SEARCH

SEARCH BY CITATION

Keywords:

  • biocide resistance;
  • MRSA;
  • MSSA;
  • susceptibility

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

Aims:  To analyse population minimum inhibitory concentrations (MICs) data from clinical strains of Staphylococcus aureus and Pseudomonas aeruginosa for changes over a 10-year period and to look for correlations between the antimicrobials tested.

Methods and Results:  Data from the MIC study of 256 clinical isolates of Staph. aureus [169 methicillin-sensitive Staph. aureus (MSSA), 87 methicillin-resistant Staph. aureus (MRSA)] and 111 clinical isolates of Ps. aeruginosa against eight antimicrobial biocides and several clinically relevant antibiotics was analysed using anova, Spearman-Rho correlation and principal component analysis. Comparisons suggest that alterations in the mean susceptibility of Staph. aureus to antimicrobial biocides have occurred between 1989 and 2000, but that these changes were mirrored in MSSA and MRSA suggests that methicillin resistance has little to do with these changes. Between 1989 and 2000 a sub-population of MRSA has acquired a higher resistance to biocides, but this has not altered the antibiotic susceptibility of that group. In both Staph. aureus and Ps. aeruginosa several correlations (both positive and negative) between antibiotics and antimicrobial biocides were found.

Conclusions:  From the analyses of these clinical isolates it is very difficult to support a hypothesis that increased biocide resistance is a cause of increased antibiotic resistance either in Staph. aureus or in Ps. aeruginosa.

Significance and Impact of the Study:  The observation of negative correlations between antibiotics and biocides may be a useful reason for the continued use of biocides promoting hygiene in the hospital environment.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

Although seen as a frontline defence against the spread of infection, concern has been expressed about the overuse of antimicrobial biocides leading to the ‘selection’ of antibiotic resistant organisms (Moken et al. 1997; Levy 1998a,b; Russell et al. 1999; Russell 2000).

Recently Joynson et al. (2002) have shown that some strains of Pseudomonas aeruginosa resistant to specific antibiotics (aminoglycosides) have an increased minimum inhibitory concentration (MIC) towards specific biocides (chlorhexidine and benzalkonium chloride). However, the converse was not found: biocide-tolerant organisms (with respect to MIC) were more susceptible towards a range of antibiotics. A similar finding was described by Loughlin et al. (2002). These findings mirrored the conclusions of Lambert et al. (2001) that the clinical environment was responsible for the emergence of antibiotic- and biocide-resistant bacteria, but that antibiotic resistance was required first.

Previous laboratory studies have also described this phenomenon: Brown and Tomlinson (1979) showed that cultures of Ps. aeruginosa trained by passage to be resistant towards polymyxin were more resistant to quaternary ammonium compounds (QACs), whereas Adair et al. (1971) showed that a Ps. aeruginosa strain trained by passage to be resistant towards a quaternary ammonium surfactant, was less resistant to polymyxin. Unfortunately the observations are not consistent, with Loughlin et al. (2002) describing an increased resistance to QACs coupled to an increase in resistance to polymyxin.

Russell et al. (1998) had shown that strains of Ps. stutzeri adapted to chlorhexidine were more resistant to QACs and to other common antimicrobials and antibiotics, whereas Loughlin et al. (2002) showed no change in resistance to chlorhexidine with strains of Ps. aeruginosa adapted to QACs. However, the level of resistance found by Russell et al. was strain variable, and this formed part of the conclusions of Loughlin et al. (2002) that the existence of cross-resistance to QAC adapted Ps. aeruginosa was strain dependent. They also concluded that the physiological changes giving protection against QACs would offer little protection against nonmembrane-active antibiotics, and therefore do not ‘represent a serious obstacle to therapy’.

In a paper by Lambert et al. (2001) a number of strains of Ps. aeruginosa were analysed for their sensitivities towards several common biocides and several clinically relevant antibiotics. Although the main aim of the work was to look for differences in antimicrobial resistance patterns between industrial and clinical isolates, the methodology used allowed a discussion of the cross-resistance between the various antimicrobials: many more correlations being found between the antimicrobials with clinical strains than with the industrial isolates.

Data sets of the MIC of clinically relevant antibiotics and common antimicrobials against methicillin-sensitive and methicillin-resistant Staphylococcus aureus (MSSA and MRSA, respectively) and Ps. aeruginosa were made available from the University of Iowa [home of the Sentry Antimicrobial Surveillance (SAS) programme]. This allowed more detailed studies of the clinical correlations between antimicrobials to be undertaken for a larger number of strains than had been previously examined in this way, and herein is presented the results of such analyses.

Organisms

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

Clinical isolates were all obtained from banked specimens that had been stored at the University of Iowa. Staphylococcal isolates from 1989 to 1991 were obtained from the nares or hands of healthcare workers or the nares of surgical patients. Pseudomonas aeruginosa strains included isolates obtained from clinical cultures or fomites in the hospital environment. Isolates obtained from 2000 represented the US clinical strains sent to the University of Iowa during various surveillance projects.

Susceptibility data from a total of 256 clinical isolates of Staph. aureus [87 (34%) were identified as MRSA (16 were isolated in 1989, 71 in 2000) and 169 (66%) were identified as MSSA (124 were isolated in 1989, 45 in 2000)] and 111 Ps. aeruginosa (54 were isolated in 1989, 57 in 2000) were used in this study. The MIC analyses were performed by the University of Iowa using standard techniques.

Antimicrobial agents

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

Minimum inhibitory concentrations of the antimicrobial biocides benzethonium chloride (Benzeth), benzalkonium chloride (BK), two bleach agents – Bleach-regular (Bleach R) and Bleach-Ultra (Bleach U), p-chloro-m-xylenol (PCMX), chlorhexidine gluconate (CHX), triclosan (TCS) and o-phenyl phenol (OPP) were obtained. The MICs for the therapeutic antimicrobials ampicillin (Amp), amoxicillin/clavulanic acid (A/C), cefepime hydrochloride (Cefep), sodium cefazolin (Cefaz), ciprofloxacin (Cipro), clindamycin (Clind), ceftriaxone (Ceftrix), ceftazidime (Ceftaz), erythromycin (Erythro), gentamicin (Gent), oxacillin (Oxa), piperacillin (Pip), piperacillin/tazobactam (P/T) and vancomycin (Vanco) were also obtained.

