What are the antibacterial susceptibility breakpoints? They are a set of values through which scientists define susceptibility and resistance of bacterial strains to various antibacterial agents. These breakpoints are expressed either in concentration (mg/liter or µg/ml) or in a zone diameter (mm), and are established by many international organizations and by using different methods. Setting the antibacterial breakpoints depends usually on four different data sources that have to be taken into consideration, while the final goal of every breakpoint is classify tests results as susceptible, intermediate, or resistant. The huge variety among breakpoints have made it possible that the same pathogenic strain causing the same damage to body tissues can be identified resistant in one country and susceptible in another to the same antibacterial agent. The fact that the data sources might be collected using experimental procedures that does not resemble the in “vivo” environment questions the validity of these breakpoints. Problems with susceptibility breakpoints extended toward reliability and even toward the economic impact of wrong breakpoints.
The four data sources used to set the antibacterial susceptibility breakpoints are the minimum inhibitory concentration (MIC) distributions, the pharmacokinetics and the pharmacodynamics consideration, the clinical and bacteriological response rates, and the phenotypic and genotypic resistance markers.
MICs are the lowest concentration that will inhibit the growth of a test organism over a defined interval of time that is related to the organism's growth rate. MICs are detected in different laboratories using different techniques (broth microdilutions, disk diffusion, antimicrobial gradient, or automated-instrument methods). Because of the availability of a numerous number of antibacterial agents and the diversity of antibiotic formulas used in different institutions have made it difficult for manufacturers to provide standardized instruments that fit everybody's' needs, and because the commercial broth microdilutions is cost inefficient and time consuming, the disk diffusion has been a preferred method over the two-fold microdilutions that time and recourses-consuming. However, MICs might vary between countries and institutions due to the inoculum size and concentration, incubation temperature and duration, and due to media used for incubation. If the antibacterials used to detect the MIC were not prepared precisely, i.e. their concentrations were not accurate; the MIC values will not be correct.
The major critique that is addressed to the MIC methods is that it does not mimic the in vivo conditions and that the pathogen in the human body would react differently to an antibacterial dose than an in vitro pathogen would do. Scientists tried to solve this issue by considering the pharmacokinetics (PK) and the pharmacodynamics (PD) of the drug used. The PK refers to the absorption, distribution, accumulation, and elimination of the antibacterial inside the human body. Such entries are usually established in healthy volunteers. The PD is the study of the time course of action of the antibacterial used. PDs are usually established in vitro by studying three predictors of efficacy, which are the time above the MIC (T>MIC), the ratio of the area under the curve over 24h to the MIC (AUC24/MIC), and the peak level-to-MIC ratio (Cmax/MIC). Here, the Cmax represents the maximum plasma concentrations a drug can achieve following each dose, while the AUC represents the area under the curve of concentration as a function of time. These values are better illustrated in figure 1.
PDs have divided antibacterials into two types: those that are time-dependent and those that are concentration-dependent. Time dependent antibacterials are able to eradicate bacteria if their concentration in the body tissue remains above the MIC for a specific period of time (this period is usually the time between the administrations of two consecutive doses). Therefore, these drugs have a saturated killing capacity directly linked to exposure time, and their main determinant is the T>MIC. Good examples of time-dependent antibacterials are the β-lactams, macrolides, and glycopeptides. In concentration-dependent antibacterials, bacterial killing becomes more rapid and profound by increasing drug concentrations. In other words, these drugs have bacterial killing capacities covering a wide range of concentrations and have effects that are proportional to concentration, and their mains determinants are the AUC/MIC ratio and/or Cmax/MIC ratio. Aminoglycosides and fluoroquinolones are concentration-dependent antibacterials.
Studying the PD also reveals other parameters like the postantibiotic effect, (PAE), which is the period of delayed bacterial growth following drug removal after a brief exposure to the antibacterial in vitro. Here, antibacterials can be grouped into either moderate to long periods of delayed growth (prolonged PAE), or immediate regrowth (minimal PAE). The PK/PD values can be used to predict how an antibacterial will react in vivo and set breakpoints that are will increase the chance of therapeutic success.
