Antibiotic consumption and resistance selection in Streptococcus pneumoniae

Journal of Antimicrobial Chemotherapy (2002) 50, Suppl. S2, 27 37 DOI: 10.1093/jac/dkf504 Antibiotic consumption and resistance selection in Streptococcus pneumoniae Fernando Baquero 1 *, Gregorio Baquero-Artigao 2, Rafael Cantón 1 and César García-Rey 3 1 Department of Microbiology, Hospital Ramón y Cajal, Carretera de Colmenar Viejo, km 9.1, 28034 Madrid; 2 Microbial Sciences Foundation, 28024 Madrid; 3 Medical Department, GlaxoSmithKline, Tres Cantos, Madrid, Spain Introduction Selection of antibiotic-resistant Streptococcus pneumoniae is an inescapable consequence of antibiotic use. The correlation between antibiotic consumption and selection of resistant organisms can be shown at every ecological level: patient, community, region or country. In the case of multiple resistance, the intensity of antibiotic selection is increased. However, different antibiotics may exert different selective powers. Because of this, co-selection of macrolide and β-lactam resistance is an asymmetrical phenomenon: macrolides select more efficiently strains resistant to both macrolides and β-lactams than aminopenicillins. The difference in rates of antibiotic resistance is also influenced by the local spread of susceptible or resistant clones; it is suggested that under mild antibiotic selection, the susceptible organisms that are more fit for host-to-host transmission could be favoured. Subsequent acquisition of resistance in these clones may rapidly increase the prevalence of resistance, and that may lead to an increase in the use of antibiotics. The reasons for antibiotic resistance are mainly the reasons explaining antibiotic consumption. A number of possible sociobiological determinants of antibiotic consumption can be identified: genetic factors in the human populations (for instance, involving different symptomatic types of infection), cultural factors and attitudes of patients towards antibiotics, sociological factors or public health factors, including incidence of other infectious diseases in the human population. Only excess in the use of antibiotics should be controlled; for such a purpose, the concept, appropriate demand for antibiotics (ADA) is proposed. If antibiotics are active against microorganisms, and there is a diversity of bacterial phenotypes with respect to the susceptibility of bacteria to the antibiotic, selection of the more resistant among these phenotypes is an inescapable practical consequence of antibiotic consumption. Only inactive agents do not select for resistance. In recent years a number of papers have discussed the possibility of using antibacterial agents at sufficient dosage over sufficient time to eradicate in the host even the bacterial populations with resistant phenotypes (which are not resistant to such intensive exposure to the drug). This approach, when applied to Streptococcus pneumoniae, is limited to a number of antibiotics (for instance, some β-lactams such as amoxicillin), but not applicable to others such as macrolides (since the limitations in the pharmacokinetics of these drugs mean that they will never succeed in reaching the eradication level of exposure for resistant organisms). Forced increase in doses or time of exposure makes the application of such a strategy to drugs with dose-related limited tolerability somewhat doubtful. In such cases, selection of resistant populations will occur. We could succeed in eradicating resistant organisms (minimizing selection) in accurately treated, well supervised individual patients. Nevertheless, we should not forget that selective effects of antibiotics on bacterial populations are complex, because the biological, sociological and cultural background greatly influences the interface between antibiotic exposure and bacteria. Indeed, the selection pressure exerted by antibiotics differs greatly between countries, regions, groups and types of patient. Theoretically, education of prescribers and patients should reduce these differences, but we should not expect such education to have an identical effect in different cultural settings.... *Corresponding author. Tel: +34-91-336-8330; Fax: +34-91-358-0614; E-mail: fbaquero.hrc@salud.madrid.org... 27 2002 The British Society for Antimicrobial Chemotherapy

F. Baquero et al. In addition to bacterial factors (prevalence of antibioticsusceptible and -resistant strains of S. pneumoniae), both human genetic and cultural factors interact to influence the emergence, spread and control of antibiotic resistance. Clearly, the total amount of antibiotics (either in total number of grams or number of prescriptions) used per capita differs between countries, even considering only the developed ones. Recent publications have highlighted the differences existing in European countries, 1 with a clear north south gradient of antibiotic consumption, and this gradient also applies to antibiotic resistance in bacterial pathogens prevalent in the community, as in S. pneumoniae. In most countries, resistance to macrolides and β-lactam antibiotics dominates the current resistance problems, and we will review in this manuscript some of the factors that underpin such problems. Antibiotic use and antibiotic resistance: the ecological links A correlation between antibiotic use and bacterial resistance has been demonstrated repeatedly. In the case of S. pneumoniae, links between antibiotic use and resistance have been consistently found at every ecological level: in patients, 2 small human communities (for instance day-care centres, see later), different geographical areas of the same country, 3 nationally 4 and internationally. 