556
Prophylaxis against Wound Infection following Herniorrhaphy or Breast Surgery R. PiaU, J. R. Zucker, D. F. Zaleznik, C. C. Hopkins, E. P. Dellinger, A. W. Karchmer, C. S. Bryan, J. F. Burke, M. A. Wikler, S. K. Marino, K. F. Holbrook, and T. D. Tosteson
Departments of Medicine and Surgery. Harvard Medical School. Departments of Medicine and Infection Control Units. Channing Laboratory. Brigham and Women's Hospital. Beth Israel Hospital. and New England Deaconess Hospital. and Departments of Medicine and Surgery and Infection Control Unit. Massachusetts General Hospital. Boston; Department ofSurgery. University of Washington. and Harborview Medical Center. Seattle; Department ofMedicine. University ofSouth Carolina. Columbia; Anti-Infective Division. Smith Kline and French Laboratories. Philadelphia. Pennsylvania
The effect of perioperative antibiotic prophylaxis on definite wound infections was assessed for 3202 herniorrhaphies or selected breast surgery procedures. Patients were identified preoperatively and monitored for ~4 weeks. Thirty-four percent of patients (1077/3202) received prophylaxis at the discretion of the surgeon; 86 definite wound infections (2.7%) were identified. Pro-
Perioperative antibiotic prophylaxis prevents postoperative wound infections following many types of surgical procedures [1,2]. Most studies have investigated the role of prophylaxis in one of two types of procedures-those with intrinsically high infection rates such as colorectal surgery, and those with lower rates but for which the consequences of infection are often catastrophic, such as implantation ofprosthetic material. There is relatively little information on the use of prophylaxis for "clean" surgery that involves soft tissue only and does not involve contamination of the operative site by the flora of the host's viscera or mucous membranes, even though these are common procedures. Recent evidence indicates that prophylaxis prevents ,...., 50% of all postoperative infections, taken together, after elective herniorrhaphy and selected types of breast surgery [3]. However, no extant study was designed to demonstrate the effectiveness of prophylaxis for wound infection alone following
Received 22 November 1991; revised 27 March 1992. Presented: annual meeting of the American Epidemiological Society. Ann Arbor. March 1992. The study was approved by the institutional review boards at each participating institution. Grant support: Smith Kline and French Laboratories. M.A.W. is a stockholder in and an employee of Smith Kline Beecham. Reprints or correspondence: Dr. R. Platt. Channing Laboratory. 180 Longwood Ave.• Boston. MA 02115.
The Journal of Infectious Diseases 1992;166:556-60 © 1992 by The University of Chicago. All rights reserved. 0022-1899/92/6603-0015$01.00
these procedures. Assessment of this question requires data from several thousand patients. We addressed this issue by studying the effect of prophylaxis against definite wound infections in a prospective validation cohort comprising patients who underwent these same procedures but who did not participate in the randomized clinical trial. Some of these patients received prophylaxis at their surgeon's discretion.
Methods Patient population. This study was done in conjunction with a randomized clinical trial of perioperative prophylaxis [3]. We attempted to identify all patients undergoing eligible procedures at participating institutions. This report includes information from the patients who did not participate in the randomized trial; some patients were ineligible because of reported allergy to a {Hactam antibiotic. Most were eligible but did not participate in the randomized trial because the study nurse was unable to talk with the patient or because the patient or the patient's surgeon refused to participate. Patients included were ~ 18 years old and underwent elective inguinal or femoral herniorrhaphy, mastectomy, lumpectomy, excisional breast biopsy, axillary node dissection for breast cancer, or reduction mammoplasty between 1 April 1985 and 30 September 1987. Participating hospitals were Beth Israel Hospital, Brigham and Women's Hospital, Massachusetts General Hospital, and New England Deaconess Hospital (Boston); Harborview Medical Center, University of Washington Hospital, and Pacific Medical Center
Downloaded from http://jid.oxfordjournals.org/ at Emory University on August 2, 2015
phylaxis recipients were at higher risk for infection, with a higher proportion of mastectomies, longer procedures, and other factors. Patients who received prophylaxis experienced 41% fewer definite wound infections (odds ratio [OR], 0.59; 95% confidence interval [Cl], 0.35-0.99; P = .04) and 65% fewer definite wound infections requiring parenteral antibiotic therapy (OR, 0.35; 95% CI, 0.15-0.88; P = .02) after adjustment for duration of surgery and type of procedure. Additional adjustment for age, body mass index, the presence of drains, diabetes, and exposure to corticosteroids did not change the magnitude of this effect meaningfully. The effect of prophylaxis was similar for all procedures studied. In the absence of formal guidelines, surgeons at these institutions administered prophylaxis preferentially to patients at highest risk.
