The mechanism of epidural infection associated with epidural block is not clearly understood. Resident organisms in skin specimens were studied after skin was prepared with disinfectants.
Sixty-nine paired skin specimens were excised at incisional sites after skin disinfection with 10% povidone-iodine (10% PVP-I) or 0.5% chlorhexidine in 80% ethanol (0.5% CHE) from 60 patients having back surgery. One of the specimen pairs was placed in 10 ml brain-heart infusion broth and incubated in air at 37 degrees C for 96 h. The other specimen was sectioned at 3 microns and prepared with Gram's stain for examination with the microscope.
Thirteen gram-positive staphylococcal species (Staphylococcus epidermidis, 69.2%; S. hyicus, 15.4%; and S. capitis, 15.4%) were isolated from cultures. The isolates were found in a significantly greater proportion of the skin specimens disinfected with 10% PVP-I than in those disinfected with 0.5% CHE (11 of 34 cultures [32.4%] vs. 2 of 35 cultures [5.7%]; P < 0.01). Many gram-positive cocci were observed with the microscope in 4 (11.8%) and 5 (14.3%) of 34 and 35 skin specimens disinfected with 10% PVP-I and 0.5% CHE, respectively. The cocci formed a dense colony in each follicle and in the stratum corneum. No organism was present in any of 17,584 sweat glands examined.
In a large proportion of patients, isolation of viable organisms from excised skin specimens after disinfection with 10% PVP-I suggests that contamination of the epidural space by the skin flora may be a potential mechanism of epidural infection associated with epidural block.
Key words: Anesthetic technique: epidural. Complications: epidural abscess. Disinfectants: 10% povidone-iodine; 0.5% chlorhexidine in 80% ethanol. Bacteriology: Staphylococcus epidermidis; Staphylococcus aureus.
Epidural abscess is a serious complication of epidural block. [1-6]Proposed mechanisms of epidural abscess include the invasion by skin bacteria through a needle track, [2-4]contaminated syringes, contaminated local anesthetics, and hematogenous spread from a distant focus of infection. Detection of organisms on the skin surface after preparation with 10% povidone-iodine (PVP-I) [5,8,9]has suggested the possibility that the epidural needle may insert organisms already existing on the skin into the epidural space. [10,11]Although organisms were detected, colony counts were generally small in these studies [8,9]and these organisms were not frequently isolated from the epidural abscess, suggesting that the organisms detected on the skin surface after preparation with a disinfectant may not be the most common source of epidural infection. The mechanism of epidural infection has not been established in most reported cases of epidural abscess associated with epidural block. To understand the role of resident organisms in the deeper skin tissue in the pathogenesis of epidural abscess, we studied, by cultures and examination with the microscope, organisms in surgically excised skin specimens after the skin was prepared with 10% PVP-I or 0.5% chlorhexidine in 80% ethanol (0.5% CHE).
Materials and Methods
The study was approved by our institutional review board, and informed consent was obtained from each patient. We studied 60 patients classified as American Society of Anesthesiologists physical status 1 and 2 who were having elective laminectomy or other operations on the back. None of the patients had apparent skin infection in the area of operation. Under general anesthesia, the patient was placed in the prone position. The operative field and a wide area surrounding it were prepared by surgeons with three-layer applications of either 10% PVP-I (Isodine; Meiji-seika, Tokyo, Japan) or 0.5% CHE (Maskin; Maruishi Pharmaceutical Co. Ltd., Osaka, Japan). The field was then draped. Approximately 10 to 12 min later, at the start of operation, one or more skin specimens measuring 1 x 0.5 cm2were excised from the site of intended incision. The specimen was immediately transferred to a sterile plate and divided into two pieces for bacterial culture and for examination with the microscope, respectively.
One of the halves was immediately placed in 10 ml brain-heart infusion broth (BHIB) and was incubated in air at 37 degrees Celsius for 96 h. Specimens excised from the skin prepared with 10% PVP-1 (n = 34) were incubated in BHIB without neutralizers. Specimens excised from the skin prepared with 0.5% CHE (n = 35) were incubated in BHIB without neutralizers (n = 18) or in BHIB containing neutralizers of chlorhexidine; that is, 5% Tween-80 (n = 7) or a combination of 1% lubrol, 0.5% lecithin, and 1% Tween-80 (n = 10). Organisms in the broth that became turbid at the end of the incubation period were identified using Gram's stain, coagulase test, catalase test, and identification kits (API Staph and API Strept; bioMerieux sa, Marcey-l'Etoile, France). Other specimens were placed promptly in 20% formalin and kept at room temperature for 48 h for study with the microscope. Tissues were washed in water for several hours and embedded in paraffin. Each of five sequentially obtained 3-micro meter histologic sections was mounted onto a glass slide prepared with albumin. The slides were warmed to facilitate adherence of the sections. Deparaffinization and hydration treatment of tissue sections were facilitated by xylene, absolute ethanol, 95% ethanol, and water. Gram staining was performed according to the method of Brown and Hopps. 
