Rethinking the WHO AWaRe classification: the need for more detailed guidance

Summary

The World Health Organization's Access, Watch, and Reserve (AWaRe) classification promotes appropriate antibiotic use and is widely adopted as a global indicator of antimicrobial use (AMU) quality. This paper highlights key limitations of the AWaRe classification system, using Japan as a case study. The global targets for the proportional use of “Access” antibiotics (60%–70%) lack epidemiological justification and may not fully capture regional disease patterns to reflect regional disease patterns. Furthermore, inconsistencies in drug classification and the influence of long-term prescriptions can distort Access category percentages. As most countries do not monitor prescription duration in AMU surveillance, global comparisons using the current AWaRe framework may be misleading. This paper suggests a careful reassessment of AMU surveillance methodology or the AWaRe classification itself to ensure meaningful evaluation of antibiotic stewardship efforts worldwide.

Funding

This article was funded by a research grant from the Ministry of Health, Labour and Welfare (JP23HA2002) and JSPS KAKENHI grant number JP23K18396 and the funders did not play any role in writing the manuscript.

Introduction

Antimicrobial resistance (AMR) has long been recognised as a major global public health issue.^1,2 Considering the disease burden and societal impact of AMR, the World Health Organization (WHO) established the AMR Global Action Plan in 2015.¹

The appropriate use of antibiotics is an important basic measure against AMR³ that is often monitored by measuring the amount of antimicrobial use (AMU). However, the appropriate use of antibiotics should be based on both the quantity and quality of AMU. At present, the quantitative evaluation of AMU quality at the national level appears to be somewhat limited. This is likely due to the challenges in assessing the quality of antibiotic administration using population-level data.

In response to this situation, the WHO proposed the Access, Watch, and Reserve (AWaRe) classification in 2017.⁴ This is a relatively new classification system that classifies antibiotics into these three categories, with the aim of promoting the appropriate use of antibiotics by prioritising the use of antibiotics classified in the “Access” category (defined as antibiotics widely available, generally used for common infections, and with lower resistance risk) and reserving those classified in the “Watch” category (defined as antibiotics with higher resistance risk or more side effects that should be used more judiciously) and “Reserve” category (defined as antibiotics that are last-resort options for difficult-to-treat infections)⁴ for use only when necessary.

As the concept of AMR has become more widely recognised, the AWaRe classification system has also gained recognition as an indicator of the quality of AMU.⁵ However, several challenges need to be addressed to enhance the effectiveness of the AWaRe classification. Herein, we discuss the issues that arise when evaluating appropriate AMU based on AWaRe classification, using Japan as an example.

Main section

Initially, the WHO recommended that the “Access” category should account for more than 60% of all antimicrobial prescriptions at the national level.⁴ This target has now been raised to 70%.⁶ However, one of the major challenges of the AWaRe classification system is that evidence supporting these figures is not well established. In general, “Access” antibiotics are narrow-spectrum antibiotics that should be prioritised when circumstances permit. However, it is unclear what benefits would be gained by setting the “Access” category proportion at 60% or 70%, or what disadvantages would arise if the proportion fell below 60%. Although it is necessary to set a target value, the setting of a single value without clear evidence as a target for all countries may require careful consideration.

The classification of each antibiotic into the “Access”, “Watch”, or “Reserve” categories is also a subject of ongoing debate. As the AWaRe classification is based on the WHO's Essential Medicines List,⁷ drugs that are essential for treating diseases that are prevalent worldwide and that are easily accessible in low-income countries tend to be classified in the “Access” category.⁸ Unfortunately, such classification may not always align with the goals of appropriate AMU. For example, the antibiotic sulfamethoxazole–trimethoprim combination (TMP/SMX) is commonly used for urinary tract infections,⁹ but is also used to treat many other diseases such as Pneumocystis jirovecii pneumonia, Nocardia infection, and methicillin-resistant Staphylococcus aureus (MRSA) infection.¹⁰ In addition, it is used as prophylaxis especially for HIV-infected and HIV-tuberculosis co-infected patients.¹¹ Compared with amoxicillin, which is also in the Access category, TMP/SMX has an considerably broader spectrum than many drugs in the “Watch” category.¹² The WHO also notes that “Some Access antibiotics, such as cotrimoxazole, have broader spectrum activity or increasing resistance but are retained in the Access category due to their essential use in high-burden settings and low-cost availability,” indicating that these drugs were not classified into “Access” based on their spectrum alone.¹³ Under the current definitions of the AWaRe classification, prescribing TMP/SMX to patients suspected of having any infection without clinical consideration could increase the percentage of “Access” category drugs. Such practices may not be ideal when considering resistance concerns of TMP/SMX at the global level.¹⁴