JIO 1992; 166 (September)
Prophylaxis against Wound Infection
Results Patient population. We identified 3221 patients who underwent an eligible procedure; 19 were excluded because no information was available after the day of surgery. Thus, 3202 patients were included in this analysis. The status was known for 96% of patients at least 6 days after surgery, for 91%at 14 days, and for 87% at > 28 days after surgery. Sixtytwo percent of patients (1981) underwent one of the breast procedures and the remaining 38% (1221) underwent a herniorrhaphy. The median duration of surgery was 60 min for lumpectomy, 68 min for herniorrhaphy, 115 min for modified radical mastectomy, and 190 min for reduction mammoplasty. Other characteristics of the study population are shown in table 1. Use ofperioperative prophylaxis. Of 3202 patients, 1077 (34%) received parenteral antibiotic prophylaxis. For those, the median interval between administration of the antibiotic and start of surgery was 20 min, with 10%of patients receiving antibiotic at least 15 min after incision, Cefazolin was the most commonly used agent, accounting for 796 courses (74%). In addition, 77 patients received another cephalosporin, 103 received another ,B-Iactam antibiotic, 41 received clindamycin, and 21 received vancomycin. No other agent was used for> 10 individuals. The use of prophylaxis varied considerably according to procedure and other characteristics (table I). It was used least often for herniorrhaphy, in 239 (20%) of 1221 patients; it was used most often for reduction mammoplasty, in 198 (85%) of 233 patients. The duration of surgical procedures was also longer for prophylaxis recipients; their median duration was 106 min, compared with 73 min for those without prophylaxis (P < .001). Infections. Eighty-six wound infections (2.7%) were identified, ofwhich 71 were observed to have purulent drainage. Twenty-nine of these prompted readmission and several others extended the original hospitalization. Forty-four infections were treated with incision and drainage or debridement, and 34 patients were treated with parenteral antibiotics. Staphylococcus aureus was identified in 37 of the 46 cases for which a pathogen was noted. Prophylaxis and infection. Because the patients who received prophylaxis differed from those who did not in several important respects, we adjusted for these differences to assess the effect of prophylaxis. The relationship between prophylaxis and infection is shown in figure I. There was important confounding by several factors. The unadjusted OR for infection comparing prophylaxis recipients with nonrecipients was 0.85 (26/1077 vs. 60/2125). After adjustment for procedure duration and type, the OR was 0.59 (95% confidence interval [Cl], 0.35-0.99; P = .04). The adjusted risks of the various procedures and time are shown in table 2, and the predicted risks of infection are illustrated in figure 2. There was no significant interaction between antibiotic prophylaxis
Downloaded from http://jid.oxfordjournals.org/ at Emory University on August 2, 2015
(Seattle); and Baptist Medical Center and VA Hospital (Columbia, SC). Hospitals initiated surveillance at different times and suspended surveillance during brief periods when study personnel were unavailable (e.g., during vacations). Some centers did not include patients undergoing ambulatory surgery because of logistic constraints. Patients were excluded if they had an active infection at the time ofsurgery, ifthey had received an antibiotic during the week before surgery, if prosthetic material was implanted during surgery, or if they underwent a concomitant surgical procedure that involved a separate skin incision. Patients received prophylaxis at the discretion of their attending physician, with an agent of the physician's choosing. There were no formal recommendations regarding the use of prophylaxis at any of the institutions during this period. Definitions. Definite wound infection, referred to subsequently as wound infection, was defined as a wound with erythema plus drainage, a wound with purulent drainage, or a wound that was opened and not reclosed for treatment of presumed infection. Erythema or a physician's diagnosis alone was not sufficient to establish a diagnosis. Perioperative antibiotic prophylaxis, referred to subsequently as prophylaxis, was defined as parenteral antibiotic administered any time on the day of surgery before the end of the procedure, as determined by review of operating room records. Identification of postoperative infection. A member of the study team assessed each patient's status on each weekday during hospitalization. Information was obtained by direct observation of the wounds when this was possible, from relevant hospital personnel (nurse, attending physician, house staff), or from the medical record. A questionnaire was administered by telephone twice to all patients, once 6-15 days after surgery and again 4-6 weeks later. The questionnaire elicited information about difficulties with wound healing, new infections, new antibiotic courses, nonroutine visits to the physician, and hospitalization. Information suggesting the presence of infection was verified by reference to available medical records. When patients could not be contacted, the physician's office notes were used. All cases that were provisionally identified by the study teams plus those that included erythema or drainage or received a physician's diagnosis of infection were reviewed and formally classified by a committee composed of all investigators and research nurses. The patient's prophylaxis status was not made available to this group, although the individual patient's investigators did have access to this information. Statistical analysis. The Wilcoxon rank sum test was used to compare the distribution of continuous variables between prophylaxis recipients and nonrecipients (SAS version 6.03; SAS Institute, Cary, NC). Infections in patients who received prophylaxis were compared with those in patients who did not by computing odds ratios (OR) for infection. Potential confounders were adjusted for by logistic regression (BMDP Statistical Software, 1990 revision; University of California Press, Berkeley). ORs were computed in all cases to permit direct comparison of the results of logistic regression analyses to the unadjusted analyses.
557
Platt et al.
558
Table 1. Characteristics of patients according to exposure to prophylaxis.
Characteristic
;;.2
General anesthesia Diabetes mellitus Preoperative disinfectant for skin site Iodine-containing compound Soap or alcohol Other or unknown Corticosteroids Duration of surgery (min) Age (years) Body mass index (kg/rn'') Prophylaxis-to-surgery interval (min) Follow-up interval (days)
Nothing
No prophylaxis n = 2125
Prophylaxis n = 1077
Total n = 3202
42 (2) 547 (26) 83 (4)
15 (I) 150 (14) 65 (6)
57 (2) 697 (22) 148 (5)
402 (19) 35 (2)
362 (34) 13 (I) 198(18)
764 (24) 19 (I) 233 (7)
28 (I) 982 (46)
35 (3) 239 (22)
63 (2) 1221 (38)
6 (0)
Adjusted for:
Procedure Duration
Procedure Duration and Procedure Type I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
.2
.4
.6
.8
1.0
1.2
1.4 1.6
1.8
ODDS RATIO 405 (19) 117 (6) 2 (0)
200 (19) 240 (22) 28 (3)
605 (19) 357 (II) 30 (I)
1413 (66) 449(21) 263 (l2) 1641 (77) 85 (4)
311 356 410 960 56
(29) (33) (38) (89) (5)
1724(54) 805 (25) 673(21) 2601(81) 141 (4)
1613 463 49 30
716 (66) 334(31) 27 (3) 24 (2)
2329 (73) 797 (25) 76 (2) 54 (2)
(76) (22) (2) (I)
73 56 24.2
106 55 24.6
83 56 24.3
NA 38
22 40
NA 39
NOTE. Data are no. of patients (%) unless stated otherwise; values for continuous variables are medians. NA, measure does not apply for procedures without prophylaxis. * Includes inguinal and femoral herniorrhaphies plus inguinal canallipectomy done for suspected inguinal hernia.
