Corresponding author and reprint requests: Keryn Christiansen, Department of Microbiology and Infectious Diseases, Royal Perth Hospital, Perth, Western Australia, Australia Tel: +61 89224 2210 Fax: +61 89224 1989 E-mail: Keryn.Christiansen@health.wa.gov.au
There are multiple interventions available that may help to control the development and spread of resistance to antimicrobial agents in bacteria implicated in community-acquired respiratory tract infections. Unfortunately, very few studies have assessed the effectiveness of these interventions using objective end-points, such as reduction in resistance rates and improvement in clinical outcomes. Most interventions are centered on reducing inappropriate or unnecessary use of antibiotics; others focus on reducing disease burden and bacterial colonization. With regard to antibiotic use, efforts should be concentrated at both the prescriber and consumer levels. Interventions that target prescribers include: provision of educational materials; strategies and tools to improve diagnosis; implementation of practice guidelines; personalized interactive sessions with feedback on the practice profile; and use of delayed prescription and alternative prescribing strategies. Optimal results are usually obtained when these interventions are combined with consumer education. Regulatory interventions (e.g. licensing regulations and controlled access to drugs), restrictions in the use of agents for growth promotion in animals, and use of nonantimicrobial therapies (e.g. probiotics) may help further to reduce inappropriate antibiotic use and thereby decrease the selective pressure for development of resistance. Infection-control strategies, public health measures, vaccination programs, and new antibiotics all have a role in minimizing the spread of resistant organisms. Ideally, resistance-control programs should include predefined criteria for success and integral audit processes based on objective end-points (antibiotic use, resistance trends, and health outcomes). Standardization of data collection is imperative so that the relative merits of various interventions can be compared. Effective implementation and audit of interventions is often difficult in developing countries owing to poor health-care infrastructures, lack of resources, poor education/training, and minimal regulatory controls on the supply and quality of antimicrobials. Substantial support from governments and health-care organizations across the globe is required to initiate and sustain effective intervention programs to control antimicrobial resistance.
Antimicrobial resistance poses a significant global health problem. To try to tackle this problem, a large body of knowledge, encompassing areas as diverse as molecular mechanisms and epidemiology to behavioral science, is being accumulated. The ultimate aim of this research is to identify means by which resistance can—as a minimum—be controlled and—at best—be eliminated. The determination of successful intervention strategies is pivotal to this aim.
Most commonly, interventions are centered on a reduction in the volume of antimicrobial use. However, the relationship between antimicrobial use and resistance is complex, and although there is evidence for a correlation, the threshold at which clinically significant resistance occurs remains unknown. A quantitative relationship has been proposed by mathematical modeling using population genetic methods and epidemiologic observations . It would be simplistic, however, to conclude that the absolute volume of antibiotic use is solely responsible for resistance selection. Other factors that may also be important in the selective process include: the transmissibility of resistance determinants within and between species and genera; varying bacterial mutation rates; dose and duration of antibiotic therapy; and variations in clonal spread through cultural and lifestyle differences between populations. The challenge for those researching in this area is to determine which of these factors play pivotal roles and, thereby, to identify the most productive directions for intervention strategies to control resistance.
Many interventions to control resistance have been suggested and implemented, some without direct evidence that the target activity does in fact lead to resistance. This paper first outlines the tools used to audit the impact of these interventions. It then reviews a wide range of interventions, assessing the means by which they have been audited and the results obtained. These interventions are broadly divided in to those concerning antibiotic use and those concerning disease burden and colonization. Finally, this paper discusses problems associated with implementing and auditing interventions in low-resource countries.
It is not the intention of this paper to cite all research published to date. Rather, selected studies that are representative of the overall body of work are reviewed.
Monitoring the impact of antimicrobial resistance control interventions
Audit is essential to measure the success of interventions designed to reduce antimicrobial use and resistance, and improve health outcomes. Many interventions are initiated before the end-points for success, the magnitude of the response that is considered clinically meaningful, or the time over which the response is expected have been established. Moreover, measurement of specific resistance determinants before and after the intervention is rarely performed. Without these measurements, comparison of the relative merits of various interventions is not possible. Reasons for the failure of an intervention can also be understood if the correct data are collected.
Interventions designed to control antimicrobial resistance can be audited for their effects on:
• Antibiotic use;
• Resistance rates;
• Health outcomes.
Standardization of data collection on antibiotic use is essential. The World Health Organization (WHO) has established both a classification system and a unit of measurement, the defined daily dose (DDD)/1000 population/day to describe antibiotic use . The DDD is often a compromise because dosing schedules differ between countries. However, it does provide a standard that allows comparisons between and within countries . In hospitals, where a different denominator is necessary, antibiotic usage is expressed as DDD/100 occupied bed days. Some studies have used raw import or institutional use volumes, and although some impression can be obtained regarding the direction of change following an intervention, quantification is not possible where these data are used. Other studies have used the number of prescriptions/office visit or the prescription rate/1000 population. These measurements contain variables, such as dose size and duration of therapy, and hence provide less meaningful results.
The DDD classification system is not perfect as it is based on the average standard dose for the major indication in question and may not reflect the actual or recommended dose or that used for other indications. In addition, it does not take into account loading or prophylactic doses and it assumes patient compliance. Specific data collection is necessary to establish use by age group or indication. If antibiotic use in control and intervention groups is measured during different time periods, correlation with disease incidence or prevalence must be considered.
Standardization of antibiotic use data in animals is much more difficult and an acceptable system has yet to be devised.
Antimicrobial resistance surveillance, the principles of which are reviewed in detail in another chapter of this Global White Paper , is an essential tool in the measurement of the effectiveness of any intervention strategy. There are, however, considerable difficulties in setting up an optimum surveillance program .