Inocula from bacterial isolates in this study and three strains of Staphylococcus aureus isolated in the past at our hospital were incubated in 5 ml BHIB overnight at 37 degrees Celsius. The suspension containing 5 x 107to 1.5 x 109colony-forming units (CFU) per milliliter was used as a test inoculum to study the bactericidal effects of disinfectants.
The disinfectant adhered on skin specimens (approximate surface area, 25 mm2) may inhibit bacterial growth in BHIB. The effect of disinfectants over skin specimens on bacterial growth in BHIB was assessed using a sterile vinyl chloride tube with an outer diameter of 3 mm soaked in either 10% PVP-I or 0.5% CHE, dried at room temperature, and cut into 3-mm lengths. The resulting pieces had an approximate surface area of 28 mm2. The tubes soaked in 0.5% CHE were placed in 10 ml BHIB containing no neutralizer or 10 ml BHIB containing either 5% Tween-80 or a combination of 1% lubrol, 0.5% lecithin, and 1.0% Tween-80. The tubes soaked in 10% PVP-I were placed in 10 ml BHIB with or without the neutralizer, 0.5% sodium thiosulfate. The 10-micro liter test inoculum from bacterial isolates from skin specimens and three previous isolates of S. aureus was added to 10 ml of the BHIB with or without neutralizers, and the suspension was incubated at 37 degrees Celsius for 96 h for colony counts.
Bactericidal activity of 10% PVP-I and 0.5% CHE was studied by adding a 10-micro liter test inoculum from each of the isolates to 5 ml of each disinfectant solution. The suspension was vortexed for 5 s. Control suspensions for each of the strains of bacteria were prepared by adding the inoculum to 5 ml distilled water. After the inocula were exposed to the disinfectants for 15, 30, 60, 120, and 240 s at room temperature (23 degrees Celsius), the antimicrobial activity of 10% PVP-I and 0.5% CHE in 10 micro liter of the suspension was inactivated by diluting the suspension with 10 ml of 0.5% sodium thiosulphate and 5% Tween-80, respectively. One hundred microliters of the resulting suspensions were transferred to nutrient agar plates in quadruplicate and were incubated at 37 degrees Celsius for 72 h for colony counts.
To test the validity of inactivation of antimicrobial activity of disinfectants by 1:1,000 dilution with neutralizers, 10 micro liter of the test inoculum were added to 10 ml of a mixture consisting of 10 micro liter of a disinfectant and 10 ml of a respective neutralizer. For control, the test inoculum was added to 10 ml of distilled water. After the inocula were exposed to the disinfectant-neutralizer mixture for 30 min at room temperature, 100 micro liter of these suspensions were transferred to agar plates in quadruplicate and incubated at 37 degrees Celsius for 72 h for colony counts. Antimicrobial activity was considered to be inactive when the decrease in the colony count, compared with the control, was less than 5%. When the neutralizers failed to inactivate the antimicrobial activity completely, the ratio of the difference in colony count in the control cultures and that in the cultures taken from the disinfectant-neutralizer mixture to the control count was calculated as an index of inactivation.
Bactericidal activity of the disinfectants was presented as percentage changes from controls. The differences in percentage changes in colony count between the disinfectants and the differences in survival rate between S. aureus and other staphylococcal species after exposure to 10% PVP-I at each period of exposure were both analyzed by one sample test of proportions. Differences in the incidence of positive culture from the skin specimens and in the incidence of organism-positive hair follicles and sweat glands examined with the microscopic were assessed using the chi-squared test. Probability values less than 0.05 were considered statistically significant.
Thirteen strains of organisms were isolated from positive cultures. The organisms were all gram-positive staphylococcal species; that is, Staphylococcus epidermidis (69.2%), S. hyicus (15.4%), and S. capitis (15.4%). The bacteria grew in a significantly higher proportion of the skin specimens that were disinfected with 10% PVP-I than in those disinfected with 0.5% CHE (10% PVP-I, 11 of 34 cultures [32.4%] vs. 0.5% CHE; 2 of 35 cultures [5.7%]; P < 0.01).