Data analyses

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

Minimum inhibitory concentrations data were obtained using half-fold diluted test samples inoculated with the microbe, with a growth-no growth criterion used for the estimation of MIC. The MIC ranges from the highest concentration used to that of the greatest dilution, this means that there is an upper (X mg l−1) and lower (Y mg l−1) bound to the MIC data. Where the organism grew in a concentration higher than X or failed to grow in a concentration as low as Y, (out of range values) then >X and <Y mg−1 are normally quoted for the MIC. In the analyses performed here, such labels were replaced by MIC = X, or MIC = Y mg l−1, for the particular antimicrobial against the isolate.

MIC data were transformed into log10 MIC. From the distributions of log10 MIC, a mean MIC was obtained for each antimicrobial against Staph. aureus and Ps. aeruginosa. MIC10, MIC50 and MIC90 values were obtained from a normal quantile plot. Comparisons of subgroupings within the data sets were carried out using one-way anova analyses. These allowed comparisons of the distributions between isolates obtained in 1989 and those in 2000, and also between MSSA and MRSA subgroupings. Statistical analyses were carried out using either Microsoft Excel, the JMP statistics package (SAS Institute, Cary, NC, USA) or with Mathematica (Wolfram Research Inc., Champaign, IL, USA).

Spearman-Rho.

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

To investigate the degree of cross-resistance between antimicrobials, correlation coefficients, which give a measure of the strength of any linear relationship between the MIC of two antimicrobials, were calculated using the Spearman-Rho nonparametric method (Hollander and Wolfe 1973). This is a correlation coefficient computed on the ranks of the data values instead of on the values themselves. For an exact relationship, the correlation is 1 or −1, depending on the relationship. If there is no linear relationship, the correlation tends to zero.

Principle component analysis.

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

Principal component analysis is a method to reduce the number of variables in a data set to a few significant variables – the principal components (PCs) (also known as artificial variables; Gabriel 1971; Gabriel 1981; Gabriel and Odoroff 1990). The principal axis method was used to extract the components, and this was followed by a varimax (orthogonal) rotation. The eigenvalue-1 criterion (Kaiser 1960) and/or the results of a Scree test (Cattell 1966) were used to determine the number of retained components. The rotated factor pattern was interpreted using a factor loading component of >0·4 as significant. The results were analysed for groupings of antimicrobials.

Distributions.

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

The MIC distributions of the isolates for the various antimicrobials against Staph. aureus are given in Table 1. Although analysed on the assumption of a normal distribution (of log10 MIC) many of the distributions are poor fits to a normally distributed population. There are two main reasons for this: the presence of sub-populations of resistant or sensitive isolates and the presence of MIC values ‘out-of-range’. This inability to fit a normal distribution leads to the quotation of values such as MIC50 and MIC90, which are the MIC whereby 50 and 90% of isolates, respectively, are inhibited. In Table 1, the mean of the normal distribution is given. Comparison with the MIC50 values shows a relationship between them, but is sensitive to the numbers quoted as being outside the concentration range.

Table 1.  The MIC population distribution data for Staphylococcus aureus and Pseudomonas aeruginosa
MIC0·060·1250·250·512481632No.<LCNo>HCMIC10MIC50MIC90Mean
(a) Distribution data for Staphylococcus aureus
Benzeth  491045238121  4900·250·502·00·69
BK   461117622  1000·501·002·01·24
Amp 189315214121128 171020·2510·9616·05·10
A/C 42328106208859 4280·251·0016·01·82
Cefep   21611141167 1552·004·0016·04·91
Cefaz     1805467 180552·002·0016·03·57
Cipro 2170761318 2023155110·250·502·00·68
Clind22145276  155  22540·240·128·00·32
Erythro13123282  99  1970·250·508·01·03
Gent  18545411317 111110·250·251·00·41
Oxa 1191222421078  0760·300·508·01·26
Vanco   342175    000·501·001·00·92
MIC0·0050·010·020·040·080·160·32No. <LCNo. >HCMIC10MIC50MIC90Mean
TCS1763432183216000·0050·0050·0200·0077
MIC326412525050010002000No. <LCNo. >HCMIC10MIC50MIC90Mean
PCMX1451341858  0063·0125·0500·0158·5
OPP  2780565934340125·0500·02000·0509·3
MIC0·390·781·563·16·25No. <LCNo. >HCMIC10MIC50MIC90Mean
CHX16430160111600·3900·3903·1000·699
MIC188375750No. <LCNo. >HCMIC10MIC50MIC90Mean
Bleach R111014500328·0656·0656·0484·4
Bleach U51817000375·0375·0750·1446·7
(b) Distribution data for Pseudomonas aeruginosa
MIC0·250·51248163264128256512No. <LCNo. >HCMIC10MIC50MIC90Mean
Benzeth2618013411930  2192·0031·99127·9429·65
BK  3121274617923 1232·064·0511·792·5
Pip  3141938184510  392·008·0063·979·35
P/T 2313302817612   292·008·0063·977·60
Ceftrix 1 34111577     668·0031·9931·9921·04
Ceftaz1 1528361777     71·004·0016·003·71
Cefep 41137281597     71·004·0016·003·55
Gent65313615558    2 0·572·0016·002·06
MIC0·0150·0300·0600·1200·2500·51·02·0No. <LCNo. >HCMIC10MIC50MIC90Mean
Cipro138381214332 200·060·252·000·33
MIC13162·5125250500No. <LCNo. >HCMIC10MIC50MIC90Mean
TCS52 3299595250500500348
MIC64125250500100020004000No. <LCNo. >HCMIC10MIC50MIC90Mean
PCMX12472228831 2505001742497
OPP   10324524  1000200020001679
MIC1·563·16·2512·525No. <LCNo. >HCMIC10MIC50MIC90Mean
CHX211204830 163·1012·5025·0010·89
MIC1883757501312No. <LCNo. >HCMIC10MIC50MIC90Mean
  1. The number of isolates with a specific MIC are given below the MIC (mg l−1) of the half-fold dilution series used for each antimicrobial. The number of isolates with a recorded MIC less than or greater than the lowest concentration or highest concentration of the dilution series used are given under the headings ‘No. <LC’ and ‘No. > HC’, respectively. For example in (a), of the 128 isolates with a recorded ampicillin MIC of 32 mg l−1, 102 isolates had an MIC > 32 mg l−1. The standard MIC10, MIC50 and MIC90 measurements (quantal measurement of 10, 50 and 90% of the population) along with the mean MIC (calculated from the log10 MIC values) are given.