Ideally, surveillance needs to encompass both human and animal bacterial isolates. However, the latter are more difficult to monitor because veterinary laboratories do not perform ‘routine’ cultures and susceptibility testing regularly. Recommendations have been made for the key indicator organisms that should be included in animal surveillance studies. The Office International des Epizooties (OIE) has proposed that specific animal pathogens, zoonotic pathogens (Salmonella spp., Campylobacter spp.) and commensal organisms (Escherichia coli, enterococci) should be monitored .
Human health surveillance programs can be passive or targeted. The use of passively acquired data (obtained from routine testing in clinical laboratories) has a number of potential weaknesses. For the data to be valid, it is necessary that:
• Standardized susceptibility testing methods, such as those recommended by the National Committee for Clinical Laboratory Standards (NCCLS) , British Society for Antimicrobial Chemotherapy (BSAC) , Deutsches Institut für Normung (DIN) , or Société Française de Microbiologie (SFM) , are used. So that results from the different methods can be compared, external quality assessment is necessary to correlate the data  and to establish the proficiency of the submitting laboratories .
• The data do not contain repeat isolates from individual patients.
• Clinical specimen data are separated from ‘screening’ data.
• Indicator isolates are identified and an appropriate range of antimicrobials is tested.
• Specific patient groups (e.g. patients in intensive-care units [ICUs], long-term care patients, pediatric patients) are identified.
• The denominator is established and standardized. Ideally, this should be the number of infected patients. However, most often it is the total number of organisms tested by the laboratory.
Some of the disadvantages of passive surveillance are that :
• The data may be skewed because patients with failed therapy are more likely to be tested.
• There may be an age bias (the elderly and very young being disproportionately represented).
• Isolates from particular disease states may be included disproportionately, e.g. patients with chronic obstructive pulmonary disease are more likely to provide a specimen than patients with community-acquired pneumonia.
• The data are more likely to be based on predefined susceptibility breakpoints (i.e. reported as ‘sensitive’ or ‘resistant’), rather than minimum inhibitory concentrations (MICs), which permit comparisons to be made across studies. This means that more subtle changes in susceptibility will not be identified.
Some studies have been conducted to assess the reliability of data obtained by passive surveillance. In a UK study, susceptibility profiles of E. coli isolates from 220 laboratories obtained using a variety of methods were compared with those obtained at a specialized laboratory. Resistance trends were similar, but there were some minor discrepancies in annual resistance figures that could be explained by geographic bias and differences in the breakpoints used . An external quality assessment program conducted in the UK and Europe to challenge laboratory testing methods revealed problems with the detection of low-level resistance in some laboratories. These problems may be due, in part, to the selection of bacterial susceptibility breakpoints . This highlights the fact that breakpoint measurements must be standardized if national resistance trends are to be compared following international intervention measures. Targeted surveillance is able to address some of these problems associated with passive surveillance. However, targeted surveillance is more difficult to conduct and to maintain.
Surveillance studies tend to focus on measuring resistance phenotypically, i.e. according to the expression of resistance by isolates. For some interventions, it may also be necessary to establish the genetic basis of resistance to demonstrate clonal spread or to detect first-stage mutants. Genotypic analysis may also explain failure of the intervention because of co-selection by other antibiotics or because of persistence of the resistant determinant in the population under study . If all these factors are taken into consideration, and data collection is tailored for a specific intervention, surveillance can be an effective audit tool.
The standard health outcomes measured in respiratory tract infections (RTIs) are mortality and morbidity. When measuring mortality as an end-point, the natural history of the disease and co-morbidities need to be considered. Morbidity measurements are more common and include quality of life, time off work/school, complication rates, requirements for alternative therapy, specialist referral, hospitalization, number of physician visits, number of investigations, length of hospital stay, ICU admissions, infection-control costs, and days to defervescence . Standardization between the intervention groups is essential for severity of illness, prior exposure to antibiotics, age, presence of invasive devices, and underlying medical conditions .
Resistance-control interventions: antibiotic use
Most interventions to reduce bacterial resistance to antibiotics and improve health outcomes are targeted at antibiotic use. The ultimate aim of all of these interventions is to increase the appropriateness of prescribing and thereby reduce the volume of antibiotic use. Figure 1 shows the pathways linking antibiotic use, bacterial resistance, and health outcomes, and indicates the areas where most interventions have been attempted. Most strategies are targeted at inappropriate or unnecessary use of antibiotics, particularly for the treatment of RTIs as this accounts for 75-80% of antibiotic prescribing in the community [17,18]. The options for intervention to reduce antibiotic use are numerous and include:
• Prescriber interventions, e.g. improved diagnosis, implementation of practice or prudent antibiotic-use guidelines, ‘practice profiling’ and feedback, education programs, and use of delayed prescription;
• Consumer interventions, which provide education on antibiotic therapy and resistance;
• Regulatory interventions, e.g. drug licensing regulations and controlled access to antibiotics;
• Restricting use of antibiotics in animals for growth promotion, prophylaxis, and therapy through regulatory control, improved disease management, use of alternatives to growth promoters, implementation of prudent use guidelines and improved dosing schedules.
Where antimicrobial therapy is essential, prescribing strategies can be modified to reduce the selective pressure for resistance. The modifications include the determination of optimum dosing regimens, restrictions applied to particular antibiotics, and use of antibiotic cycling.
Many different bodies, including government organizations, learned societies, hospital administrations, and health-management organizations, can initiate interventions targeted at prescribers. These interventions can take various forms.
Improved diagnosis of bacterial infections can be achieved using laboratory tests, including those that provide a rapid same-day/office answer, or by applying prediction rules or clinical scoring systems. The latter approach has been evaluated in a study that assessed the validity of a scoring system to predict Streptococcus pyogenes sore throat in family practice in Canada . The scoring system used was found to be more sensitive than physician judgement for identifying infection with Group A streptococci. Indeed, if patients had been managed according to these guidelines, a 52.3% reduction in antibiotic prescriptions would have resulted.
Rapid and reliable diagnostic tests that can be performed easily at the point of prescribing (with an accompanying quality assurance program) are not yet readily available. In Denmark, microscopy on urine or genital specimens and antigen testing for Group A streptococci are performed by general practitioners, but there has been no formal assessment of whether this reduces antibiotic prescribing or the prevalence of resistance . This is an area deserving further research and audit.
