Resistance patterns and implications for therapy
Experience from the World Health Organization's External quality assurance system for antimicrobial susceptibility testing has shown that disk diffusion testing is suitable for routine surveillance [
10]. However, disk diffusion is not optimal for testing of certain important resistance traits, such as penicillin-resistance in pneumococci. The lack of international standardization of methods and interpretive criteria causes concern, but there are indications that routine susceptibility testing data are suitable for surveillance even if obtained with different methods [
11].
Consistent with observations from a number of other countries in the region [
12‐
15] and elsewhere [
16], Gram-negative bacilli displayed high rates of resistance to common inexpensive antibiotics. Reasonably priced antibiotics such as ampicillin, tetracycline, trimethoprim-sulfamethoxazole and sulfonamides are now of limited benefit in the treatment of infections caused by important Gram-negative bacteria such as
E. coli,
Klebsiella spp.,
Proteus spp. and
Salmonella. Chloramphenicol may fail to cure as much as a quarter of infections caused by
Salmonella and half or more of infections caused by
E. coli,
Klebsiella spp. and
Proteus spp. Fluoroquinolones appear to be the only reliable drugs for oral treatment of infections caused by common Gram-negative bacilli, whereas gentamicin and third-generation cephalosporins remain useful for parenteral therapy.
The study showed a very low prevalence of methicillin-resistant
S. aureus, consistent with previous data from the same hospital [
17,
18]. While the results should be interpreted with some caution since confirmatory nucleic acid based techniques were not available, the data support the current use of isoxazolyl penicillins, such as cloxacillin for the treatment of staphylococcal infections at the hospital. There were few isolates of enterococci compared to studies from high-income countries [
19]. It is reassuring that the current study showed a low rate of ampicillin-resistant enterococci, indicating that nosocomial infections caused by these micro-organisms is a minor problem compared with many high-income countries. Low consumption of broad-spectrum antibiotics such as third-generation cephalosporins, fluoroquinolones, imipenem and vancomycin may explain this finding [
19‐
21]. While other countries in the region have been affected by penicillin-resistant pneumococci [
22,
23], the current study indicates that pneumococcal disease in Dar es Salaam can safely be treated with penicillin or erythromycin. However, the results should be interpreted with some caution since the number of isolates tested was small. More than a quarter of the gonococcal isolates (11/41) were obtained from the neonatal ward, and most or all of these isolates probably represent gonococcal conjunctivitis. Amoxicillin-clavulanate, spectinomycin, fluoroquinolones and erythromycin appear to be good alternatives for the treatment of gonococcal infections. An apparent increase in resistance to trimethoprim-sulfamethoxazole (from 18% to 59%) is noted since the study by Mbwana [
24] from 1993 to 1995, however, this may be due to the use of different methodology for susceptibility testing.
Applicability of data to guide treatment of serious infections
Recommendations for antibiotic treatment of serious bacterial infections such as bloodstream infections and meningitis should preferably be based on knowledge of the prevalence and antimicrobial susceptibility patterns of pathogens isolated from blood and spinal fluid. While a fair number of bacterial isolates were tested in the current study, the number of blood culture isolates was limited (n = 329, excluding the CoNS isolates). As shown in Table
5, there appears to be differences in resistance between isolates obtained from blood cultures and those from other specimen types, but these are difficult to assess because of the low number of blood culture isolates. Thus, the data from the current surveillance should be interpreted with caution with regards to the treatment of serious infections. The CoNS isolates obtained from blood were recorded in the WHONET database, since they may represent clinically important infections such as bacteremia in patients with compromised immunity, patients with indwelling intravascular devices and the newborn [
25]. The study showed that a high proportion (21.9%) of blood culture bottles yielded CoNS isolates. However, the conventional way to distinguish pathogenic isolates of CoNS from contaminants, by requiring growth of a similar CoNS isolate in a separate blood culture, could rarely be used, since follow-up cultures were seldom available. Consequently, the susceptibilities of these isolates were not evaluated further.
Relevance of data for outpatient and rural populations
It is important to specify for which population the surveillance data are valid. At our hospital, specimens from both inpatients and outpatients were examined. The hospital is to a great extent used as a primary hospital for the population in the Dar es Salaam area. However, among the cases coming to the hospital, there may be a degree of selection of patients with infections caused by resistant microbes, since many patients rely on health centers and pharmacies to cure simple ailments, and only come to the hospital when primary treatment fails. The study found that a few resistance traits, such as ampicillin resistance in E. coli and gentamicin and trimethoprim-sulfamethoxazole resistance in Klebsiella spp. were more frequent in urinary isolates from inpatients than from outpatients. Apart from that, there were no dramatic differences between isolates from inpatients and outpatients. The data from the study should be representative for both the hospital setting and to some degree the population in Dar es Salaam. However, the majority of the population of Tanzania lives in rural areas, where resistance patterns may be substantially different. Thus one should be cautious to extrapolate the results of the current study to be valid for populations in the countryside.