Bleach R 30774  328656656557
Bleach U36543   375375750481

Not withstanding the lack of fit to a normal distribution the one-way anova statistical test allows a direct comparison between the mean values from sets of data, especially where there are subsets available e.g. MRSA and MSSA. Table 2 lists the results from anova tests on three subsets of the Staph. aureus data: (i) isolates labelled as MRSA vs those labelled MSSA; (ii) MSSA isolated in 1989 vs 2000 and (iii) MRSA isolated in 1989 vs 2000.

Table 2. anova comparison of mean values of MIC data (mg l−1) for MRSA and MSSA
AntimicrobialMRSA/MSSA (difference significance)MSSA 1989/MSSA 2000 (difference significance)MRSA 1989/MRSA 2000 (difference significance)
  1. NS, no significant difference between the two groupings (P > 0·01).

  2. S, significant difference between the two groupings (P ≤ 0·01).

  3. anova was carried out on the log10 MIC values and the mean values obtained transformed back to mg l−1.

  4. MSSA, methicillin-sensitive Staph. aureus; MRSA, methicillin-resistant Staph. aureus.

Benzeth1·024/0·563; (S)0·465/0·955; (S)0·91/1·05; (NS)
BK1·29/1·213; (NS)1·36/0·884; (S)2·708/1·091; (S)
Bleach R528·4/456·6; (S)445·6/512·9; (NS)425·6/555·9; (S)
Bleach U439·5/447·7; (NS)433·5/502·3; (NS)409·3/446·7; (NS)
TCS0·009/0·007; (NS)0·006/0·012; (S)0·006/0·009; (NS)
PCMX225·4/132·1; (S)111/211; (S)136·1/252·3; (S)
OPP726·1/424·6; (S)313/984; (S)201·4/970·5; (S)
CHX1·061/0·564; (S)0·463/0·965; (S)0·528/1·242; (S)
Amp13·2/3·128; (S)3·707/1·954; (NS)12·34/13·4; (NS)
A/C8·664/0·778; (S)0·826/0·659; (NS)11·81/8·078; (NS)
Cefep11·54/3·162; (S)3·236/2·985; (NS)14·05/11·04; (NS)
Cefaz10·24/2·075; (S)2·068/2·094; (NS)14·05/9·537; (NS)
Cipro1·009/0·581; (S)0·418/1·442; (S)1·297/0·954; (NS)
Clind1·237/0·156; (S)0·152/0·169; (NS)4·955/0·904; (S)
Erythro3·802/0·523; (S)0·382/1·24; (S)6·73/3·342; (NS)
Gent0·562/0·344; (S)0·308/0·463; (S)4·365/0·354; (S)
Oxa7·388/0·51; (S)0·5/0·538; (NS)7·656/7·328; (NS)
Vanco0·909/0·933; (NS)0·956/0·871; (NS)1·139/0·864; (S)

MRSA vs MSSA: as expected MRSA had a higher mean MIC relative to MSSA in nearly all cases. Only one antibiotic (Vanco) and three antimicrobial biocides (benzalkonium chloride, bleach-ultra and triclosan) showed no significant difference between MRSA and MSSA. No antimicrobial biocide gave more than a doubling of MIC with MRSA over MSSA, whereas the mean values obtained for the antibiotics ranged from 1·6 to 14·5 times higher with MRSA than the corresponding values for MSSA.

The relevance of (or lack of) doubling the MIC can be surmised from the MIC methodology itself. Half-fold dilutions mean that the MIC is the highest concentration of inhibitor in a dilution series, which prevented growth. Therefore the ‘true’ MIC is almost certainly a concentration between this value and the concentration of the next highest dilution. A small change in susceptibility could result in a large, apparent, change in MIC i.e. an apparent doubling, where the reality was a much smaller change.

MSSA 1989 and 2000: from Table 2, the majority of the antibiotics tested showed no statistically significant difference between 1989 and 2000 populations. Only Cipro, Erythro and Gent had significantly higher MICs. The majority of antimicrobial biocides had elevated mean population MICs in 2000 relative to 1989. Only one (OPP) had an MIC higher than a factor of one doubling dilution. Benzalkonium chloride, a much-used biocide, had a significantly reduced mean MIC in 2000 relative to 1989.

MRSA 1989 and 2000: from Table 2, the majority of the antibiotics tested showed no statistically significant difference between 1989 and 2000 populations. Interestingly, none of the antibiotics had mean MICs that were significantly higher in 2000 compared with 1989, in fact three had significantly lower mean MIC values. The pattern of change with the antimicrobial biocides mirrors very closely the changes found with MSSA: the majority have elevated MIC, with OPP having a mean MIC higher by a factor of >3, and with BK showing a decrease in mean MIC. Of interest to many current debates, with these MRSA isolates triclosan showed no significant change in its mean MIC between 1989 and 2000.

These comparisons suggest that alterations in the mean susceptibility of Staph. aureus to antimicrobial biocides have occurred between 1989 and 2000, but that these changes were mirrored in MSSA and MRSA suggests that methicillin resistance has little to do with these changes.

Pairwise correlations within the MSSA and MRSA data.

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

Correlation analysis describes the presence or absence of any linear relationship between two data sets (note – it does not suggest a causal relationship, only that a relationship exists). With antimicrobials, the correlation, if present, may be positive indicating that an elevated MIC to one antimicrobial correlates with an elevated MIC of the other antimicrobial, or negative indicating that elevation of one correlates with a decrease in the other. Given data with a non-normal distribution a nonparametric test, such as Spearman's Rho (as performed here), calculates the degree of correlation based on the ranks of the data.

Table 3 lists the significant pairwise correlations (P < 0·01) along with the Spearman-Rho statistic and its probability for the MSSA and MRSA isolates. Of the possible 153 pairwise correlations, MSSA had 27 that were statistically significant, whereas MRSA had 42. MSSA shared 13 of these significant correlations with MRSA. The majority of significant correlations were between antibiotics or between biocides. Of the correlations found with MSSA only two were between antibiotics and biocides: a positive correlation between erythromycin and triclosan, but a negative correlation between ampicillin and o-phenyl phenol. With MRSA, although the number of antibiotic–biocide correlations was much greater (12/42), eight were positive and four were negative. The positive correlations occurred solely with the quaternary ammonium surfactants. Three of the negative correlations occurred with o-phenyl-phenol, the fourth was the erythromycin/triclosan pairing i.e. a reversal of the correlation found with MSSA.