5,6 Figures 1 and 2 reflect the essential coincidence of patterns of use and consumption. To construct these figures, different patterns have been applied to countries with one or two standard deviations above or below the average rates of use and consumption. Clearly, there is in both cases a north south split in use and resistance. A very elegant way of determining the relative responsibilities of the expansion of resistant clones and the selective intensity of the antibiotics in a given environment is to compare the ability of antibiotics to select resistance simultaneously in more than a single bacterial species. This approach was applied recently in comparing the 1 year regional prevalence of macrolide resistance in Streptococcus pyogenes and S. pneumoniae in Spain. A positive correlation was found in site-by-site comparisons, clearly suggesting the link with local antibiotic consumption. 7 Theoretically this type of parallel resistance response in different species, i.e. the resistance syndrome, may also occur because of the trans-species transfer of genetic determinants encoding resistance. Nevertheless, in the case of S. pyogenes and S. pneumoniae, macrolide resistance determinants frequently differ [high prevalence of erm(tr) and erm(b), respectively], whereas they may explain parallel increases of resistance in S. pyogenes and Peptostreptococcus spp. 8 It should be noted that when comparing rates of resistance and rates of consumption we should not necessarily expect a parallelism in the resulting curves. The application of the principles of population genetics to the study of antibiotic resistance indicates that there is probably a critical level of drug consumption required to trigger the emergence of resistance at significant levels. At present, it is unclear whether there is a threshold of antibiotic exposure that must be reached before resistance increases within the community, what the threshold is for the different antibiotics, and whether this threshold varies between countries. Moreover, a sigmoid rise in resistance over time may be the result of a constant rate of antibiotic consumption. 9 Conversely, resistance rates may not clearly decrease in the short or medium term even when antibiotic consumption has been significantly reduced. As both the current use of drugs and the prevalence of resistance are historical phenomena (localities with high levels of use or resistance during a period of time also tend to have high levels during the next period), it has been suggested recently that the best correlation might be obtained by considering not the prevalence of antibiotic resistance in a given pathogen, but the rate of change. 10 For the same reasons, the time series methodology of analysis of the correlation between use and resistance is becoming a promising research tool. 11 Structure of the selective landscape for antibiotic resistance Not all classes of antibiotics are equally selective for drugresistant S. pneumoniae populations. Even within a welldefined group of antibiotics, some members may exert a different selective power from others on particular resistant variants. This phenomenon is a consequence of the different mechanisms of action, and of the efficiency of bacterial mechanisms of resistance. It should be noted that the pharmacokinetics and pharmacodynamics of the drugs may imbue different antibiotics with different selective qualities. 12 Eradication potency of antibiotics Different antibiotics have quite different eradication power for S. pneumoniae. Reduction in the number of oropharyngeal carriers after conventional therapy ranges in different studies between 48% and 80% for co-amoxiclav (conventional dosage), 22% and 53% for cefuroxime, 41% and 45% for cefpodoxime, 12% and 18% for cefixime, and 14% and 17% for cefaclor. 13 The eradication of susceptible organisms, together with the absence of total eradication of the resistant variants, will lead to resistance selection. Because of this, antibiotics should be able to eradicate both the susceptible target organisms (to prevent recurrences) and the resistant variants that may be present as part of the population(s) harboured in the patient (to prevent replacement of the susceptible populations by the resistant ones). Note that both phenomena are linked: increases in antibiotic resistance reduce the probability of 28

Antibiotic consumption and resistance selection in S. pneumoniae Figure 1. (a) Consumption of broad-spectrum penicillins in 1997 (standard deviation, defined daily doses/1000 inhabitants/day). Source: Cars et al. 1 (b) Penicillin non-susceptible pneumococci in 2000 (standard deviation). Source: European Resistance Surveillance (EARSS) database; Alexander project data; PROTEKT resistance surveillance. achieving eradication, and failure to eradicate bacteria may promote the emergence and dissemination of antimicrobialresistant clones. 14 Low dosages and long duration of treatments with many β-lactam drugs increase the risk of selection (and also increase the rate of carriage) of penicillin-resistant S. pneumoniae. 15 This effect was predicted by in vitro experiments, showing that low concentrations of penicillins selected lowlevel resistance more efficiently than high-level resistance in mixed cultures with predominance of low-level resistant 29

F. Baquero et al. Figure 2. (a) Consumption of macrolides in 1997 (standard deviation, defined daily doses/1000 inhabitants/day). Source: Cars et al. 1 (b) S. pneumoniae erythromycin resistance in 2000 (standard deviation). Source: European Resistance Surveillance (EARSS) database; Alexander project data; PROTEKT resistance surveillance. variants. 16 In contrast, short-course, high-dose amoxicillin therapy reduces pneumococcal carriage and minimizes the impact of therapy on the spread of drug-resistant pneumococci. 17 In some cases, increased dosages or improved galenic formulations ensuring optimal lengths of contact of the antibiotic with the target bacterium may be able to eradicate both susceptible and resistant populations. 18 20 In that case we 30

