- Email: [email protected]
An economic analysis of antimicrobial prophylaxis against urinary tract infection in patients undergoing transurethral resection of the prostate
CLINICAL THERAPEUTICYYVOL.
21, NO. 9. 1999
An Economic Analysis of Antimicrobial Prophylaxis Against Urinary Tract Infection in Patients Undergoing Transurethral Resection of the Prostate
Gordon G. Liu, PhD,’ Trisha Nguyen, MS, PharmD,2 and Michael B. Nichol, PhD’ ‘University of Southern California, Los Angel& California
and 2Department of Veterans Affairs,
ABSTRACT Despite the high level of safety and low incidence of mortality associated with transurethral resection of the prostate (TURP), urinary tract infections (UTIs)the most common complication associated with this procedure+ontinue to be an important source of postoperative morbidity and costs. However, there is controversy about whether antimicrobial agents should be usedasUT1 prophylaxis in patients undergoing TURP and, if so, which agentsshouldbe usedand for what duration. This retrospective study used multivariate regressionanalysis to evaluate the different types and durations of antibiotic prophylaxis in 222 patientswho underwent TURP at a Veterans Affairs hospital between January 1. 1995, and March 30, 1998. The primary outcome measures were total medical costs (ie, Accepted for publication Printed in the USA. Reproduction in whole
0149-2918/99/$19.00
March
26,
1999.
or part is not permitted
medication use, clinic office visits, and hospital care in the 4 weeks after the procedure), length of hospital stay (total days in hospital due to the procedure), and probability of UT1 (incidence of infection in the 4 weeks after the procedure). Results showedthat there was no difference in the length of hospital stay regardlessof the regimen or duration of pre-TURP antibiotic therapy. Patients who received pre-TURP ampicillin plus ceftizoxime incurred moderately higher total medical costs than did patients who received the least costly drug, cefazolin (P = 0.10). Similarly, patients who received postTURP quinolones incurred a significantly higher total medical cost than did patients who received co-trimoxazole (P = 0.06). We found no evidence of a relationship between use of specific parenteral or oral antibiotic prophylaxis for UT1 in patients undergoing TURP and the rate of UT1 in such patients. Thus there is no justification for the use of more expensive antibiotic regimens.At our institution, the preferred pre-TURP prophylactic regimen 1589
CLINICAL THERAPEUTICS”
would be cefazolin, whereas co-trimoxazole would be the most cost-effective post-TURP prophylactic regimen. Because duration of post-TURP prophylaxis does not appear to influence the rate of UTI, 24 hours would seem adequate. Key words: antibiotic therapy, prophylaxis, transurethral resection of the prostate, urinary tract infection. INTRODUCTION Transurethral resection of the prostate (TURP) is a procedure for relieving bladder outlet obstruction caused by benign prostatic hyperplasia (BPH) or cancer of the prostate. Among the most commonly performed surgical procedures, TURP is the gold standard for treating BPH and an accepted therapy for prostate cancer.’ Despite the relative safety and low incidence of mortality with TURP compared with other high-risk procedures such as coronary bypass surgery, urinary tract infections (UTIs)-the most common complication associated with the procedurecontinue to be an important source of postoperative morbidity and cost.? However, there continues to be controversy about whether to use antimicrobial prophylaxis against UT1 in patients undergoing TURP. We conducted a MEDLINE@ search of the literature for articles relating to the prophylactic use of antimicrobial agents for UT1 in association with TURP. In previous studies in TURP patients,sm5 antibiotics significantly decreased the number of infectious complications and eliminated bacteriuria in approximately 50% of cases. According to the findings of the European Collaborative Study of Antibiotic Prophylaxis for Transurethral Resection of the Prostate, “patients, even those with cul1590
ture-negative urine preoperatively, benefit from antibiotic prophylaxis and...antibiotic prophylaxis should be considered for all men undergoing prostatic surgery as a routine procedure.“” This randomized, controlled study in 157 patients with no bacteriuria before undergoing TURP found that patients who received no prophylaxis had a UT1 rate of 33.9%. compared with a rate of 11.6% in patients who received intravenous (IV) ceftazidime until removal of the urethral catheter (<5 days). In their 1979 literature review, Chodak and Plaut6 found no proof that prophylactic use of antimicrobial agents for UT1 prophylaxis in patients undergoing TURP was beneficial; nevertheless, a later study’ reported that prophylactic antimicrobial agents were administered to 160% of urologic patients in the United States and were given for >7 days in 49% of patients. In an open-label, prospective. randomized, controlled study in 1IO patients with sterile urine preoperatively who received either placebo or cefoperazone every 8 hours for 24 hours after surgery, only 1 patient in the control group developed infection (ie, positive culture with signs and symptoms of infection), compared with none in the cefoperazone group.” Even if clinicians were to agree on the need to administer prophylactic antibiotics. the ideal duration of such treatment is debatable. Some investigators have administered a single preoperative dose of antibiotic. others recommend 5 days of antibiotics after surgery, and still others believe in starting antibiotic therapy a day or so before surgery and continuing it for 2 to 3 weeks after surgery.9-tZ Mebust and colleagueslo reported on the use of antibiotics (unknown types) for UT1 prophylaxis in 61% of 3885 patients undergoing TURP in 13 institutions; 49% of the anti-
G.G. LIU ET AL.
