Antibiotic Resistance in Pacific Island Countries and Territories: A Systematic Scoping Review

Several studies have investigated antimicrobial resistance in low- and middle-income countries, but to date little attention has been paid to the Pacific Islands Countries and Territories (PICTs). This study aims to review the literature on antibiotic resistance (ABR) in healthcare settings in PICTs to inform further research and future policy development for the region. Following the PRISMA-ScR checklist health databases and grey literature sources were searched. Three reviewers independently screened the literature for inclusion, data was extracted using a charting tool and the results were described and synthesised. Sixty-five studies about ABR in PICTs were identified and these are primarily about New Caledonia, Fiji and Papua New Guinea. Ten PICTs contributed the remaining 21 studies and nine PICTs were not represented. The predominant gram-positive pathogen reported was community-acquired methicillin resistant S. aureus and the rates of resistance ranged widely (>50% to <20%). Resistance reported in gram-negative pathogens was mainly associated with healthcare-associated infections (HCAIs). Extended spectrum beta-lactamase (ESBL) producing K. pneumoniae isolates were reported in New Caledonia (3.4%) and Fiji (22%) and carbapenem resistant A. baumannii (CR-ab) isolates in the French Territories (24.8%). ABR is a problem in the PICTs, but the epidemiology requires further characterisation. Action on strengthening surveillance in PICTs needs to be prioritised so strategies to contain ABR can be fully realised.


Introduction
The prevention and treatment of infectious diseases is increasingly being challenged by the growing spread of antimicrobial resistance (AMR) [1]. AMR is responsible for an estimated half a million deaths each year in Europe and the USA and thousands more globally [2,3]. If AMR is not contained, experts estimate the excess mortality rate will rise to 10 million deaths per year by 2050 [3], with nearly half of these occurring in the Asia Pacific region [4].
In 2002, the World Health Organization (WHO) Western Pacific Region Office (WPRO) identified AMR as a public health concern for the region [5]. In 2014, countries in the region agreed to strengthen their capacity to respond to AMR, and the Action Agenda for AMR was drawn up and endorsed by member countries [6]. In 2015, WHO WPRO undertook a situational review of surveillance and health systems response to AMR in the region [7]. Their findings indicated wide variation amongst countries with respect to AMR, including the capacity to participate in regional surveillance networks; regulations around the purchase and use of antimicrobial agents; support provided by health systems in the containment of AMR; and in their understanding and awareness of AMR [6,7].
Pacific Islands Countries and Territories (PICTs) are amongst WHO WPRO's low-and lower-middle-income groups, except for PNG and Tuvalu which are considered middle-income [8]. PICTs face unique challenges when it comes to addressing AMR. Their small size, remoteness, limited resource bases, fragile health infrastructures and susceptibility to natural disasters make them particularly vulnerable to AMR [7].
Several research studies have investigated AMR in low and middle-income countries [1,9], but to date, little attention has been paid to PICTs. Many PICTS have limited clinical diagnostic and laboratory capacity, which affects the availability and quality of culture and susceptibility testing data. Therefore, in most PICTs the resources and capacity to conduct good quality research into AMR is limited and published evidence on AMR in the region is scarce [10]. A scoping study was undertaken to map the evidence about antibiotic resistance (ABR) in PICTs to gain an overall understanding of ABR in the region. The findings may also inform future policy initiatives and strengthen health system adaptive capacity to contain AMR.
A scoping study was chosen for two reasons: because the literature about ABR for this region is limited and because the scoping study design allows for the capture of a broad range of results regardless of study design.

Geographic Setting
The PICTs included in this scoping study are the 22

Search Strategy
Literature searches were conducted using the following databases: PubMed, Embase, SCOPUS and Web of Knowledge accessed through the Australian National University Library system. The online systems of the National Library of Australia and the WHO Western Pacific Region (WPR) International Research Information Service (IRIS), the WPR Index Medicus (WPRIM) and Google Scholar were searched for additional articles and grey literature. Select key terms used included antimicrobial resistance, antibiotic resistance, antibacterial agents or drug or multi-drug combined with resistance or susceptibility. These terms were combined with the associated database descriptors and searched across each named PICT. The reference lists of articles and reports retrieved were searched manually for additional citations. All retrieved items were entered in an Endnote library. The details of the databases accessed, and the final search strategies used for two databases can be found in Supplementary Material S1. (Databases and grey literature sources accessed and search strategies for two databases).
Reports on ABR in humans in PICTs; 2.
Published in English or French between 1950 and 2018; 3.
Available in full text.
Reports about tuberculosis; (in view of solid literature base already known about drug resistant tuberculosis, and globally funded TB program); 2.
Literature which did not provide details about antibiotic susceptibility in PICTs; 3.
Conference abstracts and posters; and newspaper articles.