There are also variations in categorisation within the same class of drugs. For instance, Doxycycline is a highly versatile drug used to treat most infections even including atypical pneumonia, sexually transmitted infections, and malaria. Similar to TMP/SMX, doxycycline is classified as “Access” due to its ease of access and clinical indispensability, although this classification may not fully align with its broad spectrum given its broad spectrum.¹⁵ By contrast, minocycline, which belongs to the same class and has a nearly identical spectrum as doxycycline, is classified as “Watch”. This is likely due to differences in their availability in low- and middle-income countries (LMICs). However, from the perspective of AMR countermeasures, both doxycycline and minocycline have an equally broad spectrum.¹⁰ Although both should be prescribed with caution, prescribing doxycycline increases the percentage of “Access”, while prescribing minocycline decreases it. These inconsistencies are likely to pose a challenge when evaluating it.

Some antibiotics are not intended for the treatment of infectious diseases but are considered appropriate for certain uses. For example, rifaximin is an antibiotic frequently used to treat hepatic encephalopathy by suppressing ammonia production.¹⁶ There is robust evidence for the efficacy of rifaximin in treating hepatic encephalopathy,^16,17 and there are currently limited concerns about AMR.¹⁸ Therefore, there limited rationale for discouraging the use of rifaximin to treat hepatic encephalopathy. Nevertheless, as rifaximin is classified as “Watch” in the AWaRe classification,¹⁵ an increase in its prescription will result in a decrease in the “Access” percentage. This is an example of potentially misleading evaluations stemming from causes unrelated to appropriate AMU.

A significant limitation of the AWaRe classification is that there are various common diseases for which the standard treatment involves antibiotics outside the "Access" category. For example, the standard treatment for Mycoplasma pneumonia is macrolides,¹⁹ and the standard treatments for diseases such as tuberculosis and pulmonary nontuberculous mycobacteriosis (NTM) are not “Access” drugs.²⁰ Therefore, physicians may inadvertently reduce the percentage of “Access” antibiotics in their country by providing “appropriate” medical care.

One of the main challenges in using the AWaRe classification system that is often overlooked is the impact of long-term prescriptions on overall statistics. In this viewpoint, we define a long-term prescription as one that is taken for 14 days or longer, in accordance with a previous study from Japan.²¹ This is because infectious diseases are generally acute diseases, and their standard treatment period is often within 2 weeks.¹⁰

Fig. 1 shows the amount of outpatient oral antibiotics (excluding antituberculosis drugs, limited to drugs with ATC-code “J01”. The results are based on the sum of community-based and hospital-based prescriptions) prescribed in Japan calculated using the National Database of Health Insurance Claims, which covers more than 95% of the total Japanese population.²² We used defined daily dose (DDD) per 1000 inpatients per day (DID) as a unit, because this can take both the duration of therapy and population size into consideration. In Fig. 1, the lower blue portion represents short-term prescriptions (less than 14 days), and the upper orange portion represents long-term prescriptions (14 days or more). In 2019, AMU of long-term prescriptions reached 3.66 DID, accounting for 33.1% of the total AMU. In 2020, when the impact of COVID-19 was substantial, the proportion of long-term prescriptions increased to nearly half of the total AMU (4.28/8.61, 49.7%).

Fig. 1.

The breakdown of short- and long-term prescriptions in: (A) total antimicrobial use, (B). hospital-based antimicrobial use, and (C) community-based antimicrobial use.

The problem becomes more evident when we look at the proportion of long-term prescriptions for each antibiotic. Previous studies have reported that macrolide antibiotics, represented by clarithromycin, are frequently prescribed in Japan,²³ and long-term prescriptions account for more than half of the DID for macrolides. Furthermore, although the absolute quantity of prescribed macrolides is lower than that of clarithromycin, it is important to note that most prescriptions for TMP/SMX are long-term prescriptions. Additionally, a substantial proportion of the DID comprises rifaximin, which is almost always used as a long-term prescription (Fig. 2).

Fig. 2.