Ability to monitor trends of antimicrobial susceptibility
Certain trends in antimicrobial susceptibility could be identified by comparison with data from other studies. While resistance to ampicillin, tetracycline and sulfonamides in Gram-negative bacteria was frequent already in the seventies [
26,
27], it is worrying that resistance to trimethoprim-sulfamethoxazole, chloramphenicol, nitrofurantoin, nalidixic acid and amoxicillin-clavulanate appear to have increased compared to previous studies [
27,
28]. The extensive use of chloramphenicol for the treatment of presumed cases of typhoid fever and the use of trimethoprim-sulfamethoxazole for the ambulatory treatment of chest infections, malaria and other infections, may have contributed to the high prevalence of resistance to these two drugs. Although still low, it is of concern that the rate of gentamicin-resistance in
E. coli has increased from zero in 1978–79 [
27] to 2% in 1995 [
28] and 8% in the current study. In neighboring Kenya, the rate of gentamicin-resistance in
E. coli has increased from 2% in the late seventies [
29] to 20% and above in recent studies [
12].
Resistance to gentamicin is common in Gram-negative bacteria with extended-spectrum beta-lactamases (ESBL), sometimes in as much as 96% of isolates [
30]. Such an association cannot be investigated in the current study, since less than half of the isolates of
E. coli and
Klebsiella spp. were tested for susceptibility to third-generation cephalosporins and other methods for detection of ESBL (double disk synergy test, Etest, PCR) were not available.
Also in
P. aeruginosa the rate of gentamicin-resistance has increased, from zero in the seventies [
27] to 4% in the current study. Resistance to penicillin and erythromycin was common among
S. aureus isolates in this study. However, the rate of tetracycline resistance (49%) was lower than reported from the same hospital in 1979 (57%) and 1982 (74%) [
17]. In the late seventies, tetracycline was used in great quantities in Tanzania to prevent and treat cholera; as much as 1788 kilograms of the drug were used during a period of only 5 months [
31]. Due to the rapid emergence of tetracycline-resistant
Vibrio cholerae, the use of the drug was subsequently greatly reduced, and this may have contributed to a concurrent decline in the rate of tetracycline-resistance in an unrelated species such as
S. aureus.
For meaningful comparison of data from different studies, whether from the same or different laboratories, the same method of susceptibility testing should preferably be employed. In our laboratory, the same method has been used for a number of years. The WHONET software features a number of sophisticated ways to analyze susceptibility information based on the measurements of inhibition zone diameters. Recording the diameter of the inhibition zones in disk diffusion testing is generally recommended [
32], and may increase the accuracy of results and enable the detection of gradual shifts in antibiotic susceptibility over time. It also makes the data independent of the current breakpoints. With the WHONET software, data can easily and rapidly be re-analyzed with reference to new breakpoints. However, the Stokes' method for susceptibility testing [
9], which is used in our laboratory, is based on visual interpretation of the difference in inhibition zones between the clinical isolate and the control strain. The interpretation is recorded as interpreted values, i.e. either susceptible "S", intermediate susceptible "I" or resistant "R". The WHONET software also accepts susceptibility data to be entered and analyzed as "interpreted values", i.e. "S", "I" and "R". The use of such interpreted values enables most of the analysis features of WHONET, but not all. Foremost, analyzing data based on zone diameters (or MIC values) is superior for the early detection of subtle shifts in antimicrobial resistance over time, which may alert clinicians about emerging resistance trends at an early stage. However, one asset of the Stokes' method, particularly under tropical conditions, is that unsuspected poor antibiotic disk quality will be discovered quickly since a control strain is tested on every plate.
Furthermore, variations over time in the battery of antibiotics tested makes comparison of data less useful. Laboratories in low-income countries are sometimes vulnerable to this because of unreliable supplies of antibiotic discs.
Ability to detect emerging resistance traits
Disk diffusion testing may give indications of emerging resistance traits such as methicillin-resistance in
S. aureus and ESBL in Gram-negative bacteria. The current surveillance indicated that methicillin-resistance is rare in
S. aureus at the hospital. Ideally this should be confirmed with PCR-based methods to detect the
mecA gene. Likewise, the disk diffusion testing showed the presence of resistance to ceftazidime in Gram-negative isolates, albeit at a low rate, which calls for further investigation with regard to the possible presence of ESBL. Our laboratory did not employ molecular methods for detection of resistance genes on a routine basis, but a recent study showed low prevalence of methicillin-resistant
S. aureus (MRSA) [
18]. Resistance surveillance should be coupled with awareness of signs of various resistance traits and, preferably, the possibility of using molecular methods to verify emerging resistance traits.
