A systematic review of antibiotic-resistant Escherichia coli and Salmonella data obtained from Tanzanian healthcare settings ( 2004-2014 )

Antibiotic-resistant Escherichia coli and Salmonella are an increasing challenge to global health. In Tanzania reliable data is limited for trends of resistance in major hospital-acquired pathogens. Data on the prevalence of antibiotic-resistant E. coli and Salmonella from Tanzanian sources (2004-2014) was extracted from PubMed and Google Scholar databases (April -June, 2015). Descriptive statistics and logistic-regression analysis were used to estimate the prevalence and trends for resistant E. coli and Salmonella to selected antibiotics using R software. A total of 24 articles were availablefor review, of which 21/24 (87.5%) and 7/24 (29.2%) reported the prevalence of antibiotic-resistant E. coli and Salmonella, respectively. Across all studies the average prevalence of resistance to ampicillin and cotrimoxazole was higher for E. coli (81.6 and 77.7%, respectively) than for Salmonella (64.7and 59.3%, respectively). Both groups of pathogens were also resistant to ciprofloxacin (20-22%) and 3 rd generation cephalosporins (2.5-27.8%). A logistic-regression model for published data (2004-2014) indicated that during this period of time there has been a significant increase to amoxicillin/clavulanate, ceftazidime, ciprofloxacin and gentamicin in E. coli (P< 0.001), and a significant increase in resistance to ampicillin for Salmonella (P < 0.05).Decreased E. coli and Salmonella susceptibility to critical antibiotics threatens the effective treatment of these infections in Tanzania. Proactive strategies are needed to preserve these antibiotics that remain largely active against bacterial pathogens in Tanzania.


2013
). Bacteria that are resistant to ≥ 3 antibiotic classes are conventionally referred to as "multidrug-resistant" andsuch microbes challenge existing treatment regimens for bacterial infections (Laxminarayan and Heymann, 2012;Thu et al., 2012).Multidrug-resistant bacteria often cause chronic diseases in people leading to long-term hospitalization, high morbidity and mortality (Feasey et al., 2012).Escherichia coli and Salmonella sp.(S. enterica subspecies enterica) commonly cause septicemic infections in Africa (Feasey et al., 2012;Anago et al., 2015).Multidrug-resistant E. coli and Salmonella often express extended spectrum beta-lactamases (ESBLs) that favour increased resistance to broad-spectrum betalactam antibiotics.These genetically encoded traits are usually located on plasmids that are transferable between bacterial strains and species (Sweta Gupta et al., 2013;Anago et al., 2015).Data on antibiotic resistance for pathogens is generally limited in sub-Saharan Africa (Leopold et al., 2014).In Tanzania, a situational analysis report by Global Antibiotic Resistance Partnership Working Group (GARP) called for a coordinated response to AMR problem and reveals baseline data for presence of antibiotic-resistant E. coli and Salmonella sp.in nosocomial infections (GARP-Tanzania, 2015), but there is no systematic mechanism for tracking trends in major hospital-acquired pathogens (WHO, 2014).This review focused on the prevalence and trends of antibiotic resistance for nosocomial E. coli and Salmonella as reported in the literature between 2004 and 2014.The analysis focused on antibiotics that are considered critical to Tanzanian healthcare settings by the WHO-Advisory Group on Integrated Surveillance of Antibiotic Resistance (AGISAR, 2011).
According to AGISAR, a critically important antibiotic (CIA) is the sole, or only one of limited available therapies to treat serious human disease such as pneumonia.Antibiotics are also considered critical when they are important for treating diseases caused by either (1) organisms that may be transmitted to people from nonhuman sources or, (2) human diseases caused by organisms that may acquire resistance genes from nonhuman sources.Antibiotics that meet criterion 1 or criterion 2 are referred to as highly important antibiotics (HIA) (WHO-AGSAR, 2011).
Results from this review focus on these important antibiotics with the goal of improvingtreatment guidelines for hospital-acquired infections and address the need for enhanced antibiotic stewardship strategies in Tanzania.

Article search strategy and selection criteria
Search w ords "resistance" or "antibiotic resistance" or "multidrug resistance" and/or "Salmonella "or "Escherichia", or "antibiotic susceptibility", or "antibiotics", or "antibiotic" or "bacteraemia" or "bacteriuria" and *Tanzania* w ere used w ith PubMed and Google Scholar electronic databases.Boolean operators, proximity search and mapping techniques (Boell and Cecez-Kecmanovic., 2010); Boell and Cecez-Kecmanovic, 2014) w ere employed to identify relevant articles.All articles published betw een 2004 and early 2015 that reported prevalence of antibiotic-resistantE.coli and Salmonella isolates from Tanzanian clinical specimens in healthcare settings w ere retrieved and analysed if antibiotic resistance data w as reported based on Kirby-Bauer disc diffusion assays.

