Antibiotic resistance

Antibiotic resistance

ANTIBIOTICS Antibiotic resistance organisms have not yet acquired resistance despite huge pressures; these include Streptococcus pyogenes, Gram-posi...

403KB Sizes 1 Downloads 26 Views


Antibiotic resistance

organisms have not yet acquired resistance despite huge pressures; these include Streptococcus pyogenes, Gram-positive anaerobes such as Clostridium spp. and Peptostreptococcus spp. and Neisseria meningitidis (to penicillin). Almost all anaerobes were considered to be sensitive to metronidazole until recent reports from Spain, France and the USA suggested that resistant Bacteroides fragilis may soon become a serious problem.

Geoff Scott

Abstract Antimicrobial resistance is a serious increasing problem for man in all micro-organisms, from HIV through all bacterial species to Plasmodium spp. At present, we are managing to keep one step ahead of this resistance. However, some patients are dying with untreatable infections such as extreme-resistant tuberculosis. Others in hospital tend to acquire more resistant flora and this can cause troublesome delay in recovery from illness and surgery. Resistance is driven by antimicrobial use and some approaches to reducing this use are discussed. Nevertheless, our priority as physicians is to adequately treat people with infection, and we are having to reach more for newer and more expensive agents to do this successfully. A large investment in finding newer targets for antimicrobials and constructing new molecules has been disappointing. There may be some advantage in returning to old antibiotics. This article deals with current knowledge about resistance: molecular biology methods have been very useful in revealing mechanisms. However, whatever is written today will be out of date when you read this because evolution in bacteria happens at a phenomenal rate.

Ecology Resistance is driven by antibiotic use in human and agricultural/ veterinary practice. Resistance selection occurs by spontaneous mutations occurring at a rate of 109e106 driven by the presence of antimicrobials. Resistance elements appear in saprophytic bacteria as a result of antibiotic use for growth promotion and protecting crops. When these are eaten, resistance may be transferred to human bacteria that are occasionally pathogenic. The likelihood of finding resistant organisms in the gut is related to the tonnage of antibiotic use in the country in which the individual resides, and depends on the ease with which a resistance mechanism can arise. Novel resistance is usually detected in different places in the world at about the same time, implying that antibiotic pressures are similar worldwide and that bacteria have limited means of dealing with the problem of survival. Following the appearance of one mutantresistant progeny, rapidly replicating bacteria can recolonize carrier sites in less than 24 h. A resistant mutant of Mycobacterium tuberculosis, which replicates once every 24 h, can recolonize diseased lung within two weeks. Once resistance has been selected, organisms transferred from person to person continue to be resistant. When the antibiotic pressure is removed, novel colonizing and infecting flora tend to revert to the sensitive phenotype. Some bacteria containing large resistance plasmids are considered unfit and seem to disappear from the clinical environment more easily than others.

Keywords antibiotic-modifying enzymes; ecology; evolution; mutation; plasmids

In an environment containing vast numbers of micro-organisms saturated with or repeatedly exposed to antibiotics, Darwinian theory predicts the inevitable selection of resistant organisms. Following the introduction of a new antibiotic, resistant strains may be reported within as little as one year, and after a latent interval of some years, resistance increases dramatically and the value of the antibiotic for empirical therapy is reduced. (Antibiotic is a term technically used for natural products of fermentation and the synthetic molecules are called antimicrobials: however, the terms now seem to be used interchangeably). Staphylococcus aureus is now almost universally resistant to penicillin e a phenomenon first observed in hospital outbreaks of surgical sepsis in the late 1940s. This organism has shown a remarkable facility to become resistant to every antibiotic introduced, even vancomycin. Reduced sensitivity to vancomycin (vancomycin-intermediate S. aureus) is associated with a thick peptidoglycan cell wall. The complex gene cassette for vancomycin resistance in Enterococcus spp. (vanA) can easily be transferred to S. aureus in vitro and has now been detected in meticillin-resistant S. aureus (MRSA), making the organism fully resistant to glycopeptides. It is only a matter of time before strains are seen that are sensitive to only a small number of novel antibiotics in development and perhaps some of the older antibiotics (e.g. tetracycline, chloramphenicol). In contrast, some