The PK/PD values along with the correct MIC might be a good approach toward in vivo mimicking, but considering the clinical response rates greatly increases the accuracy of the susceptibility breakpoints. Clinical response rates are based on clinical trials, whether for humans or animals, through which eradication of bacteria is monitored after administration of an antibiotic. At least 80% of susceptible organisms in vitro should show a success in the in vivo therapy, but this number might be lower depending on the type and site of infection. Some scientists prefer to follow the 90-60 rule, which states that infections due to susceptible isolates respond to therapy approximately 90% of the time, whereas infections due to resistant isolates respond approximately 60% of the time.
The final data source that should be taken into consideration when creating antibacterial breakpoints is the ‘phenotypic and genotypic resistance markers'. These markers are usually helpful is detecting resistance mechanisms without referring to the bacterial MIC. Phenotypic markers embrace factors like direct detection of degrading enzymes, detection of β-lactamase by adding clavulanate or EDTA, modification of the inoculation medium to enhance resistance expression (like using brain heart infusion to detect vancomycin-resistant enterococci), and others. Genotypic markers are usually used to ensure the presence of the phenotypic markers. The use of both genotypic and phenotypic markers is still narrow due to the complexity of the methods, still, and if simple special methods are to be developed and full genotypic knowledge about resistance mechanisms is to be acquired, then the use of genotypic and phenotypic markers could actually replace the use of MICs, as they are more sensitive and can detect resistance mechanisms that cannot be detected using MIC. Resistance markers along with MICs, PK/PD data, and clinical response rates will help in creating antibacterial breakpoints that gives physicians and clinicians a high level of confidence that their susceptibility interpretations are right.
However, the major cause of antibacterial breakpoints variety, and thus the loss of confidence, is that the breakpoints are set by different committees and organizations over different countries. Some countries might be more conservative in assessing the susceptibility of an organism to an antimicrobial agent and place greater emphasis on the detection of emerging resistance. Technical factors that concern different data sources (like inoculum size for MIC, variation of PD/PK factors between different patients, difficult differentiation between a successful and a failed one…) will also affect the breakpoints' values, without being able to accept one value as true and reject all other values. Another cause for variety is that different committees and organizations weighs the four data sources differently, most probably according to the training skills and tasks of the members of each committee.
In the United States, two organizations have the right to set breakpoints. The Food and Drug Administration (FDA) is responsible for the approval of the new antimicrobial agents and for providing its breakpoints. However, the FDA does not have a formal process of reviewing breakpoints, so once the breakpoints are established, they are not reassessed and probably changed. This can result in the use of inappropriate breakpoints after a time as organisms usually start acquiring resistant mechanisms, which will decrease their actual susceptibility to the antibacterials. The other American organization that has the right to set breakpoints is the Clinical and Laboratory Standard Institute (CLSI), formerly the National Committee for Clinical Laboratory Standards (NCCLS). Nonetheless, the CLSI establishes initial breakpoints, then reviews them periodically, and can publish changed or updated breakpoints on regular bases. In Europe, a newly established committee, called the European Committee of Antimicrobial Susceptibility Testing (EUCAST) is responsible for setting susceptibility breakpoints for new antibacterials and for harmonizing of antibacterial breakpoints that were previously set by different European countries.
Outside Europe and the USA, many institutions do establish breakpoints. These institutions rarely stick to the standardized methods for establishing breakpoints, leading to breakpoints that are invalid and probably destructive for any antibiotic therapy. In the article entitled “On the validity of setting breakpoint minimum inhibition concentrations at one quarter of the plasma concentration achieved following oral administration of oxytetracycline” , scientists tried to validate the breakpoints that are set following the theory that says: it is reasonable to predict that the outcome of a therapy will be beneficial for the host if the maximum plasma concentration achieved by that therapy (plasma Cmax) is four times greater than the MIC determined in the laboratory.
Scientists tested this hypothesis (also known as the 4:1 ratio) in a clinical trial on fish after ten days of administration of oral oxytetracycline (OTC) in response of increasing mortalities due to infection with Aeromonas salmonicida. An excess OTC dosage of 75mg/Kg body weight was introduced per day, which insured that the maximum plasma concentration of this antimicrobial could be achieved. The OTC was able to reduce the mortality rate from 0.8%/day at day zero to 0.05%/day at day ten of the experiment. To detect the OTC concentration in the tissues of healthy fish, sufficient plasma was obtained from 26 healthy fish was analyzed. The mean OTC concentration was 0.25mg/L with standard deviation of 0.06 mg/L. Because the OTC was at its maximum concentration in the plasma, and because the significant decrease in the mortality rate ensures that the therapy was a success, then, and according to the 4:1 ratio rule, the estimated MIC for OTC should be , which reflects the susceptibility breakpoint estimated for this drug.
