Journal of Hospital infection (1991) 18 (Supplement B), 13-22

Computerized image analysis of full-hand touch plates: a method for quantification of surface

bacteria on hands and the effect of antimicrobial agents

J. J. Leyden, K. J. McGinley, M. S. Kaminer, J. Bakel, S. Nishijima*, M. J. Grove7 and G. L. Grove?

University of Pennsylvania, School of Medicine, Department of Dermatology, Philadelphia, PA 19104, “Department of Dermatology, Kansai Medical University, Moriguchi, Osaka, Japan and tSimon Greenberg Foundation, 835

Sussex Blvd., Broomall, PA 19008, USA

Summary: A method is described for quantification of the bacterial flora on the hand surface. Computer-assisted image analysis of bacterial growth of large full-hand touch plates provides a quantifiable measure of the bacterial flora on the hand surface. Image analysis pixel intensity values showed a significant correlation (P < 0.0001) with colony forming unit values deter- mined by the glove juice method. Image analysis of impressions from hands treated with various antimicrobial agents in detergent bases showed that 4% chlorhexidine gluconate produces a-96% reduction after a 30 s washing and 98% reduction after a 3 min washing while 7.5% povidoneiodine and 1% triclosan produce a 77% and 70% reduction after 3 min respectively, and 70% isopropanol produces a 98% reduction after a 30 s wash.

Keywords: Image analysis; hands; bacteria.

Introduction

Resident bacteria and transient pathogens on the hands of health care personnel have been recognized as potential sources of nosocomial infections. Frequent handwashing, usually with antimicrobial agents in detergent vehicles, is standard practice for health care personnel. Currently, the methodologies used to determine the efficacy of antimicrobial agents primarily involve variations of the ‘glove juice’ test in which the number of resident bacteria or inoculated pathogens are recovered by removal with a detergent inside a sterile glove or by utilization of touch plates methods before and after treatment with a test agent.14 Many investigators have demonstrated that antimicrobial agents in detergent vehicles reduced bacteria to a greater degree than a plain detergent. The magnitude of the effect on the resident bacteria is low, particularly in view of the demonstrated greater effect for these’agents on large populations of bacteria on other body sites.’ Results from testing the effect of chlorhexidine

0195%6701/91/06BO13+10 SOS.OO/O

13 0 1991 The Hospital Infection Smety

14 J. J. Leyden et al.

gluconate and povidone-iodine usually show approximately a one logarithm reduction after a 3 min scrubbing. Application to higher densities of bacteria on other body sites demonstrates reduction of several orders of magnitude.’ Recently, we demonstrated that the subungual spaces contain large populations of bacteria and that these bacteria contribute significantly to the number of organisms recovered by the glove juice test.6*7 For example, use of. 4% chlorhexidine and 10% povidone-iodine on hands in which the subungual spaces have been obliterated by cyanoacrylate glue results in up to a four logarithm reduction compared to hands in which the subungual spaces are not occluded which show only a one logarithm reduction.7 These studies demonstrate the need for a methodology other than the glove juice method for evaluating the degerming efficacy of an agent intended to rapidly reduce bacteria on the surface of the hand, e.g. the hand of a nurse or physician between patient-patient contact. The glove juice method seems more appropriate for evaluating the degerming effect for surgeons who have hands inside gloves for prolonged periods of time during which time subungual bacteria can find their way into the glove and potentially to the patient. Subungual bacteria are unlikely to be of significance in the casual, brief contact which commonly occurs between health care personnel and patients.

In this work, we describe a methodology for quantifying the effect of antimicrobial substances on the surface flora and demonstrate the relative efficacy of commonly used antimicrobial agents for both brief washing and standard preoperative surgical scrubbing.

Methods

Subjects Healthy adults who were not using any topical antimicrobial agent for at least 1 month served as test subjects after giving informed consent. Potential subjects were screened and only those showing a pattern of bacterial colonization of the hands and finger surface were used (Figure 1). Subjects with dry, chapped hands commonly have very low numbers of bacteria and are not acceptable for this procedure. In our studies, we avoided health care personnel because of their frequent use of antimicrobial agents. Our subjects were primarily university students and non-health care employees of the University.

