. Sahm, and G. S. Tillotson. 2006. Geographically-based evaluation of multidrug resistance trends among Streptococcus pneumoniae in the USA: findings of the FAST surveillance initiative (2003-2004). Int. J. Antimicrob. Agents 28:525-531.[CrossRef][Medline]
5
5. Farrell, D. J., K. P. Klugman, and M. Pichichero. 2007. Increased antimicrobial resistance among nonvaccine serotypes of Streptococcus pneumoniae in the pediatric population after the introduction of 7-valent pneumococcal vaccine in the United States. Pediatr. Infect. Dis. J. 26:123-128.[CrossRef][Medline]
6
6. Felmingham, D., R. N. Grüneberg, et al. 1996. A multicentre collaborative study of the antimicrobial susceptibility of community-acquired, lower respiratory tract pathogens 1992-1993: the Alexander Project. J. Antimicrob. Chemother. 38(Suppl. A):1-57.[Free Full Text]
7
7. Felmingham, D. 2004, Comparative antimicrobial susceptibility of respiratory tract pathogens. Chemotherapy 50(Suppl 1):3-10.[Medline]
8
8. García-Rey, C., L. Aguilar, F. Baquero, J. Casal, and R. Dal-Ré. 2002. Importance of local variations in antibiotic consumption and geographical difference of erythromycin and penicillin resistance in Streptococcus pneumoniae. J. Clin. Microbiol. 40:159-164.[Abstract/Free Full Text]
9
9. Goossens, H., M. Ferech, R. Vander Stichele, M. Elseviers, et al. 2005. Outpatient antibiotic use in Europe and association with resistance: a cross-national database study. Lancet 365:579-587.[Medline]
10
10. Goossens, H., D. Guillemot, M. Ferech, B. Schlemmer, M. Costers, M. van Breda, L. J. Baker, O. Cars, and P. G. Davey. 2006. National campaigns to improve antibiotic use. Eur. J. Clin. Pharmacol. 62:373-379.[CrossRef][Medline]
11
11. Gruneberg, R. N. 2002. Global surveillance through PROTEKT: the first year. J. Chemother. 14(Suppl 3):9-16.[CrossRef]
12
12. Hanage, W. P., S. S. Huang, M. Lipsitch, C. J. Bishop, D. Godoy, S. I. Pelton, R. Goldstein, H. Huot, and J. A. Finkelstein. 2007. Diversity and antibiotic resistance among nonvaccine serotypes of Streptococcus pneumoniae carriage isolates in the post-heptavalent conjugate vaccine era. J. Infect. Dis. 195:347-352.[CrossRef][Medline]
13
13. Jacobs, M. R., H. J. Koornhof, R. M. Robins-Browne, C. M. Stevenson, Z. A. Vermaak, I. Freiman, G. B. Miller, M. A. Witcomb, M. Isaacson, J. I. Ward, and R. Austrian. 1978. Emergence of multiple resistant pneumococci. N. Engl. J. Med. 299:735-740.[Abstract]
14
14. Klugman, K. 2001. Antibiotic selection in multiply resistant pneumococci. Clin. Infect. Dis. 33:489.[CrossRef][Medline]
15
15. Lipsitch, M. 2001. The rise and fall of antimicrobial resistance. Trends Microbiol. 9:438-444.[CrossRef][Medline]
16
16. Mera, R. M., L. A. Miller, J. J. Daniels, J. G. Weil, and A. R. White. 2005. Increasing prevalence of multidrug-resistant Streptococcus pneumoniae in the United States over a 10-year period: Alexander Project. Diagn. Microbiol. Infect. Dis. 51:195-200.[CrossRef][Medline]
17
17. Mera, R. M. 2005. Predicting the future Streptococcus pneumoniae resistance landscape. Curr. Opin. Pharmacol. 5:459-464.[CrossRef][Medline]
18
18. Mera, R. M., L. A. Miller, and A. White. 2006. Antibacterial use and Streptococcus pneumoniae penicillin resistance: a temporal relationship model. Microb. Drug Resist. 12:158-163.[CrossRef][Medline]
19
19. Musher, D. M. 1992. Infections caused by Streptococcus pneumoniae: clinical spectrum, pathogenesis, immunity, and treatment. Clin. Infect. Dis. 14:801-807.[Medline]
19
20. NCCLS. 2003. Performance standards for antimicrobial disk susceptibility tests, 8th ed. Approved standard M02-A8. NCCLS, Wayne, PA.
19
21. NCCLS. 2004. Performance standards for antimicrobial susceptibility testing; 14th informational supplement. CLSI/NCCLS M100-S14. NCCLS, Wayne, PA.
20
22. Oteo, J., E. Lazaro, F. J. de Abajo, F. Baquero, J. Campos, and Spanish Members of the European Antimicrobial Resistance Surveillance System. 2004. Trends in antimicrobial resistance in 1,968 invasive Streptococcus pneumoniae strains isolated in Spanish hospitals (2001 to 2003): decreasing penicillin resistance in children's isolates. J. Clin. Microbiol. 42:5571-5577.[Abstract/Free Full Text]
21
23. Poole, M. D. 1995. Otitis media complications and treatment failures: implications of pneumococcal resistance. Pediatr. Infect. Dis. J. 14(Suppl.):S23-S26.[Medline]
22
24. Whitney, C. G., M. M. Farley, J. Hadler, L. H. Harrison, C. Lexau, A. Reingold, L. Lefkowitz, P. R. Cieslak, M. Cetron, E. R. Zell, J. H. Jorgensen, and A. Schuchat. 2000. Increasing prevalence of multidrug-resistant Streptococcus pneumoniae in the United States. N. Engl. J. Med. 343:1917-1924.[Abstract/Free Full Text]

---

Antimicrobial Agents and Chemotherapy, October 2007, p. 3491-3497, Vol. 51, No. 10
0066-4804/07/$08.00+0     doi:10.1128/AAC.01581-06
Copyright © 2007, American Society for Microbiology. All Rights Reserved.

This article has been cited by other articles:

• Patrick, D. M. MD MHSc, Hutchinson, J. MD (2009). Antibiotic use and population ecology: How you can reduce your "resistance footprint". CMAJ 180: 416-421 [Full Text]  
• Amrine-Madsen, H., Van Eldere, J., Mera, R. M., Miller, L. A., Poupard, J. A., Thomas, E. S., Halsey, W. S., Becker, J. A., O'Hara, F. P. (2008). Temporal and Spatial Distribution of Clonal Complexes of Streptococcus pneumoniae Isolates Resistant to Multiple Classes of Antibiotics in Belgium, 1997 to 2004. Antimicrob. Agents Chemother. 52: 3216-3220 [Abstract] [Full Text]  
• Flamaing, J., Verhaegen, J., Vandeven, J., Verbiest, N., Peetermans, W. E. (2008). Pneumococcal bacteraemia in Belgium (1994 2004): the pre-conjugate vaccine era. J Antimicrob Chemother 61: 143-149 [Abstract] [Full Text]  

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Reprints and Permissions Copyright Information Books from ASM Press MicrobeWorld Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Van Eldere, J. Articles by Amrine-Madsen, H. Search for Related Content PubMed PubMed Citation Articles by Van Eldere, J. Articles by Amrine-Madsen, H.