Practice or prudent-use guidelines
In general, clinical guidelines have been shown to improve medical practice . Many learned societies, government agencies and other bodies have provided guidelines for the ‘appropriate’ use of antimicrobials. Many of these guidelines were initially developed to control the expanding cost of health care. Now the focus is on ‘evidence-based’ best practice where possible, with a greater emphasis being placed on minimizing resistance selection by reducing antimicrobial prescribing. Many of the guidelines are designed to limit antibiotic use for infections that are not bacterial or that are self-limiting. In 1995, the US Centers for Disease Control (CDC) initiated the publication of educational materials for patients and physicians treating RTIs in children, and later published these guidelines . An audit carried out by the CDC showed that the annual rate at which office-based physicians prescribed antibiotics for infectious RTIs (otitis media, common cold, bronchitis, sinusitis and pharyngitis) in children aged <15 years decreased by 44% (P < 0.001) for prescriptions/1000 children and by 14% (P < 0.001) for prescriptions/1000 office visits between 1989/1990 and 1999/2000 . This reduction in antibiotic prescribing coincided with increased awareness of antibiotic resistance among physicians. Although encouraging, this study was retrospective and therefore unable to control for population variation, Haemophilus influenzae type B vaccine uptake and day-care participation rates.
The CDC recently convened a panel of physicians to formulate guidelines for the treatment of upper RTIs in adults  (including nonspecific infections [25,26], sinusitis [27,28], acute pharyngitis [29,30], and acute bronchitis [31,32]). Only recently published, these have not yet been audited for their effects on prescribing or resistance. Other organizations that have produced RTI guidelines include the American Thoracic Society [33,34], the Infectious Diseases Society of America , the Canadian Infectious Diseases Society together with the Canadian Thoracic Society , the European Respiratory Society , and the British Thoracic Society . Audit is essential if the value of these guidelines is to be assessed. Despite this, audit is not routinely incorporated into guidelines. Some assessments of adherence and the effect of guidelines on clinical outcome have been reported , but these have not specifically looked at antibiotic prescribing or resistance prevalence.
Other audited prescriber interventions have been reported. A prospective German study on 1010 patients examined the guideline-influenced effect on prescribing, and clinical outcome, for two primary-care physicians . The intervention, received by one physician consisted of management guidelines for sore throat. The second physician received no intervention. Retrospective data were also examined for both physicians. The physician subjected to the intervention showed decreases in the frequency of throat swabs (from 32% to 7%) and in antibiotic use (from 51% to 31%). Interestingly, clinical outcome (in terms of success and relapse rates) remained the same before and after the intervention and between the two physicians. In a second study, an antibiotic management program, which used local clinician-derived consensus guidelines embedded in a computer-assisted decision-support system, was assessed for its impact on clinical outcomes, patterns of antimicrobial resistance, and antibiotic usage (measured in DDD/100 occupied bed days) . Measures of antibiotic use and clinical outcomes improved during the 7-year study period. Antibiotic use decreased by 22.8%, appropriate timing of presurgical prophylaxis increased from 40% to 99.1%, the rate of antibiotic adverse events decreased by 30%, and the mortality rate decreased from 3.7% to 2.7%. Antibiotic resistance patterns remained stable during the study period. The same group reported a reduction of antibiotic use from 23.6 to 11.4 DDD/100 occupied bed days following the application of a computer-assisted management program in an ICU; resistance was not studied . Although these data appear convincing, these studies were observational, and it is possible that the changes reported could have been a result of other institutional practices.
Practice profiling and feedback
Practice profiling involves the acquisition of individual practitioner prescribing data, usually from national databases, which are then compared with peer practices or the group norm. The information is then fed back to the prescriber with the intention of highlighting deviations from the normal prescribing pattern. Studies have shown a lack of effect of practice profiling performed in isolation on prescribing practices [43,44]. However, when combined with an educational component (e.g. targeted mailed material and outreach visits), an improvement in prescribing practice has been demonstrated compared with control groups [43,45,46].
Educational programs for medical practitioners include provision of office-based material (e.g. posters, information sheets, evidence-based practice guidelines, alternative nonprescription pads), academic detailing (i.e. personal, interactive educational visits), problem-oriented case studies, seminars, and small group discussions managed by opinion leaders. Combinations of some or all of these educational tools have reduced or improved the quality of antibiotic prescribing in primary care [47-50], reduced antibiotic prescription rates in the treatment of specific infections such as acute bronchitis , and decreased the use of specific agents in the hospital setting .
The strategy of providing a prescription to be dispensed if clinical improvement has not occurred within 24-48 h has been used in the management of acute otitis media. Acute otitis media is mainly self-limiting and symptomatic relief is often all that is required [53,54]. In one study, introduction of this delayed-prescription approach resulted in a 19% reduction in antibiotic use compared with a similar period during the previous year . This approach has also been assessed as a means of ensuring patient satisfaction in the treatment of mainly viral upper RTIs, where patients often expect to receive a prescription . Patient satisfaction was high (96%) and the dispensing rate of the delayed prescriptions was 50%, resulting in an overall reduction in antimicrobial therapy for these infections.
Multifaceted physician-directed interventions
Two systematic reviews have concluded that the most effective approach for improving antimicrobial use is a combination of all of the preceding strategies. A Cochrane review was conducted to assess the effects of interventions to improve the delivery of preventive services in primary care. This review was not specifically directed at strategies to reduce antimicrobial prescribing, but rather at the effectiveness of the various intervention models . The second review was restricted to implementation methods to improve antimicrobial use and assessed both multifaceted and individual intervention strategies .