Selection and Screening
The titles and abstracts of all results retrieved underwent an initial screen guided by minimum inclusion and exclusion criteria by one member of the team. Duplicates and candidate studies not meeting the criteria were excluded. The full text of all remaining studies was obtained and stored in the Endnote library. Four members of the team working independently carried out a second screening of the studies using the full inclusion and exclusion criteria. In situations where there was uncertainty, the team members reached a decision through discussion. If published reviews included data that were available from original articles, the data were extracted from the original articles. A flow chart ( Figure 1) provides details of the number of items screened and assessed for eligibility. 3. Conference abstracts and posters; and newspaper articles.

Selection and Screening
The titles and abstracts of all results retrieved underwent an initial screen guided by minimum inclusion and exclusion criteria by one member of the team. Duplicates and candidate studies not meeting the criteria were excluded. The full text of all remaining studies was obtained and stored in the Endnote library. Four members of the team working independently carried out a second screening of the studies using the full inclusion and exclusion criteria. In situations where there was uncertainty, the team members reached a decision through discussion. If published reviews included data that were available from original articles, the data were extracted from the original articles. A flow chart (Figure 1) provides details of the number of items screened and assessed for eligibility.

Data Extraction
A charting tool in the form of an Excel spreadsheet was developed by two members of the team to extract and record the following information about each article: first author; publication date; country of focus; date research conducted; sample size; study design; age group; type of infection; bacteria isolated; specimen types; antibiotic susceptibility testing method; and antibiotics tested; acquisition being community or healthcare associated. Publications were broadly categorised under the following types: journal articles and published reports. Journal articles were further classified into randomised control trial, clinical trial, prospective and retrospective cohort, case control study, cross sectional survey, case study, descriptive study, laboratory study and antibiotic guideline. Reports included surveillance and outbreak reports. Guidelines included antibiotic guidelines which included antibiograms. Age groups were categorised as neonate (<1 year ), infant (≥1 to <3 years), children (3 to ≤18 years), and adult (>18 years). The results of the charting process are detailed in Supplementary Material S2. (Characteristics describing studies reporting gram-negative and grampositive pathogens in PICTs), which describe the studies reporting gram-negative bacteria and grampositive bacteria respectively.

Data Extraction
A charting tool in the form of an Excel spreadsheet was developed by two members of the team to extract and record the following information about each article: first author; publication date; country of focus; date research conducted; sample size; study design; age group; type of infection; bacteria isolated; specimen types; antibiotic susceptibility testing method; and antibiotics tested; acquisition being community or healthcare associated. Publications were broadly categorised under the following types: journal articles and published reports. Journal articles were further classified into randomised control trial, clinical trial, prospective and retrospective cohort, case control study, cross sectional survey, case study, descriptive study, laboratory study and antibiotic guideline. Reports included surveillance and outbreak reports. Guidelines included antibiotic guidelines which included antibiograms. Age groups were categorised as neonate (<1 year ), infant (≥1 to <3 years), children (3 to ≤18 years), and adult (>18 years). The results of the charting process are detailed in Supplementary Material S2. (Characteristics describing studies reporting gram-negative and gram-positive pathogens in PICTs), which describe the studies reporting gram-negative bacteria and gram-positive bacteria respectively.

Synthesis of Results
The studies were divided according to whether the bacterial pathogens analysed are gram-negative or gram-positive. Within these two broad categories, the findings reported about each bacterium, including the PICT(s) in which it was reported, the infections the bacterium caused, the population involved, and the level of susceptibility reported for each antibiotic tested is summarised. Gram-negative bacteria causing healthcare associated infections are grouped together and summarised before other gram-negative bacteria. The summary is supported by seven tables, which describe the study characteristics and report the prevalence of ABR in selected organisms by PICT and study date. Further details can be found on the PRISMA-ScR checklist [11] (Supplementary Material S3: Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) Checklist).