Cultures and hand surface-hand stamping Cultures of the surface bacterial flora were obtained by pressing the palmar surface on 245 x 245 x 20 mm plates containing 200 ml of trypticase Soy agar and broth (BBL-Becton Dickinson, Cockysville, MD) which contained 2% polysorbate-80, 0.1% lecithin and 0.1% sodium thiosulphate as neutralizers, and 4% added sodium chloride to give a total of 4.5% to prevent spreading of aerobic spore-forming organisms and Gram-negative

Image analysis of hand bacteria

Figure 1. Large touch plate of hand demonstrating colony forming units as seen after being digitized by computer. Left panel shows pre-treatment and right panel shows a 65% reduction after treatment with triclosan.

bacteria, and 0.04% cycloheximide to prevent fungal contamination. The addition of 4% sodium chloride was done after preliminary studies demonstrated no difference in the density of colony forming units of bacteria from plates not containing sodium chloride. The suitability of the neutralizing agents was tested by washing hands with 4% chlorhexidine gluconate and immediately applying the hand to large touch plates seeded with a lawn of M~crococcus Zuteus (Figure 2). Effective neutralization was demonstrated for all antimicrobial test agents. The reproducibility of serial imprints of the hand was determined in 20 subjects who were sampled three

Figure 2. Left panel shows carry over of chlorhexidine inhibiting growth of Micrococcus luteus; right panel shows effective neutralization of potential carry over effect.

16 J. J. Leyden et al.

times in succession. In 60 subjects, the right and left hands were compared. In 20 subjects the composition and density of microorganisms on the

hands was determined by immersing each hand in a sterile glove containing 50 ml of 0.1% Tween 80 and scrubbing the palmar surface for 1 min. The subungual spaces were obliterated with cyanoacrylate glue in order to prevent sampling of that reservoir of organisms. The number of bacteria recovered by the glove juice procedure, expressed as log,, colony forming units (cfu) from these 20 subjects (40 hands), was compared to the image analysis data obtained from large hand touch plates of these subjects.

Image analysis The image analysis system consisted of a central computer (Dell Computer PCs Limited Systems 200), colour video monitor (Mitsubishi), video camera (Southern Micro Instruments, Inc.), and densitometry software package (Southern Micro Instruments, Inc.). Images of the full-hand touch plates (FHTP) were transmitted (digitized) onto a video monitor in black/white mode. The system employed grey scale values (GSV) from 0 to 255 (0 = black; 255 = white), with the range between black and white sub-divided based on relative shades of grey. Working area on the video screen was 466 X 466 pixels (which represented cm2 after calibration), generating 217 156 pixels per field, The system applied a GSV to each pixel. Maximum cumulative GSV for each field was 55.6 X 106, minimum was zero (black). Pseudocolour enhancement was then performed. In brief, the computer assigned each GSV a colour (0 = dark purple --, 255 = light pink) based on established relationships of the colour spectrum. The computer then changed the image on the video monitor so that the GSV were represented by colours rather than shades of grey. Pseudocolour served as a visual enhancement to aid in defining borders of objects of interest on the video screen.

The image of the FHTP was then re-acquired in the pseudocolour mode to allow for background calibration. Background of the FHTP was calibrated by adjusting the gain control (S/N ratio) on the video camera. To ensure reproducibility and allow comparison between subjects, FHTP were calibrated so that all colonies were between dark purple and light pink. The bacterial colonies that were closest to the white end of the grey scale in the black/white mode were calibrated to be light pink, taking care to eliminate white (in the pseudocolour mode) from the colonies of interest. This was done because white is assigned a value greater than 255, and represents the limit of the system’s ability to analyse grey scale variations. By calibrating the colonies so those that appeared closest to white were light pink in pseudocolour (as close to white as the software’s resolution allowed), we maximized the use of grey scale values between black and white. Background colour was normalized to a value of 0 intensity units (i.u.) by editing the pseudocolour spectrum.

Following background calibration, ‘area’ and ‘intensity’ calculations were

Image analysis of hand bacteria 17

IO 14.