Consumer interventions are educational in nature and can be introduced at various levels. Types of intervention include the use of written material (posters, pamphlets, educational sheets); prescriber explanation and increased consultation time; and presentations at day-care centers, community organization meetings, schools, and primary health-care centers. In addition, coverage of the issues in both print and television media can achieve wide penetration of the community. A recent study has shown that a multifaceted consumer education program changed public awareness about antibiotic resistance and indications, with the result that the proportion of parents who expected antimicrobial therapy for their child declined in the intervention area while it increased in the control population . Other studies reported a decrease in antibiotic prescribing following a program aimed at both the prescriber and the consumer [50,51,59]. For example, implementation of a public education program consisting of consumer booklets, prime-time television advertising, websites, letters to general practitioners, and pharmacists conducted in Belgium in 2000 decreased total antibiotic sales by 10% compared with the corresponding period in the previous year . The mean monthly reduction in antibiotic sales was significantly larger than any decrease caused by variations in the incidence of acute RTIs. However, the decrease was transient and antibiotic prescribing rates returned to previous levels within a short period. A planned repetition of the program should be able to assess whether these changes can be sustained.
Guidance issued by the European Agency for the Evaluation of Medicinal Products (EMEA) recommends that assessment for registration of antibiotics should include resistance selection data, resistance prevalence data applicable to the target population, a requirement for regular updating of resistance prevalence data in the product information document, and restriction of prescribing indications . The use of pharmacodynamics to establish optimum dosing regimens is also recommended . Although the European Union (EU) and Australian regulatory authorities have accepted these recommendations in principle, they have yet to be fully implemented. In addition, there is currently no formal mechanism for evaluating the impact of these requirements on antimicrobial use or resistance patterns. Other licensing issues relevant to resistance control include the need for a mechanism to evaluate generic antimicrobials for acceptable bioavailability and potency and to detect and eliminate counterfeit antimicrobials.
Access to antimicrobials
Access to antimicrobials varies greatly between countries. In some countries, access is restricted through a physician prescription system, while in others the consumer can buy antimicrobials directly over the counter from pharmacies. Controlling access to antimicrobials is an important way of limiting inappropriate use of these agents and regulation to ensure better control is required in many developing countries. A control mechanism used by some countries is the method of payment. Direct government cost subsidies influence the availability and total usage of antimicrobials and may thereby affect resistance. In two countries, actions taken through drug subsidies have affected antibiotic use. In Iceland, the removal of government subsidies for antimicrobials was followed by a decrease in the volume of antibiotic use and in the level of resistance in Streptococcus pneumoniae. Education programs aimed at prescribers and consumers were implemented concomitantly and hence it is difficult to determine the relative contributions of the interventions to this effect. In Australia, the government subsidizes some antimicrobials in community practice. For RTIs these are penicillin, amoxicillin, first-generation cephalosporins and macrolides. These are the antibiotics to which S. pneumoniae is most commonly resistant in Australia . In contrast, there is a low level of resistance to the newer fluoroquinolones, which have a very restricted license that does not allow widespread use for community-acquired infections. Other than in methicillin-resistant Staphylococcus aureus (MRSA), fluoroquinolone resistance in other organisms is also low . Antibiotic use data (DDD/1000 population/day) confirm high use of the subsidized drugs and low consumption of fluoroquinolones in Australia .
The widespread use of antimicrobials in animals for growth promotion, prophylaxis and therapy has led to the emergence of resistance. This resistance has the potential to affect human health. Multiply resistant salmonellae and fluoroquinolone-resistant Campylobacter jejuni are recognized causes of zoonotic infection. Resistance in the normal intestinal flora of poultry and swine has also been documented. In particular, vancomycin-resistant enterococci (VRE), streptogramin-resistant enterococci, and ciprofloxacin-resistant E. coli have been associated with the use in animals of avoparcin, virginiamycin and enrofloxacin, respectively. Transfer of genes conferring resistance to antimicrobials not used in human medicine (e.g. nourseothricin, apramycin and hygromycin) has also been shown to occur from animal bacteria to organisms that are commensal or pathogenic in humans . Various organizations [WHO, OIE, Committee for Veterinary Medicinal Products (CVMP), House of Lords] have assessed the available data and have concluded that antimicrobial use in animals is creating a reservoir of resistant organisms and genes that may have a significant impact on human health. Many strategies have been put forward to decrease the use of antimicrobials in animals. However, audit data have been reported for very few specific interventions.
Use of avoparcin as a growth promoter was banned in Sweden (1986), Denmark (1995), Germany (1996), and subsequently in all EU member states (1997). Further bans were introduced in the EU on the use of virginiamycin, tylosin, spiramycin and bacitracin in 1998. There are now some data showing a decrease in vancomycin resistance in enterococci isolated from poultry following this action. In the 18 months following the ban there were reductions in Denmark (from 82% in 1995 to 12% in 1998) , Germany (from 100% in 1994 to 25% in 1997, coupled with a decrease in VRE in human fecal samples from 12% to 3%) , and Italy (from 14.6% to 8%) . Comparisons of resistance rates in fecal indicator bacteria (E. coli and enterococci) in pigs between The Netherlands (1995/1996) and Sweden (1997) showed a lower prevalence for most antibiotics, in particular for vancomycin (39% vs. 0%), in Sweden. These variations reflected differences between the countries in antibiotic usage .
Improved disease management
Growth promotion provided by antimicrobials is due, in part, to the control of intestinal infections, e.g. necrotic enteritis in poultry and postweaning diarrhea in pigs. Prevention of these diseases by alternative strategies can provide the same, if not greater, economic advantages. Alternative strategies include total or partial exclusion of pathogens (including specific pathogen-free production), vaccination, and improvements in animal husbandry . Indeed, employment of all or some of these measures in Sweden and other Scandinavian countries has resulted in reduced antimicrobial consumption. However, it is very difficult to compare the relative benefit between countries as there is no standard means of comparing consumption. The use of the ratio of volume of antimicrobials used to estimated live weight of animals slaughtered is of partial use, but this does not take into account antimicrobial use in companion animals and other variables .