biotic prophylaxis group received therapy for >8 days. The authors did not assess the correlation between infection rate and duration of prophylaxis. In a randomized study in urologic surgery patients, Grabe et alI3 compared the efficacy of short-term ciprofloxacin (3 to 4 days), long-term ciprofloxacin (8 to 9 days), and no antibiotic in 222 patients. At discharge, no patient in the short- or long-term prophylaxis group had bacteriuria, whereas 38.5% of the patients receiving no prophylaxis had bacteriuria at discharge. At -14 days after discharge, bacteriuria was detected in 20% of the short-term prophylaxis patients, 4.3% of the long-term prophylaxis patients, and 44.9% of those who had received no prophylaxis. In a prospective, randomized study of 171 patients, Hall et alI4 evaluated the efficacy of a single oral dose of fleroxacin, a single IV dose of fleroxacin, and an extended regimen consisting of an initial IV dose of fleroxacin followed by oral therapy until removal of the urinary catheter (~6 days); there were no differences in the rates of infection between the 3 groups. In another study, Goldwasser et alI5 administered 3 different regimens to 81 TURP patients who had sterile urine preoperatively: group A received parenteral co-trimoxazole, 1 dose preoperatively and 1 dose postoperatively, and then continued on oral antibiotics for 10 days; group B received only the parenteral cotrimoxazole regimen; and group C received nothing. The investigators found that by day 7, the rate of infection in groups A and B was 3.8%, compared with 32% in group C. Therefore, these investigators recommended a short course of antimicrobial prophylaxis. The choice of antibiotic is another area of controversy. Some clinicians use an-
tibiotics known to penetrate prostate tissue well (ie, the quinolones), whereas others prefer to use antibiotics that are known to be effective against the organisms likely to cause UTIs (eg, Escherichia coli. Proteus spp).“,t6 Prophylactic ampicillin, sulfonamides, tetracycline, nitrofurantoin, cephalosporins, and quinolones have all been studied in TURP patients.4~6*‘6~19 Most of these studies were placebo controlled. Crawford et al,17 however, randomized patients to a single dose of oral lomefloxacin or IV cefotaxime and found no significant difference in the rate of infection between the 2 groups. A prospective, randomized, double-masked study by Iversen and Madsen’O compared 3 regimens of cephalosporin in 116 men who had sterile urine preoperatively. One group received 1 g cefotaxime IV 30 to 90 minutes before surgery, every 60 to 90 minutes during surgery, and immediately after surgery; of the other 2 groups, 1 received 1 g cefotaxime and the other received 1 g cefazolin, following the same regimen used in the first group and every 8 hours for 24 hours postoperatively. The 3 groups had a similar incidence of infection 7 days postoperatively. In summary, our literature review found little consensus on the effectiveness of antibiotic prophylaxis for UT1 in patients undergoing TURP. The weaknesses of some of these studies included lack of a double-masked design, small number of study subjects, lack of placebo control, and variations in definition of UTI. There are currently few data on the comparative effectiveness of different antibiotics for the prophylaxis of UT1 in TURP patients and even fewer data defining the appropriate duration of such therapy. The objective of the present study was to evaluate the costs and effectiveness of 1591
CLINICAL
different antibiotics and durations of antibiotic prophylaxis against UT1 in patients undergoing TURP. Two specific issues were addressed:(1) whether the rate of infection differs significantly between patients prescribed different prophylactic regimens before and after TURP, and (2) whether and how the length of hospital stay and total medical cost differ between patients prescribed different prophylactic antibiotics for different durations. Our perspective was that of a Veterans Affairs (VA) hospital, and our results may help similar hospitalsinvestigating ways of providing TURP in the most cost-effective manner. MATERIALS
