Journal of Antimicrobial Chemotherapy (1990) 25, 175-181

Predictive value of susceptibility tests for the outcome of antibacterial therapy

Victor Lorian and Lillian Burns

From the Department of Epidemiology <t Infection Control, The Bronx Lebanon Hospital Center and Department of Laboratory Medicine, The Albert Einstein College of Medicine, 1650 Grand Concourse, Bronx, NY 10457, USA
A total of 510 charts of patients who received antibacterial agents were examined for clinical outcome and microbiology findings. A total of 382 patients (75%) had one or more specimens submitted for culture and susceptibility tests before the administration of the drugs; 298 (78%) of these had positive cultures and susceptibility tests were done. A total of 18 species were isolated. Of the 298 patients with organisms of known susceptibilities, 271 (91%) received antibacterial agents to which the respective organisms were susceptible and 219 of these patients (81%) improved (P < 0-05). This high rate of good infectious diseases practice is probably due to two factors: (1) susceptibility tests results were available in most cases the next day after the submission of a specimen; (2) the medical board distributed guidelines for the use of antibiotics and monitored the compliance closely. The patients treated with antibacterial agents to which the bacteria were resistant improved in 3% and did not improve in 82% (P < 0-05) of the patients. This study shows that choosing an antibacterial agent in accordance to the susceptibility test resulted in a high rate of improvement. When the choice of agent disregarded bacterial resistance in vitro, therapy almost always ended in failure. Therefore, susceptibility tests in vitro have a good predictive value for the outcome of antibacterial therapy.

Introduction

The antibiotic susceptibility of bacteria isolated from patients with infections is probably the most pertinent information provided by a medical microbiology laboratory influencing patient care. An enormous number of publications (over 5000 in the English language alone) during the last ten years have been dedicated to the performance of in-vitro susceptibility tests. In contrast, there is a scarcity of studies that report on the relevance of in-vitro bacterial susceptibility to the outcome of antibacterial therapy. The successful outcome of antibacterial therapy is conditioned by a complexity of factors other than the sensitivity of the organism to the antibiotics being used. The pharmaco-dynamics of the drug, the physical and chemical conditions at the site of infection, the permeability of the drug into the site of infection and, most of all, the level of the host defences and the nature of an underlying disease, are factors that significantly influence the outcome of anti-infectious therapy (McCabe & Treadwell, 1986). Nevertheless, most studies recognize that successful antibacterial therapy 175
0305-7453/90/010175 + 07 S0Z00/0 1990 The British Society for Antimicrobial Chemotherapy

176

V. Lorian and L. Burns

depends on the in-vitro susceptibility of the organism that causes the infection (Abboud & Waisbren, 1959; Berk & McCabe, 1980; Sanders & Sanders, 1982). The aim of the present study was to compare clinical data on the outcome of treatment of infection with laboratory data on in-vitro susceptibility of organisms isolated from the patients.
Materials and methods

Patients A total of 510 charts was examined. One chart in four was selected among consecutively admitted patients receiving antibacterial therapy who were hospitalized during the period from March through September 1986. Charts of paediatric patients, ICU patients, patients with AIDS and those treated for tuberculosis were excluded. The majority of the patients included in this study were admitted to the medical service (79%); the other patients were either admitted to the surgery or gynaecology services. Among the 510 charts examined, 77 patients (15%) received antibacterial agents prophylactically, before, during and after surgery. The remaining 433 patients studied received a mean of seven days (98% > 3 days) antibacterial therapy and 98% of patients received more than one drug, a practice aimed toward broader cover and the prevention of bacterial resistance. Dosage and frequency of administration were within the limits indicated in the Physicians' Desk Reference (1988). The diagnoses were bacteraemias (12%), pneumonias (12-7%), endocarditis (1-6%), pyelonephritis (1-6%), urinary infections (11-5%), pelvic inflammatory disease (6-9%), wound infections (22-6%), and others (total 16%, < 1% for each individual infection which included meningitis, peritonitis, pleural effusion or pancreatitis). Microbiology One or more specimens from the site of infection, if accessible, were submitted to the microbiology laboratory for microscopy and culture. Isolation and identification of the organisms were done by routine methods (Lennette et al., 1985). The in-vitro susceptibility tests were done by a modified Bauer-Kirby disc susceptibility test and results were recorded only as susceptible or resistant (Lorian, 1977). When two, or rarely three, aerobic organisms were isolated, the susceptibility of each strain was recorded. Anaerobic species were not tested. Pneumococci and Group A /?-hemolytic streptococci (when sole isolate) legionellae and chlamydia were also not tested and were assumed, and so recorded, to be susceptible to antibacterial agents to which they are naturally susceptible. While technical and clerical errors occur in a limited number in most clinical microbiology laboratories (Jones & Edson, 1982), it is assumed for the purpose of this study that the results of in-vitro tests were 100% accurate. Antibiotics used The antibiotics tested are: amikacin, ampicillin, aztreonam, cefotaxime, cefotetan, cefoxitin, ceftazidime, cephalothin, chloramphenicol, ciprofloxacin, clindamycin, erythromycin, gentamicin, imipenem, oxacillin, penicillin, piperacillin, trimethroprimsulphamethoxazole, tetracycline, ticarcillin and vancomycin.