Alternatives to growth promotants
Probiotics (see later section) have been investigated as an alternative to antimicrobials to provide the equivalent weight gain , and to reduce disease burden  and carriage of pathogenic bacteria prior to slaughter [75,76]. Studies showing the effect of probiotics on antimicrobial use and resistance reduction in animals have not been found.
Implementation of guidelines for antimicrobial use in animals as recommended by the WHO  has not been reported. It is reasonable to conclude that this process will be subject to the same adherence problems seen in human health.
Improving dosing schedules
As with human health, pharmacodynamic studies (see later section) can provide a rational basis for antibiotic dosing in animals . Given the different pharmacokinetics in the various animals requiring treatment, much research is necessary before any controlled intervention studies can be performed.
Many interventions to control resistance are aimed at reducing the use of antimicrobials in situations where there is little or no clinical benefit to be gained. There are, however, many clinical situations where antimicrobial therapy is essential for the well-being of the patient. It is therefore important to identify and encourage strategies whereby appropriate antimicrobial usage exerts the least selective pressure. In these situations, audit should be directed at resistance prevalence or at the level of resistance gene detection.
Pharmacodynamic dosing studies have advanced our understanding of the factors that determine appropriate antibiotic doses and dosing intervals to achieve optimal clinical efficacy [79,80]. For β-lactam and macrolide antibiotics, efficacy is associated with the time that serum antibiotic concentrations remain above the MIC of the organism. For fluoroquinolones and aminoglycosides, efficacy most closely correlates with the ratio between the area under the concentration-time curve (AUC) and the MIC. Furthermore, in vitro and in vivo pharmacodynamic models have been used to establish dosing regimens that will minimize resistance selection and mutation frequencies [81-86]. For fluoroquinolones, a maximum plasma concentration (Cmax)/MIC ratio >10 reduces the selection of resistant organisms . These findings have been confirmed clinically in a study in acutely ill patients . However, further clinical studies are required in this area.
Pharmacodynamic studies can also be used to determine optimal therapeutic regimens to reduce the carriage of resistant organisms. For example, nasopharyngeal colonization with resistant S. pneumoniae during therapy for RTIs has been described [88,89]. Eradication of these bacteria during therapy has been suggested as necessary not only for optimum clinical outcome but also to reduce the spread of resistant organisms [90,91]. The use of low doses and long treatment durations with β-lactam antimicrobials contributes to the selective pressure for pharyngeal carriage of resistant S. pneumoniae. Conversely, short-course, high-dose amoxicillin therapy for RTIs in children has resulted in a small, but significant reduction in the carriage of resistant strains 28 days after treatment . Differences between classes of antimicrobials have also been studied, with amoxicillin—clavulanate being superior to azithromycin in eradicating resistant S. pneumoniae.
Optimization of antimicrobial doses through pharmacodynamics is a promising area for future intervention studies using agents currently in use, and these studies should be incorporated into the discovery and development phases of drug research and in the registration of new agents .
Restricting use of antibiotics
Restricting or reducing the use of a single antibiotic has failed to result in the loss of resistance. Resistance can persist for many months after the cessation of use of an antibiotic , and for many years [97,98] if the resistance determinant is linked to genes or transposons conferring resistance to other agents in continued use. Nevertheless, restriction coupled with a requirement for authorization to use certain parenteral antibiotics resulted in decreased resistance to β-lactam and fluoroquinolone antibiotics in one hospital-based study. The clinical outcomes of mortality, time to receipt of appropriate antibiotics, and discharge from hospital were not adversely affected .
More success has been achieved by restricting whole classes of antibiotics during outbreaks of diseases caused by specific resistant organisms. Restriction of third-generation cephalosporins and substitution with a β-lactam/β-lactamase inhibitor was effective in controlling VRE colonization [100,101] and cephalosporin resistance in nosocomial Klebsiella. Similarly, substitution of cephalosporins with a carbapenem was shown to control another nosocomial Klebsiella outbreak. However, this intervention resulted in the emergence of carbapenem-resistant Pseudomonas. In a further study, restriction of third-generation cephalosporins was associated with the termination of a multiresistant Acinetobacter baumannii outbreak . In all these studies, additional infection control measures were introduced, making the relative contribution of class restriction and infection control difficult to assess.
The concept of rotating empiric antibiotic regimens as a means of reducing and/or preventing resistance selection was initiated by Gerding [105,106], who published 10 years' experience of alternating between amikacin and gentamicin. At the start of the study period, levels of gentamicin resistance were high. After a switch to amikacin, aminoglycoside resistance levels fell, only to rise rapidly on the re-introduction of gentamicin. This increase in resistance did not occur after a second cycle of amikacin, when gentamicin was re-introduced at a much slower rate. In this study, the reduction and maintenance of low resistance levels coincided with the loss of the resistance determinant (a plasmid carrying one of the aminoglycoside-inactivating enzymes) from Gram-negative organisms in the environment and in patients. Following this report, the use of antimicrobial ‘cycling’ as a successful intervention has gathered momentum [107,108].
Despite these promising results, some of the studies used to support the concept of antibiotic cycling [109-112] leave questions unanswered. Indeed, two of the studies practised antimicrobial class substitution rather than rotation. The outcomes measured in these studies were the incidence of ventilator-associated pneumonia as a result of antibiotic-resistant bacteria [109,111] and inadequate treatment as a result of organism resistance . These studies did not report full resistance trends before, during, or after the study, and resistance mechanisms were not investigated. Therefore, it is possible that selection of organisms with multiple resistance mechanisms could occur if this strategy was employed without careful monitoring [113,114]. Models for studying antimicrobial cycling have been proposed [106,115]. Thus, detailed, carefully controlled, multicenter studies are required before this intervention can be considered successful.
Resistance-control interventions: disease burden and colonization
Although most resistance-control interventions aim to improve antibiotic use, resistance can also be targeted using other interventions concerned with disease burden and colonization. These interventions include:
• Public health measures, e.g. day-care exclusion and improved sanitation.