AND METHODS
This retrospective study was conducted at a VA hospital in West Los Angeles. California. It was designedto analyze the experience of patients during 1 episode of hospitalization for TURP and for 4 weeks after discharge, a sufficient period for the development of possible infectious complications of TURP and for assessingthe benefits of antimicrobial prophylaxis.” The fact that a patient may have undergone >I TURP procedure during the study should not affect the validity of the study results, since each TURP event was assessedindependently. Our analysis applies to all patients being considered for TURP, whetherfor BPH or prostatecancer. From the hospital database.we identified 222 patients who had undergone TURP between January 1, 1995, and March 30, 1998. Patients were excluded if they had severe infection causing their transfer to the intensive care unit before the procedure. Data recorded from patients’ charts included date of surgery; date of discharge; birth date and age; in1592
THERAPEUTICS@
dication for TURP; comorbidities; presence of an indwelling Foley catheter; presence of bacteriuria on urinalysis before surgery; antibiotics received for an existing infection at the time of admission; prophylactic antibiotic regimen and duration of therapy before and after TURP; operative time (defined asthe time in minutes from operating table to recovery room); and postsurgicalinfection (defined as any additional post-TURP intervention for which the diagnosiswas UTI, including additional antibiotic prescriptions, genitourinary clinic or emergency department visits, or hospitalizations for UTI). The number of postsurgical infections and number of clinic visits were calculated for the 4-week period after the procedure. According to general guidelines for the evaluation of new anti-infective therapies for the prophylaxis of surgical infections, any unexplained use of anti-infective agentsin the 4-week period after surgery should be regarded as failure of prophylaxis, as should development of postoperative bacteriuria. with or without systemic signsof infection. I5 However, our review of patients’ charts found that due to the possible transient nature of these events, physicians had not treated some patients having a single increase in white blood cell count, temperature, or bacteriuria without symptoms.In accordancewith the physicians’judgment, we did not consider these events infections. Patient comorbidities included in our chart analysis were hypertension, coronary artery disease, chronic obstructive pulmonary disease,cancer, dementia,diabetes mellitus, degenerative joint disorder, psychiatric illness (eg, depression, bipolar disorder, and schizophrenia), and Parkinson’s disease.These comorbidities
G.G. LIU ET AL.
Table I. Description
of variables
used in the study.
Variable
Age
Type
Operation time Race Use of urethral catheter before TURP Positive bacteriuria test or culture Receiving antibiotics at time of admission No. of patients undergoing TURP Indication for TURP Pre-TURP antibiotic regimen Amp + gent Van + gent Amp + ceft Cefazolin (control option) Dose Post-TURP antibiotic regimen Cephalexin Quinolones Co-trimoxazole (control option) None Duration of prophylaxis Comorbidities (diagnoses recorded in medical history) Hypertension Coronary artery disease Chronic obstructive pulmonary disease Cancer Parkinson’s disease Dementia Diabetes mellitus Degenerative joint disorder Psychiatric illness (depression or schizophrenia) Year of TURP 1995 1996 1997 1998 = transurethral resection of the prostate; BPH gentamicin; van = vancomycin; ceft = ceftizoxime.
