There is scant evidence describing antimicrobial (AM) usage in companion animal primary care veterinary practices in the UK. The use of AMs in dogs and cats was quantified using data extracted from 374 veterinary practices participating in VetCompass. The frequency and quantity of systemic antibiotic usage was described.
Overall, 25 per cent of 963,463 dogs and 21 per cent of 594,812 cats seen at veterinary practices received at least one AM over a two-year period (2012–2014) and 42 per cent of these animals were given repeated AMs. The main agents used were aminopenicillin types and cephalosporins. Of the AM events, 60 per cent in dogs and 81 per cent in cats were AMs classified as critically important (CIAs) to human health by the World Health Organisation. CIAs of highest importance (fluoroquinolones, macrolides, third-generation cephalosporins) accounted for just over 6 per cent and 34 per cent of AMs in dogs and cats, respectively. The total quantity of AMs used within the study population was estimated to be 1473 kg for dogs and 58 kg for cats.
This study has identified a high frequency of AM usage in companion animal practice and for certain agents classified as of critical importance in human medicine. The study highlights the usefulness of veterinary practice electronic health records for studying AM usage.
- Companion animals
- Clinical practice
Statistics from Altmetric.com
There is scant evidence describing the extent of antimicrobial (AM) usage in companion animal species attending veterinary practices in the UK, and these species have received limited attention as a reservoir of antimicrobial resistance (AMR). AM usage in companion animals is potentially of considerable importance to the efforts to control AMR, a growing problem in human and animal medicine (Prescott 2008), since companion animals are often in close contact with the human population (Guardabassi and others 2004, Cain 2013). Central to addressing AMR in companion animal veterinary practice is the need for a clear understanding of current levels and patterns of AM usage. The frequency of AM use in companion animals may be growing as a result of increased population, better availability of veterinary services and use of AMs for a range of health conditions (Guardabassi and others 2004). In Norway, there were 338 prescriptions per 1000 dogs per year in 2004, increasing by 13.3 per cent by 2008 (Kvaale and others 2013). Previous studies have described the use of AMs in UK companion animal practices (Mateus and others 2011, Radford and others 2011) and agents most frequently used were amoxicillin-clavulanate, cephalexin, clindamycin and cefovecin (Mateus and others 2011, Radford and others 2011, Summers and others 2012). However, these studies were limited by small sample sizes and were therefore not able to fully quantify AM usage across the UK. Annual antibiotic sales data have been used as a proxy for the assessment of antibiotic use (VMD 2015), with the recognition that sales data are likely to overestimate actual usage and provide little information on the species and dosages used. There is a need to develop ongoing systematic capture of AM usage data by either utilising existing technologies or creating new systems to improve data capture in order to more fully estimate the scale of usage and potential role of companion animals in AMR and zoonotic transmission. There may be geographical differences in patterns of usage, for example, in urban versus rural areas, with potentially important implications for tackling AMR (Kvaale and others 2013). A greater understanding of the absolute quantities dispensed and administered in companion animal practice would aid policymakers in their assessment of the relative contribution of this veterinary subgroup to overall AM levels of usage and AMR.
In order to establish a current baseline for AM usage in the UK, this study aimed to characterise the frequency and quantity of AM usage in cats and dogs over a two-year study period in a large sample of practices participating with the VetCompass programme (www.rvc.ac.uk/VetCompass), which collects deidentified clinical data from veterinary practices across the UK. The objectives of the study were to identify the frequency of AM events and the respective quantities of AM product as indicators of the magnitude of usage, especially those classified as critically important in human medicine, and to evaluate variation in spatial distribution in AM usage across the UK. These data can be used to support policy on AM usage in companion animal species.
The study was approved by the Ethics and Welfare Committee of the Royal Veterinary College (reference number: 2010 1076k). VetCompass is a collaborative research programme that shares veterinary clinical information to support the improvement of clinical services and animal health and welfare (O'Neill and others 2014a). Data were extracted centrally from veterinary practice management software (PMS) systems via a bespoke clinical reporting query (Microsoft Access, 2011). Data from primary care practices only were included in the study (i.e. practices engaged mainly in referral and emergency care were excluded). VetCompass data management is underpinned by the Venom Coding platform (www.venomcoding.org), which provides open-access terminology enabled in the PMSs of participating VetCompass practices.