• Infection control in hospitals, e.g. improved hand hygiene, use of barrier precautions, and implementation of decontamination protocols to reduce carriage of resistant organisms.
• Vaccination programs.
• Use of alternative nonantimicrobial therapies.
Public health measures
A number of studies have identified prior antibiotic use [88,116], and attendance at a day-care center [117-119] or other ‘overcrowded’ circumstances  as major risk factors for the carriage and spread of resistant S. pneumoniae. Using mathematical modeling, a critical review of the relationship between antibiotic use and carriage of infection with resistant pneumococci has provided further support for the importance of the selective pressure of antimicrobial therapy . This model showed that antibiotic treatment substantially increases a patient's risk of carrying/being infected by penicillin-resistant S. pneumoniae compared with penicillin-susceptible strains. Furthermore, superinfection with resistant S. pneumoniae from the nasopharynx has been shown to follow antibiotic treatment for otitis media and to result in treatment failure . Recognizing that carriage and spread of resistant pneumococci is related to day-care attendance has led to a unique intervention program in Sweden . Children in day-care centers found to be colonized or infected with penicillin-resistant S. pneumoniae were excluded from attendance. All contacts were followed up and, if cultures proved positive, they were also excluded. The cost of exclusion (including costs of alternative day-care placement and parents' absence from work) was funded by the government. The median duration of carriage was found to be 19 days. In a limited number of cases, decolonization was achieved by the use of rifampicin in combination with another agent. Unfortunately, the end-point of the study, as reported, is not defined. The monthly number of resistant S. pneumoniae cases and contacts is provided, but the denominator is unclear. The final evaluation of the intervention has not yet been published. However, the simultaneous introduction of a program to reduce antimicrobial use, which resulted in a 24% reduction in consumption (DDD/1000 inhabitants for selected respiratory antibiotics), will make interpretation of the outcome difficult.
Multidrug-resistant strains of Salmonella typhi and Vibrio cholerae have been responsible for outbreaks of disease in the developing world. Resistance to ampicillin, chloramphenicol, trimethoprim, sulfonamides and tetracyclines has become common in these organisms and more recently fluoroquinolone resistance has been reported in both species [126,127]. As these infections are food- and water-borne, improved sanitation and food hygiene may be the most important resistance control mechanisms. For example, a surveillance study in Peru showed that patients with cholera were less likely to have access to municipal water, sewage systems, or adequate toilet facilities than those with acute diarrhea not caused by cholera . Nevertheless, a reduction in antimicrobial resistance as a direct result of improved sanitation has not been reported.
Infection control in hospitals
There are two important aspects to the control of antimicrobial resistance in hospitals and long-term-care facilities. Firstly, the emergence of resistance is related to the use of antimicrobials. The interventions described above, aimed at prescribers and the way in which antimicrobials are used, are equally applicable in the hospital setting and in the community. Secondly, once resistance has emerged or entered a hospital from the community [129,130], it extends through clonal spread or genetic transfer via transposons and plasmids. The implementation of basic infection control practices (improved hand hygiene and the use of barrier precautions and decontamination protocols to eliminate carriage of resistant organisms) is central to the prevention of this dissemination.
The efficacy of hand washing in preventing the transmission of organisms between patients is well established . However, hand washing is still not performed adequately in routine practice . Interventions designed to improve the frequency and acceptability of hand washing for every patient contact include education programs and the use of alcohol-based hand-washes . The routine use of surgical gloves has also been implemented as an alternative approach to reduce bacterial transmission between patients . These hand-hygiene approaches have been evaluated using a variety of end-points, including adherence to the intervention, hand colonization rate, and rate of nosocomial infection. The results of educational programs have been disappointing [134,135]. In one ICU-based study, no significant improvement in hand-washing compliance was shown before and after an educational intervention (22% vs. 25% in the medical ICU and 13% vs. 14% in the cardiac ICU). The use of an alcohol-based hand-wash has been shown to increase compliance, but only to a maximum of 48% . None of the published studies has defined the rate of hand washing required for effective infection control. This rate may be difficult to establish as it may vary between different patient populations (e.g. the severely immunocompromized, those with intravascular lines, and those in the ICU), and the number of patients colonized with a resistant organism . Mathematical modeling suggests that the level of compliance with hand washing needs to be considerably higher than reported levels to prevent nosocomial transmission of VRE in endemic settings . Therefore, based on the studies to date, the optimum intervention for limiting bacterial transmission via hand contact has yet to be determined.
Barrier precautions and separation of patients and staff into cohorts have been shown to be effective if all those infected and colonized are identified. Geographic isolation is widely practiced and has been demonstrated to reduce rates of bacteremia in a burns unit  and the rate of MRSA in clinical specimens from an ICU . The use of surgical gloves has been successful in reducing the rate of transmission of MRSA within a neonatal ICU setting . In addition, universal surgical glove usage has been shown to reduce the rate of nosocomial Clostridium difficile-associated diarrhea .
Patients are the main reservoir of resistant organisms in hospitals and specific screening programs are required to detect all carriers. Warren & Fraser  recently reviewed studies of interventions designed to eliminate the carriage of resistant organisms and thereby limit the spread of resistance. The use of topical antibiotics to eliminate enteric carriage through selective digestive tract decontamination has had mixed results. Some studies have shown a decrease in respiratory and urinary tract infections while others have shown no significant difference in overall nosocomial infection rates compared with controls. A negative result with this approach was increased colonization with MRSA, enterococci and other Gram-positive bacteria, as well as with aminoglycoside-and fluoroquinolone-resistant Gram-negative organisms. Attempts have been made to decolonize patients carrying VRE with varying results. Studies have mainly used bacitracin, alone [143,144] or in combination with other agents such as doxycycline , and more recently ramoplanin . Most of the studies were conducted on a very small number of patients, were not double-blind, and used different treatment regimens for relatively short courses of therapy. The studies to date indicate a lack of efficacy of bacitracin when used as a 14-day course. Topical treatment to remove MRSA carriage has also had mixed results. Topical mupirocin has been shown to be effective in health-care workers, but less so in patients . Systemic antibiotics to remove MRSA carriage have also had varied success .