Statistical analysis
Extracted data w ere entered into a spreadsheet (Excel 2013, Microsoft Corp., Redmond, WA, USA).Tables and descriptive statistics w ere used to summarize data.Average prevalence (Number of resistant/total number isolates tested) for a 10-year period and the proportion of antibiotic-resistant E. coli and Salmonella [number of resistant/(number of sensitive +number of resistant isolates)]w as computed for each antibiotic across all studies.Logistic-regression w as used to assess trends in resistance for E.coli and Salmonella to selected antibiotics for the data, published betw een 2004and 2014, using R softw are (v3.2.5, stats package).All results at P< 0.05 w ere considered statistically significant.

Description of search results
A total of 1,136 articles was retrieved and screened from PubMed (n=616) and Google Scholar (n=520) electronic databases between April and June, 2015 (Figure 1).Twenty-four articles (n=24) passed inclusion criteria set for this review (Table 1).The majority of the articles (16/24; 67%) consisted of cross-sectional, hospital-based studies (Table 1).
A hospital-based infection was defined as(1) an infection that was acquired by neonates within 10 days of birth in a hospital, or (2) inpatients showing symptoms of new infection >48 h following admission, or (3) community-acquired infections involving septicemic infection with the growth of pathogenic bacteria in a blood-culture that was obtained within the first 48 h of admission.Only one study by Blomberg et al.( 2007), examined data from both community-and hospital-based infections.For cross-sectional studies the presence of bacterial pathogens (exposure) and antibiotic resistant infections (disease) were determined at the same point in time in a given population and the prevalence of exposure and/or diseases was assessed.A few studies (8/24; 33%) were either retrospective or prospective cohort studies (Table 1).
For retrospective studies bacteria were isolated from a cohort of individuals prior to the onset of the study and assessed for antibiotic resistance.prospectivestudies are rare in the field of antibiotic resistance, but typically involve a cohort of individuals that are identified and examined for the presence ofantibiotic-resistant bacteria relative to risk factors for carriage of antibiotic-resistant strains during a defined study period (Euser et al., 2009).

Prevalence of antibiotic-resistant E. coli
In the last two decades there has been an increasing number of reports about antibiotic-resistant E. coli isolates from tertiary hospitals.In the selected studies, antibiotic-resistant E. coli from septicaemia (BSI) and urinary tract infections (UTI) was reported in seventeen studies (17/24; 70.8%), while four studies (4/24; 16.7%) reported other E. coli-associated infections such as surgical site infections (SSI) and diarrhoea (Table 1).When data were pooled from the 21 published reports, E. coli indicated high resistance to ampicillin (81.6%), tetracycline (74.9%) and co-trimoxazole (77.7%) (Table 2).