and procedure duration or between prophylaxis and procedure type. The actual occurrence of wound infection among those with and without prophylaxis by procedure type and duration of surgery is shown in table 3. Further adjustment for age, diabetes, body mass index, and use of incisional drains did not meaningfully change this ratio (OR, 0.57; 95% CI, 0.34-0.97; P = .03). The effect of prophylaxis did not vary significantly for the different procedures. For modified radical mastectomies, the adjusted OR for infection among prophylaxis recipients was 0.67, for lumpectomies it was 0.61, and for herniorrhaphies it was 0.54. Antibiotic prophylaxis was also associated with signifi-
Figure 1. Prophylactic effect before and after adjustment for duration of procedure and type of surgery. Odds ratio (OR) falls from 0.88 (unadjusted) to 0.66 (adjusted for procedure duration) to 0.59 (adjusted for both). Unadjusted OR is 0.88 rather than 0.85 as noted in text because 37 cases, including 2 infections that had partially missing data, were excluded from logistic regression analysis.
Table 2. Adjusted risks ofinfection in patients who did or did not receive prophylaxis before breast surgery.
Prophylaxis vs. no prophylaxis Procedure* Local excision of breast lesion Lumpectomy Simple mastectomy Modified radical mastectomy Radical mastectomy Reduction mammoplasty Axillary node dissection Procedure duration (min)
OR
95% confidence interval
0.59
0.35-0.99
0 2.09 2.71
0 1.07-4.09 0.99-7.36
3.03 4.40 1.69 4.10 1.004
1.60-5.75 0.54-35.8 0.58-4.93 1.15-14.6 1.00-1.01
NOTE. Values are taken from logistic regression model that included variables shown plus additional terms for 23 femoral herniorrhaphies and I inguinal canal lipectomy. Constant for this model was -4.478. Coefficients for variable shown are natural logarithms of odds ratios (OR) shown. * OR are compared with those of inguinal herniorrhaphy. t OR is increment/min of procedure duration.
cantly lower adjusted risks of wound infection with purulent drainage (OR, 0.57; 95% CI, 0.32-1.00; P = .05) and of subsequent parenteral antibiotic treatment of wound infection (OR, 0.35; 95% CI, 0.14-0.88; P = .02). There was a nonsignificantly lower risk of wound infection with incision and drainage (OR, 0.66; 95% CI, 0.30-1.46; P = .3) and of
Downloaded from http://jid.oxfordjournals.org/ at Emory University on August 2, 2015
Procedure Local excision Lumpectomy Simple mastectomy Modified radical mastectomy Radical mastectomy Reduction mammoplasty Axillary node dissection alone Herniorrhaphy* Axillary node dissection (with or without other procedure) Bilateral procedure Flap reconstruction Drains at operative site None I
lID 1992; 166 (September)
JID 1992; 166 (September)
30
Prophylaxis against Wound Infection
60 1~0 Duration of procedure (minutes)
1~0
readmission for wound infection (OR, 0.82; 95% CI, 0.351.96; P = .7).
Discussion Several findings emerged from these analyses. Prophylaxis was used relatively often, even though there were no formal guidelines recommending it for any of these procedures and no compelling published data on its efficacy when these procedures were done. Surgeons used prophylaxis more often in cases with commonly accepted risk factors, such as more extensive surgery or longer operative times. The use of prophylaxis was associated with a statistically significant 41 % reduction in the adjusted risk of wound infection and a 65% reduction in the adjusted risk of parenteral antibiotic therapy for treatment of these infections. The reduction in the risk of wound infection was consistent with the 48% reduction that we observed in the concur-
rent randomized clinical trial done in the same institutions [3]. This finding indicates that after appropriate adjustment for confounding, an observational study can yield the same result as a randomized clinical trial. Observational data may be particularly useful in assessing efficacy of prophylaxis when logistic requirements preclude study of enough patients to allow sufficient power to identify effects of interest [4]. This constraint has been a particular problem in the investigation of "clean" procedures, such as breast surgery and herniorrhaphy [5]. Two recent randomized trials ofprophylaxis for these procedures identified reductions of --40% in the number of postoperative wound infections, but these reductions were not statistically significant [3, 6]. Similarly, this population is not large enough to provide conclusive assessments of the individual procedures considered in isolation. However, the proportional reduction in risk ofinfection was about the same for all of the procedures studied, and there was no statistically significant difference in prophylactic effect among the procedures. These findings, coupled with our belief that the mechanism of operative contamination by the host's skin flora is similar, suggested that findings may be similar in each procedure. The criteria we used to define wound infection are similar to those currently recommended by the Centers for Disease Control [7], and we chose these definitions at the beginning of our data collection. However, the effect of prophylaxis was nearly identical in reducing the occurrence of wound infection with pus and somewhat stronger in reducing the occurrence of wound infection requiring parenteral antibiotic therapy. We do not know what criteria surgeons used in deciding to administer prophylaxis. It is possible that they based their decision on type and anticipated duration of surgery, but they may also have used other risk factors that tend to occur with certain types of procedures or during long procedures, such as the use of incisional drains, which have been identified as a risk factor for herniorrhaphy [8]. It is important to
Table 3. Occurrence of wound infection among procedures with or without antibiotic prophylaxis (P), categorized by procedure type and duration. Duration of surgery (min) 31-60
~30
Procedure type* Axillary node dissection Excisional breast biopsy or lumpectomy Mastectomy Reduction mammoplasty Herniorrhaphy
With P
Without P
61-120
>120
With
Without
With
Without
P
P
P
P
With P
0
a
0/12
1/7
1/19
0/16
0/4
0/13 0/4 0 0/6
3/208 0/2
1/34 0/36 0/1 0/76
5/139 2/43 0/1 5/366
2/85 8/207 1/8 1/127
3/165 9/240 0/4 6/453
0/32 8/191 3/188 1/30
a
1/60
* Procedures were combined into 5 groups to reduce no. of groups with no infections.