Antibiotic consumption and resistance selection in S. pneumoniae could have an apparently paradoxical effect of antibiotic consumption: low dosages (low consumption) may eventually select more resistance than high consumption. 21 Despite their bacteriostatic mechanism of action, macrolides are excellent eradicator agents for macrolide-susceptible pneumococci. Conversely (and in contrast to the behaviour of some penicillins), when used on macrolide-resistant organisms, macrolides have an eradication power similar to that of placebo. 22 Co-selection of macrolide and β-lactam resistance: an asymmetrical phenomenon Not only the amount, but also the usage of the different families of antibiotics differs between individual countries. As pneumococcal clones with different blends of antimicrobial resistance determinants may invade different geographical areas, selection will depend on the interaction of the consumption of multiple antibiotics and the multi-resistant organisms. For instance, in many countries S. pneumoniae clones are simultaneously resistant to β-lactams and macrolides. In Spain, there was a strong correlation between the prevalence of high-level penicillin resistance and that of erythromycin resistance (r = 0.903). 4 The use of macrolides will select for both macrolide resistance and β-lactam resistance. This has been documented in countries colonized with β-lactam/macrolide-resistant strains. 13 We should therefore expect the converse to be true: β-lactams should act as selectors of macrolide resistance. However, it has been shown that the respective co-selection by macrolides and penicillins cannot be represented as a symmetrical phenomenon. Multivariate models heve shown that, in relative terms, in Spain, a country where co-resistance to β-lactams and macrolides is extremely frequent, at equal levels of consumption, macrolides were almost five-fold more likely to select for erythromycin resistance, and threefold more likely to do so for penicillin resistance, than were β-lactams. 3 In other words, macrolides efficiently select for strains with erythromycin resistance, including those with erythromycin and penicillin resistance. Penicillins (typically high-dosage amoxicillin) select less efficiently for penicillinresistant strains (as a proportion of organisms are inhibited), including strains resistant to both penicillin and erythromycin. That means that in a population in which doubleresistant strains are frequent, and where the use of amoxicillin is prevalent, there is a possibility of finding a proportion of erythromycin-susceptible clones, but not if the use of macrolides is prevalent. Obviously, maximal strength of selection will occur if both types of antibiotic (even more if the use of oral cephalosporins is prevalent) are intensively used, as is the case in France or Spain. Interestingly, differences in the ability to select resistant populations can be found among different members of the same family of antibiotics. Differences in selective power of different β-lactams and macrolides In vitro modelling studies have been performed by challenging mixtures of S. pneumoniae strains with different levels of penicillin susceptibilities (MICs 0.015, 0.5, 1 and 2 mg/l) with different concentrations of β-lactam agents. Amoxicillin was a poor selector for the strains with higher MICs in comparison with oral cephalosporins, particularly cefixime. 22 Indeed, co-amoxiclav reduced nasopharyngeal carriage of penicillin-susceptible S. pneumoniae by 90%, and of penicillin-resistant strains by 60%. In contrast, cefixime showed a reduction of only 30% of penicillin-susceptible isolates, without any impact on the carriage of resistant strains. 23 Confirming these observations, ecological studies based on data recovered in the Alexander Project have also suggested that a decrease in the prescription ratio of aminopenicillins compared with cephalosporins per 1000 habitants could be correlated with the increase in penicillin resistance in France. 24 Totally consistent results for the higher selective value of oral cephalosporins were obtained by sequential in vitro selection. 25 In addition, we showed in an ecological study in Spain that the prevalence of high-level penicillin resistance correlated with consumption of oral cephalosporins (adjusted r 2 = 0.877). 4 The reasons for such differences are related mainly to the effect of different pharmacokinetics of these drugs on susceptible and resistant pneumococcal populations. 26 Among macrolide agents, an ecological study showed that the relationship between erythromycin resistance and macrolide consumption was probably due mainly to consumption of long-acting macrolides (adjusted r 2 = 0.886) rather than to the consumption of erythromycin. 4 This is probably due to the long periods of exposure of bacterial populations to selective concentrations of antibiotics for resistant organisms. 4,27 Confirming these findings, a recent study showed that once-a-day macrolides are 50% more prone than other macrolides to cause erythromycin resistance at a given site. 3 Clonal dynamics and antibiotic resistance The clonal dissemination of certain resistant S. pneumoniae organisms is essential in understanding local resistance rates. For instance, an epidemic strain harbouring mechanisms of antibiotic resistance may invade a geographical area for reasons unrelated to selection of resistance. In that case, an antibiotic resistance epidemic may occur in total independence of antibiotic consumption. Conversely, epidemics may occur involving clones susceptible to different antibiotic groups. If that were the case, we should expect more susceptibility in the invaded regions. The probability of developing (or acquiring) antibiotic resistance seems not to be equivalent among the different pneumococcal clones. Because of this, it is fully conceivable that in some regions resistance could 31