Prevention of disease by vaccination reduces the overall burden of disease and consequently the requirement for antimicrobial therapy. There is also evidence that vaccines may reduce carriage of resistant organisms. Potentially, this intervention strategy has a very significant effect in resistance reduction. The conjugate pneumococcal vaccine has been demonstrated to be effective in reducing invasive disease  and otitis media  in infants and children. Moreover, in a double-blind, randomized, placebo-controlled trial in 500 infants in South Africa, administration of a nonavalent conjugate vaccine reduced the nasopharyngeal carriage of vaccine S. pneumoniae serotypes 6 months after vaccination compared with placebo (18% vs. 36%, respectively); carriage of nonvaccine serotypes increased. More importantly, carriage of penicillin-resistant S. pneumoniae, regardless of serotype, was significantly reduced compared with placebo (21% vs. 41%, respectively; P < 0.001), as was carriage of cotrimoxazole-resistant pneumococci (23% vs. 35%, respectively; P < 0.003) .
Other developmental vaccines for use against organisms in which resistance is currently a problem include those against Mycobacterium tuberculosis, Helicobacter pylori, V. cholerae, and S. typhi. Vaccines under study against animal pathogens may reduce the requirement for antimicrobials in food-producing animals [156-158].
Alternative nonantimicrobial therapies
The use of alternative modalities to prevent or treat infectious diseases can potentially reduce the selective pressure of antimicrobial therapy for resistance. One option is the use of probiotics or bacteriotherapy, whereby nonpathogenic organisms are the therapeutic agents [159,160]. For example, the organisms, Lactobacillus rhamnosus GG, Bifidobacterium spp., and Saccharomyces boulardii, have been used with some success in the treatment of infectious diarrhea in children (particularly rotavirus-associated diarrhea), antibiotic-associated diarrhea, and candidal vaginitis. The mechanisms by which probiotic therapy is thought to work include immunomodulation at the mucosal surface, competition with pathogenic organisms, production of antimicrobial agents, and lactose digestion.
Few studies have researched the role of these agents in reducing antimicrobial-resistant infections or colonization. A study comparing cranberry-lingonberry juice, Lactobacillus GG drink and placebo for the prevention of urinary tract infections in women found only a benefit from the cranberry-lingonberry juice . A second double-blind, placebo-controlled trial in Finland examined the effect of long-term consumption (7 months) of milk containing L. rhamnosus GG on gastrointestinal infections and RTIs in 571 children attending day-care centers . The results showed a modest reduction in the incidence and severity of RTIs and antibiotic use in the study group. However, statistical significance was not reached. Bacteriologic studies were not performed, so no conclusions can be made regarding the potential usefulness of Lactobacillus in reducing carriage of resistant organisms. In a third study, a mixture of commensal organisms (Streptococcus sanguinis, Streptococcus mitis, and Streptococcus oralis) was effective in reducing recurrences of otitis media in children . Children with recurrent otitis media were randomized to receive either the streptococcal nasal spray or placebo. During the 3-month follow-up period, 42% of the study group did not have a recurrence compared with 22% of the placebo group (P = 0.02). Cultures were performed at the time of inclusion into the study, but were not repeated and antimicrobial susceptibilities of the isolates were not reported. As the trends in the last two studies were in the direction of disease reduction, this is an area for future research. Importantly, studies should have well-designed end-points to measure antimicrobial resistance reduction.
Issues specific to low-resource countries
Antimicrobial resistance is as much of a problem in developing countries as in the developed world. However, less is known about the extent of the problem in developing countries. Furthermore, the burden of infectious diseases—particularly RTIs, gastrointestinal infections, tuberculosis, and malaria—is greater in these regions. Resistance in M. tuberculosis is a well-recognized problem  and there are some data demonstrating resistance in shigella, salmonella, campylobacter, V. cholerae and S. pneumoniae. The interventions for resistance control outlined above are of relevance in the industrialized, high-income sections of developing countries, but these account for only a fraction of antimicrobial use. In poorer regions, there are specific economic, social and behavioral problems that must be taken into account in any intervention program.
Two recent reviews provide insight into many of the problems [166,167]. In brief, health-care infrastructure is often poorly established in these countries, with minimal regulatory controls on the supply of antimicrobials . Antimicrobials are often prescribed by unskilled practitioners or purchased over the counter by patients [169,170]. Over the counter availability may not necessarily be the major contributor to inappropriate use as only a small proportion of patients self-medicate . Low antibiotic doses, short courses, and poor compliance all contribute to reduced clinical efficacy and increased selective pressure for resistance. These factors are often the result of incorrect dosing schedules or poverty, whereby only a few units of the antimicrobial are purchased . Counterfeit antimicrobials [172,173], poor storage conditions, and poor-quality generic antibiotics with low bioavailability can also contribute to low dose exposure and selective pressure. In addition, the presence of dispensing physicians in developing countries may play a role in the increased use of antimicrobials [174,175]. Within hospitals, poor hygiene, overcrowding, lack of resources for infection control, and the absence of trained infection-control personnel contribute to increased levels of nosocomial infections and dissemination of resistant organisms.
The use of audit tools to measure the effectiveness of resistance-control interventions also has problems specific to low-resource countries. Resistance surveillance, as a means of identifying problems and measuring the impact of intervention programs, is hampered by the lack of trained staff, the cost of consumables and equipment, poor access to standard antimicrobial disks, and the absence of quality assurance programs . The establishment of sentinel laboratories  and data analysis using computer programs, such as WHONET  may partially resolve this problem. However, good antimicrobial-usage data are hard to obtain in developing countries, which when compounded by a lack of resources, counterfeit antimicrobials, drug smuggling, and limited regulation, the collection of reliable data is extremely difficult. Similarly, health outcome data are very limited.