Examination of 345 sections with the microscope showed many organisms in 4 (11.8%) and 5 (14.3%) of 34 and 35 skin specimens disinfected with 10% PVP-I and 0.5% CHE, respectively. The organisms were all gram-positive cocci. They were found in 14 (3.7%) and 17 (2.3%) of 379 and 734 hair follicles of the specimens disinfected with 10% PVP-I and 0.5% CHE, respectively. The cocci formed a dense colony in each follicle and horny layer (Figure 1). No organisms was found in any of 17,584 sweat glands studied.
All 14 strains of organisms grew when a 3-mm vinyl chloride tube soaked in 10% PVP-I was added to the mixture of a 10-micro liter inoculum and 10 ml of BHIB without the neutralizer. With the addition of a vinyl chloride tube soaked in 0.5% CHE, bacterial growth was suppressed in 4 of the 14 strains of organisms without neutralizers. All 14 strains of organisms grew when neutralizers of disinfectants were present in BHIB (Table 1).
Bactericidal activity of 0.5% CHE was seen significantly earlier and was more potent than that of 10% PVP-I against all 14 strains of staphylococcus species we studied (P < 0.0001 at 15-, 30-, 60-, and 120-s exposures; Table 2). No strain of organism grew after a 15-s exposure to 0.5% CHE. Whereas exposure of the organisms to 10% PVP-I for 15, 30, 60, and 120 s reduced the colony counts by 55.8%, 82.1%, 94.4%, and 98.5%, respectively. The three strains of previously isolated S. aureus were significantly more resistant to 10% PVP-I than were the other staphylococcal species isolated from the skin specimens (P < 0.0001 for 15-, 30-, 60-, and 120-s exposures).
The 1:1,000 dilution of 10% PVP-I with 0.5% sodium thiosulphate and 0.5% CHE in ethanol with 5% Tween-80 inactivated the antimicrobial activity of these disinfectants by 91.7 +/- 16.6% and 83.5 +/- 13.6%, respectively (Table 3). Each control colony count in Table 3was adjusted accordingly.
Surgical infections were observed in 3 of 34 patients whose operative field was disinfected with 10% PVP-I. Two were wound infections with dehiscence that developed on the eighth and ninth postoperative days. S. epidermidis was isolated from one wound and no organism was isolated from the other. The third infection was meningitis that developed on the eighth postoperative day. Acinetobacter baumannii was isolated from the cerebrospinal fluid. No surgical infection developed in 35 patients whose operative fields were disinfected with 0.5% CHE. The difference was not statistically significant (according to Fisher's exact test).
Bactericidal activity of 0.5% CHE in 80% ethanol was significantly more potent than that of 10% PVP-I against all 11 strains of S. epidermidis and S. hyicus isolated from the skin specimens and three strains of S. aureus isolated at our hospital in the past. Many strains grew colonies even after 120-s exposure to 10% PVP-I, but no organisms grew in any of the strains after 15-s exposure to 0.5% CHE in 80% ethanol. The results in this study were identical to those in our recent report in which the bactericidal effects of 0.5% CHE, 0.5% CHE in 80% ethanol, 10% PVP-I, and 80% ethanol on six strains of S. aureus were compared. The results suggest that the potent bactericidal activity of 0.5% CHE in 80% ethanol was due to its ethanol content. 
Cultures of the skin specimens, although the exposure time to disinfectants was 10 min or longer, grew colonies of gram-positive staphylococcal species in a surprisingly large proportion (32%) of the specimens after disinfection with 10% PVP-I. On the other hand, organisms grew in a significantly smaller proportion (6%) of specimens after disinfection with 0.5% CHE in 80% ethanol. However, 0.5% CHE in 80% ethanol adhered on vinyl chloride tubes with surface areas similar to the skin specimens suppressed bacterial growth in approximately 30% of the isolates. These findings thus suggest that 0.5% CHE in 80% ethanol on the specimens may have had a similar suppressive effect on the 18 BHIB cultures that did not contain disinfectant neutralizers, increasing the detection rate from 6% to approximately 7%. The difference in the detection rate, however, remained large and significant.
Resident bacteria hidden in hair follicles and orifices of sebaceous glands may be difficult to eradicate. Selwyn and Ellis, in a study of cultures of skin specimens excised from cadavers and patients, showed that disinfection of the skin with 1.5% iodine in 70% ethanol or with 0.5% CHE in 70% ethanol reduced colony counts by 95.5% and 87.6%, respectively, and showed, by examination with the microscope, the presence of bacteria deep in the larger hair follicles. Many bacteria were viable and seen exuding from the orifices of follicles after incubation. More superficially located bacteria formed many colonies on incubated sections of the specimens even after strict disinfection. They suggested that resident organisms are protected from disinfectants by lipids at the orifices of the follicles or overlying portions of the stratum corneum. Subsequently, Zamora and colleagues showed that the presence of organic substances in the skin decrease the bactericidal activity of povidone-iodine markedly. On the other hand, the detection rate of organisms in cultures of the skin surface has been reported to be low. [8,9,16]Thus the disparity in detection rate between skin-surface sampling methods and skin biopsy methods appears to be due largely to the difference in accessibility of these methods to resident organisms.