F. Baquero et al. remain low despite substantial increases in antibiotic use. Only occasionally, if a resistant clone arrives in this region, would the selective activity of antibiotics (selecting the clone against the susceptible indigenous clones) be fully exerted, and resistance rates might increase rapidly. The natural history of the emergence and spread of pneumococcal resistance in different countries mimics a sigmoid growth curve. A possible cryptic phase is followed by an emergence phase, and once a threshold of prevalence of resistance is crossed, a rapid (log-like) penetration phase tends to occur, followed finally by a stationary phase. 28 Mathematical modelling accounts for these dynamics. 9 Under a high antibiotic intensity of selection in a human population, it is possible that variants of susceptible (or lowlevel resistant) bacteria that are particularly fit for host-tohost transmission have a higher probability of survival (moving to non-treated hosts). If resistance develops among these hyper-transmissible variants (as an effect of its growing density in the population), we would expect a rapid penetration of resistance. When the intensity of the antibiotic effect on the bacterial clone decreases (as it is resistant), features ensuring hyper-transmissibility may no longer constitute an advantage in comparison with better colonizing abilities in each particular host. The current lack of animal models for the study of in vivo transmissibility of S. pneumoniae makes it difficult to prove (or disprove) this hypothesis. Perhaps this approach should be tested with other streptococci able to colonize laboratory animals, eventually harbouring (selectable) resistance determinants from S. pneumoniae. Studies on the population biology (clonal structure) of S. pneumoniae may help to explain the discrepancies between resistance rates in different regions of a wide geographical area. Indeed, at national level, comparison of means and variances of resistance among regions or towns (or hospitals within a town) may be useful to describe the phenomenon, and understand the possible discrepancies with antibiotic consumption. In that respect, the possibility of creating maps of selective environments at national or sub-national level (even at hospitals) for particular dangerous (resistant) clones could be of great interest. Sociobiological determinants of antibiotic consumption The reasons for antibiotic resistance are mainly the reasons that explain antibiotic consumption. It is essential to investigate the reasons that may explain the huge local and countryto-country differences in antibiotic consumption. 1 Very few studies have been conducted to document the determinants of antibiotic consumption. Many factors have been suggested to influence antibiotic consumption, some of which are listed below. Genetic factors Genetic factors may influence the host susceptibility to infectious diseases. 29 The bacteria host relationship has been refined during millions of years of co-evolution and the subtle specificity arising from these interactions could be determined by the host, resulting in different infection-response patterns in different populations. For instance, northern human populations may have evolved a less symptomatic response to infection (for instance, involving weaker induction of biological response modifiers), which could influence the personal demand for antibiotics. Eventually these populations, to compensate for the obvious environmental risks (such as very cold and humid winters) for respiratory tract infections (RTIs), could have developed certain types of mucosal immunity, decreasing the ability of certain bacterial populations to produce heavy colonization. Less dense bacterial populations will decrease the possibility of emergence of antibiotic resistance. It may be hard to test these hypotheses, as the sociological environment (for instance, crowding in winter in igloos in Alaska) may provide confounding factors. 30 The years to come should add important data on the influence of host genetics on colonization, transmission and infection by microorganisms. For instance, some human populations could be more genetically prone than others to harbour and disseminate particular S. pneumoniae clones (see later). It is a well-known fact that the prevalence of resistance is strongly influenced by the local spread of a relatively low number of resistant bacterial clones. Herd immunity, decreasing the dissemination of a resistant clone in a particular geographical area, will certainly influence the resistance rates of a given pathogen. In general, genetic diversification in human populations will decrease the spread of a given bacterial clone, but at the same time may increase the risk of importation of foreign clones. To what extent both the immigration of resistance and the stabilization plateau of resistant S. pneumoniae in different countries are influenced by the local genetic structure of the host populations remains to be investigated. Pharmacogenomic studies may show that human populations can also differ in their ability to absorb, distribute, metabolize and excrete antibiotics. By affecting the pharmacokinetics of antibiotics, these differences may also influence the emergence and selection of antibiotic-resistant microorganisms. It is interesting to note that European populations speaking Germanic languages have lower antibiotic consumption and less antibiotic resistance than populations speaking Mediterranean or Ugro-Altaic languages, and there is a known correlation between genetic and linguistic diversity. It should be noted that antibiotic resistance is not only the effect, but also an essential cause, of consumption of antibiotics (resistant organisms require higher dosages to be controlled, or the use of alternative antibiotics). 32