Unknown
Without P 1/5 5/72 13/203 2/29 2/79
With
Without
P
P
0
a
0/1 0/2 0/1 0
0/5 1/3 0/1 1/24
Downloaded from http://jid.oxfordjournals.org/ at Emory University on August 2, 2015
Figure 2. Variation of predicted risk of infection with duration of procedure for modified radical mastectomy. Solid line, predicted values for patients with prophylaxis; dashed line, patients without prophylaxis. Regression lines are derived from values in table 2.
559
560
Platt et al.
Members of Study Teams Beth Israel Hospital: D. F. Zaleznik, M. B. Collins, and L. Robertson; Brigham and Women's Hospital: R. Platt, M. Albano, S. Petrycki, and S. Fischer;Massachusetts General Hospital: C. C. Hopkins, J. F. Burke, and P. Kelleher; New England Deaconess Hospital: A. W. Karchmer and M. Christensen; Harborview Medical Center, University of Washington Hospital, and PacificMedical Center: E. P. Dellinger, E. Dever, J. Taylor, and M. Wertz; Baptist Medical Center and VA Hospital: C. S. Bryan and J. Wells; Channing Laboratory: R. Platt, J. R. Zucker, S. K. Marino, K. F. Holbrook, M. R. Segal,T. D. Tosteson, A. Munoz, S. Graham, and G. Campbell; and Smith Kline and French Laboratories: M. A. Wikler.
References I. Meakins JL. Prophylactic antibiotics. In: Wilmore DW, Brennan MF, Harken AH, Holcroft JW, Meakins JL, eds. Care of the surgical patient. New York: Scientific American, 1988:VI,3,19. 2. Kaiser AB. Antimicrobial prophylaxis in~urgery. N Engl J Med 1986;315: 1129-38. 3. Platt R, Zaleznik OF, Hopkins CC, et al. Perioperative antibiotic prophylaxis for herniorrhaphy and breast surgery. N Engl J Med 1990;322: 153-60. 4. Platt R. Methodologic aspects of clinical studies of perioperativeantibiotic prophylaxis. Rev Infect Dis 1991;13(suppl 1O):S81O-4. 5. Hopkins Cc. Antibiotic prophylaxis in clean surgery: peripheral vascular surgery, noncardiovascular thoracic surgery, herniorrhaphy and mastectomy. Rev Infect Dis 1991;13(suppl 1O):S869-73. 6. Wagman LD, Tegtmeier B, Beatty JD, et al. A prospective, randomized double-blind study of the use of antibiotics at the time of mastectomy. Surg Gynecol Obstet 1990; 170:12-6. 7. Garner JS, Jarvis WR, Emori TG, Horan TC, Hughes JM. CDC definitions for nosocomial infections, 1988. Am J Infect Control 1988;16: 129-40. 8. Simchen E, Wax Y, Pevsner B. The Israeli Study of Surgical Infections (ISSI): II. Initial comparisons among hospitals with special focus on hernia operations. Infect Control Hosp Epidemiol 1988;9:241-9.
Downloaded from http://jid.oxfordjournals.org/ at Emory University on August 2, 2015
recognize that the type of procedure is not a risk factor itself but is a convenient surrogate for risk factors that playa direct role in the pathogenesis of wound infection. Similarly, duration of procedure might also be a marker for other risk factors. Regardless of the actual algorithm that surgeons used, the use of prophylaxis was effectively concentrated among highrisk patients. Thus, in the absence of formal guidelines for use of prophylaxis, surgeons tended to administer it to patients who were most likely to benefit in terms of their absolute decrease in risk of infection. Another effect of use of prophylaxis in high-risk patients was to obscure the prophylactic effect in the unadjusted analysis, since prophylaxis reduced the occurrence ofinfection among higher-risk patients to about the level of risk of the lower-risk patients who did not receive prophylaxis. Given the apparent efficacy of antibiotic prophylaxis, it is still necessary to decide who should receive it. Although prophylaxis prevented about the same proportion of infections in each major group of procedures we studied, the fact that the absolute risk was different for these procedures means that the number of courses of prophylaxis required to prevent an infection is larger for low-risk procedures than for high-risk procedures. Ultimately, the decision will best be made through cost-benefit analyses that consider a patient's risk ofinfection , the morbidity and cost ofinfection, the cost of prophylaxis, and the adverse consequences of prophylaxis, including drug reactions and selection of antibiotic resistance. When other factors are constant, the predicted risk of infection will be the best guide to use of prophylaxis. In summary, these analyses provide independent support for a protective role of perioperative antibiotic prophylaxis to prevent postoperative wound infection after either breast or hernia surgery. Further analysis is necessary to identify optimal clinical applications of these findings.