Resistance and choice of antibiotics The term ‘antibiotic resistance’ implies that a particular antibiotic is ineffective in a clinical infection. This may be because the organism is inherently resistant to the antibiotic or because it is inaccessible. In vitro, resistance is defined by measuring the minimum inhibitory concentration (MIC) and the minimum bactericidal concentration (MBC) of the antibiotic against an organism, under ideal laboratory conditions, using appropriate controls to define the cut-off points between ‘resistant’, ‘intermediate’ and ‘sensitive’. Methods of testing in the laboratory are problematic, however, and any result that does not correlate with clinical experience should be challenged. Low MIC and high MBC imply that an antibiotic is bacteristatic; that is, it inhibits the growth of an organism but is unable to kill it. Even with bactericidal antibiotics, killing in vivo normally requires an intact immune system. Thus, though bacteristatic antibiotics are ineffective in neutropenic sepsis, bactericidal antibiotics may suppress infection only for it to recrudesce when the antibiotic is removed. This observation governs strategies for the management of such infections. Paradoxically, even when an organism is declared resistant to an antibiotic by in vitro tests, it may appear to be effective

Geoff Scott MD is retired Consultant in Clinical Microbiology at University College London Hospitals, London, UK. Competing interests: none declared.



Ó 2009 Elsevier Ltd. All rights reserved.


in clinical use; this may be because MIC/MBC can be exceeded in vivo by giving a sufficiently large dose. This is often seen in lower urinary tract infection and in the treatment of, for example, non-meningeal penicillin-resistant pneumococcal infection with high-dose penicillin.

site changes include topoisomerases II and IV (quinolones), b subunit of DNA-dependent RNA polymerase (rifamycins) and methylation of 23S target (14/15-membered macrolides). Exclusion of antibiotic: porins are protein structures embedded in the outer bilipid membrane only of Gram-negative organisms.1 They control what passes into and out of the cell on the basis of molecular size and charge. Mutations in the genetic elements encoding porins (permeability mutations) may exclude one antibiotic or, more commonly, multiple antibiotics. Pseudomonas aeruginosa has two outer lipid membranes, produces b-lactamases constitutively, and therefore tends to be more resistant than coliforms. Alternatively, organisms may actively excrete antibiotics, notably tetracyclines, macrolides and quinolones. Efflux pumps may be up-regulated, usually by exposure to the antibiotic in question. S. aureus resistant to erythromycin by active excretion remains sensitive to clindamycin, but the more common target site mutation affects susceptibility to both antibiotics.

Empirical treatment: management of an infection before bacteriology results are available depends on making the correct diagnosis and assessing the antibiotic sensitivities of the suspected organism or organisms from local epidemiological knowledge. Once culture results are known, it is usually another day before in vitro antibiotic sensitivities are available. This is a critical period in the management of an infected patient and explains why, in severe infections, broad-spectrum antibiotics are often chosen initially when the definitive diagnosis is uncertain e and why withholding treatment may be life-threatening. This period is generally more protracted when an organism such as M. tuberculosis is slow to grow in vitro. Delay in the treatment of tuberculous meningitis may lead to permanent neurological damage. Whether the correct antibiotics were chosen initially is known only weeks later.

Acquisition of resistance

Mechanisms of resistance (Figure 1)

Direct mutation of chromosomal genes may lead to resistance, and non-fatal mutations are passed to all progeny. Alternatively, small, mobile, circular pieces of DNA termed ‘plasmids’, which exist in the cytosol separately from the chromosome, may be passed from one bacterium to another (even to bacteria of different species) by various mechanisms such as direct transfer by type II pili (Figure 2) and phage transfer. Plasmids reproduce each time the organism divides and can probably be lost as easily as gained, given the correct environment. Transposons are large genetic elements often containing multiple genes necessary to confer phenotypic resistance. They may encode pheromones, which attract bacteria to each other and can be transferred on plasmids, generally being incorporated into the chromosomal DNA of the new host. Free DNA from dead bacteria and even mammalian cells can be incorporated into bacterial chromosomes. A remarkable mechanism of resistance is seen in penicillin-resistant Streptococcus pneumoniae, which have incorporated a mosaic of genes into the chromosome from other resistant a-haemolytic streptococci (e.g. Streptococcus milleri).