For the current study, AMs were defined as those medicines that destroy or inhibit the growth of bacterial microorganisms (i.e. antibacterials; Giguère 2013) authorised for systemic use (i.e. injectable, tablets/capsules (‘tablet’) and oral suspensions (‘other oral’)). Other AM agents (e.g. antiviral, antifungal, biocides) were not included in this study. AM agents were categorised into AM groups (e.g. macrolides, penicillins) based on their mechanism of action, chemical structure or spectrum of activity. Data were presented for AM groups and AM agents; specific AM products were not reported. AM agents were categorised as critically important AM agents (CIAs) according to the most recent World Health Organisation (WHO) publication (WHO 2012). A master table of AM agents was created based on data from recent British Small Animal Veterinary Association (BSAVA) formularies (BSAVA 2014, 2013), National Office for Animal Health (NOAH) and Veterinary Medicines Directorate (VMD)-authorised veterinary product databasesi and previous published work on AM usage in small animals (Mateus and others 2011). For each specific AM agent listed in the master table, information was collated on agent group (e.g. cephalosporin), recommended International Non-proprietary Name (cefalexin), authorised UK trade names (Cephacare) and specific item names (Cephacare Flavour 250 mg Tablets for Dogs). Product formulation, method of administration and base active ingredient strength (i.e. removing the molecular weight of associated water and/or salt) were also recorded for each unique AM product. The master table included veterinary products licensed for use in companion animals, other species and human beings. Products licensed for other animal species and human beings may be used in companion animals under the Cascade principle, a regulatory framework that allows such use when there are no available alternative licensed veterinary products to appropriately treat an animal patient (EU Commission Regulation no. 37/2010).
The study population included all dogs and cats that had at least one electronic patient record (EPR) entry (clinical note, VeNom term, bodyweight or treatment record) within the VetCompass programme database within the two-year study period from June 1, 2012, to May 31, 2014. EPRs were analysed for recorded dispensing and administration of AM events. Written prescriptions accounted for a minority (0.10 per cent) of all AM events and were excluded because it could not be ascertained whether these scripts were subsequently fulfilled. AM agent usage was reported by episode of care, that is, each independent record in the PMS was defined as a single event. The number of single and repeated AM therapies given to animals within the study period, identified via unique patient identification numbers, was reported separately but without reference to whether these repeated events were for distinct conditions, recurring conditions or repeated prescriptions. AM products that combined multiple agents within the same preparation (e.g. amoxicillin plus clavulanic acid) were classified as a single AM substance (potentiated amoxicillin). If potentiated AM products contained multiple AM agents (e.g. Stomorgyl, Merial Animal Health, containing spiramycin and metronidazole), the quantity of active ingredients was calculated for each AM substance.
The usage of AM agents was identified from the VetCompass database by querying the AM terms from the master AM table against the clinical records (Microsoft Access, 2011) in order to extract relevant treatment records. Additional data from relevant records were extracted from the following data fields: clinic (identification code and postal code location), unique patient identification code, species (dog or cat), the date of AM event, product item name, units purchased by client and dosage (free text). Details, where available, identified from the product item name and dosage fields included the product name and size of product (i.e. strength), and label information referencing method of administration (systemic: injection, oral (tablet) or other oral (e.g. liquid/powder)). The total quantity of AM product per event was calculated using the number of units purchased by clients associated with each event, as recorded in the PMS. The quantity of each AM product per event was calculated using the base strength per mg/ml of active ingredient of the product multiplied by the mg or ml of product dispensed or administered, where this information was available. Quantities were reported separately for AM groups and individual AM agents for dogs and cats.
The geographical distributions of the study population and frequency of AM events were plotted for dogs and cats, separately. Clinic postcode data were used to generate coordinates for mapping geographical locations of all UK veterinary clinics registered with the Royal College of Veterinary Surgeons (RCVS), including those VetCompass clinics from which AM data were derived. Clinic sites and UK postcode area boundaries were projected onto shapefiles representing the UK (with polygons for each UK postcode area) according to the British National Grid system. Figures for geographical data description were produced using ArcMap version 10.2 geographic information system software (ESRI 2015. ArcGIS Desktop: Release 10.2. Redlands, California, USA, Environmental Systems Research Institute). The frequency maps were based on mean number of AM events divided by the total number of practice-attending dogs or cats in the study population for participating VetCompass clinics in that area. Density categories were defined using a system of equal intervals: darker-coloured regions represented greater density (i.e. more animals or higher AM use). Moran's I test of spatial autocorrelation (GeoDA Software, School of Geographical Sciences and Urban Planning, Tempe, Arizona, USA) was used to determine whether there was significant clustering of postcode areas with similar mean AM events. The spatial weights matrix was based on queen contiguity (common border and/or corner) and significance was based on 499 permutations (P<0.05).