The A. salmonicida colonies that were isolated from dead and moribund fish were tested for their susceptibility to the OTC using disk diffusion and MIC broth dilution methods. The disk diffusion zones values for the isolates ranged between 36 to 55 mm but all isolates had MIC values of 0.5 mg/L OTC. This actual MIC value was higher than the estimated breakpoint value by eight times, although the therapy was a success. Thus, the application of this ratio would predict that the administration of OTC studied here would have had no beneficial effect and that the isolates of A. salmonicida should be classified as resistant.
Although this is only a clinical trial on fish, but the experiment clearly shows the invalidity of many susceptibility breakpoints that are set using this ratio rule. An indirect conclusion form the experiment would be the invalidity of all breakpoints that are established without following the standardized rules, more such examples are to come later on.
Some antibacterial breakpoints might be classified as invalid if there was a difference between the in vitro bacterial susceptibility to the drug and the in vivo result. In another article entitled “Reassessment of cefaclor breakpoint for Streptococcus pneumonia” , the differences between proven clinical success of cefaclor and its relatively poor activity in vitro were investigated against isolates of Streptococcus pneumonia. In this experiment, the scientists used the pneumococcal isolates to determine the cefaclor susceptibility breakpoint by following the standard CLSI methods for broth micro-dilutions MIC. This methodology states that MICs are calculated by preparing doubling dilutions of an antibiotic, dispensing them in 50μl amounts in microtiter trays, and then adding 50μl of the bacterial inoculum to them. Sterility and growth controls are also included in the methodology, and the plates should be incubated for 20-24 h at 37oC. The mean MIC for all the isolates was 1.4mg/L. The scientists then studied the antibiotic stability in vitro and the time kill curves. The antibiotic stability study showed that cefaclor had a half-life of approximately 9 hours with less than 11.6% active antibiotic left after 24 hours. These numbers show that cefaclor is not a stable drug, especially when compared to cefuroxime, another orally administered cephalosporin, which has a half-life greater than 24 hours and maintains more than 90% of its activity for more than 12 hours. The time kill curves showed that the bacteriostatic MIC for cefaclor was between 0.28 and 0.42 mg/L, which is significantly lower than the CLSI MIC calculated earlier.
This article presents a possible explanation for the differences between clinical and laboratory data. The explanation is that the cefaclor -currently in use- Breakpoints values are inaccurate and invalided. These breakpoints do not account for an important pharmacokinetic factor that is the drug's instability, and as a result, in vitro isolates would be classified as resistant to cefaclor following the CLSI susceptibility breakpoint, what is actually going on is that the drug's concentration is dropping down during the period of incubation of the MIC test, leading to a false-high MIC value. However, and in vivo, the dosing concentration and relatively short time interval between two consecutive antibiotic administrations will solve this problem and the antibiotic therapy will have a high chance of success. One conclusion of this article is that the PK factors should be underestimated, or else we will be depriving a patient from a potentially successful therapy.
Reliability of susceptibility breakpoints that are set using computer methods was the question of the article entitled as “Data Mining Validation of Fluconazole Breakpoints Established by the European Committee on Antimicrobial Susceptibility Testing” . Although fluconazole is an antifungal and not an antibacterial, but still this article sheds some light on antimicrobial breakpoints determined by machinery methods. The computer software's job is to determine the antimicrobial breakpoints by calculating the MICs that will split the population into susceptible and resistant. The experiment is a way of comparing these breakpoints with the breakpoints set by the Antifungal Susceptibility Testing Subcommittee of the EUCAST. The EUCAST subcommittee takes into account the PD/PK data and other factors, such as dosing regimens, toxicology, resistance mechanisms, and clinical outcome data. The clinical data were analyzed by the correlation of the MICs and the proportion between the dose administered and the MIC of the isolate (dose\MIC) to the clinical outcome seen in patients, whether a success or a failure.