The WHO has recognized that drug accessibility is of major importance in developing countries. Through its Action Program on Essential Drugs (DAP) , the WHO is helping these countries to develop and implement a national drug policy, which ensures access to essential drugs, rational use of drugs, and drug quality. Assessment of the success of the DAP initiative in an individual country is by a standardized field-tested set of basic drug use indicators , as devised by the International Network for the Rational Use of Drugs (INRUD). Antimicrobial use is only a part of the program and the specific issue of resistance prevention is not included in the assessment.
It is not surprising, given all the difficulties, that the number of well-monitored resistance-control interventions with outcome assessment is minimal in low-resource countries. However, interventions to improve antimicrobial prescribing practices have been suggested . A number of papers were presented at the International Conference on Improving the Use of Medicines in Chiang Mai, Thailand, in 1997. Although they were not directed toward reducing antimicrobial resistance, most of the interventions were designed to reduce inappropriate antimicrobial use. The outcome measures chosen were those devised by INRUD . In one study in Peru, provision of a broad education program aimed at the community, carers and physicians resulted in a significant decrease in the percentage of children receiving antimicrobials for diarrhea . Similarly, education of physicians reduced antibiotic use in the treatment of acute RTIs in Haiti  and training of health-care workers in Vietnam increased the number of patients receiving appropriate dosing schedules of antibiotics . Conversely, the introduction of standard treatment guidelines alone did not change prescribing practices in Uganda  or Tanzania, where the introduction was coupled with educational and managerial strategies . More promisingly, focus group discussions with drug-selling storekeepers in the Philippines reduced the sales of penicillin and amoxicillin . The education of drug retailers has been in practice in Nepal since 1981, but no assessment has been made of the effect of this education program on the appropriateness of prescribing, or the dispensing or storage of antimicrobials . There are few reports on regulatory action (deregistration, relabeling, or restriction) taken by developing countries to improve drug usage. However, at the Chiang Mai conference, regulation to promote generic prescribing was described. Regulatory action in a number of countries was reported following the release of WHO advice on the use of inappropriate therapeutic agents for diarrhea  and, in Nepal, the introduction of a fee structure that discouraged consumer demand has led to a decrease in antimicrobial prescribing .
In summary, implementation of interventions to limit antimicrobial resistance, particularly in respiratory and alimentary tract pathogens and in hospitals, is as important in low-resource countries as in industrialized ones. Implementation of these measures is associated with many problems in low-resource countries. Monitoring the success of any intervention program will be difficult without first establishing the laboratory infrastructure or augmenting existing facilities to perform resistance surveillance. However, regulatory action to ensure appropriate access to antimicrobials and to act against the manufacture of counterfeit and substandard generic agents should be considered as a matter of urgency. Assistance from the WHO and professional societies could be carefully targeted to maximize the training of local personnel so that any results can be sustained.
Many interventions for the control of antimicrobial resistance are available. Some have been evaluated, to varying degrees, while others are only of theoretical value at present. It is difficult to prioritize these interventions owing to the paucity of studies that have measured objective end-points, i.e. reduction in resistance rates and improvement in health outcomes. Nonetheless, certain general points can be made. Firstly, 50% of antibiotic consumption in developing countries is in humans and 50% is in animals . This suggests that efforts should be made to optimize antibiotic use in both humans and animals. Human antibiotic consumption can be categorized further, with 80% occurring in the community and 20% in hospitals . Within the community, 80% of antibiotic use is for the treatment of RTIs [17,18], many of which do not warrant antibiotic therapy. Therefore, interventions aimed at community RTI prescribing should provide maximum benefit in reducing the overall volume of antimicrobial use.
Efforts should be concentrated at both the prescriber and consumer level. Programs that have been shown to have the greatest impact in reducing antibiotic use are those that are comprehensive. These programs have targeted prescribers using a variety of methods, such as: educational materials; strategies to improve diagnosis; practice guidelines; personalized, interactive sessions with feedback on the practice profile; use of delayed prescriptions; and provision of alternative nonantimicrobial prescribing strategies. The best results are usually obtained when these measures are combined with consumer education on the self-limiting nature of many community-acquired infections (including many RTIs), the role of antibiotics in the treatment of these infections, and the problem of antimicrobial resistance.
With regard to animal use, reductions in resistance have already been seen following regulatory bans on the use of certain antimicrobials for growth promotion. Further reductions are possible, and the prescribers and end users should benefit from similar educational programs and interventions, although the evidence for this is not yet available.
Intervention at the hospital level requires good infection-control programs to limit the clonal spread of resistant genetic elements and organisms, as well as modification of prescribing practices. Although the latter is difficult to address, reducing unnecessary antibiotic usage by effective practice guidelines will certainly be helpful. It is also important to understand how to reduce the selective pressure for resistance arising from clinically warranted antimicrobial therapy. At present, we do not have sufficient information regarding the relative selective pressure of different classes of antimicrobials, nor do we know if strategies such as antimicrobial restriction or cycling will reduce the prevalence of resistance. However, accumulating evidence from pharmacodynamic studies suggests that optimal dosing of antibiotics (in terms of dose size, dosage interval, and treatment duration) plays an important role in reducing selective pressure. Improvements in public health, vaccination programs, and the development of new antimicrobials may also help to reduce the development and spread of resistance.
There are many gaps in our knowledge of the dynamics of resistance development and spread and the effectiveness of interventions. Much research in this area is necessary, particularly in developing countries. However, our lack of understanding should not prevent us from implementing those strategies already shown to be of benefit and from introducing others with careful monitoring. These interventions must be audited using the best means possible to determine the ‘best value for money’. Surveillance of antimicrobial resistance and use are essential tools that must be integral to any program. Provision of reliable surveillance and health outcome data will allow us to concentrate our efforts and resources on the most cost-effective approaches to resistance control.
To initiate and sustain such intervention programs requires substantial government support and regulatory control. Many governments and other health-care organizations are now offering support, but for a sustainable, effective approach there must be a strong political will and commitment to succeed.