Predictive rotae of susceptibility tests

177

Clinical outcome The outcome of antibacterial therapy was classified according to the physician's notes in the chart at the end of therapy. The outcomes were grouped into three categories: (a) improved; (b) not improved or, (c) equivocal. The criteria used to determine clinical improvement were the following: (1) negative culture after an initial positive culture; (2) improved or negative chest roentgenogram; (3) decrease from an initial WBC count of more than 12,000; (4) decrease in body temperature from at least 102 degrees; (5) healed or healing wound infection; (6) clinical improvement such as decrease in coughing, wheezing, decrease of local redness and swelling. Two or more of these criteria were required to classify the outcome of therapy as improved. The category 'not improved' included all patients who died during treatment. Statistical analysis In order to assess whether a proportion differed significantly from zero, exact 95% CI confidence interval using the binomial distribution were constructed about the proportion. Differences between two proportions were tested using a two-tailed Fisher Exact test. Differences in frequency distributions were tested using x1 statistics. All tests were performed using alpha = 0-05 (Snedecor and Cochran, 1967). Results A total of 382 patients (75%) of the 510 patients studied had one or more specimens submitted for culture (Table I). Specimens were not submitted for culture in 128 patients (25%). A total of 298 patients (78%) among those who had specimens submitted for culture had positive cultures; 84 patients (22%) had negative cultures. Two hundred and fifty different aerobic isolates were identified from the 298 cultures tested for susceptibility to antimicrobial drugs. The remaining 48 isolates consisted of species that were not tested for susceptibility such as chlamydia or pneumococci. A total of 18 different aerobic and anaerobic species was isolated. Each of the following species represented over 5% of all the aerobic isolates identified: Escherichia coli, Klebsiella pneumoniae. Staphylococcus aureus, Staph. epidermidis. Pseudomonas

Table I. Bacterial species which each accounted for at least 5% of culture-positive infections Species most frequently identified ( > 5%) number per cent Most frequent species from patients who improved number per cent 37 18 20 20 18 14 22 149 24-8 121 13-4 13-4 121 9-4 14-8 100-0 Most frequent species from patients who did not improve number per cent 11 5 10 6 13 0 6 51 21-5 9-8 19-6 11-8 25-5 0 11-85 100O

Species Esch. coli K. pneumoniae Staph. aureus Staph, epidermidis Ps. aeruginosa Pr. mirabilis Enterobacter spp.

48 23 30 26 31 14 28 200

24-0 11-5 15-0 13-0 15-5

7-0
14-0 10(H)

178

V. Lorian and L. Boras

Table II. Details of prescribing in relation to in-vitro susceptibility of the infecting bacteria Specimens submitted for culture No specimen submitted Total Culture done 382 culture positive culture negative susceptibility tested susceptibility not tested but known by species Total susceptibilities known by test or species In-vitro susceptibility tested or known by species 298 prescribed in accord with susceptibility prescribed not in accord with susceptibility Prescription in accord with in-vitro susceptibility 271 improved not improved equivocal Prescription not in accord with in-vitro susceptibility* 27 improved not improved equivocal Prescription empirical (negative cultures) 84 improved not improved equivocal Specimens not submitted for culture 128 prophylactic pelvic inflammatory disease pneumonias other 382 (75%) 128 (25%) 510 298 (78%) 84 (22%) 250 (84%) 48 (16%) 298 (78%)