Antibiotic consumption and resistance selection in S. pneumoniae Cultural factors and attitudes Cultural factors and attitudes of patients among different human populations may also contribute to the relationship between tolerance of infective symptoms, severity of symptoms and antibiotic use, thereby affecting resistance rates. This is reflected by significantly higher community antibiotic consumption, and pneumococcal resistance rates, in southern Europe. For example, the level of symptom severity at which antibiotic therapy is demanded may be lower in southern European populations than in northern European populations. Psychological tolerance to symptoms versus ask-for-help attitudes may be influenced by culture and tradition. As a way of considering the psychological influence of long historical patterns of education (for instance, priming self-responsibility in looking for help versus confidence in the group) we recently indicated that antibiotic consumption in countries with predominantly Protestant populations is consistently lower than in those with predominantly Catholic populations. This suggestion has been illustrated recently by Dominique Monnet (ESAC Meeting, Brussels 2001). Protestant countries have a mean consumption of antibiotics near to 13 defined daily doses (DDDs)/1000 inhabitants/day, and this figure is doubled in countries of Catholic majority, this difference being statistically significant. A frequent patient perception in southern European countries is: A similar (mild) infection (or any other illness) to that I am now suffering from was successfully treated on a previous occasion with a particular drug. As I have such a drug stored at home, I will start with it, or try to get more of it at the pharmacy, or ask my doctor for it. Such an attitude is the result of public awareness of antibiotics perceived as a cure for infections, often available overthe-counter (OTC), unlike for example angiotensin II antagonists. Patients frequently stop self-medication (and even prescribed drugs) when they feel better, unused antibiotics being kept for the next episode. Domestic antibiotic pollution has been clearly demonstrated in some surveys. Telephone interviews with 1000 households in Spain showed that in 42% of the homes, one (88.1%) or more antibiotic packets were present (80% of these were the remainder of a prescription). 31 Economic pressures also play a role, particularly when both parents are working. The parent will have to take time off work, certainly if children with infection are not allowed in day-care centres until the episode resolves. Sociological factors Sociological factors also play a role in the spread of resistance. For example, the proportion of children within the population and the frequency of day-care centre use both influence the prevalence of resistance in the community. The proportion of oropharyngeal carriers of S. pneumoniae reaches a maximum ( 50 60%) at 1 2 years of age, a figure that decreases by 50% when children are 8 10 years old. 32 A number of recent papers have highlighted the role of daycare centres in the transmission of S. pneumoniae local resistant clones among children. 33,34 Elderly residential care populations are also important, as they may spread hospitalacquired resistant pathogens. Variation in family structures, and differences between urban and rural populations, are also important. In view of these social influences, policies on antibiotic usage must be tailored for individual countries. Epidemiological factors To understand prescribing practices, it is important to consider how the epidemiology of viral RTIs, as well as bacterial RTIs, and even other types of infection, drives antibiotic usage. In a study performed in Spain, a country with high levels of antibiotic consumption and bacterial resistance, the use of β-lactam and macrolide antibiotics during the years 1995 1997 correlates faithfully with the incidence of cases diagnosed as influenza-like illness. 35 The use of quinolone antimicrobial agents, not used in RTIs, did not show such a correlation. In support of this observation, Dominique Monnet 36 recently presented data suggesting that countries declaring in 1997 widespread or regional influenza-like illness, as did France and Spain, were much higher consumers of antibiotics than countries with no activity, such as Denmark or the Netherlands. It seems probable that a number of febrile viral diseases are treated with antibiotics, particularly among older people, on a preventive basis; at the same time, viral respiratory tract diseases may contribute to the spread in the population of resistant bacterial organisms. It is interesting that children suffering from rhinopharyngitis are more frequently carriers of S. pneumoniae. 30 The selective landscape may change with the introduction and wide dissemination of pneumococcal conjugate vaccines. In a recent 2 year followup study, it has been shown that the rate of carriage of vaccinetype pneumococci was lower among toddlers attending day-care centres who had received a nine-valent conjugate vaccine than among control subjects. The effect is due to a lower acquisition rate in the vaccinated group. 37 Even the frequency of non-rtis with cross-therapy with RTIs could be considered as a driver of resistance, for instance, the use of aminopenicillins and macrolides to control Helicobacter pylori in peptic ulcer, the use of respiratory fluoroquinolones for UTIs or the use of macrolides to prevent the presumed Chlamydia pneumoniae infections that are supposed to play a role in coronary artery diseases. The future wide application of new, effective, safe and non-expensive drugs for respiratory viral diseases could represent a major contribution to reducing the consumption of antibiotics and perhaps the long-term decrease in antimicrobial resistance. 33