JlD 1992; 166 (September)
Prophylaxis against Wound Infection following Herniorrhaphy or Breast Surgery R. PiaU, J. R. Zucker, D. F. Zaleznik, C. C. Hopkins, E. P. Dellinger, A. W. Karchmer, C. S. Bryan, J. F. Burke, M. A. Wikler, S. K. Marino, K. F. Holbrook, and T. D. Tosteson
Departments of Medicine and Surgery. Harvard Medical School. Departments of Medicine and Infection Control Units. Channing Laboratory. Brigham and Women's Hospital. Beth Israel Hospital. and New England Deaconess Hospital. and Departments of Medicine and Surgery and Infection Control Unit. Massachusetts General Hospital. Boston; Department ofSurgery. University of Washington. and Harborview Medical Center. Seattle; Department ofMedicine. University ofSouth Carolina. Columbia; Anti-Infective Division. Smith Kline and French Laboratories. Philadelphia. Pennsylvania
The effect of perioperative antibiotic prophylaxis on definite wound infections was assessed for 3202 herniorrhaphies or selected breast surgery procedures. Patients were identified preoperatively and monitored for ~4 weeks. Thirty-four percent of patients (1077/3202) received prophylaxis at the discretion of the surgeon; 86 definite wound infections (2.7%) were identified. Pro-
Perioperative antibiotic prophylaxis prevents postoperative wound infections following many types of surgical procedures [1,2]. Most studies have investigated the role of prophylaxis in one of two types of procedures-those with intrinsically high infection rates such as colorectal surgery, and those with lower rates but for which the consequences of infection are often catastrophic, such as implantation ofprosthetic material. There is relatively little information on the use of prophylaxis for "clean" surgery that involves soft tissue only and does not involve contamination of the operative site by the flora of the host's viscera or mucous membranes, even though these are common procedures. Recent evidence indicates that prophylaxis prevents ,...., 50% of all postoperative infections, taken together, after elective herniorrhaphy and selected types of breast surgery [3]. However, no extant study was designed to demonstrate the effectiveness of prophylaxis for wound infection alone following
Received 22 November 1991; revised 27 March 1992. Presented: annual meeting of the American Epidemiological Society. Ann Arbor. March 1992. The study was approved by the institutional review boards at each participating institution. Grant support: Smith Kline and French Laboratories. M.A.W. is a stockholder in and an employee of Smith Kline Beecham. Reprints or correspondence: Dr. R. Platt. Channing Laboratory. 180 Longwood Ave.• Boston. MA 02115.
The Journal of Infectious Diseases 1992;166:556-60 © 1992 by The University of Chicago. All rights reserved. 0022-1899/92/6603-0015$01.00
these procedures. Assessment of this question requires data from several thousand patients. We addressed this issue by studying the effect of prophylaxis against definite wound infections in a prospective validation cohort comprising patients who underwent these same procedures but who did not participate in the randomized clinical trial. Some of these patients received prophylaxis at their surgeon's discretion.
Methods Patient population. This study was done in conjunction with a randomized clinical trial of perioperative prophylaxis [3]. We attempted to identify all patients undergoing eligible procedures at participating institutions. This report includes information from the patients who did not participate in the randomized trial; some patients were ineligible because of reported allergy to a {Hactam antibiotic. Most were eligible but did not participate in the randomized trial because the study nurse was unable to talk with the patient or because the patient or the patient's surgeon refused to participate. Patients included were ~ 18 years old and underwent elective inguinal or femoral herniorrhaphy, mastectomy, lumpectomy, excisional breast biopsy, axillary node dissection for breast cancer, or reduction mammoplasty between 1 April 1985 and 30 September 1987. Participating hospitals were Beth Israel Hospital, Brigham and Women's Hospital, Massachusetts General Hospital, and New England Deaconess Hospital (Boston); Harborview Medical Center, University of Washington Hospital, and Pacific Medical Center
Downloaded from http://jid.oxfordjournals.org/ at Emory University on August 2, 2015
phylaxis recipients were at higher risk for infection, with a higher proportion of mastectomies, longer procedures, and other factors. Patients who received prophylaxis experienced 41% fewer definite wound infections (odds ratio [OR], 0.59; 95% confidence interval [Cl], 0.35-0.99; P = .04) and 65% fewer definite wound infections requiring parenteral antibiotic therapy (OR, 0.35; 95% CI, 0.15-0.88; P = .02) after adjustment for duration of surgery and type of procedure. Additional adjustment for age, body mass index, the presence of drains, diabetes, and exposure to corticosteroids did not change the magnitude of this effect meaningfully. The effect of prophylaxis was similar for all procedures studied. In the absence of formal guidelines, surgeons at these institutions administered prophylaxis preferentially to patients at highest risk.