With examination with the microscope, we found many dense colonies of gram-positive cocci in the hair follicles and the overlying portions of the stratum corneum in a substantial number of specimens. The detection rate was similar regardless of the disinfectant used (12% and 14%). However, the size of specimens used for examination with the microscope was much smaller, approximately 1/300 th the size of specimens used for culture study. Therefore the real detection rate of organisms using the microscope in specimens with an equivalent size would have been much higher. The organism detection rate by culture of skin specimens after disinfection by 0.5% chlorhexidine in 80% ethanol in our study was similar to that found after skin preparation by alcoholic disinfectants in the study by Selwyn and Ellis. A significantly higher isolation incidence, in our study, of gram-positive staphylococcal species after disinfection by 10% PVP-I suggests that a large proportion of the organisms may have derived from resident organisms. The markedly lower detection rate after disinfection with 0.5% CHE in 80% ethanol may have been due to its more potent bactericidal activity and, possibly, high permeability into the hair follicles and the stratum corneum due to its alcohol content.
The most frequently detected organism in the normal skin flora is S. epidermidis (65% to 69%), whereas the prevalence of S. aureus ranges from only 1% to 2%. The detection rates after skin preparation of S. epidermidis (69.2%) and S. aureus (0%) in this study were consistent with previous findings. A recent review of 16 published epidural abscesses has shown that S. aureus was identified in 9 (82%) and S. epidermidis in 2 of 11 (18%) cases in which causative organisms were isolated. Seven of these patients had either diabetes with renal failure, cancer, steroid medication for rheumatoid arthritis, or herpes zoster, suggesting that impaired host immune activity may be an important risk factor. The mechanism of the infection was identified in only 3 of these 16 cases: hematogenous spread, contaminated drug vials, and an infection possibly transmitted by the anesthesiologist. This discrepancy between the low prevalence of S. aureus in the human skin flora and its high isolation rate in the abscesses also suggests that S. aureus may be more resistant to bactericidal effect of disinfectant than are other skin flora. Three strains of S. aureus tested in this study were significantly more resistant to the bactericidal effect of 10% PVP-I than were the isolates of S. epidermidis.
Demonstration with the microscope of a high prevalence of resident organisms and viable organisms cultured in a large proportion of the specimens suggests the possibility that epidural space contamination by the skin flora during insertion of an epidural needle or by subsequent spread along the epidural catheter may be a mechanism of epidural infection. Isolation of S. epidermidis from epidural abscess, detection of the skin flora after skin preparation, [5,8,9]and the lack of identification of the mechanism in most reported cases all appear to support this possibility. The low prevalence of S. aureus in the skin flora, host immune competence, the antimicrobial property of local anesthetics, and concomitant use of antibiotics may explain the apparently low incidence of reported epidural abscess. 
The concentrations of bacteria used in this study (105CFU/ml to 3 x 106CFU/ml) may have been higher than those clinically encountered. The number of organisms in the normal skin flora varies widely. Bacterial counts in excised skin specimens may range from 4,400/cm sup 2 in the skin over the breast to 395,000/cm2in the skin in the axilla. Disinfectants must be effective in a broad range of bacterial concentrations. Higher levels of bacterial concentration have been used frequently to study disinfectants. 
Our results show that many resident organisms are present in the hair follicles, and viable organisms can be isolated from skin specimens in substantial proportions after strict preparations with disinfectants. Although 0.5% CHE in 80% ethanol was significantly more potent than 10% PVP-I, staphylococcal species grew after preparation of the skin with both disinfectants. These results suggest that contamination of the epidural space by the skin flora may be a mechanism of epidural infection.
The authors thank Professor A. Nagayama, Department of Microbiology; Professor M. Kikuchi, The First Department of Pathology; and N. Murakami and the staff of the microbiology and pathology laboratories, Fukuoka University, for their assistance in this study. They thank Dr. K. Tsueda, Department of Anesthesiology, School of Medicine, University of Louisville, for help with the revision, Dr. R. L. Vogel for performing the statistical analysis, and Mrs. P. Bensinger for preparing the manuscript.