271 (91%) 27 (9%)

219 (81%) 41 (15%) 11 (4%)

1 (3%) 22 (82%) 4 (15%)

22 (26%) 58 (69%) 4 (5%)

71 (56%) 30 (23%) 20 (16%) 7 (5%)

The organism(s) were resistant to the agents) prescribed.

Predkdre value of susceptibility tests

179

aeruginosa, Enterobacter spp. and Proteus mirabilis (Table II). There were 298 patients with infections where the bacterial species was identified and judged to be the agent causing the infection. The susceptibility of antibiotics was known for all organisms isolated; 271 patients (91%) received antibacterial agents to which the respective organisms were susceptible. Twenty seven patients (9%) were treated with antibacterial agents to which the respective strains were resistant. There were 84 patients who had negative cultures and who received antibacterial therapy empirically based on clinical symptoms. A high proportion (58 of 84, 69%) of these patients did not improve with antibacterial treatment. In all probability, the aetiological agent was not susceptible to the respective antibacterial agent. There were 298 patients who had at least one positive bacterial culture. Of these 271 were sensitive and 27 were resistant in vitro to the drug used to treat the infection. The proportion of patients who responded to treatment in these two groups was 219/271 (organism sensitive to antibacterial drug) and 1/27 (organism resistant to antibacterial drug). The difference between these proportions is statistically significant (j2 test with Yates' continuity correction = 71-6; P value with 1 degree of freedom < 0-001). One hundred and twenty-eight patients 25% of the 510 patients investigated did not have a specimen submitted for culture. This group contained 77 patients who received antibiotics prophylactically, 30 patients with pelvic inflammatory disease where the site of infection was inaccessible for culture, and 21 patients with pneumonia or upper respiratory infections where therapy had to be initiated on the basis of clinical findings alone. The microbiological profile of the most common species isolated from patients in whom antibacterial therapy was successful or in patients in whom antibacterial therapy failed is shown in Table II. Patients who either improved or who did not improve were compared with regard to the distributions of species found. The difference between the distributions was tested using a Chi-Square statistic, and the results indicated that the difference was not statistically significant {P = 0-09). The x2 has a value of 10-8 with a 6 degree of freeaom. The distributions of diagnoses were compared with regard to those patients who improved and those who did not improve. A multiple comparison procedure proposed by Brunden (1972) and Everitt (1977) was used to assess whether the proportions of patients with each of the six most common diseases differed significantly between the groups of patients who either improved or did not improve. Differences were not found to be significant (P > 0-05). Discussion A majority of patients had specimens submitted for culture and almost all antibiotics selected for therapy were in accord with available susceptibility data. This high rate of good infectious diseases practice is probably due to two factors: (1) susceptibility tests results were available in most cases the next day after the submission of a specimen; (2) during the last six years the medical board distributed guidelines for the use of antibiotics and monitored compliance closely. A majority of specimens that were submitted had positive cultures indicating proper collection, handling and processing of specimens. There was a variety of infections which established the relevance of the patients sampled as being representative of a patient population from a general community hospital.