Table 3.  Significant Spearman Rho correlation coefficients for antimicrobial pairs against MSSA and MRSA
 MSSAMRSA
Antimicrobial-1Antimicrobial-2Spearman Rho statisticProb>|Rho|Antimicrobial-1Antimicrobial-2Spearman Rho statisticProb>|Rho|
  1. Of the total 153 possible correlations only those with P < 0·01 are listed. Bold lettering signifies antibiotic-biocide pairing; bold italic lettering signifies negatively correlated pairs. Spearman Rho statistic is the correlation coefficient between antimicrobial-1 and antimicrobial-2, the probability of the correlation occurring by chance is given by Prob >|Rho|.

  2. MSSA, methicillin-sensitive Staph. aureus; MRSA, methicillin-resistant Staph. aureus.

1A/CAmp0·7396<0·0001CHXPCMX0·923<0·0001
2PCMXTCS0·5619<0·0001CefazCefep0·9213<0·0001
3OPPTCS0·5379<0·0001CefazA/C0·8866<0·0001
4CHXTCS0·4939<0·0001CefepA/C0·8643<0·0001
5CHXPCMX0·4324<0·0001CHXTCS0·8349<0·0001
6OPPPCMX0·4268<0·0001PCMXTCS0·8163<0·0001
7CefepA/C0·4199<0·0001BKBenzeth0·7418<0·0001
8CHXOPP0·406<0·0001ErythroClind0·5237<0·0001
9ErythroClind0·3873<0·0001OxaCipro0·5208<0·0001
10OxaA/C0·3669<0·0001CiproA/C0·5119<0·0001
11Bleach UBleach R0·3428<0·0001ErythroCipro0·4974<0·0001
12OxaAmp0·3148<0·0001OxaA/C0·4908<0·0001
13CefepAmp0·3069<0·0001CiproBK0·4869<0·0001
14OPPBenzeth0·295<0·0001A/CAmp0·4846<0·0001
15ErythroTCS0·28350·0002OxaCefaz0·4842<0·0001
16OPPBK−0·24220·0015OxaCefep0·4717<0·0001
17Bleach UBenzeth0·2350·0021CiproBenzeth0·4652<0·0001
18Bleach RBenzeth0·23150·0025CiproCefaz0·448<0·0001
19OPPBleach U0·21290·0055CefazAmp0·4458<0·0001
20TCSBenzeth0·21030·0061CefepAmp0·4222<0·0001
21ErythroCipro0·20850·0065ErythroBK0·38930·0002
22AmpOPP0·20660·007OxaBK0·38750·0002
23OPPBleach R0·20420·0077CiproCefep0·38720·0002
24VancoClind0·20290·0082GentOPP0·38340·0002
25PCMXBenzeth0·2010·0088ErythroA/C0·37190·0004
26VancoGent0·19960·0093OPPBK−0·36650·0005
27GentErythro0·19920·0094TCSBleach R0·36170·0006
28    ClindOPP−0·36090·0006
29    OxaErythro0·35840·0007
30    OxaBenzeth0·35720·0007
31    CHXBleach R0·34460·0011
32    Bleach UBleach R0·3410·0012
33    ErythroCefaz0·33470·0015
34    ClindBK0·33050·0018
35    CHXOPP0·32650·002
36    OPPPCMX0·32580·0021
37    ClindCipro0·3120·0033
38    ErythroTCS0·30320·0043
39    A/CBK0·29880·0049
40    CefazBK0·28550·0073
41    OxaOPP0·28140·0083
42    TCSBK−0·28050·0085

Although not shown in Table 3, the nonsignificant correlations are also of real importance: they show what pairings have no statistically significant effect on each other. For example oxacillin and chlorhexidine show no correlation at all. Where no correlation exists between antibiotics, it may be possible that these observations could be used as a basis for combined antibiotic therapy. Although it must also be noted that the presence of zero correlation does not imply that there will always be zero correlation.

Principal component analysis of Staphylococcus aureus.

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

Principal component analysis is a useful technique to look for clusters of data related by some underlying phenomenon. The technique transforms a multivariate set of data, (in this case MIC data from a variety of antimicrobials) possibly correlated, to a new set of uncorrelated variables the PCs. The new variables are linear combinations of the original variables and the order in which the variables are derived is in decreasing order of importance. For example, the first PC accounts for the most variation in the original data. Often the majority of the variation in a data set can be described by only a few of the PCs and published methods are used to set the criteria for deciding how many are truly useful. Often a scatter plot of the first two PCs is more informative than pairwise scatter plots of all the original variables. Furthermore such a plot often reveals clusters of data from which information on the underlying reasons can be obtained.

A PC analysis of the Staph. aureus MIC data revealed five PCs, which met the eigenvalue 1 criterion (Kaiser 1960). Combined these five components accounted for 68% of the total variation within the data (Table 4). The first component, rPC1 (r because the PC has been rotated), which contains the greatest variance in the data, consists of major contributions from the penicillins, the cephalosporins along with erythromycin and clindamycin, with only minor contributions from the antimicrobial biocides. The second component (rPC2) consists of major contributions from TCS, PCMX, OPP and CHX with little contribution from the antibiotics. The third component consists of loadings from the two quaternary ammonium surfactants together with loadings from clindamycin, erythromycin and ciprofloxacin. The two bleach formulations are the major contributors to the fourth component. Gentamycin along with vancomycin forms the fifth component.

Table 4.  Rotated factor pattern for Stapylococcus aureus
 rPC1rPC2rPC3rPC4rPC5
  1. Major components making up a group are given in bold (>0·4 criterion).