The breakdown of short- and long-term prescriptions for each antibiotic based on: (A) total antimicrobial use, (B) antimicrobial use in hospitals, and (C) antimicrobial use in communities. DOXY: doxycycline; TC: tetracycline; CP: chloramphenicol; ABPC: ampicillin; AMPC: amoxicillin; BAPC: bacampicillin; DBECPCG: benzathine benzylpenicillin; CVA/AMPC: amoxicillin and beta-lactamase inhibitor; SBTPC: sultamicillin; CEX: cefalexin; CXD: cefroxadine; ST: sulfamethoxazole and trimethoprim; CLDM: clindamycin; MNZ: metronidazole; TNZ: tinidazole; KM: kanamycin; VCM: vancomycin; RFX: rifaximin; FDX: fidaxomicin; DMCTC: demeclocycline; MINO: minocycline; CXM-AX: cefuroxime; CCL: cefaclor; CFIX: cefixime; CPDX-PR: cefpodoxime; CETB: ceftibuten; CFDN: cefdinir; CDTR-PI: cefditoren; CFPN-PI: cefcapene; CFTM-PI: cefteram; TBPM-PI: tebipenem pivoxil; EM: erythromycin; AC-SPM: spiramycin; RXM: roxithromycin; JM: josamycin; CAM: clarithromycin; AZM: azithromycin: LCM: lincomycin; OFLX: ofloxacin; CPFX: ciprofloxacin; NFLX: norfloxacin; LFLX: lomefloxacin; LVFX: levofloxacin: MFLX: moxifloxacin; PUFX: prulifloxacin; GRNX: garenoxacin; STFX: sitafloxacin; TFLX: tosufloxacin; PPA: pipemidic acid; FOM: fosfomycin; RFP: rifampicin; RBT: rifabutin; PL-B: polymyxin B; CL: colistin; FRPM: faropenem; LZD: linezolid; TZD: tedizolid; ABPC/MCIPC: ampicillin and cloxacillin∗. ∗Classified as “Not recommended”.

First, let us confirm the extent to which long-term prescriptions affect the calculation of the percentage of “Access” theoretically. Community-acquired pneumonia is treated with amoxicillin in accordance with the Infectious Diseases Society of America guidelines.²⁴ Assuming an average of 1 week treatment duration, one case of community-acquired pneumonia would require 7 DDD of “Access”. By contrast, tuberculosis is a representative example of an infectious disease requiring long-term treatment. The standard treatment for tuberculosis requires four antibiotics: isoniazid, rifampicin, pyrazinamide, and ethambutol. Among these, only rifampicin is classified as being in “Watch”, while the other three are “Unclassified”. The standard treatment for tuberculosis involves the administration of rifampicin for 6 months, meaning that treating one case of tuberculosis would require 180 DDD of “Watch”.^15,25 This means that to maintain an “Access” category rate of 70% or higher, approximately 60 cases of community-acquired pneumonia must be treated for every one tuberculosis case (180/0.3 = 600, 600 × 0.7/7 = 60). Furthermore, this figure assumes that no other drugs in the “Watch” or “Reserve” categories are used.

Although there are fewer NTM infections than tuberculosis cases, the presence of NTM infections must not be overlooked. The standard treatment period for lung NTM disease is 12 months from the confirmation of a negative sputum culture, and it typically takes 3–4 months for the sputum smear-negative conversion, resulting in a total treatment period of over 1 year.²⁰ Even if it takes only 3 months for the sputum culture conversion, this requires the administration of the three drugs (clarithromycin, rifampicin, and ethambutol) for 450 days. Ethambutol is “Unclassified”, but clarithromycin and rifampicin belong to the “Watch” category. Therefore, treating one case of lung NTM disease would require 900 DDD from “Watch”. Moreover, this is assuming that treatment progresses smoothly; it is not uncommon to encounter cases requiring prolonged treatment for reasons such as failure to confirm sputum culture conversion.²⁰ Even if treatment with 900 DDD is successful, maintenance of the “Access” rate of 70% or higher requires the treatment of 300 cases of community-acquired pneumonia for a single case of lung NTM disease (900/0.3 = 3000, 3000 × 0.7/7 = 300). Of course, lung NTM disease cannot be treated with “Access” category drugs but should instead be treated with “Watch” category drugs.

In other words, the use of DDD or DID as units means that diseases requiring long-term prescriptions will have a substantial impact on the overall percentage of “Access” category drugs for AMU. The situation is even more challenging because it is known that “Access” category drugs cannot treat tuberculosis or NTM infections. Furthermore, the WHO does not define how to calculate the proportion of “Access”. The availability of data varies significantly by region, making it difficult to specify a single calculation method. This makes it challenging in practise to compare the proportion of “Access” across different countries. Therefore, a high percentage of “Access” category drugs in one country does not necessarily mean that appropriate antibiotic therapy is being provided compared with other countries. Therein lies the risk associated with setting a single numerical target for the percentage of Access category drugs, whether it be 60% or 70%.