F. Baquero et al. Culture and perceptions among community prescribers and suppliers Many prescribers in community practice do not fully appreciate the rationale for restricting the use of antibiotics. Owing to the self-limiting nature of many community-acquired RTIs, and the fact that antibiotics can often eradicate RTI pathogens, even those that are defined as resistant by standard in vitro testing, prescribers may not directly experience treatment failure attributable to bacterial resistance. Spontaneous improvement or resolution tends to be attributed to medical intervention (prescription). On the other hand, many prescribers have a preventive feeling: their antibiotic prescription may have prevented complications of the RTIs or meningitis (and any possibly associated legal challenge), or may have contributed to the eradication of some important diseases (rheumatic fever). These attitudes (defensive medicine) are expected to be particularly relevant in areas in which the mean time per consultation is low. It is also clear that prescription is expected to be proportional to the number of prescribers. In countries with a free-market health environment, a high number of prescribers could produce competition for patients, and therefore more tolerance of the desires expressed by them. The same is true for the number of suppliers (pharmacists), particularly in countries with high tolerance to providing OTC pharmaceuticals. Figure 3 illustrates this point. Comparison of this figure with Figures 1 and 2 reveals the influence of these factors on the north south pattern of antibiotic use and antibiotic resistance. Appropriate demand for antibiotics (ADA) It is essential to understand that antibiotics are absolutely needed to maintain high health quality in modern societies. In the cases in which antibiotics are indicated, the benefits outweigh by far the risks, including antibiotic resistance. Because of that, we should propose that only the excessive use of antibiotics constitutes a dangerous practice for public health. The question is how to quantify and control this excess. In general, we should be able to construct, for each particular area and period of time, a theoretical line of appropriate use of antibiotics, correlating (in the abscissa) with the number of infections in which antibiotics are indicated (there is an expectation about a positive intervention in the outcome of the infection), and the rate of use of antibiotics. For instance, during epidemics, the use of antibiotics may also increase in a totally justified way. If that increase does not occur, there may be under-prescription of antibiotics. If an increase occurs without any associated increase in the number of cases, other factors unrelated to justified demand are influencing the prescription trends, and measures should be taken to control them. The use of the parameter number of infections in which antibiotics are indicated is just an approximation that should be further refined. For instance, the indication for therapy of pneumococcal pneumonia by a macrolide may exist, but this indication should be modulated by the local prevalence of macrolide resistance in S. pneumoniae, increasing the expected failure rate (the indication exists in general, but may be not in a particular location). The severity of infection (including chronic infections) should also be considered: patients with severe or chronic infections receive (in a justified way) more antibiotics than do patients with acute, milder infections. It could be useful to build an integrated parameter considering these facts, to better correlate antibiotic consumption with real justified needs. This parameter should reflect the ADA in a particular time/space frame. ADA should be defined in a more specific way in relation to: (i) intent-to-treat organism (for instance, S. pneumoniae, eventually, only macrolide-resistant S. pneumoniae). Clearly in countries with a high prevalence of resistance, high consumption of a given antibiotic may be justified but not in another country with a different resistance pattern; (ii) location of the patient: sub-country locations for communityacquired infections, but also for hospitals, ICUs, elderly facilities or day-care centres; (iii) age and gender of the patients. For instance, in the absence of paediatric DDDs, the correlations between use in grams and appropriate consumption may be misleading; (iv) type of infection are we using antibiotics appropriately in acute otitis media or chronic bronchitis? The way of representing antibiotic use is certainly a major issue. The classic methods based on DDDs require further European standardization. The main difficulty is the discrepancies between countries of antibiotic dosage and length of therapy in different types of infection. In some cases, this is due to differences in antibiotic susceptibilities of frequent pathogens. Perhaps a local correction index could be introduced to obtain a more faithful representation of the real consumption. This index could be obtained on the basis of realistic prescribed daily doses (PDDs). Cosentino et al. 38 have suggested that DDD/day/ADU (apparent drug users, that is, individuals for whom at least one prescription of the drug has been dispensed during a given time period) = PDD days. Correlation studies of antibiotic use and antibiotic resistance in particular pathogens can be carried out with each one of these values representing antibiotic use. For many pathogens that are members of normal microbiota, we can theoretically consider the total amount of antibiotics used (grams) in a period of time as a collective prescription of a given selective intensity exerted on a collective microbial flora. Indeed, the main difficulty of these studies is based on the current international classification of antimicrobial agents in families (J n -groups). This classification is most frequently used for usage analysis and it is not always helpful to our 34