Benzeth0·2070·0790·7580·248−0·208
BK0·015−0·3450·674−0·0670·077
Bleach R0·1790·196−0·0200·7610·010
Bleach U−0·1140·0800·0840·797−0·010
TCS−0·0570·6710·0580·1420·047
PCMX0·1390·8740·059−0·015−0·011
OPP0·1170·6630·0540·251−0·209
CHX0·1480·870−0·0410·0210·009
Amp0·762−0·093−0·1990·019−0·012
A/C0·9620·0720·1430·0160·033
Cefep0·9040·0700·1580·0400·019
Cefaz0·8680·1200·2730·0130·070
Cipro0·0510·2810·5440·1050·356
Clind0·4150·1730·523−0·1900·270
Erythro0·4330·3040·519−0·0570·266
Gent0·1280·0010·273−0·0040·620
Oxa0·8280·2700·2610·0370·052
Vanco−0·023−0·126−0·088−0·0100·805
Cumulative percent23·8339·9652·1760·0567·87

Principal components are made up of contributions from the data from all the variables used. All components have zero correlation with each other (principle of orthogonality). This means that a linear regression plot of one component against another gives a gradient of zero with an r2 = 0. A component, which consists of (almost) exclusively one type of variable e.g. antibiotics, is suggesting there is little influence on it by the other variables e.g. antimicrobial biocides. rPC1 consists almost exclusively of antibiotics, and rPC2 consists almost exclusively of antimicrobial biocides. Because of the principle of orthogonality this means that there is little correlation between antibiotic and antimicrobial biocide MIC values for this data set.

Figure 1a–d show plots of the first two rotated PCs. Figure 1a,b also show the normal ellipses for the year 1989 and 2000 at the P = 0·9 level (90% of samples fall on or within the ellipse). Figure 1c,d also describe the normal ellipses for the designations of MRSA and MSSA at the P = 0·9 level. Superimposed on each figure are density contours, which help reveal some interesting patterns.

image

Figure 1. A plot of all Staphylococcus aureus isolates using the first two rotated principal components. (a,b): 90% density ellipses for both year groupings are shown in red (1989) and black (2000). The majority of 1989 isolates fall within the 2000 90% ellipse. The position of MSSA and MRSA isolates are given by black rectangles and red circles, respectively. Density contours are given, based on the number of isolates in 1989 (a) and 2000 (b). (c,d): 90% density ellipses for the two groupings MSSA (black rectangles) and MRSA (red circles) are shown in black and red, respectively. Density contours are based on the number of MSSA (c) and MRSA (d). (e): A plot of all Staphylococcus aureus isolates using the first and third rotated principal components. Density contours are based on the number of isolates (MSSA and MRSA) with the highest resistance to benzalkonium chloride

Download figure to PowerPoint

In 1989 (Fig. 1a), essentially two major groupings are observed – one associated with MSSA and one with MRSA (note the contours are restricted to the year designated). Although having different values for rPC1, the two groupings have similar values for rPC2. Two minor groupings, associated with MSSA are also present having similar values of rPC1 to the main group, but higher values of rPC2. This means that although of similar susceptibility towards antibiotics (the major factors of rPC1), a few isolates have higher MIC values towards biocides.

In 2000, (Fig. 1b) a more complex pattern has emerged. There are now several new clusters, with MRSA showing three major groupings. The MRSA grouping observed in Fig. 1a is still present but there is also now a group associated with biocide resistance, but with equivalent antibiotic resistance to the original group. Another group associated with MRSA has lower antibiotic resistance but similar biocide resistance to the former grouping. For MSSA, a similar pattern emerges – the presence of the original 1989 grouping, but with two new ones of higher biocide resistance, one of which has similar antibiotic resistance and one that is lower.

Figure 1c,d show the contour densities of MSSA and MRSA, respectively. What appears remarkable is that the pattern for MSSA is similar to MRSA, only occurring at a lower value of rPC1. MRSA clearly has three major groupings, possibly with a fourth, which overlaps with MSSA; MSSA has similar groupings but the higher rPC2 groupings are less well-defined.

These figures appear to show that between 1989 and 2000 a subpopulation of MRSA has acquired a higher resistance to biocides, but this has not altered the antibiotic susceptibility of that group. When these two groups were analysed for the difference between them, the group with higher biocide resistance showed greater resistance to TCS, PCMX and CHX. Another subpopulation has also acquired biocide resistance, but has reduced antibiotic resistance. A similar conclusion was found with MSSA.

The principle factors of rPC3 are from the two QACs and from three of the antibiotics (Table 4). A plot of rPC1 against rPC3 followed by density maps of Benzeth, BK (Fig. 1e), Cipro, Clind and Erythro were produced. The highest levels of resistance to both BK and Benzeth occur in both MSSA and MRSA, but MRSA has a higher number of isolates with high-level resistance to both BK and Benzeth. However, in both cases the MSSA and MRSA share ca the same value for rPC3. High level resistance to both Erythro and Clind were equally distributed between the MSSA and MRSA, with high level Cipro resistance associated only with MSSA. Figure 1e shows that the high resistance BK isolates have similar rPC1 values to each other and to those with lower BK resistance. This suggests that although quaternary ammonium resistance is associated more with MRSA than with MSSA, there is no evidence to suggest that BK resistant MRSA are any more resistant to antibiotics in general.

Distributions.

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

The MIC distributions of the isolates for the various antimicrobials against Ps. aeruginosa are given in Table 1. The mean values were obtained from the distribution of the log10 MIC distributions.

An analysis, using one way anova, of the Ps. aeruginosa data was performed to examine any significant differences between the isolates of 1989 (54) vs those of 2000 (57). The results are given in Table 5 and show the majority of antimicrobial biocides had significantly lower mean MIC in 2000 relative to 1989. Of the seven antibiotics, three had significantly lower mean MICs in 2000, three showed no change and only ciprofloxacin showed a statistically significantly higher mean MIC in 2000 relative to 1989.

Table 5.  Mean population MIC (mg l−1) in 1989 and 2000
AntimicrobialPs. aeruginosa 1989Ps. aeruginosa 2000Significance
  1. NS, no significant correlation between the pairings (P > 0·01).

  2. S, significant,positive, correlation between pairings (P ≤ 0·01).

  3. −S, significant, negative, correlation between pairings (P ≤ 0·01).

Benzeth18·945·0S
BK13365·6−S
Bleach R615508−S
Bleach U503461NS
TCS428285NS
PCMX977263−S
OPP21571321−S
CHX14·48·36−S
Pip13·46·67−S
P/T10·35·69−S
Ceftrix21·520·6NS
Ceftaz3·433·98NS
Cefep3·753·37NS
Gent2·751·57−S
Cipro0·170·62S

Pairwise correlation analysis.