180

V. Lorian and L. Buns

The seven most frequently isolated species in our study were found at 20-80% higher rate than those identified in a sample of 10 million strains isolated in American hospitals during 1971-82 (Atkinson & Lorian, 1984). The presence of Ps. aeruginosa is a cause of therapy failure regardless of whether the patient is treated according to the sensitivity data or not. The proportion of patients with Ps. aeruginosa among those who improved is notably different from the proportion among patients who did not improve (121% vs 25-5%). It is also notable that nine of the 13 patients with Ps. aeruginosa who did not improve were treated according to the susceptibility results. Similar disappointing results were obtained in another study of patients with Ps. aeruginosa bacteraemia (Freid & Vosti, 1968). A high percentage of infections due to Staph. epidermidis, observed in our study, 13% and 11-8%, has been described previously (Centres for Disease Control, 1986). In-vitro susceptibility tests are not always absolute guidelines for the use of antiinfectious drugs. Salmonella typhi and Nocardia sp. are usually susceptible in vitro to a wide range of agents. Only chloramphenicol, ampicillin or co-trimoxazole have been shown to be effective in the treatment of typhoid fever (DuPont & Pickering, 1980). Only sulphonamides have been shown to cure Nocardia infections (Palmer et al., 1974). The predictive value of in-vitro susceptibility tests in our study is good; when indicating susceptibility and acknowledged in prescriptions it resulted in a high rate of success; when indicating resistance and disregarded, it resulted in failure in 26 of 27 patients.

Acknowledgement The authors wish to thank Dr Katherine freeman for statistical analysis. References
Abboud, F. M. & Waisbren, B. A. (1959). Correlation between results of the tube dilution method for determining bacterial sensitivity to antibiotics and the results of the administration of these antibiotics to patients with staphylococcic bacteremia. Antibiotics Annual1958-9. 748-56. Atkinson, B. A. & Lorian, V. (1984). Antimicrobial agent susceptibility patterns of bacteria in hospitals from 1971 to 1982. Journal of Clinical Microbiology 20, 791-6. Berk, S. L. & McCabe, W. R. (1980). Meningitis caused by Gram-negative bacilli. Annals of Internal Medicine 93, 253-60.
Brunden, M. N. (1972). The analysis of non-independent 2 x 2 tables from 2 x C tables using rank sums. Biometrics 28, 603-7. Centers for Disease Control (USA). (1986). Nosocomial Infection Surveillance, 1984. Morbidity

and Mortality Weekly Report 35, Suppl. 1SS, 17SS-29SS. DuPont, H. L. & Pickering, L. K. (1980). Infections of the Gastrointestinal Tract: Microbiology, Pathophysiology, and Clinical Features. Plenum, New York. Everirt, B. S. (1977). 77K Analysis of Contingency Tables, p. 45. Chapman & Hall, London. Freid, M. A. & Vosti, K. L. (1968). The importance of underlying disease in patients with Gramnegative bacteremia. Archives of Infernal Medicine 121, 418-23. Jones, R. N. & Edson, D. C. (1982). Interlaboratory performance of disk agar diffusion and dilution antimicrobial susceptibility tests, 1979-1981. A summary of the microbiology portion of the College of American Pathologists (CAP) surveys. American Journal of Clinical Pathology 78. Suppl. 4, 651-8. Lennette, E. H., Balows, A., Hausler, W. J. & Shadomy, H. J., Eds. (1985) Manual of Clinical Microbiology, 4th edn. American Society for Microbiology, Washington, DC.

Predfctire ralne of cnsceptfbfllty tests

181

Loriaa, V. (1977). A five-hour disk antibiotic susceptibility test In Significance of Medical Microbiology in the Care of Patients (Lorian, V., Ed.), pp. 203-12. Williams & Wilkins, Baltimore, MD. McCabe, W. R. & TreadwelL T. L. (1986). In vitro susceptibility tests: correlations between sensitivity testing and clinical outcome in infected patients. In Antibiotics in Laboratory Medicine, 2nd edn. (Lorian, V., Ed.), pp. 925-37. Williams & Wilkins, Baltimore, MD. Palmer, D. L., Harvey, R. L. & Wheeler, J. K. (1974). Diagnostic and therapeutic considerations in Nocardia asteroides infection. Medicine 53, 391-401. Physicians' Desk Reference 42nd edn. (1988). Medical Economics Company, Inc, OradelL NJ. Sanders, W. E. & Sanders, C. C. (1982). Do in vitro antimicrobial susceptibility tests accurately predict therapeutic responsiveness in infected patients? In Significance of Medical Microbiology in the Care of Patients. (Lorian, V., Ed.), pp. 325-40. Williams & Wilkins, Baltimore, MD. Snedecor, G. W. & Cochran, W. G. (1967). Statistical Methods, 6th edn. Iowa State University Press, Ames, IA.
{Received 9 February 1989; revised version 30 August 1989)