Resistance may be inherent (e.g. vancomycin against Gramnegative organisms, nitrofurantoin against Proteus spp.) or acquired via genetic elements encoding three fundamental mechanisms:  production of inactivating enzymes  change in the target site  exclusion of the antibiotic from the target site. The last may occur by restriction of access through porins (only in Gram-negatives) or by active excretion. Antibiotic-inactivating enzymes may be produced in vast excess, surrounding the organism (e.g. b-lactamase from S. aureus), or in limited amounts in the periplasmic space of Gram-negatives. The effect in vivo is similar, but the latter organisms may appear sensitive in vitro. Classically, this is seen in Enterobacter spp. resistant to extended-spectrum b-lactam antibiotics such as piperacillin and cefotaxime; exposure of the organism to inducers such as penicillin, clavulanate, certain cephalosporins and imipenem may ‘switch on’ the production of chromosomal (AmpC) b-lactamases. The classical TEM (Escherichia coli) and SHV (Klebsiella spp.) b-lactamases (which inactivate ampicillin) have shown a remarkable ability to mutate to extended-spectrum forms (which inactivate cefoxitin and cefotetan) and to inhibitor resistance (not inhibited by clavulanate or tazobactam). ctx-M cefotaximase arose by escape from the chromosome of Kluyvera spp. and is now widespread in Enterobacteriaceae including E. coli. Other classical inactivating enzymes include chloramphenicol acetyltransferase and aminoglycoside-modifying enzymes.

Problematic resistant bacteria Gram-positive organisms MRSA has replaced meticillin-sensitive S. aureus as the more common cause of hospital-associated sepsis in most parts of the world. The targets of the b-lactam antibiotics are penicillinbinding proteins (PBPs), carboxypeptidases and transpeptidases, which catalyse bridging of pentapeptide subunits of peptidoglycan. All strains of MRSA already produce penicillinase, but they now have a new penicillin-binding protein (PBP20 ) that is not inhibited by meticillin and its congeners oxacillin and flucloxacillin. They are resistant to all b-lactam antibiotics. The mecA gene encodes PBP20 , but requires expression of several ancillary genes within a transposon for full expression. Thus, some strains with mecA may appear to be sensitive to meticillin in vitro. In addition, a plasmid encodes resistance to a variable number of other antibiotics.

Target site change may be a structural alteration preventing binding of an antibiotic, or a mechanism whereby the metabolic pathway that is normally inhibited is bypassed by an alternative one. This is seen in sulphonamide resistance and in meticillinresistant S. aureus (MRSA), which has acquired a novel penicillin-binding protein from Staphylococcus scuiri. Important target



Ó 2009 Elsevier Ltd. All rights reserved.



Sites of action of antibiotics and some resistance mechanisms Porin Glycopeptide too large

Aminoglycosides β-lactams Mutation in porin

Gram- Gram-positive negative Peptidoglycan

Aminoglycosides Macrolides Chloramphenicol

Substrate change

Supercoiled DNA Topoisomerases


Periplasmic space (β-lactamase activity)

Folate synthesis

Penicillin-binding proteins

Sulphonamides Trimethoprim



Inactivating enzymes β-lactams Active excretion β-lactamases Chloramphenicol acteyltransferase Aminoglycoside-modifying enzymes

Macrolides Quinolones Tetracyclines

Figure 1

Strains with a tendency to spread easily and to predominate in hospitals are termed ‘epidemic MRSA’ (EMRSA). The current epidemic of EMRSA15 or EMRSA16 in the UK started in 1994. In the laboratory, these are detected by their antibiogram, but other typing methods (e.g. pulsed-field gel electrophoresis of chromosomal DNA) are needed to show that two strains are

indistinguishable, thus implying a common parent organism and that cross-infection might have occurred. When a few patients in one or two wards acquire a novel strain, such an outbreak can easily be tracked and then controlled. When EMRSA strains become endemic (Figure 3), more general measures are needed to reduce the risk to patients admitted to the hospital. MRSA may

DNA can be transferred, a on direct contact, b, c via type II pili or, d by phages (bacterial viruses). (b and c by courtesy of Astra Zeneca.) Figure 2



Ó 2009 Elsevier Ltd. All rights reserved.