Deidentified clinical data from 374 UK companion animal practices were accessed via the VetCompass programme. Participating practices comprised principally of two large practice groups as well as several independent practices and reflected a geographically widely dispersed cohort of practices. Three clinics (0.8 per cent) did not have a usable postcode and were therefore not included in any map figures throughout the report, but were included for all other AM event and quantity analyses. The majority of practices were located in England, particularly the Midlands and East England, and few practices were located in northern Scotland. In total, across the practices, 963,463 dogs and 594,812 cats had at least one EPR recorded within the two-year period and these animals comprised the study population, shown in Fig 1.
Frequency of AM usage events
Over the two-year period, 242,736 of the 963,463 study dogs (25.19 per cent; 95 per cent confidence interval (CI) 25.11 to 25.28) and 122,594 of the 594,812 study cats (20.61 per cent; 95 per cent CI 20.51 to 20.71) were given at least one AM event. A total of 676,712 dispensing and administering events related to 211 AM products with a unique Marketing Authorisation (MA) number. Of these, 472,159 (70.48 per cent) and 196,923 (29.52 per cent) AM events were for dogs and cats, respectively (Table 1). The method of administration of AM events was principally oral tablets for dogs (381,532 events, 80.81 per cent of total dog events) and injections for cats (109,187 events, 55.45 per cent, Table 1). For both species, the main agents dispensed or administered were penicillin types (dogs: 254,394 events, 53.88 per cent of total; cats: 91,318 events, 46.37 per cent) and cephalosporins (dogs: 80,982 events, 17.15 per cent; cats: 63,505 events, 32.25 per cent; Fig 2). WHO CIAs were used in 284,721 events in dogs (60.30 per cent of total) and 159,433 events in cats (80.96 per cent of total). Of CIAs, agents of highest importance were used in 30,241 of events in dogs (6.40 per cent of total) and 68,084 events in cats (34.57 per cent; Table 1). Fluoroquinolones and third-generation cephalosporins were the most commonly used CIAs of highest importance in dogs and cats, respectively. Potentiated agents were used for 229,919 (49 per cent) and 59,496 (30 per cent) events in dogs and cats, respectively, and potentiated amoxicillin (amoxicillin-clavulanate) was the most commonly used potentiated agent for both species. For both dogs and cats, pleuromutilin (e.g. tiamulin), dihydrofolate reductase inhibitor (e.g. trimethoprim) and aminoglycosides (e.g. amikacin) events were low in frequency (<0.01 per cent of total AM quantity) and these groups were therefore not reported further (including for quantity), but are included in total frequency and quantities for AM usage.
Of dogs that received at least one AM event, 139,920 dogs (57.64 per cent) received a single AM event over the two-year period, while 102,816 dogs (42.36 per cent) received multiple AM events. For cats, 84,175 cats (68.66 per cent) received a single AM event, while 38,419 cats (31.34 per cent) received multiple AM events. The median number of events per animal was 1 (IQR 1–2; range 1–60) for dogs and 1 (IQR 1–2, range 1–216) for cats. The maximum of 216 events for one cat was an outlier (though appeared biologically possible), and after removal, the next highest maximum was 75 events for a single cat over the two-year period.
Spatial distribution of AM events
Figure 3 displays the geographical distribution of the total number of AM events dispensed or administered for dogs and cats. Statistical comparisons were not made but dogs appeared to receive a higher frequency of AM events per animal than cats in many regions, with a density range of 0.10–1.09 AM events per dog per postcode area versus 0.04–1.00 AM events per cat per postcode area. Both dogs and cats appeared to have a similar disparate pattern of regional AM events, with greater proportional use in the south of England and southwest Scotland regions. There was statistically significant positive spatial autocorrelation of mean AM events in dogs (Moran's I: 0.221; P=0.002) but not in cats (Moran's I: 0.062; P=0.100).