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

Table 6 lists the significant pairwise correlations (P < 0·01) along with the Spearman-Rho statistic and its probability for the Ps. aeruginosa isolates. Of the possible 105 pairwise correlations, 28 were statistically significant. The 15 significant correlations between the antibiotics and eight between the antimicrobial biocides were all positively correlated. However, four of the five significant correlations found between antibiotics and antimicrobial biocides were negatively correlated i.e. an increase in MIC with respect to an antibiotic is correlated to a decrease in MIC towards the relevant biocide, and vice versa. A significant positive correlation between gentamycin and PCMX was found.

Table 6.  Significant correlations between antimicrobials for Pseudomonas aeruginosa
Antimicrobial-1Antimicrobial-2Spearman Rho statisticProb >|Rho|
  1. Of the total 105 possible coefficients only those with P < 0·01 are listed. Bold lettering signifies antibiotic–biocide pairing; bold italic lettering signifies negatively correlated pairs. Spearman Rho statistic is the correlation coefficient between antimicrobial-1 and antimicrobial-2, the probability of the correlation occurring by chance is given by Prob >|Rho|.

BKBenzeth0·4185<0·0001
CefepPip0·6728<0·0001
CefepP/T0·6603<0·0001
CefepCeftrix0·4183<0·0001
CefepCeftaz0·663<0·0001
CeftazPip0·7059<0·0001
CeftazP/T0·6534<0·0001
CeftazCeftrix0·5663<0·0001
CeftrixPip0·5908<0·0001
CeftrixP/T0·5786<0·0001
CHXPCMX0·411<0·0001
CiproCefep0·375<0·0001
GentPip0·4602<0·0001
GentP/T0·4572<0·0001
GentCefep0·6293<0·0001
OPPPCMX0·558<0·0001
P/TPip0·9329<0·0001
PCMXBK0·3936<0·0001
PCMXTCS0·3981<0·0001
Bleach UBleach R0·30710·001
OPPTCS0·29640·0016
GentCeftaz0·27690·0033
GentPCMX0·26650·0039
PCMXBR0·24870·0085
CiproPCMX0·475<0·0001
CiproOPP0·35050·0002
CiproTCS0·30730·001
CiproBK0·24950·0083

Principal component analysis of Ps. aeruginosa data.

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

The MIC values were subjected to a principal component analysis. The first five PCs gave eigenvalues greater than one and were retained, the results of a Scree test (Cattell 1966) also suggested that only the first five components were meaningful. Combined, the five components accounted for 70% of the total variance (Table 7).

Table 7.  Rotated factor patterns for Pseudomonas aeruginosa
 rPC1rPC2rPC3rPC4rPC5
  1. Major components making up a group are given in bold (>0·4 criterion).

Benzeth−0·070·100·870·130·04
BK0·04−0·200·82−0·17−0·03
Bleach R−0·01−0·150·030·70−0·03
Bleach U−0·040·04−0·020·810·01
TCS−0·01−0·330·040·48−0·30
PCMX0·290·81−0·04−0·240·19
OPP0·070·710·11−0·140·12
CHX0·010·620·060·14−0·32
Pip0·72−0·05−0·02−0·070·58
P/T0·67−0·110·00−0·120·61
Ceftrix0·07−0·010·06−0·170·85
Ceftaz0·440·15−0·140·110·67
Cefep0·830·140·020·15−0·30
Gent0·87−0·10−0·030·000·07
Cipro0·260·740·040·100·01

The five rPCs essentially separate the antibiotics from the antimicrobial biocides. The first principal component (rPC1) consists of major contributions from Pip, P/T, Ceftaz, Cefep and Gent with little contribution from the antimicrobial biocides. The second component (rPC2) consists of major contributions from PCMX, OPP, CHX and Cipro. However, the contribution from Cipro is in a direction negative to the antimicrobial biocides i.e. shows a negative relation to them. The two quaternary ammonium biocides (rPC3) form a component almost devoid of any contribution from other antimicrobials. The two bleach products and triclosan form the fourth component (rPC4). The fifth rotated component (pRC5), which accounts for only 7% of the total variation, is a cluster of four antibiotics (Pip, P/T, Ceftrix and Ceftaz).

The results show that for Ps. aeruginosa the major factors of the PCs are either groups of antibiotics or groups of antimicrobial biocides: there are no major groups of antibiotics with antimicrobial biocides.

Figure 2a,b show plots of the first two rotated PCs. The figures also show the normal ellipses for the year 1989 and 2000 at the P = 0·9 level (90% of samples fall on or within the ellipse). Superimposed on each figure are density contours. Unlike Staph. aureus where distinct groups were observed, Ps. aeruginosa shows a greater degree of scatter, with perhaps one major grouping centred at ca−0·25, −1·25 in 1989 (Fig. 2a) and a more diffuse group centred around −0·5, 0·25 in 2000 (Fig. 2b). In this case as rPC2 has a positive contribution from Cipro but negative factors from the antimicrobial biocides, a positive value for the rPC2 reflects a high MIC for Cipro but low MIC for the antimicrobial biocides. The overall patterns therefore suggest a decreased resistance to biocides between 1989 and 2000 and a small decrease in antibiotic resistance (the centres of the 90% density ellipses have moved down the rPC1 axis and up the rPC2 axis between 1989 and 2000. These results reflect well the mean MIC values found (Table 5).

image

Figure 2. A plot of all Pseudomonas aeruginosa isolates using the first two rotated principal components. The 90% density ellipses for both year groupings are shown in red (1989) and black (2000); individual isolates within these groupings are given as 1989 (red circles) and 2000 (black circles). Density contours for 1989 (a) and 2000 (b) are used to suggest possible cluster positions

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References

The relationship between antibiotic and antimicrobial biocide (disinfectants) resistance is currently considered a hot topic (Levy 1998a,b). In cases where antibiotic-resistant organisms have become a serious problem with respect to nosocomial infection, the widespread use of antimicrobial biocides and the introduction (or enforced compliance) of handwashing and general hygiene measures normally leads to its amelioration (Seal 1983; Schmitz et al. 1998; Dancer and Crawford 1999). If, as has been suggested, antimicrobial biocides may be aiding the incidence of the same antibiotic-resistant organisms, then there appears to be a dichotomy of reasoning.