Scientists entered the following information in the databases: the MIC of the isolate, the dose\MIC values, and the treatment outcome for each patient. The sum was 258 isolates from 258 different patients suffering from candidemia or oropharyngeal candidiasis. The scientists used five different computer classifiers that are able to analyze the data entered and are known for their suitability for intuitive interpretation. They then evaluated these classifiers by their sensitivity, specificity, false-positive rate, and other factors.
The results showed that all the classifiers had good results; the sensitivity and the specificity percentages where high (88.6 and 90% respectively), while the false-positive rates had an average of 10% which the scientists said it was acceptable. As a result, the breakpoints determined were very close to the breakpoints set by the EUCAST, which proves that the computerized methods are good enough to be incorporated into the process of developing breakpoints because such an approach reduces time and mainly avoids researchers' bias during breakpoints' estimation. However, one drawback of these methods is that the classifiers express the breakpoints' categories as susceptible or resistant, with any mentioning of the intermediate category.
Antibiotic breakpoints that are set using different standards than those mentioned earlier should be questioned for their validity, revised, and reset following the standard guidelines. This issue is discussed in the article entitled as “Evaluation of wild-type MIC distribution as a tool for the determination of clinical breakpoints for Mycobacterium tuberculosis” . The scientists used M. tuberculosis isolates to assess the effectivity of breakpoints based on the critical concentrations. A critical concentration is the lowest concentration of an antibacterial that will inhibit 95% of wild-type strains of M. tuberculosis that have never been exposed to this antibacterial before.
The scientists collected 90 M. tuberculosis wild-type isolates, sub-cultured them for about three weeks, and then determined the antibacterial susceptibility breakpoints according to the EUCAST methodology. This methodology includes determining the MICs (of Ethambutol, Rifampicin, and Isoniazid) using the double micro-dilution series, calculating the PD and the PK factors, and taking into consideration the clinical data. Then the scientists are to compare the new breakpoints to the old ones set using the critical concentration rule.
The experimental results showed that the two sets of breakpoints did not match, but on the contrary, MIC-based breakpoints showed to be more effective as they did clearly divide the bacterial population into wild-type strains and resistant ones. This solved the main problem caused by using the critical concentrations method, because this method always yielded breakpoints that fall inside the range of wild-type strains. In other words, some of the susceptible strains were identified as resistant, thus preventing a number of patients with tuberculosis from being treated with a potentially active drug. An example is the anti-tuberculosis drug ‘ethambutol'. The breakpoint for this drug was between 4 and 8mg/L (depending on the institution that is determining the breakpoint). The new breakpoint is 5mg/L, a specific breakpoint that draws a clear borderline between resistance and susceptibility. Following the CLSI standards have an indirect advantage of harmonizing the breakpoints as the methods to establish them are clear and easy to follow. This is another example of the invalidity and unreliability of breakpoints that are set without abiding by the standardized methods.
When committees recognize that bacterial susceptibility to a certain antibiotic is decreasing, they will try to reassess and reevaluate existing breakpoints, and new values will emerge with questions being asked about the necessity of these values and about the differences they will make in any antibacterial therapy. These questions were tackled in an article entitled as “New penicillin susceptibility breakpoints for Streptococcus pneumoniae and their effects on susceptibility categorisation in Germany” , where scientists tried to evaluate penicillin susceptibility to S. pneumoniae isolates and to assess the differences in susceptibility categorization when applying the new and the old CLSI breakpoints.
Scientists managed to isolate 12,137 samples of S. pneumoniae from diagnostic microbiology laboratories throughout Germany that refer to cases from 1-1-1992 to 31-12-2008. The samples were classified as from patients with meningitis or from non-meningitis patients. After isolation and identification, the samples were tested for their MICs using broth micro-dilution method, the method that is recommended by the CLSI.
Applying the former CLSI breakpoints, the non-meningitis cases were classified as 4.1% intermediate and 1.2% resistant, while the meningitis cases were 5.5% intermediate and 1.0% resistant. However, and when applying the new breakpoints, 6.5% of the meningitis cases were resistant (no intermediate category for meningitis cases according to the new guidelines of the CLSI), while 0.0% of non-meningitis cases are classified as resistant and 0.3% as intermediate.
The numbers show that for meningitis cases there was an elevation in the percentage of resistant strains following the new breakpoints, but actually, and because the new guidelines added the intermediate category to the resistant one, we can say that the percentages did not change, because the intermediate category for brain infections is meaningless as we cannot increases the concentration of the antimicrobial that reaches there.