Quantity of AM used
The quantity of AM administered/dispensed was calculated for 470,159 events for dogs (99.57 per cent of the total dog events) and 195,128 events for cats (99.09 per cent of the total cat events), after removing events where strength of the active ingredient could not be determined from the data available (2000 for dogs; 1795 for cats).
The overall quantity of AMs used in dogs was 1472.910 kg and 58.383 kg for cats over the two-year study period (Table 2). The agent groupings with the highest quantities administered/dispensed were penicillin types (dogs: 568.197 kg; cats: 33.810 kg) and cephalosporins (dogs: 505.004 kg; cats: 5.746 kg; Fig 4; Table 2), though there was also relatively high quantities of nitromidazoles for dogs (170.237 kg) and of lincosamides for cats (8.565 kg). For individual AM agents in dogs, the highest quantities were given for potentiated amoxicillin (538.473 kg), cefalexin (503.672 kg), metronidazole (152.004 kg) and potentiated sulfadiazine (114.004 kg; Table 2). For individual agents in cats, the highest quantities were given for potentiated amoxicillin (31.206 kg), clindamycin (8.331 kg), metronidazole (3.763 kg) and cephalexin (3.556 kg; Table 2).
A broad evaluation of AM usage within veterinary practices is now possible with the development of practice-based research programmes such as VetCompass. This study reports on the usage of AM products for dogs and cats across a large sample of UK veterinary practices. The results highlight the relatively high frequency of usage of AMs per animal in veterinary practice and of those considered of critical importance to human health (CIA) and provide a valuable evaluation of current veterinary prescribing activity and a baseline from which to evaluate future usage.
Of those animals which attended a veterinary practice during the two-year study period, 25 and 21 per cent of dogs and cats, respectively, were dispensed or administered at least one AM. Mateus and others (2011) reported administration or prescription of AMs in 45 and 33 per cent of dogs and cats that presented for consultations, respectively, over a one-year period in 2007. Radford and others (2011) reported a similar proportion of AM use per consultation, 35 per cent of consultations for dogs and 49 per cent consultations for cats. These earlier AM usage figures reflect different methods of reporting AM events per consultation or AM usage within dogs presenting for consultations and represented much smaller sample sizes of 16 and 11 practices, respectively. The current study reported usage per animal under practices' veterinary care and included animals deemed to be actively registered at participating practices as indicated by EPR evidence of presenting for a consultation as well as for other services or recorded communication with the practice during the study period. The current study utilised data from 374 small animal veterinary practices, which is the largest sample used to investigate AM usage in the UK to date, representing approximately 7 per cent of all UK RCVS registered veterinary premises (RCVS 2014). The participating practices came from all regions of the UK, and although the data derived principally from two large practice groups, a number of independent practices were also involved and therefore the data were likely to give a reliable example of the level of usage across UK companion animal general practice. Emergency and referral hospitals or charity-based practices were not included in this study, and thus the level of AM usage reported here may be less generalisable to those practice types. The representativeness of data in the current study according to the total UK dog and cat population is difficult to determine, and although the values are accurate in terms of AM events per animals attending study practices, the data may overestimate usage per animal in the population due to the absence of animals that are not registered at a veterinary practice or that did not attend a veterinary practice within the two-year study period. For example, it is known that not all dogs and cats are registered with veterinary practices; Asher and others (2011) reported that approximately 17 per cent of owners in a public survey had not registered their dog with a veterinary practice.
In the current study, there was widespread use of broad-spectrum agents such as aminopenicillins plus clavulanic acid, cephalosporins and fluoroquinolones across the participating VetCompass practices for both dogs and cats, in agreement with previous UK data (Mateus and others 2011, Radford and others 2011) and data from other European countries, for example, France (Anses 2014). Many of these agents, and up to 81 per cent of the AM events together across both species, were considered to be critically important for human health (WHO 2012), including those deemed of highest importance (approximately 40 per cent of AMs used). This raises concerns about potential horizontal transmission of resistance determinants and resistant bacteria to CIAs through companion animals. Mateus and others (2011) reported a similar proportion of CIA usage, 61 per cent in dogs and 83 per cent in cats from 2007, which suggests usage of CIAs has not decreased in the UK from 2007 to 2014. A recent survey study reported 30 and 15 per cent of antibiotics used for dogs and cats, respectively, in Europe were highest importance CIAs, with more AMs prescribed being considered of lower importance (e.g. tetracyclines, De Briyne and others 2014), suggesting that there may be differences in prescribing behaviour between EU member states. Only a small proportion of veterinarians were surveyed and the methodological differences in how these data were collected may also contribute to these differing results.