9 12.

a

7 IO

6 - ‘;

8

: 5 ‘: 2

o 4 ” 6

3

2

I Ir - IILL 1 4

2

0 I -t 4 0 2000 4000 6000 81 0 10000 12000 14 00 0 b 3 2 4 5 =i--- 7

Arithmetic scale Logarithmic scale

Figure 3. Distribution of grey level pixel intensity values indicates that their values are log normally distributed.

performed. Area refers to the area of objects selected for data analysis, and intensity is the summation of the intensity value (pseudocolour correlate of GSV) for each pixel in the objects selected for data analysis. Objects of interest (bacterial colonies) were identified by selection of pseudocolours that correspond to the colonies on the plate. Pseudocolour selection allowed refinement of colony analysis by choosing only those pseudocolours that corresponded to bacterial colonies, so eliminating background contamination. Area and intensity measurements were displayed on the video monitor upon completion of this process.

Hand area for each subject was determined by outlining the hand on paper, acquiring the hand image onto the video screen, tracing the hand image on the video screen and calculating the area.

Effect of antimicrobial agents The effect of antimicrobial agents and a plain detergent were determined for both a 30 s handwashing and a 3 min scrub. In the former, hands were wetted with warm water, 2 ml of the test agent was added and the hands washed vigorously for 30 s. Hand scrubbing was done by wetting hands, applying 2 ml of the test agent then scrubbing the hands for 1 min with a sterile brush, Following this the debris from the subungual areas was removed by a technician and then 2 ml of the test agent was applied, the hands scrubbed for 1 min, rinsed and the process repeated again. Cultures of the hand were obtained prior to washing and after air drying of the hands following use of test agents. Carry over of antimicrobial agents could be demonstrated by scrubbing for 3 min, air drying for 5 min and then placing the hand on a plate seeded with M. Zuteus (A.T.C.C.27141). Plates containing the neutralizing agents described above did not show carry over effects for any of the antimicrobial agents tested. The test agents used were:

18 J. J. Leyden et al.

‘Hibiclens’ (Stuart Pharmaceuticals, Wilmington, DE)-4% chlorhexidine gluconate in detergent base;

‘Hibistat’ (Stuart Pharmaceuticals, Wilmington, DE)-0.5% chlorhexidine gluconate with 70% isopropyl alcohol;

‘SoftCide’ (Stahman, Weston & Co., Inc., N. Hampton, NH)-1% chloroxylenol (PCMX) in detergent base;

‘Phisoderm’ (Winthrop Consumer Products, New York, NY)-non-medicated detergent;

‘Betadine’ (Purdue Frederick Co., Norwalk, CT)-7.5% povidone-iodine surgical scrub;

70% Isopropyl alcohol; ‘Bat Down’ (Decon Laboratories Inc. Wayne Pa.) 1% Triclosan in

detergent base.

Results

The results of the qualitative and quantitative analysis of microorganisms recovered from the hand are found in Table I. The dominant organisms were staphylococci which composed 88.5% of the total flora and coryneforms which made up 11% of the total flora. Gram-negative bacteria and aerobic spore-forming organisms were frequently found (72% and 90% respectively) but were present in very low numbers. The use of 4.5% sodium chloride prevented swarming of these organisms and addition of 0.04% cycloheximide prevented fungal overgrowth, all of which would have inhibited image analysis.

No significant difference was found between the grey level intensity of colony forming units of bacteria on the right and the left hand of 60 subjects, a correlation coefficient of O-82, P < 0~0003. Serial sampling showed no significant differences in the area covered by bacteria on successive samples with logarithm means of 4.39 f 0.10, 4.44 f 0.13 and 4.38 f 0.12 log intensity cmp2. There was a significant correlation between the log cfu and the log pixel grey level intensity on image analysis (P < O*OOOl; Figure 4).