This article shows that reassessment of antibacterial breakpoints should not be the end, but it should be the mean through which higher rates of therapeutic success can be achieved. As a result, reassessment should only be done when there is a need for it, as in a case of pathogenic organisms acquiring new mechanisms of resistance in high rates, or when new clinical situations suggest a need for this, as in the treatment of meningitis caused by Streptococcus pneumonia where it was essential to set new breakpoints for cofotaxime and ceftriaxone. Moreover, and if new formulas and doses are available for old antibiotics, new breakpoints should be created. However, setting new antibacterial breakpoints will lead to over-use to these antibacterials, and that is not recommended especially for antibacterials where their increased consumption is directly linked to resistance, as in the case of the traditional extended-spectrum cephalosporin breakpoints that remains unchanged in order to reduce the use of carbapenems in cases where the organisms are producing ESBLs.
Another form of breakpoints' variety is the way these breakpoints are categorized, whether as susceptible and resistant, or following the SIR system. The intermediate category has been the debate of the EUCAST during the process of harmonization of antibacterial breakpoints all over Europe. This issue was discussed in the supplementary article “Harmonization of antimicrobial testing breakpoints is Europe: implications for reporting intermediate susceptibility” , where the different advantages of reporting an intermediate category have been explained. Many international committees consider the intermediate category as of limited value because of the implicit uncertainty of the clinical response, and as a result, they merged the intermediate and the resistant results together. However, classification of a potential pathogen into the intermediate category may be viewed in a variety of ways. First, it means that an infection due to a bacterium stain classified as susceptible may be cured if the antibiotic can be concentrated in the tissue, as in any urinary tract infection UTI, where the bacteria can be an in vivo susceptible to the antibiotic especially in uncomplicated infections. Antibiotic prescribers, then, will have to increase the antibiotic dose by either giving the initial dose more frequently or give the same dose by prolonged or continuous infusion. Another advantage of the intermediate category in that it provides a buffer zone that greatly reduces the interpretative errors by microbiology clinicians. Interpreting a susceptible bacterium as resistant or a resistant bacterium as susceptible is considered as a major error and would reduce the chance of a successful therapy. As a result, one of the new recommendations by the EUCAST and other standardizing institutions is to publish breakpoints using the SIR system (Sensitive, Intermediate, Resistant). However, reporting the intermediate category should be accompanied by proper tutoring about its meaning, because not all antibiotic prescribers are specialized in infection control, and because different antibiotics in different body tissues have different PK parameters.
Breakpoints, but to lesser extent, have an economic impact that should be taken into consideration. When the breakpoints of cefotaxime and ceftriaxone against non-meningitis infection were reevaluated by the CLSI, there was a 10 to 20% increase in the strains identified as susceptible, which led to a great increase in the revenues of the pharmaceutical companies that synthesized these antibacterials. On the other hand, increasing the resistance rates and thus therapy failures in organisms identified as susceptible in vitro above a predefined thresholds will stimulate the officials to recommend changing the empirical therapy, thus reducing the use of certain antibiotics. For such reasons, and to prevent the huge pharmaceutical companies from obtaining antimicrobial breakpoints that suit its products, setting these breakpoints should be done in the most evenhanded approach in order to provide the most effective and safe treatment for the patients.
Disk diffusion methods, that are used both when setting the breakpoints and later when testing the bacterial susceptibility, might not be very reliable themselves. The best way to transform the MIC results into disk diffusion criteria is to plot a scattergram of the zone diameter versus MICs for strains tested by both methods, and earlier methods assumed that drawing a regression line between through the data points would lead to good estimation of zone diameter for each MIC value. However, the nowadays-asked question is that if the relation between the MICs and the zone diameter is linear or not. The answer is, and as figure 2 shows, the MIC and zone diameter are not evenly distributed across the continuum but tend to accumulate. As a result, several articles tried to evaluate the relationship between MIC and zone diameter and its effect on susceptibility breakpoints. In brief, an article's experiments  showed that the disk diffusion methods are reliable, except when isolates where cultured on CBA, where the zone diameter for the control strain was reduced by 15mm, the article leave this problem highlighted, without giving neither a clear cause nor a solution. What is known for sure that the disk methods are easy to use, flexible, cheap, and do not consume a lot of time. However, the increasing doubt about their integrity should motivate clinicians to reintroduce the usage of the serial microdilutions if routine lab testing and most importantly in breakpoint estimations, where any small inaccuracy will count.