AM usage patterns differed substantially between dogs and cats in the current study. Administration of AM agents used for dogs was principally by oral tablet (81 per cent), whereas the majority of AM agents for cats were administered via the injectable method (55 per cent). This likely reflects general differences in methods of administering medicines between the species, with oral tablets being perceived as more difficult to effectively administer in cats (Traas and others 2010). Dogs received proportionally more aminopenicillin types (54 per cent v 46 per cent) and nitromidazoles (11 per cent v 5 per cent) than cats. Cats received proportionally higher usage of third-generation cephalosporins, critically important agents of highest importance, and this was largely explained by the more frequent use of cefovecin-injectable products in cats (54 per cent of total cat events v 1.31 per cent of dog events). This is in agreement with other studies, for example, Murphy and others (2012) found higher use of cefovecin in cats and amoxicillin-clavulanic acid in dogs. A higher frequency of cats received a single AM event over the two-year period, 69 per cent of cats versus 58 per cent of dogs. However, the range of counts for repeat AM events was more diverse for cats. These differences may reflect differences in the deliverance of AMs for the two species, for example, preference for a single long-lasting injectable in cats. No attempt was made to evaluate the underlying disorders requiring AMs in these animals, or the clinical appropriateness of dosages, though this is possible with further analysis of the data contained with VetCompass. Murphy and others (2012) showed overuse of cefovecin and fluoroquinolones for the treatment of common diseases in dogs and cats (feline upper respiratory tract disease, feline lower urinary tract disease and canine infectious tracheobronchitis) in Ontario, where 67–74 per cent of disease events were treated with AMs and 65 per cent of AMs prescribed were β-lactams.
For both dogs and cats, spatial analysis suggested that there was a trend towards higher AM event frequency in southeast England, south Wales and southwest Scotland, with significant spatial clustering observed in dog but not cat AM usage overall, though the latter lack of statistical significance may have reflected the smaller sample size for cats. Spatial distribution accounted for variation in animal density distributions, and so any differences identified were unlikely to be explained by regional differences in the numbers of animals attending practices. Possible explanations for the geographical variation seen could be differences in animal demographics, associated diseases and/or regional variations in prescribing and administering behaviours of the veterinary practices. The spatial clustering undertaken was exploratory only in nature, and any underlying reasons for differences in regional distribution of AM events would need to be investigated further. Similar regional differences in AM prescriptions were described in Norway (Kvaale and others 2013), although here, the differences were likely to correspond to the density of dogs and veterinary clinics. Geographical differences in AMR have also been reported; dogs in an urban habitat had a higher risk of carrying isolates resistant to methicillin and other antimicrobials compared with dogs in a rural environment (Huerta and others 2011). Such differences may in part be related to differences in AM usage, exposure to other sources of AMR and/or awareness of regional resistance patterns. Nonetheless, maps of the geographical distribution of AM usage have been used in risk-based sampling approaches to monitor AM usage and AMR (Stärk and others 2006) and future work is merited to explore these geographical differences further.
For both species, the quantities of AM used corresponded approximately with the number of events recorded, for example, proportionally greater number and quantity of penicillin types were dispensed and administered in both cats and dogs. The quantity data also reflected the method of administration since tablets had a greater strength and/or longer recommended course of therapy than injection or other oral medicines. Substantially greater quantities (1472.910 kg) of AMs were administered or dispensed for use in dogs compared with cats (58.383 kg), which would be explained at least in part by the smaller average bodyweight of cats (O'Neill and others 2014a, b). The data for units purchased by clients were likely to correspond closely to the actual levels of dispensed and administered AMs as these data are derived from financial transactions. It is important to note, however, that data on AMs purchased by clients may overestimate actual use since there may be wastage due to pack sizes that exceed dosage needs and due to medicine expiry. The approximate dose rate could be calculated to validate the data, based on the total quantity sold and number of events. For example, 2 kg of AM related to 59,000 cefovecin-injectable events, which equated to approximately 34 mg per injection or around 8 mg/kg for an approximate average 4 kg cat, which is the recommended dose (Convenia, Noah Compendium).