Ability to detect nosocomial problems
The WHONET software is well suited to analyze antibiograms in order to detect suspicious nosocomial outbreaks. These functions too are dependant on the use of a consistent battery of test drugs, and also works better when results are entered as actual values for MIC or zone diameters, as opposed to the interpreted value ("S", "I" or "R"). In our hospital, comparison of resistance rates did not show dramatic variation between isolates from inpatients and outpatients. The exception was a trend for more frequent gentamicin-resistance in inpatient isolates of Gram-negative bacteria, particularly Klebsiella spp., which may suggest possible nosocomial spread. The analysis of antibiograms did not produce convincing evidence of clonal patterns spread of bacterial isolates, possibly partly due to the variations in the battery of antibiotics tested. Molecular methods for the evaluation of the genetic relatedness of bacteria were not available in this study.
Suitability for international comparison of resistance data
In 2002 a total of 880 laboratories in 76 countries across the world used the software, including 41 laboratories in 4 African countries. The WHONET system has been implemented at MNH since 1998. Unfortunately, there is no international consensus on a recommended method for antimicrobial susceptibility testing. Worldwide at least twelve different
in vitro methods are followed, and only in Europe the number is at least ten [
5]. Furthermore, there are ongoing changes in the interpretive criteria for susceptibility testing [
10]. In addition to this, there is an abundance of molecular methods to describe various genetic markers of resistance.
In vivo clinical assessment is of great importance in understanding bacterial drug resistance and the gold standard for evaluating resistance in malaria parasite. The multitude of methods employed for antimicrobial susceptibility testing has to some extent hampered the meaningful sharing and comparison of resistance data among countries. Recently, much work has been done in Europe to harmonize resistance surveillance efforts across country borders [
33,
34]. While many laboratories record inhibition zones for disk diffusion results, interpretation is usually according to national guidelines. Thus, susceptibility patterns from different countries must be compared prudently. The lack of standardization in methods is a problem that must be addressed at an international level.
The surveillance system as a quality assurance tool
The implementation of the surveillance system brought focus on methodological issues, including microbial identification and susceptibility. The WHONET software has built-in functions to alert the operator if isolates with unexpected resistance patterns are entered. During the surveillance exercise in our laboratory, it was discovered that four isolates of Streptococcus pyogenes were reported as resistant to penicillin. This was subsequently double-checked, and consulting the laboratory bench-book we found that clerical errors were the explanation for this. The use of the surveillance software enabled the easy detection, investigation and correction of such errors, and consequently may contribute to increase the attention to quality issues and generally improve the performance of the lab. The current surveillance project highlighted some methodological issues, most of which were caused by budgetary limitations, such as the occasional use of locally made antibiotic disks and limitations in the identification of organisms due to lack of reagents.
Impetus for further research
Routine surveillance makes use of available large data sets at little additional cost and may be representative for a greater part of the population. However, often it is necessary to supplement the routine surveillance with ad hoc studies aimed at investigating particular problems. While ad hoc studies generally are more expensive to conduct, they allow for the use of more advanced and expensive laboratory methods and are better at targeting particular populations of interest. The current surveillance study identified a need for more data from bloodstream infections in order to provide reliable guidance for the treatment of serious bacterial infections. As a consequence of this, we started a study of bloodstream infections with the pediatric department at the hospital. Another laboratory-based research was started to ascertain the finding that methicillin-resistance in staphylococci is still relatively infrequent at this hospital.
Cost considerations and human resources
The study suggests that laboratories, which perform susceptibility testing, can gain useful information on antimicrobial susceptibility with a minimal budget. As appropriate software can be obtained free-of-charge, the main cost of the surveillance system is associated with purchasing a computer. However, there are other, indirect costs, which may be attributed to the surveillance program depending on the situation of the laboratory, such as running costs for microbiologic procedures, including susceptibility testing. Particularly, it is important to ensure availability of antimicrobial discs of satisfactory quality. A susceptibility surveillance system also implies the need for some additional human resources for data entry and analysis. In our experience, it is recommendable to allocate approximately 50% of a laboratory technologist position to this task. While the WHONET program is excellent for entry, analysis and reporting of resistance data, the software is not intended to function as a complete patient management system for the laboratory. Data can be transferred from other databases into WHONET by the use of a complementary software called BacLink (also free-of-charge). However, in laboratories such as ours, where the management of patients' laboratory tests (i.e. receipt of specimens and laboratory forms, inscription in registers, return of test results, etcetera) is handled manually via register-books, the data must be punched into WHONET by hand. Since the WHONET database is not used directly for patient management, the surveillance activity tends to become less integrated in the clinical routine work than it should. Thus, although the program performs its task very well, in a long-term perspective, a surveillance system that is integrated with a patient management system might be more sustainable. It is difficult to quantify the potential benefits of a well-functioning surveillance system. However, we are fully convinced that the modest costs of the surveillance program are highly justified since the data generated may improve empiric therapy, help contain or prevent the further emergence of antimicrobial resistance, decrease the need for expensive second-line antimicrobial drugs and, ultimately, save lives and reduce suffering.