Antibiotic consumption and resistance selection in S. pneumoniae Figure 3. (a) Pharmacists per 1000 inhabitants in 1998 (standard deviation). (b) Over-the-counter pharmaceutical expenditure (last year available for all countries; standard deviation). Source (a and b): European Health For All database, WHO Regional Office for Europe, Copenhagen, Denmark. understanding of the relationship between use and resistance, since these groups are highly heterogeneous in terms of selective activity for particular mechanisms of resistance. Perhaps, it would be desirable for this particular purpose to build new groups of antibiotics with common selective activities on particular mechanisms or even (why not?) on particular bacterial clones that we are trying to contain. Clearly, many factors play a role in the development of antibiotic resistance and the interplay of these factors needs to be better understood. The key to achieving benefits in terms of 35

F. Baquero et al. curing infectious diseases is to be able to assess accurately the risk of antibiotic resistance in populations and then to manage it appropriately. Mathematical modelling may be a way to condense complex sets of dispersed data and produce meaningful predictions able to inspire successful interventions. Acknowledgements The authors are greatly indebted to both Lorenzo Aguilar (Medical Department, GlaxoSmithKline Spain) for his scientific support to the Consumption/Resistances Programme, and to Gerry Halls for providing data and thoughts during several discussions about antibiotic consumption and resistance. References 1. Cars, O., Mölstad, S. & Melander, A. (2001). Variation in antibiotic use in the European Union. Lancet 357, 1851 3. 2. del Castillo, F., Baquero-Artigao, F. & García-Perea, A. (1998). Influence of recent antibiotic therapy on antimicrobial resistance of Streptococcus pneumoniae in children with acute otitis media in Spain. Pediatric Infectious Disease Journal 17, 94 7. 3. García-Rey, C., Aguilar, L., Baquero, F., Casal, J. & Dal-Ré, R. (2002). Importance of local variations in antibiotic consumption and geographical differences of erythromycin and penicillin resistance in Streptococcus pneumoniae. Journal of Clinical Microbiology 40, 159 64. 4. Granizo, J. J., Aguilar, L., Casal, J., García-Rey, C., Dal-Ré, R. & Baquero, F. (2000). Streptococcus pneumoniae resistance to erythromycin and penicillin in relation to macrolide and β-lactam consumption in Spain (1979 1997). Journal of Antimicrobial Chemotherapy 46, 767 73. 5. Felmingham, D. & Grüneberg, R. N. (2000). The Alexander Project 1996 1997: Latest susceptibility data from this international study of bacterial pathogens from community-acquired lower respiratory tract infections. Journal of Antimicrobial Chemotherapy 45, 191 203. 6. Baquero, F., Barrett, J. F., Courvalin, P., Morrisey, I., Piddock, L. & Novick, W. J. (1998). Epidemiology and mechanisms of resistance among respiratory tract pathogens. Clinical Microbiology and Infection 4, Suppl. 2, S19 26. 7. Pérez-Trallero, E., Fernández-Mazarrasa, C., García-Rey, C., Bouza, E., Aguilar, L., García-de-Lomas, J. et al. (2001). Antimicrobial susceptibilities of 1,684 Streptococcus pneumoniae and 2,039 Streptococcus pyogenes isolates and their ecological relationships: Results of a 1-year (1998 1999) multicenter surveillance study in Spain. Antimicrobial Agents and Chemotherapy 45, 3334 40. 8. Reig, M., Galán, J., Baquero, F. & Pérez-Díaz, J. C. (2001). Macrolide resistance in Peptostreptococcus spp. mediated by ermtr: possible source of macrolide lincosamide streptogramin B resistance in Streptococcus pyogenes. Antimicrobial Agents and Chemotherapy 45, 630 2. 9. Austin, D. J., Kristinsson, K. G. & Anderson, R. M. (1999). The relationship between the volume of antimicrobial consumption in human communities and the frequency of resistance. Proceedings of the National Academy of Sciences, USA 96, 1152 6. 10. Lipsitch, M. (2001). The rise and fall of antimicrobial resistance. Trends in Microbiology 9, 438 44. 11. López-Lozano, J. M., Monnet, D. L., Yague, A., Burgos, A., Gonzalo, N., Campillos, P. et al. (2000). Modelling and forecasting antimicrobial resistance and its dynamic relationship to antimicrobial use: a time series analysis. International Journal of Antimicrobial Agents 14, 21 31. 12. Ball, P., Baquero, F., Cars, O., File, T., Garau, J., Klugman, K. et al. (2002). Antibiotic therapy of community respiratory tract infections: strategies for optimizing outcomes and minimizing resistance emergence. Journal of Antimicrobial Chemotherapy 49, 31 40. 13. Cohen, R. & Varon, E. (2001). Flore rhinopharyngée: impact des traitements antibiotiques conventionnels. La Lettre de l Infectiologue 16, 12 6. 14. Dagan, R., Klugman, K. P., Craig, W. A. & Baquero, F. (2001). Evidence to support the rationale that bacterial eradication in respiratory tract infection is an important aim of antimicrobial therapy. Journal of Antimicrobial Chemotherapy 47, 129 40. 