New patients with meticillin-resistant Staphylococcus aureus at University College London Hospitals, 1991–1999

Annual incidence

450 400




300 250 200 150 100 50 0










Year Figure 3

become endemic in nursing homes, creating a pool for novel introduction into hospital; transfer of patients from ward to ward and colonization of staff leads to continued exposure of patients to new strains and helps maintain endemicity. Infection with MRSA is not untreatable. Strains are often (though not predictably) susceptible to gentamicin or other aminoglycosides, rifampicin, co-trimoxazole, chloramphenicol or ciprofloxacin, and sometimes to tetracyclines, macrolides, fusidic acid and pseudomonic acid. Note that rifampicin and fusidic acid should never be used alone because resistant mutants are selected very rapidly. MRSA is almost always susceptible to glycopeptides, though strains with reduced sensitivity to vancomycin have been occasionally described in patients with chronic colonization or infection who have been treated for several weeks.2 Some very rare strains of S. aureus are dependent on vancomycin to allow them to grow.

overwhelming pneumococcal sepsis. In the future, such prophylaxis may become less effective, but it is difficult to identify an alternative simple and safe regimen. Glycopeptide-resistant enterococci have become important causes of nosocomial postoperative infection in the USA. They are likely to cause sepsis and pneumonia only in severely ill patients in the UK. Strains are resistant to vancomycin with (vanA) or without (vanB) teicoplanin and other genes have now been discovered. These genes require the action of complex accessory genes for full expression: vanA encodes a structural change in the terminal amino acid of the pentapeptide chain of peptidoglycan, from D-ala to D-lac. This substitution prevents binding of vancomycin, enabling construction of the dipeptide bridges in peptidoglycan to continue. The complex gene cassette has formed over millennia and probably entered enterococci from an unusual anaerobe. The gene can be incorporated into S. aureus, and the emergence of, say, a successful spreadable VMRSA will reflect the greatest threat to modern medicine which we can foresee.4 Typing indicates wide heterogeneity of strains of VRE, and that carriers usually harbour more than one strain. This implies that the transposon encoding vanA and the accessory genes is promiscuous and easily able to enter the host’s enterococci. These organisms are of low virulence. Infections with Enterococcus faecalis may respond to simple antibiotics such as amoxicillin but are generally resistant to all but a few new antibiotics such as linezolid. Colonization with enterococci is driven by cephalosporins and fluoroquinolones, to which enterococci are constitutively resistant.

Penicillin-resistant S. pneumoniae strains have shown a gradual phase-shift increase in MIC to penicillin over several years. MIC is about 0.001 mg/l in sensitive strains and 1 mg/l in resistant strains.3 This is achieved by changes in penicillinbinding proteins with lower affinities for b-lactam antibiotics. In the UK, such strains currently represent about 4% of those causing invasive infection, but in some areas (e.g. Spain) the rate is as high as 50%. These strains are often resistant to many other useful antibiotics, including cephalosporins, chloramphenicol and erythromycin although this pattern is not predictable. Treatment with high-dose penicillin is effective in pneumonia but not in meningitis caused by resistant strains, some of which are sensitive to second-generation and third-generation cephalosporins. The mortality from severe pneumococcal pneumonia remains relatively constant regardless of whether the strain is resistant to penicillin, but meningitis caused by a resistant strain is more likely to be fatal. Splenectomized patients are advised to take life-long low-dose oral penicillin prophylaxis against the rare possibility of


Gram-negative organisms In contrast to resistant Gram-positives, against which some old antibiotics or antibiotics in development are often active, some Gram-negative organisms (particularly non-glucose fermenters such as Pseudomonas, Stenotrophomonas and Acinetobacter


Ó 2009 Elsevier Ltd. All rights reserved.


spp.), especially in ICUs, are resistant to every useful antibiotic and there are no new agents in development.

In some countries (e.g. the states of the former USSR, and South Africa), multi-drug-resistant strains (resistant at least to rifampicin and isoniazid) are extremely common, especially in patients who have been treated previously. In the UK, lone resistance to isoniazid occurs in 5% (except in London, where there is currently an epidemic of one clone resistant to isoniazid that started in 1995 and has led to a resistance rate of 15% in affected areas), and to rifampicin, ethambutol and pyrazinamide in 1% or fewer. Multi-drug-resistant strains account for 1e2%. Strains resistant to all are most commonly seen in London. Most patients can be treated satisfactorily with a standard 6-month regimen, but the results of routine antibiotic sensitivity tests are unknown for 5e16 weeks after sending specimens to the laboratory, and patients with resistant strains need more toxic second-line drugs and very prolonged courses of treatment. Rapid liquid culture and detection of genes encoding resistance are improving the speed of laboratory diagnosis.