JIO 1992; 166 (September)
Prophylaxis against Wound Infection
Results Patient population. We identified 3221 patients who underwent an eligible procedure; 19 were excluded because no information was available after the day of surgery. Thus, 3202 patients were included in this analysis. The status was known for 96% of patients at least 6 days after surgery, for 91%at 14 days, and for 87% at > 28 days after surgery. Sixtytwo percent of patients (1981) underwent one of the breast procedures and the remaining 38% (1221) underwent a herniorrhaphy. The median duration of surgery was 60 min for lumpectomy, 68 min for herniorrhaphy, 115 min for modified radical mastectomy, and 190 min for reduction mammoplasty. Other characteristics of the study population are shown in table 1. Use ofperioperative prophylaxis. Of 3202 patients, 1077 (34%) received parenteral antibiotic prophylaxis. For those, the median interval between administration of the antibiotic and start of surgery was 20 min, with 10%of patients receiving antibiotic at least 15 min after incision, Cefazolin was the most commonly used agent, accounting for 796 courses (74%). In addition, 77 patients received another cephalosporin, 103 received another ,B-Iactam antibiotic, 41 received clindamycin, and 21 received vancomycin. No other agent was used for> 10 individuals. The use of prophylaxis varied considerably according to procedure and other characteristics (table I). It was used least often for herniorrhaphy, in 239 (20%) of 1221 patients; it was used most often for reduction mammoplasty, in 198 (85%) of 233 patients. The duration of surgical procedures was also longer for prophylaxis recipients; their median duration was 106 min, compared with 73 min for those without prophylaxis (P < .001). Infections. Eighty-six wound infections (2.7%) were identified, ofwhich 71 were observed to have purulent drainage. Twenty-nine of these prompted readmission and several others extended the original hospitalization. Forty-four infections were treated with incision and drainage or debridement, and 34 patients were treated with parenteral antibiotics. Staphylococcus aureus was identified in 37 of the 46 cases for which a pathogen was noted. Prophylaxis and infection. Because the patients who received prophylaxis differed from those who did not in several important respects, we adjusted for these differences to assess the effect of prophylaxis. The relationship between prophylaxis and infection is shown in figure I. There was important confounding by several factors. The unadjusted OR for infection comparing prophylaxis recipients with nonrecipients was 0.85 (26/1077 vs. 60/2125). After adjustment for procedure duration and type, the OR was 0.59 (95% confidence interval [Cl], 0.35-0.99; P = .04). The adjusted risks of the various procedures and time are shown in table 2, and the predicted risks of infection are illustrated in figure 2. There was no significant interaction between antibiotic prophylaxis
Downloaded from http://jid.oxfordjournals.org/ at Emory University on August 2, 2015
(Seattle); and Baptist Medical Center and VA Hospital (Columbia, SC). Hospitals initiated surveillance at different times and suspended surveillance during brief periods when study personnel were unavailable (e.g., during vacations). Some centers did not include patients undergoing ambulatory surgery because of logistic constraints. Patients were excluded if they had an active infection at the time ofsurgery, ifthey had received an antibiotic during the week before surgery, if prosthetic material was implanted during surgery, or if they underwent a concomitant surgical procedure that involved a separate skin incision. Patients received prophylaxis at the discretion of their attending physician, with an agent of the physician's choosing. There were no formal recommendations regarding the use of prophylaxis at any of the institutions during this period. Definitions. Definite wound infection, referred to subsequently as wound infection, was defined as a wound with erythema plus drainage, a wound with purulent drainage, or a wound that was opened and not reclosed for treatment of presumed infection. Erythema or a physician's diagnosis alone was not sufficient to establish a diagnosis. Perioperative antibiotic prophylaxis, referred to subsequently as prophylaxis, was defined as parenteral antibiotic administered any time on the day of surgery before the end of the procedure, as determined by review of operating room records. Identification of postoperative infection. A member of the study team assessed each patient's status on each weekday during hospitalization. Information was obtained by direct observation of the wounds when this was possible, from relevant hospital personnel (nurse, attending physician, house staff), or from the medical record. A questionnaire was administered by telephone twice to all patients, once 6-15 days after surgery and again 4-6 weeks later. The questionnaire elicited information about difficulties with wound healing, new infections, new antibiotic courses, nonroutine visits to the physician, and hospitalization. Information suggesting the presence of infection was verified by reference to available medical records. When patients could not be contacted, the physician's office notes were used. All cases that were provisionally identified by the study teams plus those that included erythema or drainage or received a physician's diagnosis of infection were reviewed and formally classified by a committee composed of all investigators and research nurses. The patient's prophylaxis status was not made available to this group, although the individual patient's investigators did have access to this information. Statistical analysis. The Wilcoxon rank sum test was used to compare the distribution of continuous variables between prophylaxis recipients and nonrecipients (SAS version 6.03; SAS Institute, Cary, NC). Infections in patients who received prophylaxis were compared with those in patients who did not by computing odds ratios (OR) for infection. Potential confounders were adjusted for by logistic regression (BMDP Statistical Software, 1990 revision; University of California Press, Berkeley). ORs were computed in all cases to permit direct comparison of the results of logistic regression analyses to the unadjusted analyses.
557
Platt et al.
558
Table 1. Characteristics of patients according to exposure to prophylaxis.
Characteristic
;;.2
General anesthesia Diabetes mellitus Preoperative disinfectant for skin site Iodine-containing compound Soap or alcohol Other or unknown Corticosteroids Duration of surgery (min) Age (years) Body mass index (kg/rn'') Prophylaxis-to-surgery interval (min) Follow-up interval (days)
Nothing
No prophylaxis n = 2125
Prophylaxis n = 1077
Total n = 3202
42 (2) 547 (26) 83 (4)
15 (I) 150 (14) 65 (6)
57 (2) 697 (22) 148 (5)
402 (19) 35 (2)
362 (34) 13 (I) 198(18)
764 (24) 19 (I) 233 (7)
28 (I) 982 (46)
35 (3) 239 (22)
63 (2) 1221 (38)
6 (0)
Adjusted for:
Procedure Duration
Procedure Duration and Procedure Type I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
.2
.4
.6
.8
1.0
1.2
1.4 1.6
1.8
ODDS RATIO 405 (19) 117 (6) 2 (0)
200 (19) 240 (22) 28 (3)
605 (19) 357 (II) 30 (I)
1413 (66) 449(21) 263 (l2) 1641 (77) 85 (4)
311 356 410 960 56
(29) (33) (38) (89) (5)
1724(54) 805 (25) 673(21) 2601(81) 141 (4)
1613 463 49 30
716 (66) 334(31) 27 (3) 24 (2)
2329 (73) 797 (25) 76 (2) 54 (2)
(76) (22) (2) (I)
73 56 24.2
106 55 24.6
83 56 24.3
NA 38
22 40
NA 39
NOTE. Data are no. of patients (%) unless stated otherwise; values for continuous variables are medians. NA, measure does not apply for procedures without prophylaxis. * Includes inguinal and femoral herniorrhaphies plus inguinal canallipectomy done for suspected inguinal hernia.