DISCUSSION
Published data (2004-2014) aboutthe prevalence of antibiotic-resistant E. coli and Salmonella from hospitalacquired infections in Tanzania suggests thatthere was a high average prevalence of resistanceto ampicillin (81.6 versus 64.7%, E. coli and Salmonella, respectively) and co-trimoxazole (77.7 versus 59.3%) (Tables 2 and 3).Comparable results were reported for E. coli in Kenya (ampicillin, 95%; co-trimoxazole, 95%) (Sang et al., 2012), Ethiopia (ampicillin, 100%; co-trimoxazole, 62.9%) (Beyene and Tsegaye, 2011;Kibret and Abera, 2011), Zimbabwe (ampicillin, 84.5%; co-trimoxazole, 68.5%) (Mbanga et al., 2010)   (ampicillin, 100%; co-trimoxazole 75.6%) (Yah et al., 2007) and South India (ampicillin, 99%; co-trimoxazole, 68.7%) (Razak and Gurushantappa, 2012).High resistance to ampicillin and co-trimoxazole is a challenge for treatment of bacterial infections in Tanzania where ampicillin is used as an empirical therapy and co-trimoxazole is used as a prophylaxis to prevent opportunistic infections among HIV-infected individuals (Hamel et al., 2008;Marwa et al., 2015).The use of robust and affordable diagnostic tools for bacterial infections in Tanzanian hospitals in accordance is highly recommended to restrict ineffectiveadministration of these antibiotics.Resistance to these "older" antibiotics is particularly unfortunate because alternatives will be increasingly expensive in a country that can ill-afford increased medical expenses.Relatively high resistance of E. coli and Salmonella to critical antibiotics such as ciprofloxacin (20 versus 22.2%, respectively) was evident (Table 2 and 3).Over the course of the review period, there was a statistically significant increase in E. coli resistance to amoxacillin/clavulanate (P < 0.001), ceftazidime (P < 0.001), ciprofloxacin (P < 0.001) and gentamicin (P < 0.01), whereas no significant trend was observed for Salmonella (Table 4 and 5).This disparity in trends suggests that there is a greater need to scrutinize treatment decisions for E.coli infections.Reduced susceptibility of nosocomial E. coli pathogens to critical antibiotics was also reported by others in Nigeria (ciprofloxacin, 15.4%) and Iran (ciprofloxacin, 16.8%) (Khameneh and Afshar, 2009;Akinkunmi et al., 2014).Conversely, the rapid spread of ciprofloxacin resistance in a widely disseminated S.typhi strain (haplotype H58) both in Africa and Southeast Asia (Berkley et al., 2001;Chiou et al., 2014)  this lineage of bacteria in Tanzania, particularly if there is heavy reliance of fluoroquinolones to treat Typhoid infections.
Detection of ESBL and carbapenamase-producing strains among E.coli isolates has been reported in several studies (Ndugulile et al., 2005;Moremi et al., 2014;Mushi et al., 2014).These isolates were highly resistant to amoxicillin/clavulanate (88.5-90.9%),ceftazidime (50-100%), ciprofloxacin (45.5-61.3%)and gentamicin (72.7-93.5%).Emerging resistance to extended beta-lactams and fluoroquinolones is an escalating public health concern for the management of infections among children and immuno-compromised individuals.Dissemination of ESBL strains in healthcare settings has been previously reported in various countries, including Kenya (Kiiru et al., 2012), Benin (Anago et al., 2015), Iran (Rezai et al., 2015), Brazil (Ferreira et al., 2011) and Bangladesh (Lina et al., 2014).Access to diagnostic tools that can detect ESBLs in local healthcare settings needs to be enhanced.In Tanzania,like many developing countries, laboratory capacity to confirm ESBL phenotypes is limited and diagnostic tools for infections are commonly unavailable or unreliable (Berkley et al., (2001).
There is evidence of increasing numbers of E. coli and Salmonella resistance to critical antibiotics in Tanzania over the past 10 years (Figures 2a and 3a).This is probably explained in part by a high prevalence of nosocomial infections and growing rates of hospitalization reported in developing countries.This increased service demand has likely increased reliance on more potent antibiotics as initial or empirical treatment because they act against a wide range of pathogens (Laxminarayan and Heymann, 2012;Thu et al., 2012).As a consequence, this practice facilitates selection and persistence of bacterial strains resistant to critical antibiotics (Mshana et al., 2009;Meremo et al., 2012).High resistance to these antibiotics in nosocomial E. coli and Salmonella infections has been reported in Cameroon (Lonchel et al., 2012), India (GARP-India, 2011) and Latin America (Salles et al.,2013).Decreased Salmonella non-susceptibility to highly important antibiotics (Figure 3b) may suggest an increased proportion of susceptible isolates to this group of antibiotics for the period between 2004 and 2014.Nevertheless, a relatively high resistance (59.3%) to cotrimoxazole may be explained by its common usage as an alternative treatment for infectious diarrhoea (Casburn-Jones and Farthing, 2004) (Table 3).In contrast, high Salmonella susceptibility to highly important antibiotics such as co-trimoxazole, has been reported in various countries like Nepal (1995-2015: co-trimoxazole, 98.8%) (Shrestha et al., 2016), Southern India (2009-2011:co-trimoxazole,95%) (Choudhary et al., 2013), and Montenegro (2005-2010: co-trimoxazole, 96.3%) (Mijovic, 2012).These findings suggest that local susceptibility testing of highly important antibiotics may be essential for timely treatment of Salmonella infections in low-income populations like Tanzania where access and/or options tomore potent antibiotics is generally limited (Laxminarayan et al., 2015).
It is important to note that the majority of the reviewed studies relied on data from hospital-acquired infections.Only one study included data from community-acquired infections and consequently, it is possible that the numbers reported in the are upwardly biased.This can happen when patients self-medicate prior to presentation at a hospital and this probably increases the possibility of isolating resistant strains.The use of Clinical and Laboratory Standard Institute (CLSI) guidelines was reported only by a subset of studies.Thus, the accuracy of any susceptibility data from studies that employed different guidelines might have caused variation in results.Finally, the lack of ESBL phenotype data in many studies might result in an underestimate of the prevalence of multidrug-resistant bacteria.Overall, high E. coli and Salmonellanon-susceptibility to ampicillin and co-trimoxazole suggests that these antibiotics can be inappropriate empirical treatment for major nosocomial infections in Tanzania.Further, decreased E. coli and Salmonella susceptibility to amoxicillin/clavulanate, ceftazidime, ciprofloxacin and gentamicin threatens the effective treatment ofthese infections in Tanzania.Implementing proactive strategies in antibiotic stewardship to preserve the effectiveness of critical antibioticsthat appear to remain largely effective against bacterial pathogensin Tanzania is crucial.Applying enhanced infection control measures would limit further spread of resistant bacteria in healthcare settings and community as well.

Figure 1 .
Figure 1.Flow diagram indicating an inclusion assessment of selected articles for systematic review .

Table 1 .
Synopsis of studies included in systematic review (n=24 articles).

Antibiotic a (*N) b Prevalence range (%) of resistant E. coli in various studies Average prevalence n
a superscripts (3-18) indicate the number of review ed studies; b *N=Total number of tested E. coli isolates; c n= number of antibioticresistant isolates.

of resistant Salmonella in various studies
superscripts (2-7) indicate the number of review ed studies; b *N=Total number of tested Salmonella isolates; c n= number of antibiotic-resistant isolates.For Meropenem no resistant Salmonella w as detected. a
a Odds ratios as estimated by logistic-regression analysis.Chloramphenicol, Nitrofuratoin and Meropenem w ere not analysed due to insufficient dat; b Categories according to WHO advisory group on integrated surveillance of antibiotic resistance (AGISAR, 2011); c * P < 0.05; ns, non-significant (P > 0.05).