The effects of a 30 s washing and a 3 min scrubbing with various agents

Table I. Microflora on hand surface

Prevalence Composition (%) Mean ?c SD*

Staphylococci Coryneforms Gram negative rods Bacillus spp. Fungi (other than yeasts) Yeasts Others

100 88.5 5.1 f 0.5 100 11.0 3.8fl.O

72 0.1 0.7f0.9 90 0.1 1.4fl.O

z: 0.02 0.9 f 0.9 0.1 1.3*1.0

70 0.1 0.5 f0.9

*Logarithmic mean and standard deviation, post 5-hour soap wash (A’ = 40, left and right hands, 20 subjects).

Image analysis of hand bacteria 19

r2= 0.44, P=0.0001

0 0.5 I I.5 2 2.5 3 3-5 4 4.5

Log CFU cm-’

Figure 4. Correlation of pixel grey intensity (INT) cme2 with the colony forming unit density determined by the glove juice procedure.

Table II. Eficacy of antimicrobial detergents

Detergents 3 min scrub 30 s wash P value

4% Chlorhexidine gluconate 7.5% Povidone-iodine 1% Triclosan 1% Chloroxylenol Non-medicated

***97.9 f 2.3 ***92.6 2~ 6.3 NS ***76,9f 13.2 **so.2 * 25.2 ***

***70.1 -f 14.2 *48.3 rk 30.2 ***

32.9 AL 22.5 25.0f21.9 31~9zt30~1 21.9f25.4

Panel of 10 subjects; N = 20 for each time period. Results expressed as % reduction pre/post image analysis intensity cm-*. P value is for the difference between 30 s and 3 min washing. Asterisks indicate statistical difference between detergent base and test agents: *P i 0.05; **P < 0.01; ***P < O-001.

are seen in Table II. Washing with a plain detergent resulted in a non-significant reduction of 3 1.9% reduction after a 3 min scrub; povidone-iodine produced a 50.2% reduction after 30 s and 76.9% reduction after a 3 min scrub both of which were significant (P < 0.01 and P < 0.001, ANOVA), while chlorhexidine gluconate produced a 92.6% reduction after 30 s and 97.9% reduction after a 3 min scrub (P < 0.001). Both the 30 s and 3 min washings with 1% triclosan produced significant reductions (P < 0.05 and P < 0.001, ANOVA), while PCMX failed to produce a significant reduction at both time points. The use of 4% chlorhexidine gluconate, 7.5% povidone-iodine and 1% triclosan in detergent bases produced a significantly greater reduction in bacteria than use of the plain detergent alone after both 30 s and 3 min washings (P -=E 0.001).

Chlorhexidine used for 30 s produced a greater reduction than a 3 min scrub with povidone-iodine (P < 0.01) and triclosan (P < 0.001). Three min

20 J. J. Leyden et al.

Table III. Effect of antimicrobial solutions on surfaceflora of the hand

Test agent % Reduction on

30 s exposure (3 ml)

70% Isopropyl alcohol 0.5% Chlorhexidine and 70% isopropyl alcohol

98.7zk2.7

97.9 * 3.4

Panel of 10 subjects; N = 20 for each testing. Results expressed as % reduction of pre/post intensity cme2 as determined by image analysis of full hand touch plates.

scrubbing with chlorhexidine was more effective than 3 min scrubbing with povidone-iodine (P < 0.01) and triclosan (P < 0.001).

Isopropanol produced a 98.7% reduction after 30 s and 0.5% chlorhexidine in 70% isopropanol produced comparable results (Table III).

Discussion

In this study we have demonstrated that computer-assisted image analysis of large touch plates provides a reproducible, quantifiable measure of the surface bacterial population on the hand. This method provides results which correlate with recovery of bacteria by the glove juice method provided that the subungual regions are obliterated with cyanoacrylate glue to ensure that only bacteria from the hand surface are being harvested.

Using this methodology, we have found that 4% chlorhexidine gluconate in a detergent base produces a profound reduction both after a 30 s wash and a 3 min scrubbing. The level of reduction after a 3 min scrub was comparable to that found with 70% isopropanol applied for 30 s. Povidone-iodine and triclosan produced significant reductions at both time points while PCMX was no better than the detergent alone. In the case of topical 70% isopropanol we have found interesting differences in efficacy between use alone and following a pre-alcohol hand wash with a non-medicated detergent (unpublished observations). The latter is associated with a significant reduction in the effect of alcohol suggesting that deposition of fatty acids or other ingredients of the detergent interfere with the efficacy of alcohols. These interesting results indicate the need for further work in view of the use of alcohols following handwashing with plain detergents as preoperative degerming techniques in some countries.