and procedure duration or between prophylaxis and procedure type. The actual occurrence of wound infection among those with and without prophylaxis by procedure type and duration of surgery is shown in table 3. Further adjustment for age, diabetes, body mass index, and use of incisional drains did not meaningfully change this ratio (OR, 0.57; 95% CI, 0.34-0.97; P = .03). The effect of prophylaxis did not vary significantly for the different procedures. For modified radical mastectomies, the adjusted OR for infection among prophylaxis recipients was 0.67, for lumpectomies it was 0.61, and for herniorrhaphies it was 0.54. Antibiotic prophylaxis was also associated with signifi-
Figure 1. Prophylactic effect before and after adjustment for duration of procedure and type of surgery. Odds ratio (OR) falls from 0.88 (unadjusted) to 0.66 (adjusted for procedure duration) to 0.59 (adjusted for both). Unadjusted OR is 0.88 rather than 0.85 as noted in text because 37 cases, including 2 infections that had partially missing data, were excluded from logistic regression analysis.
Table 2. Adjusted risks ofinfection in patients who did or did not receive prophylaxis before breast surgery.
Prophylaxis vs. no prophylaxis Procedure* Local excision of breast lesion Lumpectomy Simple mastectomy Modified radical mastectomy Radical mastectomy Reduction mammoplasty Axillary node dissection Procedure duration (min)
OR
95% confidence interval
0.59
0.35-0.99
0 2.09 2.71
0 1.07-4.09 0.99-7.36
3.03 4.40 1.69 4.10 1.004
1.60-5.75 0.54-35.8 0.58-4.93 1.15-14.6 1.00-1.01
NOTE. Values are taken from logistic regression model that included variables shown plus additional terms for 23 femoral herniorrhaphies and I inguinal canal lipectomy. Constant for this model was -4.478. Coefficients for variable shown are natural logarithms of odds ratios (OR) shown. * OR are compared with those of inguinal herniorrhaphy. t OR is increment/min of procedure duration.
cantly lower adjusted risks of wound infection with purulent drainage (OR, 0.57; 95% CI, 0.32-1.00; P = .05) and of subsequent parenteral antibiotic treatment of wound infection (OR, 0.35; 95% CI, 0.14-0.88; P = .02). There was a nonsignificantly lower risk of wound infection with incision and drainage (OR, 0.66; 95% CI, 0.30-1.46; P = .3) and of
Downloaded from http://jid.oxfordjournals.org/ at Emory University on August 2, 2015
Procedure Local excision Lumpectomy Simple mastectomy Modified radical mastectomy Radical mastectomy Reduction mammoplasty Axillary node dissection alone Herniorrhaphy* Axillary node dissection (with or without other procedure) Bilateral procedure Flap reconstruction Drains at operative site None I
lID 1992; 166 (September)
JID 1992; 166 (September)
30
Prophylaxis against Wound Infection
60 1~0 Duration of procedure (minutes)
1~0
readmission for wound infection (OR, 0.82; 95% CI, 0.351.96; P = .7).
Discussion Several findings emerged from these analyses. Prophylaxis was used relatively often, even though there were no formal guidelines recommending it for any of these procedures and no compelling published data on its efficacy when these procedures were done. Surgeons used prophylaxis more often in cases with commonly accepted risk factors, such as more extensive surgery or longer operative times. The use of prophylaxis was associated with a statistically significant 41 % reduction in the adjusted risk of wound infection and a 65% reduction in the adjusted risk of parenteral antibiotic therapy for treatment of these infections. The reduction in the risk of wound infection was consistent with the 48% reduction that we observed in the concur-
rent randomized clinical trial done in the same institutions [3]. This finding indicates that after appropriate adjustment for confounding, an observational study can yield the same result as a randomized clinical trial. Observational data may be particularly useful in assessing efficacy of prophylaxis when logistic requirements preclude study of enough patients to allow sufficient power to identify effects of interest [4]. This constraint has been a particular problem in the investigation of "clean" procedures, such as breast surgery and herniorrhaphy [5]. Two recent randomized trials ofprophylaxis for these procedures identified reductions of --40% in the number of postoperative wound infections, but these reductions were not statistically significant [3, 6]. Similarly, this population is not large enough to provide conclusive assessments of the individual procedures considered in isolation. However, the proportional reduction in risk ofinfection was about the same for all of the procedures studied, and there was no statistically significant difference in prophylactic effect among the procedures. These findings, coupled with our belief that the mechanism of operative contamination by the host's skin flora is similar, suggested that findings may be similar in each procedure. The criteria we used to define wound infection are similar to those currently recommended by the Centers for Disease Control [7], and we chose these definitions at the beginning of our data collection. However, the effect of prophylaxis was nearly identical in reducing the occurrence of wound infection with pus and somewhat stronger in reducing the occurrence of wound infection requiring parenteral antibiotic therapy. We do not know what criteria surgeons used in deciding to administer prophylaxis. It is possible that they based their decision on type and anticipated duration of surgery, but they may also have used other risk factors that tend to occur with certain types of procedures or during long procedures, such as the use of incisional drains, which have been identified as a risk factor for herniorrhaphy [8]. It is important to
Table 3. Occurrence of wound infection among procedures with or without antibiotic prophylaxis (P), categorized by procedure type and duration. Duration of surgery (min) 31-60
~30
Procedure type* Axillary node dissection Excisional breast biopsy or lumpectomy Mastectomy Reduction mammoplasty Herniorrhaphy
With P
Without P
61-120
>120
With
Without
With
Without
P
P
P
P
With P
0
a
0/12
1/7
1/19
0/16
0/4
0/13 0/4 0 0/6
3/208 0/2
1/34 0/36 0/1 0/76
5/139 2/43 0/1 5/366
2/85 8/207 1/8 1/127
3/165 9/240 0/4 6/453
0/32 8/191 3/188 1/30
a
1/60
* Procedures were combined into 5 groups to reduce no. of groups with no infections.