Study Characteristics
Sixty-five studies, reports and guidelines met the selection criteria and provided information about ABR in 12 of the 22 PICTs. The publications were primarily about PNG (n = 26), Fiji (n = 8) and New Caledonia (n = 10). Four studies were conducted in French Polynesia, two each in Cook Islands and Samoa and one each for Solomon Islands and Wallis and Futuna. Fiji, Kiribati, Marshall Islands, Micronesia, PNG, Samoa, Solomon Islands and Tonga were all mentioned in the eight studies which included multiple PICTs. Three studies reference ABR in Pacific Islanders living in countries outside of the PICTs. Most (48%) studies focused on community-acquired infections, while 16% focused on healthcare-acquired infections (HCAIs) and 17% on both. The source of acquisition was unclear in the remainder (19%). Further details can be found in Supplementary Material S2 which provide details of the studies referring to gram-negative and gram-positive bacteria, respectively.

Antibiotic Susceptibility Test Methods
Several methods were used to determine resistance patterns. The most widely used were the disk diffusion and E-test gradient diffusion methods. Disk diffusion was used alone in 11 (17%) studies and combined with other methods in 26 (40%), whilst E-test gradient diffusion was combined with disk diffusion in 18 studies and used alone in two (3%). The agar dilution and broth microdilution techniques, the replica plating method, the automated VITEK system, the ATB PNO strip and molecular detection (PCR) were all used as the sole method in seven (11%) studies or in combination with another method. The remaining studies did not report the method used (19, 29%).

Acinetobacter Baumannii
A. baumannii was reported in studies about HCAIs in French Polynesia, New Caledonia and Fiji [15,19,[22][23][24]. Carbapenem-resistant A. baumannii (CR-Ab) was isolated from clinical specimens in 2004 from hospitals in both French Polynesia and New Caledonia [22][23][24]. Twenty-four patients in French Polynesia were either colonised (19; 80%) or infected (5; 20%) with A. baumannii. This bacterium was linked to pneumonia, skin and soft tissue, surgical site or blood stream infections. The isolates were resistant to all beta-lactams and 20 (83% 20/24) were susceptible to colistin and aminoglycosides [24]. In New Caledonia, CR-Ab isolates represented 24.8% (123/202) of multidrug-resistant bacteria (n = 202); the isolates were resistant to beta-lactams, quinolones and aminoglycosides except amikacin and tobramycin, but susceptible to colistin [22,23]. The types of infections in French Polynesia were like those found in New Caledonia, but also included urinary tract infections. Both outbreaks were associated with the OXA-23 producing A. baumannii clone [22][23][24]. HCAIs were the subject of two studies conducted in Fiji at CWMH during 2011 and 2012 [15,19]. These studies involved patients in adult and neonatal ICUs. Neither study provided antibiotic susceptibility data. The adult ICU study found that 21% (92/437) of isolates cultured from blood, respiratory tract, surgical site and urinary specimens to be Acinetobacter spp. [19]. In the neonatal ICU study, A. baumannii was responsible for 14.5% (15/103) of gram-negative sepsis [15].

Pseudomonas aeruginosa
Pseudomonas aeruginosa was reported in PNG, New Caledonia, Fiji, Samoa and Cook Islands [12,14,18,19,23,25,26]. In PNG in 1999, 11 P. aeruginosa isolates were cultured from blood specimens taken from 54 children with severe sepsis in Goroka Hospital and 82% were resistant to gentamicin [1]. In 2009 at the Modilon Hospital in PNG, one HCA P. aeruginosa isolate (1/9; 11%) was found to be resistant to tetracycline, cotrimoxazole, chloramphenicol and ampicillin, but susceptible to ciprofloxacin and gentamicin [12]. A laboratory study in 2004 in New Caledonia found seven (3.5%) ceftazidime-resistant P. aeruginosa isolates amongst other MDR bacteria (n = 202) [23]. Isolates collected from all sources in 2015 and 2016 in Samoa were 24% (9/37) resistant to ciprofloxacin. Reduced susceptibility was shown to gentamicin [26]. P. aeruginosa isolates cultured from all sources between 2015 and 2017 at the Rarotonga Hospital microbiology laboratory, Cook Islands were shown to have reduced susceptibility to ceftazidime (n = 117) and, ciprofloxacin and gentamicin (n = 154), but all were susceptible to meropenem [25].
Four studies reported on Enterobacter spp. for PNG, Fiji and French Polynesia [14,19,27,28]. A 1997 investigation into severe sepsis in children in Goroka Hospital, PNG found 12% (7/61) of positive cultures to be Enterobacter spp.: three were HCAIs and four were community-acquired [14]. Whilst all were resistant to chloramphenicol, four were resistant to gentamicin. In 2007 an outbreak investigation of E. aerogenes in the neonatal ICU in Fiji's CWMH found 55.5% (10/18) of septicaemia were caused by ESBL producing E. aerogenes [28]. The isolates were resistant to ampicillin, trimethoprim-sulfamethoxazole, gentamicin, cephalothin, and ceftriaxone, but remained susceptible to meropenem, amikacin with intermediate resistance to ciprofloxacin. In 2015 a single isolate of imipenem-resistant (IMI-1 producing) E cloacae was detected in an adult male in the main hospital in Papeete, French Polynesia [27]. This isolate was resistant to aminopenicillins and carboxypenicillins, amoxicillin/clavulanic acid, first and second generation cephalosporins, as well as imipenem. The isolate remained susceptible to the ureidopenicillins, to ceftazidime, cefepime and cefotaxime, meropenem, ertapenem and doripenem, as well as the non-beta-lactams [27].