Resistance is spreading. Antibacterial susceptibility breakpoints are an important tool for monitoring the spread of resistance. Good choice of the various susceptibility testing strategies along with harmonization of efforts, methods, and information is the only way to overcome the dilemma of antibacterials and resistance.
1. Turnidge J. (2007). Setting and Revising Antibacterial Breakpoints. Clinical Microbiology Reviews, 20(3), 391-408. http://www.cmr.asm.org
2. Gums J. (2003). NCCLS perspectives in changing susceptibility breakpoints for antimicrobial drugs. International Journal of Antimicrobial Agents, S3-S13. http://www.ischemo.org
3. Mouton J. (2002). Breakpoints; current practice and future perspectives. International Journal of Antimicrobial Agents, 323-331. http://www.ischemo.org
4. Schon T. (2009). Evaluation of wild-type MIC as a tool for determination of clinical breakpoints for Mycobacterium tuberculosis. Journal of Antimicrobial Chemotherapy, 64, 786-793. http://www.jac.org
5. Pea F. (2006). Which reliable pharmacodynamics breakpoint should be advised for ciprofloxacin monotherapy in hospital setting? A TDM-based retrospective perspective. Journal of Antimicrobial Chemotherapy, 58, 380-386. http://www.jac.org
6. Cuesta I. (2009). Data Mining Validation of Fluconazole Breakpoints Established by the European Committee on Antimicrobial Susceptibility Testing. Antimicrobial Agents and Chemotherapy, 53(7), 2949-2954. http://www.aac.asm.org
7. Potz N. (2004). Reliability of routine disk susceptibility testing by the British Society for Antimicrobial Chemotherapy method. Journal of Antimicrobial Chemotherapy, 53, 729-738. http://www.jac.org
8. Coyne R. (2004). On the validity of setting breakpoint minimum inhibition concentrations at one quarter of the plasma concentration achieved following oral administration of oxytetracycline. Aquaculture, 239, 23-35. http://www.elsevier.com/locate/aqua-online
9. Imohl M. (2009). New penicillin susceptibility breakpoints for Streptococcus pneumonia and their effects on susceptibility categorisation in Germany. International Journal of Antimicrobial Agents, 271-273. http://www.elsevier.com/locate/ijantimag
10. MacKenzie F. (2004). Reassessment of the cefaclor breakpoint for Streptococcus pneumonia. International Journal of Antimicrobial Agents, 337-342. http://www.ischemo.org
11. Kelly K. (1996). Twenty-Four-Hour Area under the Concentration-Time Curve/MIC Ratio as a Generic Predictor of Fluoroquinolone Antimicrobial Effect by Using Three Strains of Pseudomonas aeruginosa and an In Vitro Pharmacodynamic Model. Antimicrobial Agents and Chemotherapy, 40(3), 627-632. http://www.aac.asm.org
12. Andrews J. (1999). Study to assess the reliability of a disc diffusion method for determining the sensitivity of Gram-positive pathogens to dalfopristin/quinuprisitn. Journal of Antimicrobial Chemotherapy, 43, 141-143. http://www.jac.org
13. Firsov A. (2000). Gemifloxacin and ciprofloxacin pharmacodynamics in an in-vitro dynamic model: prediction of the equivalent AUC/MIC breakpoints and doses. International Journal of Antimicrobial Agents, 407-414. http://www.eschemo.org
14. Ferraro M. (2001). Should We Reevaluate breakpoints? Clinical Infectious Diseases, 33(Suppl 3), S227-S229. http://www.cid.org
15. Jorgensen J. (2004). Who Defines Resistance? The Clinical and economic Impact of Antimicrobial Testing Breakpoints. Seminars in Pediatric Infectious Diseases, 15(2), 105-108. http://www.elsevier.com
16. Brown D. (2009). Harmonization of antimicrobial testing breakpoints is Europe: implications for reporting intermediate susceptibility. Journal of Antimicrobial Chemotherapy. http://www.oxfordjournals.org
17. Perea E. (2001). Clinical relevance of laboratory susceptibility data. International Journal of Antimicrobial Agents, S29-S32. http://www.ischemo.org