The need to degerm the hand is most frequently expressed in terms of the surgeon whose hands are washed before operating and health care personnel who wash between handling of patients. The former is referred to as surgical handwashing where the aim is to reduce both the resident bacterial flora as well as any transient pathogenic organisms such as Staphylococcus aureus, Gram-negative bacteria and yeasts. The latter is usually referred to as hygienic handwashing in which the primary need is to eradicate transient pathogenic bacteria in order to minimize nosocomial infections. However,

Image analysis of hand bacteria 21

in hygienic handwashing it is also desirable to reduce the resident flora which can cause infection, particularly in the immunocompromised host. In the case of surgical handwashing, variations of the glove juice or sterile bag technique are commonly employed. Hygienic handwashing is frequently evaluated by determining the ability of a test agent to reduce pathogenic bacteria such as Escherichia coli or other Gram-negative bacteria including Acinetobacter spp. and Enterobacter spp. 2,3 All of these procedures provide relevant data on the in-vivo antimicrobial activity of agents designed to produce a degerming effect on hands. The methodology described in this work evolved from our observation that the glove juice, sterile bag technique samples both the bacterial flora of the surface of the hand as well as the subungual spaces. We demonstrated that the subungual spaces contribute significantly to the recovery of bacteria in the glove juice test and that currently available antimicrobial detergents fail to eradicate that population of bacteria. Furthermore, interpretation of results obtained by the glove juice method are complicated in terms of discerning the relative effects of a test agent on the surface flora and the subungual space.7

As a result of these studies, we feel that the evaluation of degerming agents on the hand flora should take into account the finding that the surface flora is much more accessible and more easily reduced than subungual bacteria. Health care personnel who move from patient to patient and have brief contact need to minimize transfer of bacteria from the hand surface. The need to minimize transfer of bacteria encompasses not only transient pathogens but also the recent bacterial flora. Surgeons or others performing longer procedures, particularly when operating within the body, need to be more concerned with the subungual bacterial population in addition to surface bacteria. This population can find its way into the glove and thus pose a potential hazard to the patient. The evaluation of the efficacy of an antimicrobial agent designed for use on the hands needs to be tailored to the setting in which that agent will be used. For health care personnel who have brief patient contact, evaluation of the effect of an agent on the surface flora and transient pathogens is critical while agents designed to be used by surgeons should be evaluated for their effort on both the surface and subungual flora.

We acknowledge the support of Simon Greenberg Foundation.

References

Michaud RN, McGrath MB, Goss WA. Applications of a gloved-hand model for multiparameter measurements of skin-degerming activity. J Clin Microbial 1976; 3: 406-423. Rotter M, Koller W, Wewalka G, Werner HP, Ayliffe GAJ, Babb JR. Evaluations of procedures for hygienic hand disinfection: controlled parallel experiments on the Vienna test model. J Hygiene (Cambridge) 1986; 96: 27-37. Ayliffe GAJ, Babb JR, Quoraishi AH. A test for hygienic hand disinfection. J Clin Pathol 1978; 31: 923-928.

22 J. J. Leyden et al.

4. Casewell MW, Law MM, Desai N. A laboratory model for testing agents for hygienic hand disinfection: handwashing and chlorhexidine for the removal of klebsiella. J Hosp Infect 1988; 12: 163-175.

5. Levden IT. Stewart R, Kligman AM. Updated in-vivo methods for evaluation of topical antimicrobial agents on human skin. J Ikest Dermatol 1979; 72: 165-170. -

6. McGinlev KT. Larson EL. Levden TT. Comnosition and densitv of microflora in the subungual q&e of the hand. J’Clin ikicrobioi1988; 26: 950-953.’

7. Leyden JJ, McGinley KJ, Kates SG, Myung KB. Contribution of subungual bacteria to the recovery of bacteria in the glove juice test. Infect Control Hasp Epidemiol 1989; 10: 451454.