Salmonella spp.
Seven studies about resistance patterns in Salmonella spp.
Antibiotic resistance in Campylobacter spp. was reported for PNG [18]. The bacteria were isolated from blood, stool, lung and skin tissue, and urine samples between 1984 and 1986: 22 (n = 22) isolates of C. jejuni and 33 of C. coli were resistant to cotrimoxazole (100%) and ampicillin (24%), but remained susceptible to chloramphenicol, tetracycline and gentamicin [18].
Several studies focused on S. aureus in Pacific Islanders living in Hawaii [56], New Zealand [59] or Australia [52]. The findings indicated that patients from PICTs were over-represented among cases with nmMRSA. In Hawaii, where Pacific Islanders comprise 24% of the total population, 51% of nmMRSA isolates (176/346) were cultured from specimens taken from Pacific Islanders [56].
Two retrospective studies investigated ABR in invasive S. pneumoniae and H. influenzae (Hib) isolates in children in PNG, from 2006 to 2009 in Modilon Hospital and 1996 to 2005 in Goroka Hospital [65,76]. In both studies isolates from blood and/or CSF specimens were taken from hospitalised patients prior to the 2008 and 2014 roll-out of the Hib and PCV13 vaccines, respectively. The Modilon study found all Hib isolates (n = 15) were chloramphenicol-resistant, while S. pneumoniae isolates (n = 17) exhibited reduced susceptibility. All isolates were susceptible to ceftriaxone [76]. The Goroka study conducted between 1996 and 2005 revealed ABR in both pathogens was demonstrated across most antibiotics tested during the period [65]. Some 32.5% (53/165) of H. influenzae isolates type B (Hib) were beta-lactamase-positive and resistant to ampicillin and cotrimoxazole. In addition, 98% (161/165) and 96% (158/165) of the Hib isolates were resistant to chloramphenicol and tetracycline respectively. A total 21.5% (39/180) of S. pneumoniae isolates were penicillin-resistant and 4% were either cotrimoxazole, tetracycline or chloramphenicol resistant (7/180) [65]. The serotypes responsible for penicillin-resistance and associated with S. pneumoniae were examined in a study from New Caledonia in 2005 (n = 298) [70]. Invasive isolates were cultured from blood (37%), spinal fluid (6%) and respiratory specimens (41%). Findings indicate that 14.4% (43/298) of isolates had reduced susceptibility to penicillin, with five isolates being fully resistant. Some 3.7% (12/298) expressed intermediate susceptibility to amoxicillin and 1.7% (5/298) to both amoxicillin and cefotaxime. Several isolates belonged to the group of serotypes present in the Pacific, which are more likely to express reduced susceptibility to penicillin [70].
Two nasopharyngeal carriage studies in infants < 2 years reported on antibiotic resistance in S. pneumoniae in New Caledonia and Fiji in 2005 and 2006, respectively [60,71]. In New Caledonia, 52% (544/1040) of nasopharyngeal specimens were positive for S. pneumoniae and 21% (114/544) were penicillin-resistant [60]. Although, penicillin-resistance in the Fiji study was lower at 11.4% (28/246), resistance to cotrimoxazole was 20.3% (50/246). Approximately 72% (20/28) of isolates resistant to penicillin were also resistant to cotrimoxazole. Two isolates were also resistant to ceftriaxone, four to erythromycin and three isolates were multidrug-resistant. All isolates were fully susceptible to chloramphenicol [71]. The serotypes with reduced susceptibility in both Fiji and New Caledonia included those identified in the earlier 2005 New Caledonian study mentioned above [60,70,71].