The study of the mean MIC, MIC50 or the MIC90 values for a population does allow general trends in the change of resistance with time to be highlighted. However, such a study will fail to answer any questions on potential correlations of the antimicrobials. Correlation analysis allows the researcher to highlight cases where increases (or decreases) in the MIC of an antimicrobial are connected in some way with increases in the MIC of another. If we are concerned with the effects of one group of antimicrobials on another group, relative to an organism, then correlation analysis has to be performed using data on the MIC per strain of the relevant organism. Such data is rarely published, whereas general population distributions are.

The calculation of a significant correlation, however, does not prove that a causal link exists. The correlation may be happenstance or it may be a true physico/chemical interaction. Proof of causality would require further experimentation (e.g. biochemical/physiological) although the reason for many of the correlations could be hypothesized to occur because of shared mechanisms of action, if these were known a priori.

An examination of the Staphylococcal data showed that correlations between antibiotics were numerous with MRSA accounting for the majority. Many of the correlations are those expected on the basis of some shared mechanism of action. When the correlations between antimicrobial biocides and antibiotics are examined, the idea of shared mechanisms as an explanation becomes tenuous and difficult to support. The positive correlations between the QACs and certain antibiotics may be due to reduced uptake mechanisms or a general reduced permeability. The possibility of a shared efflux pump must also be considered, but if efflux pumps are associated with high level antibiotic resistance then the distribution of the QAC pump is not evenly spread across the antibiotic-resistant strains (Fig. 1e). Furthermore if the presence of pumps were the source of QAC resistance then MSSA may also share the mechanism as the rPC3 values were similar (rPC3 has major factors from both QACs, clindamycin, erythromycin and ciprofloxacin). More data may be required before a definitive answer is given.

The negative correlations, such as erythromycin with triclosan or oxacillin with o-phenylphenol also need to be understood. It would be interesting to hypothesize that a mechanism of antibiotic resistance increases the activity of a range of antimicrobial biocides or that energy requirements to maintain antibiotic resistance result in increased biocide susceptibility. That the correlation is negative, however, also means that an increased resistance to the antimicrobial biocides leads to a decrease in the resistance to the antibiotics, what single mechanism can account for both phenomena?

From Fig. 1a–d, we can be certain that this population of MRSA has developed biocide resistance. That the grouping of biocide-resistant MRSA has an essentially identical antibiotic resistance, and that MSSA appears to have developed this in a similar manner, suggests that the acquisition of biocide resistance does not alter antibiotic resistance. An alteration of antibiotic resistance within a population may obviously have a clinical consequence, but the acquisition of biocide resistance is probably of no consequence given that the usage of biocides within the clinical environment is totally different to the administration of clinical antibiotics. Biocides are normally used at concentrations far in excess of the reported MIC, while the serum level of antibiotic is usually slightly in excess of the MIC.

Bamber and Neal (1999), using 186 isolates of MRSA and MSSA, showed that although 7·5% of isolates had elevated MIC towards triclosan (>1 mg l−1), there was no significant difference in the incidence of triclosan resistance between the strains of MSSA and MRSA. Slater-Radosti et al. (2001) also showed that mutants of Staph. aureus with an over-expression of the enoyl-ACP reductase showed moderate increases in MIC to triclosan, but no increase of MIC to antibiotics (tetracycline, penicillin G, erythromycin, neomycin, mupirocin, and ciprofloxacin), suggesting that no cross-resistance was developed by these strains. Gordon and McClure (1987) showed no significant differences in the MIC of chlorhexidine when comparing MRSA with MSSA isolates, although Suller and Russell (1999) found MRSA isolates to have increased MIC towards a range of antimicrobial biocides. These latter findings are completely compatible with those discussed here, especially with reference to Fig. 1a–d.

Pseudomonas aeruginosa appears to have had a general decrease in antimicrobial resistance between 1989 and 2000. There are numerous significant correlations between the antimicrobials, again most are either between antibiotics or between antimicrobial biocides. Between antibiotics and biocides there are five significant correlations, four are negative correlations with ciprofloxacin, and one is positive (p-chloro-m-xylenol and gentamicin). Given the strong correlations between PCMX and the other antimicrobial biocides it is surprising that other significant antimicrobial biocide-gentamicin correlations were not observed – unless this is a case of happenstance correlation.

The principal component analysis of the Ps. aeruginosa data suggests that antibiotics and antimicrobial biocides form separate groupings (Table 6). The first principle component, rPC1, is factored heavily by antibiotics and is almost devoid of any antimicrobial biocide factors, apart from a small contribution from PCMX. Although rPC2 has loadings from three antimicrobial biocides (chlorhexidine, o-phenyl phenol and p-chloro-m-xylenol) and one antibiotic (ciprofloxacin), ciprofloxacin is negative to the antimicrobial biocides. This means that, in this case, a high value for rPC2 reflects high values for the MIC of ciprofloxacin but low values for the MIC of the three biocides. This reflects well the studies of Baillie et al. (1993) which showed hypersensitivity to chlorhexidine with ciprofloxacin-resistant Ps. aeruginosa, whereas no such reduction in MIC was observed with ciprofloxacin-sensitive isolates. The rotated component also suggests that any isolate resistant to chlorhexidine would be sensitive to ciprofloxacin.

Much work has been published on the acquisition of resistance to QACs by Ps. aeruginosa. Recently, Joynson et al. (2002) showed that an isolate of Ps. aeruginosa trained by subculture (passage) to be resistant to two aminoglycosides had an increased resistance to benzalkonium chloride, but that the same organism trained by subculture to be resistant to benzalkonium chloride was more sensitive to the same two antibiotics. Table 5 shows that benzalkonium chloride and ciprofloxacin are negatively correlated but no correlation was found with gentamycin. The rotated factor, rPC3, consists of factors from the two QACs and is essentially devoid of any interaction with any other antimicrobial. Although this compares well with the study of Jones et al. (1989) who showed cross resistance between QACs when Ps. aeruginosa was trained to be resistant to a single QAC, the absence of antibiotic factors and the fact that there is zero correlation between rPC1 and rPC3, suggests that, from this data, that there are no correlations between QACs and antibiotics. As this latter result was not expected, a study with a larger data set with more antibiotics is needed to help confirm or change this conclusion.