Having clarified the impact of long-term prescriptions on the ratio of Access category drugs, we will now examine each drug that is commonly prescribed as a long-term treatment. First, long-term TMP/SMX administration is used to prevent opportunistic infections in patients with cancer but has the effect of increasing the “Access” category rate. As previously mentioned, TMP/SMX has an exceptionally broad spectrum of activity for an “Access” category drug and therefore should not be prescribed indiscriminately to increase the “Access” category rate. However, the current classification system is insufficient to address this issue.

The classification of rifaximin in the “Watch” category is another issue that requires improvement in the AWaRe classification system. It may be necessary to exclude rifaximin from the data collected in each country or to reclassify it as “Unclassified”, judging from its usage mentioned above.

The high prescription rate of macrolide antibiotics, represented by clarithromycin, may be one of characteristics of AMU in Japan.²³ As low-dose long-term macrolide therapy has been proven efficacious for diffuse panbronchiolitis in Japan,²⁶ macrolides are generally used long-term for their anti-inflammatory effects rather than their antibacterial effects. Although the unrestricted long-term administration of macrolides has contributed to the development of AMR,²⁷ their anti-inflammatory effects provide benefits in the treatment of various diseases, such as asthma and community-acquired pneumonia, in Japan and other countries in recent years.²⁸ The balance between the risk of AMR development from prescribing macrolides and the benefits of their anti-inflammatory effects is difficult to quantify, making it challenging to definitively determine whether long-term macrolide administration is appropriate.

Although the number of drugs frequently prescribed for long-term treatment is limited, all such drugs involve factors that hinder the evaluation of the appropriateness of antimicrobial prescription. As an example, Oda et al. reported that the proportion of “Access” increased from 42% to 78% by limiting prescriptions for acute infectious diseases.²⁹ Although this report derives from just a single-centre study, its results suggest that the current aggregation may not be accurately evaluating the proportion of “Access”.

Although the high prevalence of long-term macrolide prescriptions may be a phenomenon unique to Japan, some countries have regions with a high prevalence of tuberculosis and/or NTM disease. Additionally, many LMICs widely use TMP/SMX and doxycycline,³⁰ which means that the impact of long-term prescriptions may alter the intended meaning of the percentage of “Access” category drugs in many cases. Furthermore, for almost all countries including high-income countries, it is difficult to perform AMU surveillance that considers the duration of prescription. To the best of our knowledge, no country has publicly released AMU surveillance data that includes the duration of prescription. As observed to date, the current AWaRe classification is substantially influenced by the duration of prescription, and this often leads to evaluations that differ from our intended purposes. It may be timely to consider revising either the AMU surveillance methodology, the AWaRe classification definitions, or both.

Outstanding questions

When considering AMR countermeasures, it is important to evaluate not only the “quantity” but also the “quality” of AMU. Given the current lack of indicators to easily evaluate the quality of AMU, the WHO AWaRe classification is considered as a potential solution. However, the AWaRe classification still presents many importnat limitations. In particular, the impact of long-term prescriptions is not yet fully understood, even in high-income countries, and implementing the AWaRe classification in healthcare policy without resolving these issues may result in an inaccurate assessment of the quality of AMU and of each country's AMR countermeasures.

Nevertheless, we should take note again that there is a lack of established population-level indicators. Which enable us to evaluate the appropriateness of AMU precisely and their accuracy depend on the quality of data and users’ understanding. Therefore, examining how we can evaluate the appropriateness of AMU more accurately would be our future challenge, and the AWaRe classification might be one of promising indicators with these issues taken into consideration. Epidemiologists and policymakers should be aware of this and aim to improve the AWaRe classification so that it can be used in practical healthcare policy making.

Contributors

ST conceptualised the manuscript. RK and KA analysed the data and created the figures. ST, RK, and YA interpreted the results. All authors critically reviewed the first draft and approved the final version of the manuscript. All authors confirm that they had full access to all the data in the study and accept responsibility to submit for publication.

Declaration of interests

Shinya Tsuzuki is a member of the Technical Advisory Group for the WHO AWaRe Classification of Antibiotics. All other authors declare that they have no conflicts of interest.