Data analysis in this study was limited by the lack of standardisation of some EPR fields. For example, >15,000 unique AM treatment items were recorded across the practice data, and this related to only 211 unique products. Data entry varied with different practitioners, practices and PMS systems, including variations in spelling, order of wording, limited/incomplete options available and/or a lack of specificity in free-text descriptions. It is possible that the master list of AM agents included in this study was not exhaustive and did not allow complete identification of all AM medicinal products available and used in companion cats and dogs. In particular, the use of AM products not licensed for animals (e.g. human medicines) would not be listed in the veterinary authorised databases. It is not known to what extent AM products are dispensed or administered to veterinary patients under the Cascade system. Neither the commercial MA number nor the Global Trade Item Number for veterinary medicinal product sales are currently routinely recorded in PMS databases, but could provide complete and automated identification of UK-authorised AM products for all species.
With the frequent use of AMs for cat and dog species, and in particular of agents considered critically important to human health, combined with evidence of the risk for AMR for AM therapy in pet species and the potential for zoonotic transmission (Guardabassi and others 2004), there is a need to reflect on current usage patterns for small animal veterinary species and to further develop protocols for responsible AM usage. Further work should focus on the assessment of appropriateness of AM therapy using more detailed analysis of the clinical condition, the relatedness of treatments and the dosage applied, and this may further identify important patterns of AM usage in both species. Baseline data on the types and quantities of AMs used in dog and cat species are essential to enable associations and trends to be identified, followed and analysed and appropriate adjustments to best-practice guidelines to be made. Such data may be used for benchmarking practices in order to help veterinarians reduce use of antimicrobials. In the Netherlands, discovery of extensive overuse of AM and significant reservoirs of AM-resistant pathogens led to a successful collaboration between government and stakeholders to reduce AM use in farm animals by as much as 56 per cent (Speksnijder and others 2015). AM use in Danish pig production significantly reduced with the introduction of the ‘yellow card’ intervention, as monitored with the national database of veterinary prescribed medicines (Jensen and others 2014). Databases such as the VetCompass programme could be used to monitor temporal and geographical trends for AM use in the UK, following compulsory or voluntary actions. Practice-based guidelines on appropriate use of antimicrobials have been developed in the UK by expert panels (e.g. FECAVA, 2014, BSAVA, 2016) though relevant policies were reported to be applied in as few as 3.5 per cent of small animal veterinary practices (Hughes and others 2012). Complex intrinsic and extrinsic factors affect veterinarians' decision-making for prescribing AM therapy, such as a veterinarian's preference for certain products, perceived efficacy, ease of administration and perceived owner compliance (Mateus and others 2014). Better understanding of these factors, and of the importance of AMR transmission between pets and human beings, could promote more conscientious AM prescribing behaviour in veterinary clinics (Beco and others 2013).
Overall, approximately a quarter of dogs and cats attending veterinary practices in the UK received at least one AM event over the two-year period 2012–2014. Dogs mainly received oral tablets while cats had AMs administered mainly as injectable preparations. The total AM quantity, by weight of active ingredient, was estimated to be 1473 kg for dogs and 58 kg for cats. In particular, the most common agents used for cats and dogs were aminopenicillin types and cephalosporins. Of the AM events, 60 per cent in dogs and 81 per cent in cats were of AMs classified as critically important (CIAs) to human health by the WHO, and this may be important to consider when addressing companion animals as a potential source or reservoir for AMR in human beings. These findings can provide a baseline for AM usage in companion animals in the UK and can support continued surveillance of AM usage and investigation of the role of companion animal veterinary practices in AMR.
The authors are grateful to the VMD, who funded the project. Thanks to Peter Dron (RVC) for VetCompass database development and Noel Kennedy (RVC) for software and programming development. Thanks also to Dr Kim Stevens, who undertook the spatial clustering analysis. The authors acknowledge the Medivet Veterinary Partnership, Vets4Pets/Companion Care, Blythwood Vets, Vets Now and the other UK practices who collaborate in VetCompass. They are also grateful to The Kennel Club, The Kennel Club Charitable Trust and Dogs Trust for supporting VetCompass.
Provenance: Not commissioned; externally peer reviewed
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.