From the analyses performed here it is very difficult to support a hypothesis that increased biocide resistance is a cause of increased antibiotic resistance either in Staph. aureus or in Ps. aeruginosa. The data suggests that although there may be a link between QACs and antibiotics with Staph. aureus, there appears to be no evidence to state that the acquisition of quaternary ammonium resistance increases the resistance to antibiotics. The observation of negative correlations between antibiotics and biocides may be a useful reason for the continued use of biocides promoting hygiene in the hospital environment.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Organisms
  6. Antimicrobial agents
  7. Data analyses
  8. Correlation analyses
  9. Spearman-Rho.
  10. Principle component analysis.
  11. Results
  12. Staphylococcus aureus
  13. Distributions.
  14. Pairwise correlations within the MSSA and MRSA data.
  15. Principal component analysis of Staphylococcus aureus.
  16. Pseudomonas aeruginosa
  17. Distributions.
  18. Pairwise correlation analysis.
  19. Principal component analysis of Ps. aeruginosa data.
  20. Discussion
  21. Acknowledgements
  22. References
  • Adair, F.W., Geftic, S.G. and Gelzer, J. (1971) Resistance of Pseudomonas to quaternary ammonium compounds. Cross resistance characteristics of Pseudomonas aeruginosa. Applied Microbiology 21, 10581063.
  • Baillie, L.W.J., Power, E.G.M. and Phillips, I. (1993) Chlorhexidine hypersensitivity of ciprofloxacin-resistant variants of P. aeruginosa PAO. Journal of Antimicrobial Chemotherapy 31, 219225.
  • Bamber, A.I. and Neal, T.J. (1999) An assessment of triclosan susceptibility in methicillin-resistant and methicillin sensitive Staphylococcus aureus. Journal of Hospital Infection 41, 107109.
  • Brown, M.R.W. and Tomlinson, E. (1979) Sensitivity of Pseudomonas aeruginosa envelope mutant to alkylbenzyldimethyl ammonium chlorides. Journal of Pharmaceutical Science 68, 146149.
  • Cattell, R.B. (1966) The scree test for the number of factors. Multivariate Behavioral Research 1, 245276.
  • Dancer, S.I. and Crawford, A. (1999) Keeping MRSA out of a district hospital. Journal of Hospital Infection 43, S19S27.
  • Gabriel, K.R. (1971) The biplot graphic display of matrices with application to principal component analysis. Biometrika 58, 453467.
  • Gabriel, K.R. (1981) Biplot display of multivariate matrices for inspection of data and diagnosis. In Interpreting Multivariate Data ed. Barnett, V. pp. 147173. London: John Wiley and Sons.
  • Gabriel, K.G. and Odoroff, C.L. (1990) Biplots in biomedical research. Statistics in Medicine 9, 469485.
  • Gordon, J. and McClure, A.R. (1987) Comparison of in-vitro activity of povidone iodine and chlorhexidine against clinical isolates of methicillin-resistant Staphylococcus aureus (MRSA). In Proceedings of the First Asian/Pacific Congress on Antisepsis ed. Selwyn, S. pp. 133140. Royal Society of Medicine Services International Congress and Symposium Series No 129.
  • Hollander, M. and Wolfe, D.A. (1973) Nonparametric Statistical Methods. New York: Wiley.
  • Jones, M.V., Herd, T.M. and Christie, H.J. (1989) Resistance of Pseudomonas aeruginosa to amphoteric and quaternary ammonium biocides. Microbios 58, 4961.
  • Joynson, J.A., Forbes, B. and Lambert, R.J.W. (2002) Adaptive resistance to benzalkonium chloride, amikacin and tobramycin: the effect on susceptibility to other antimicrobials. Journal of Applied Microbiology 93, 96107.
  • Kaiser, H.F. (1960) The application of electronic computers to factor analysis. Educational and Psychological Measurement 20, 141151.
  • Lambert, R.J.W., Joynson, J.A. and Forbes, B. (2001) The relationships and susceptibilities of some industrial, laboratory and clinical isolates of Pseudomonas aeruginosa to some antibiotics and biocides. Journal of Applied Microbiology 91, 972984.
  • Levy, S.B. (1998a) The challenge of antibiotic resistance. Scientific American 278, 4653.
  • Levy, S.B. (1998b) Antimicrobial resistance: bacteria on the defense. Resistance stems from misguided efforts to try and sterilize our environment. British Medical Journal 317, 612613.
  • Loughlin, M.F., Jones, M.V. and Lambert, P.A. (2002) Pseudomonas aeruginosa cells adapted to benzalkonium chloride show resistance to other membrane active agents but not to clinically relevant antibiotics. Journal of Antimicrobial Chemotherapy 49, 631639.
  • Moken, M.C., McMurry, L.M. and Levy, S.B. (1997) Selection of multiple-antibiotic-resistant (Mar) mutants of Escherichia coli by using the disinfectant pine oil. Antimicrobial Agents and Chemotherapy 41, 27702772.
  • Russell, A.D. (2000) Do biocides select for antibiotic resistant bacteria? Journal of Pharmacy and Pharmacology 52, 227233.
  • Russell, A.D., Tattawasart, U., Maillard, J.Y. and Furr, J.R. (1998) Possible link between bacterial resistance and use of antibiotics and biocides. Antimicrobial Agents and Chemotherapy 42, 2151.
  • Russell, A.D., Suller, M.T.E. and Maillard, J.-Y. (1999) Do antiseptics and disinfectants select for antibiotic resistance? Journal of Medical Microbiology 48, 613615.
  • Schmitz, F.-J., Verhoef, H., Idel, H., Heinz, H.P. and Jones, M.E. (1998) Impact of hygienic measures in the development of methicillin resistance among Staphylococci between 1991 and 1996 in a university hospital. Journal of Hospital Infection 38, 237240.
  • Seal, D.V. (1983) The role of antiseptics and disinfectants in the control of nosocomial infection. British Journal of Clinical Practice 23(Suppl.), 4654.
  • Slater-Radosti, C., Van Aller, G., Greenwood, R., Nicholas, R., Keller, P.M., DeWolf, W.E., Fan, F., Payne, D.J. et al. (2001) Biochemical and genetic characterization of the action of triclosan on Staphylococcus aureus. Journal of Antimicrobial Chemotherapy 48, 16.
  • Suller, M.T.E. and Russell, A.D. (1999) Antibiotic and biocide resistance in methicillin resistant Staphylococcus aureus and vancomycin-resistant Enterococcus. Journal of Hospital Infection 43, 281291.