Streptococcus pyogenes (Group A Streptococcus)
Only four studies reported on susceptibility patterns in Group A Streptococcus: PNG, New Caledonia, Fiji and Samoa [50,58,78,79]. Ninety (90) cases of Group A Streptococcus were recorded in New Caledonia during 2006. Isolates from skin and soft tissue, blood and, pleural, spinal and amniotic fluid were 10% resistant to tetracycline, but susceptible to all other antibiotics tested including; penicillin, amoxicillin, gentamycin, erythromycin, vancomycin, streptomycin and rifampin [78]. In 2016 in Samoa, 15% of Group A Streptococcus isolates (11/78) were resistant to erythromycin and 23% to (53/70) and clindamycin [50].
Two studies examining E. faecalis and E. faecium were identified [80,81]. A single case of vancomycin resistant E. faecalis in a patient hospitalised for an intestinal obstruction was reported in New Caledonia in 2006. Three isolates cultured from stool and urine samples were resistant to vancomycin and teicoplanin due to the presence of the vanA gene, as well as to streptomycin, erythromycin and kanamycin [81].

Discussion
Sixty-five studies on ABR in PICTs spanning five decades (1958-2018) were identified. Whilst the available data suggest widespread ABR across PICTs, there was scarce or a complete absence of data from several countries.
S. aureus was the predominant gram-positive pathogen, mainly in association with skin and soft tissue infections in the community setting. Rates of MRSA ranged widely: ≥50% in PNG and > 20% in other PICTs [25].
Several gram-negative bacteria listed on the WHO Priority Pathogen List [69] were reported in PICTs, including A. baumannii, P. aeruginosa K. pneumoniae and E. coli. All were reported in PICTs in association with HCAIs including pneumonia, sepsis, catheter-related bloodstream, surgical site infections and UTIs. Significant rates of resistance in K. pneumoniae (>50%) and E. coli (12% to 50%) to third generation cephalosporins were reported in several PICTs [15,21]. Reports about confirmed ESBL-producing organisms were limited to K. pneumoniae and E. aerogenes in New Caledonia and Fiji [17,19,28] and carbapenem resistance in A. baumannii isolates in the French Territories: New Caledonia and French Polynesia [23,24]. Whilst other gram-negative pathogens were reported in PICTs including P. aeruginosa and S. typhi [40], the studies were few and the antibiotic susceptibility patterns reported varied widely.
These findings are particularly concerning as antibiotic options, to treat life threatening infections caused by multidrug-resistant organisms, are limited in PICTs due to cost and availability.
Our findings suggest that for severe gram-negative HCAIs, alternative broad-spectrum therapies, such as the carbapenems will likely see an increase in use. However, if the emerging resistance to the carbapenems continues [22][23][24], last line drugs such as colistin will be required. Health systems in PICTs will find it difficult to fund these last line antibiotics, which are often more expensive, toxic and not readily available.
Whilst this scoping study confirms the presence of MDR organisms in the region, interpretation of the literature and application of the data is challenging for several reasons. Firstly, the results represent only 54.5% (12) of the 22 countries included in the review. Most studies (72%) are about PNG, Fiji and New Caledonia. Eighteen percent (18%) refer to multiple PICTs and 10% are about individual PICTs. Secondly, this is a scoping study and no attempt has been made to judge the quality of the data. Microbiology laboratories in majority of PICTs were restricted to larger urban areas and results may not be representative of the whole country. The methodologies and breakpoints used for testing antibiotic susceptibility varied across PICTs and were not consistently reported. Performing microbiology testing in PICTs is challenged by a shortage of trained staff, consumables and equipment and this limits the availability of quality diagnostic testing. Storing, transporting and organism identification also present difficulties [82]. Thirdly, several studies have limited data, which increases the possibility of results being biased and ABR being over-or underestimated.

Conclusions
The epidemiology of antimicrobial resistance in the Pacific requires further characterisation. Available data were limited to a few major centres and further studies are needed to provide greater representation of the region. Our study found that data were too heterogenous for comparisons within and between countries. Therefore, it is imperative that future studies capture local data that allow comparison between settings, including hospital verses community infections, and urban, rural and outer island populations.
Improved surveillance data from PICTs to support efforts to contain AMR at the local, national and international levels are urgently needed. At the local level, surveillance data can be used to inform infection control programs and develop locally relevant antibiotic guidelines. Although several PICTs have published their own guidelines, much work remains to be done to produce guidelines for specific infectious diseases; and to ensure recommendations remain current with local microbiology and susceptibility patterns.
The literature supports the need for a formal introduction of antimicrobial stewardship programs in PICTs, which include developing and disseminating these guidelines; improving awareness of AMR; improving antibiotic prescribing behaviours; and implementing infection prevention and control practices. This will contribute to containing AMR in PICTs.

Conflicts of Interest:
The authors declare no conflict of interest.