Unknown
Without P 1/5 5/72 13/203 2/29 2/79
With
Without
P
P
0
a
0/1 0/2 0/1 0
0/5 1/3 0/1 1/24
Downloaded from http://jid.oxfordjournals.org/ at Emory University on August 2, 2015
Figure 2. Variation of predicted risk of infection with duration of procedure for modified radical mastectomy. Solid line, predicted values for patients with prophylaxis; dashed line, patients without prophylaxis. Regression lines are derived from values in table 2.
559
560
Platt et al.
Members of Study Teams Beth Israel Hospital: D. F. Zaleznik, M. B. Collins, and L. Robertson; Brigham and Women's Hospital: R. Platt, M. Albano, S. Petrycki, and S. Fischer;Massachusetts General Hospital: C. C. Hopkins, J. F. Burke, and P. Kelleher; New England Deaconess Hospital: A. W. Karchmer and M. Christensen; Harborview Medical Center, University of Washington Hospital, and PacificMedical Center: E. P. Dellinger, E. Dever, J. Taylor, and M. Wertz; Baptist Medical Center and VA Hospital: C. S. Bryan and J. Wells; Channing Laboratory: R. Platt, J. R. Zucker, S. K. Marino, K. F. Holbrook, M. R. Segal,T. D. Tosteson, A. Munoz, S. Graham, and G. Campbell; and Smith Kline and French Laboratories: M. A. Wikler.
References I. Meakins JL. Prophylactic antibiotics. In: Wilmore DW, Brennan MF, Harken AH, Holcroft JW, Meakins JL, eds. Care of the surgical patient. New York: Scientific American, 1988:VI,3,19. 2. Kaiser AB. Antimicrobial prophylaxis in~urgery. N Engl J Med 1986;315: 1129-38. 3. Platt R, Zaleznik OF, Hopkins CC, et al. Perioperative antibiotic prophylaxis for herniorrhaphy and breast surgery. N Engl J Med 1990;322: 153-60. 4. Platt R. Methodologic aspects of clinical studies of perioperativeantibiotic prophylaxis. Rev Infect Dis 1991;13(suppl 1O):S81O-4. 5. Hopkins Cc. Antibiotic prophylaxis in clean surgery: peripheral vascular surgery, noncardiovascular thoracic surgery, herniorrhaphy and mastectomy. Rev Infect Dis 1991;13(suppl 1O):S869-73. 6. Wagman LD, Tegtmeier B, Beatty JD, et al. A prospective, randomized double-blind study of the use of antibiotics at the time of mastectomy. Surg Gynecol Obstet 1990; 170:12-6. 7. Garner JS, Jarvis WR, Emori TG, Horan TC, Hughes JM. CDC definitions for nosocomial infections, 1988. Am J Infect Control 1988;16: 129-40. 8. Simchen E, Wax Y, Pevsner B. The Israeli Study of Surgical Infections (ISSI): II. Initial comparisons among hospitals with special focus on hernia operations. Infect Control Hosp Epidemiol 1988;9:241-9.
Downloaded from http://jid.oxfordjournals.org/ at Emory University on August 2, 2015
recognize that the type of procedure is not a risk factor itself but is a convenient surrogate for risk factors that playa direct role in the pathogenesis of wound infection. Similarly, duration of procedure might also be a marker for other risk factors. Regardless of the actual algorithm that surgeons used, the use of prophylaxis was effectively concentrated among highrisk patients. Thus, in the absence of formal guidelines for use of prophylaxis, surgeons tended to administer it to patients who were most likely to benefit in terms of their absolute decrease in risk of infection. Another effect of use of prophylaxis in high-risk patients was to obscure the prophylactic effect in the unadjusted analysis, since prophylaxis reduced the occurrence ofinfection among higher-risk patients to about the level of risk of the lower-risk patients who did not receive prophylaxis. Given the apparent efficacy of antibiotic prophylaxis, it is still necessary to decide who should receive it. Although prophylaxis prevented about the same proportion of infections in each major group of procedures we studied, the fact that the absolute risk was different for these procedures means that the number of courses of prophylaxis required to prevent an infection is larger for low-risk procedures than for high-risk procedures. Ultimately, the decision will best be made through cost-benefit analyses that consider a patient's risk ofinfection , the morbidity and cost ofinfection, the cost of prophylaxis, and the adverse consequences of prophylaxis, including drug reactions and selection of antibiotic resistance. When other factors are constant, the predicted risk of infection will be the best guide to use of prophylaxis. In summary, these analyses provide independent support for a protective role of perioperative antibiotic prophylaxis to prevent postoperative wound infection after either breast or hernia surgery. Further analysis is necessary to identify optimal clinical applications of these findings.
JlD 1992; 166 (September)