Glucose fermenters: in terms of resistance and endemic/ epidemic problems in ill, hospitalized patients, Klebsiella, Enterobacter and Serratia spp. are more troublesome than E. coli. They produce SHV-type b-lactamases constitutively, and mutations lead to resistance to all b-lactams other than the carbapenems (or occasionally aztreonam). They often lose susceptibility to quinolones and aminoglycosides and become essentially untreatable. Some strains of Salmonella enterica (particularly typhimurium), and the enteric salmonellae (e.g. Salmonella typhi) have acquired stable broad resistance to all useful antibiotics. These are now pandemic. Chromosomal AmpC b-lactamases are seen in Citrobacter, Enterobacter, Serratia, Providencia, Hafnia and Aeromonas spp. associated with a repressor gene. Derepressed mutants can produce high levels of b-lactamase and plasmid-encoded AmpC genes have no associated repressor. This sort of resistance is very difficult to detect in the laboratory. Extended-spectrum b-lactamases are often plasmid encoded and can move easily from one organism to another. They are often associated with multiple resistance. (e.g. to quinolones and aminoglycosides). Some contain a suicide plasmid which implies that the antibiotic pressure does not have to be present for the organism to survive. The antimicrobials of choice for multiresistant coliforms remain the carbapenems. However, extensive use of these will result in more resistance problems through the proliferation of specific metallo-b-lactamases.

Strategies for reducing the impact of resistance Antibiotic resistance is driven by antibiotic use. When antibiotics are superseded and therefore used less, strains resistant to these tend to disappear. In the community: in the UK, more than 80% of human use of antibiotics occurs in the community, mostly for respiratory tract infections. The Standing Medical Advisory Committee, in its report The path of least resistance, made recommendations to reduce inappropriate prescribing.5  No antibiotics should be given for simple coughs and colds.  Antibiotics should not be routinely prescribed for sore throats, unless there is evidence of streptococcal infection. [It is not possible to tell clinically whether a sore throat is viral or caused by S. pyogenes.]  Antibiotics are not routinely required for acute otitis media and sinusitis-like symptoms; if given, courses can be limited to three days. In addition, three days’ treatment should suffice in otherwise healthy women with uncomplicated cystitis. This strategy has been useful in reducing prescribing by GPs, but anxiety has resulted from anecdotal reports of an increase in bacterial respiratory infections.

Non-glucose fermenters: P. aeruginosa has been replaced as a troublesome cause of nosocomial infection by other environmental and skin bacteria such as Acinetobacter baumanii var. calcoaceticus. Some strains are resistant to all available antibiotics, others are sensitive only to carbapenems and some aminoglycosides (e.g. amikacin) or colistin. In response to longterm antibiotic use over many years, Burkholderia cepacia tends to replace S. aureus and P. aeruginosa as colonizing flora in patients with cystic fibrosis. It may also show resistance to many antibiotics and can cause troublesome cross-infection. Neisseria spp.: Neisseria gonorrhoeae is interesting in that different strains may exhibit the three mechanisms of resistance to penicillin (reduced permeability, changes in penicillin-binding proteins and b-lactamase production). Each confers a stepwise increase in MIC, and there is no clear cut-off between sensitive and resistant strains. Many strains have become resistant to other oral drugs such as ciprofloxacin. The current epidemic of ciprofloxacin resistance in the UK requires the use of empirical single-dose injectable ceftriaxone. N. meningitidis has been slow to acquire resistance to penicillin, though a few strains isolated recently in Spain have slightly higher MICs than predicted. The value of sulphonamides was largely lost by the 1970s, and though chloramphenicol remains useful, second-generation cephalosporins (cefotaxime or ceftriaxone) seem to give the best results in clinical treatment.

In hospital, antibiotic use can be reduced by several means.  Routine use of antibiotics for surgical prophylaxis should be reduced to a minimum.  Antibiotics should not be started immediately in all suspected infections. Certain patients (e.g. those with febrile neutropenia or evidence of septicaemia) require urgent antibiotic therapy, but in many other cases (e.g. mild pyrexia postoperatively), it is safe and ultimately preferable to withhold antibiotics until culture results are known or there is clear evidence of bacterial infection.  An alternative is to discontinue antibiotics as soon as information is available suggesting that the problem is not bacterial or has resolved itself. This is called an ‘antibiotic-stop’ policy, and the aim is to encourage doctors to actively review the need for antibiotics after, say, the second day, given information on

M. tuberculosis: resistance to first-line antituberculosis agents has been a major problem in the re-emergence of tuberculosis.