15. Guillemot, D., Carbon, C., Balkau, B., Geslin, P., Lecoeur, H., Vauzelle-Kervroedan, F. et al. (1998). Low dosage and long treatment duration of β-lactam: risk factors for carriage of penicillinresistant Streptococcus pneumoniae. Journal of the American Medical Association 279, 365 70. 16. Baquero, F., Negri, M. C., Morosini, M. I. & Blázquez, J. (1997). The antibiotic selective process: concentration-specific amplification of low-level resistant populations. Ciba Foundation Symposium 207, 93 111. 17. Schrag, S. J., Pena, C., Fernández, J., Sánchez, J., Gómez, V., Pérez, E. et al. (2001). Effect of short-course, high-dose amoxicillin therapy on resistant pneumococcal carriage: a randomized trial. Journal of the American Medical Association 286, 49 56. 18. Kaye, C. M., Allen, A., Perry, S., McDonagh, M., Davy, M., Storm, K. et al. (2001). The clinical pharmacokinetics of a new pharmacokinetically enhanced formulation of amoxicillin/clavulanate. Clinical Therapeutics 23, 578 84. 19. Berry, V., Singley, C., Satterfield, J. & Woodnutt, G. (2001). Efficacy of a pharmacokinetically enhanced formulation of amoxicillin/clavulanate against experimental respiratory tract infection (RTI) in rats caused by Streptococcus pneumoniae (Sp.). In Program and Abstracts of the Forty-first Interscience Conference on Antimicrobial Agents and Chemotherapy, Chicago, IL, USA, 2001. Abstract B-988, p. 53. American Society for Microbiology, Washington, DC, USA. 20. File, T., Jacobs, M., Poole, M., Richard, M. & Wynne, B. (2001). Clinical efficacy of pharmacokinetically enhanced amoxicillin/ clavulanate (AMX/CA) vs. comparators against Streptococcus pneumoniae (Sp) in respiratory tract infections (RTIs). In Program and Abstracts of the Forty-first Interscience Conference on Antimicrobial Agents and Chemotherapy, Chicago, IL, USA, 2001. Abstract L-866, p. 451. American Society for Microbiology, Washington, DC, USA. 21. Lipsitch, M. (2001). Measuring and interpreting associations between antibiotics and penicillin resistance in Streptococcus pneumoniae. Clinical Infectious Diseases 32, 1044 54. 36

Antibiotic consumption and resistance selection in S. pneumoniae 22. Negri, M. C., Morosini, M. I., Loza, E. & Baquero, F. (1994). In vitro antibiotic selective concentrations of β-lactams for penicillinresistant Streptococcus pneumoniae populations. Antimicrobial Agents and Chemotherapy 38, 122 5. 23. Dabernat, H., Geslin, P., Megraud, F., Begue, P., Boulesteix, J., Dubreuil, C. et al. (1998). Effects of cefixime or co-amoxiclav treatment on nasopharyngeal carriage of Streptococcus pneumoniae and Haemophilus influenzae in children with acute otitis media. Journal of Antimicrobial Chemotherapy 41, 253 8. 24. Baquero, F. (1996). Antibiotic resistance of respiratory pathogens: an analysis and commentary on a collaborative surveillance study. Journal of Antimicrobial Chemotherapy 38, Suppl. A, 117 32. 25. Pankuch, G. A., Jueneman, S. A., Davies, T. A., Jacobs, M. R. & Applebaum, P. C. (1998). In vitro selection of resistance to four β-lactams and azithromycin in Streptococcus pneumoniae. Antimicrobial Agents and Chemotherapy 42, 2914 8. 26. Thorburn, C. E., Knott, S. J. & Edwards, D. I. (1998). In vitro activities of oral β-lactams at concentrations achieved in humans against penicillin-susceptible and -resistant pneumococci and potential to select resistance. Antimicrobial Agents and Chemotherapy 42, 1973 9. 27. Baquero, F. (1999). Evolving resistance patterns of Streptococcus pneumoniae: a link with long-acting macrolide consumption? Journal of Chemotherapy 11, Suppl. 1, 35 43. 28. Baquero, F. (1996). Epidemiology and management of penicillinresistant pneumococci. Current Opinion in Infectious Diseases 9, 372 9. 29. Qureshi, S. T., Skamene, E. & Mallo, D. (1999). Comparative genomics and host resistance against infectious diseases. Emerging Infectious Diseases 5, 36 47. 30. Cohen, R. & Varon, E. (2001). Variations de la flore oro- et rhinopharyngée en dehors des antibiotiques. La Lettre de l Infectiologue 7, 4 8. 31. Orero, A., González, J. & Prieto, J. (1997). Antibiotics in Spanish households. Medical and socioeconomical implications. URANO Study Group. Medicina Clínica 109, 782 5. 32. Loda, F. A., Collier, A. M., Glezen, W. P., Strangert, K., Clyde, W. A. & Denny, F. W. (1975). Occurrence of Diplococcus pneumoniae in the upper respiratory tract of children. Journal of Pediatrics 87, 1087 93. 33. Sá-Leao, R., Tomasz, A., Santos-Sanches, I., Nunes, S., Alves, C. R., Brito-Avô, A. et al. (2000). Genetic diversity and clonal patterns among antibiotic-susceptible and -resistant Streptococcus pneumoniae colonizing children: day care centres as autonomous epidemiological units. Journal of Clinical Microbiology 38, 4137 44. 34. Sá-Leao, R., Tomasz, A., Santos-Sanches, I., Brito-Avô, A., Vilhelmsson, S. E., Kristinsson, K. G. et al. (2000). Carriage of internationally spread clones of Streptococcus pneumoniae with unusual drug resistance patterns in children attending day care centres in Lisbon, Portugal. Journal of Infectious Diseases 182, 1153 60. 35. Moreno, R. (2000). Ph.D. Thesis, School of Medicine, Universidad Complutense de Madrid, Spain. 36. Monnet, D. (2001). European Conference on Antibiotic Consumption, ESAC November 2001. http://oms2.b3e.jussieu.fr-/flunet/ activity.html 37. Dagan, R., Givon-Lavi, N., Zamir, O., Sikuler-Cohen, M., Guy, L., Janco J. et al. (2002). Reduction of nasopharyngeal carriage of Streptococcus pneumoniae after administration of a 9-valent pneumococcal conjugate vaccine to toddlers attending day care centres. Journal of Infectious Diseases 185, 927 36. 38. Cosentino, M., Leoni, O., Banfi, F., Lecchini, S. & Frigo, G. (2000). An approach for the estimation of drug prescribing using the defined daily dose methodology and drug dispensation data. European Journal of Clinical Pharmacology 56, 513 7. 37