Ó 2009 Elsevier Ltd. All rights reserved.


cultures and surrogate markers that has by then become available.  Certain antibiotics can be withheld from the hospital formulary. This is the main benefit of an agreed antibiotic policy. However, the choice of restricted antibiotics is likely to be decided more on the basis of cost rather than on the likely selection of resistance. (Restricted antibiotics may have to be used occasionally, however, when resistance to other available drugs has been selected.)  In theory, antibiotics can be rotated such that, for example, predominantly penicillins are used at some times, and cephalosporins or quinolones at others. There is little clear scientific evidence that this has any effect, and major, complicated studies would be needed to determine the effect of change in use on both normal and infecting flora. However, it has been shown that, in a setting of heavy cephalosporin use, discontinuation of use of this class of drugs leads to a reduction in the risk of colonization with glycopeptide-resistant enterococci and Clostridium difficileassociated diarrhoea.

Doctors and vets should overcome their view that they have an inalienable right to prescribe empirical antibiotics, and should set targets for reduction of their own personal prescribing. A

REFERENCES 1 Pages JM, James CE, Winterhalter M. The porin and the permeating antibiotic: a selective diffusion barrier in gram-negative bacteria. Nat Rev Microbiol 2008; 6: 893e903. 2 Hiramatsu K, Aritaka N, Hanaki H, et al. Dissemination in Japanese hospitals of strains of Staphylococcus aureus heterogeneously resistant to vancomycin. Lancet 1997; 350: 1670e3. 3 Richter SS, Heilman KP, Dohrn CL, Riahi F, Beekman SE, Doern GV. Changing epidemiology of antimicrobial-resistant Streptococcus pneumoniae in the United States 2004e2005. Clin Infect Dis 2009; 48(3): e23e33. 4 Werner G, Strommenger B, Witte W. Acquired vancomycin resistance in clinically relevant pathogens. Future Microbiol 2008; 3: 547e62. 5 Standing Medical Advisory Committee, Department of Health. The path of least resistance. London: HMSO, 1998. 6 Bumann D. Has nature already identified all useful antibacterial targets? Curr Opin Microbiol 2008; 11: 387e92.

Other strategies: development of resistance in human pathogens might be delayed if antibiotics were not used so widely in animal husbandry, particularly for growth promotion.

The future

FURTHER READING Baquero F, Bla´zquez J. Evolution of antibiotic resistance. Trends Ecol Evol 1997; 12: 482e7. Butler CC, Rollnick S, Kinnersley P, et al. Reducing antibiotics for respiratory tract symptoms in primary care: consolidating ‘why’ and considering ‘how’. Br J Gen Pract 1998; 48: 1865e70. Jacobs MR, Felmingham D, Appelbaum PC, et al. The Alexander Project 1998e2000. J Antimicrob Chemother 2003; 52: 229e46. Lipsitch M, Bergstrom CT, Levin BR. The epidemiology of antibiotic resistance in hospitals: paradoxes and prescriptions. Proc Natl Acad Sci U S A 2000; 97: 1938e43. Pitout JD, Laupland KB. Extended-spectrum beta-lactamase-producing enterobacteriaceae: and emerging public-health concern. Lancet Infect Dis 2008; 8: 159e66.

Resistance to antibiotics is one of the greatest threats to the success of modern medicine. It has recently become more serious because we can no longer be sure that any antibiotic chosen empirically will work, and because of the emergence of totally resistant bacteria. How to reduce resistance without simply discontinuing use of all antibiotics is a dilemma. We do not know to what degree antibiotic use must be reduced to decrease the selective pressure and allow reversion to colonization with more sensitive flora, nor do we know how to protect the few remaining drugs that can be used to treat resistant infections. Despite a decade of innovative exploration of bacterial genomes to identify novel potential targets for designer antimicrobials, no new agents have yet appeared.6



Ó 2009 Elsevier Ltd. All rights reserved.