TURP
are rarely investigated in relation to TURP2 but were included in the present study for completeness and to control for extraneous variables. Other comorbidities
Continuous Continuous 1 = white; 0 = other 1 = yes; 0 = no I = yes; 0 = no 1 = yes; 0 = no Continuous 1 = BPH; 0 = prostate cancer or other 1= 1= 1= 1= 1=
amp + gent; 0 = other van + gent; 0 = other amp + ceft; 0 = other cefazolin; 0 = other 1 dose; 0 = >l dose
1 = yes; 0 = other 1 = yes; 0 = other 1 = yes: 0 = other Continuous (days) 1= 1= 1= 1= 1= 1= 1= 1= 1=
yes; 0 yes; 0 yes; 0 yes; 0 yes; 0 yes; 0 yes; 0 yes; 0 yes; 0
= = = = = = = = =
no no no no no no no no no
1 =95;O=other 1 = 96; 0 = other 1 = 97; 0 = other 1 = 98; 0 = other
= benign prostatic hyperplasia; amp = ampicillin; gent =
occurring in ~1% of patients pothyroidism, gout, and anemia) controlled for in our analysis. Length of stay was calculated
(eg, hywere not from the 1593
CLINICAL
day the patient underwent TURP to the day of discharge. If the patient was kept in the hospital for a reason unrelated to TURP (eg, admission to a psychiatric ward or delay in nursing home admission), the day the patient was discharged from the genitourinary service was considered the dischargedate, since this was the date at which the patient was considered stable enough to be discharged.The same criterion was applied to patients transferred to a nursing home or intermediate facility. However, if a patient was transferred to the intensive care unit, the period spent in the unit was included in the length of stay. The costs evaluated included acquisition costs of the prophylactic antibiotics and costs of additional clinic visits and hospitalizations associated with TURP. These costs were obtained from the VA reimbursement databasefor surgical patients and adjusted for inflation using the medical care component of the Consumer Price Index. Total medical costs were expressedin 1998 dollars. Antibiotic Regimensand Outcome Measures We considered 4 pre-TURP antibiotic regimens:ampicillin and gentamicin, vancomycin and gentamicin, ampicillin and ceftizoxime. and cefazolin alone. Prophylactic antibiotic administration started either immediately before surgery or the night before surgery.All patientswere continued on parenteral antibiotics for 1 to 2 days after surgery and before discharge. Four post-TURP antibiotic regimenswere considered: co-trimoxazole, quinolones, cephalexin, and none. These oral medications were given to patients at discharge and continued for various durations. 1594
THERAPEUTICF’
Three outcome measureswere used: meanlength of hospital stay for the TURP procedure. total direct cost, and incidence of post-TURP UTI. Indirect costssuch as lost productivity at home and lost earnings at work were not included, mainly becauseindirect costs are not significant from a payer’s perspective. Statistical Analysis Linear regressionanalysis was used to assessthe effects of the various pre-TURP and post-TURP antibiotic regimenson the 3 outcome measures:length of stay, total medical costs, and incidence of infection. Ordinary least squaresmethodology was employed to estimatethe 3 outcome models using SAS software (SAS Institute, Cary, North Carolina). The variables assessedin the study are defined in Table I. For hospital length of stay, we used a multivariate model that measuredthe effect of pre-TURP antibiotics, with cefazolin as a control. while controlling for certain other factors that may influence this variable. In particular, the model was designedto include type of parenteral antibiotics given before TURP (ampicillin and gentamicin, vancomycin and gentamicin, ampicillin and ceftizoxime, or cefazolin). age, indication for surgery (BPH or prostate cancer), number of doses of pre-TURP antibiotic (1 vs 2), number of comorbidities (~2 vs 52 diagnoses),race (white or nonwhite), and duration of the TURP operation (a continuous variable). Post-TURP antibiotics were not included in this model becausethese medications were generally given just before discharge and thus had no influence on length of stay. Becauseboth could contribute to total medical cost, the effects of pre- and post-
G.G. LIU ET AL.
Table II. Characteristicsof the study population at baseline(N = 222). No. (‘3%) of Patients
Characteristic Age (y) <65 26.5 Unknown Race White Nonwhite Black Hispanic Asian Use of urethral catheter before TURP None or unknown Yes Bacteriuria test or culture Positive Negative Undocumented urinalysis
81 (36.5) 119 (53.6) 22 (9.9) 143 79 50 23 6
187 (84.2) 35 (15.8) 37 53 132 45
Receivingantibioticsat time of admission Comorbidities Coronaryartery disease COPD Cancer Dementia Diabetes mellitus Degenerative joint disorder Psychiatric illness (depression or schizophrenia) Year of TURP 1995 1996 1997 1998 = chronic
obstructive
pulmonary
37 66 80 39
disease; TURP = transurethral
TURP antibiotic regimens on total cost were modeledin a secondmultivariate regression.Independentvariables controlled for in this model were age. indication for TURP (BPH or prostate cancer), number of dosesof pre-TURP antibiotic (1 vs 2), duration of post-TURP antibiotic, number
(16.7) (23.9) (59.5) (20.3)
110 (49.5) 49 (22.0) 26 (11.7) 22 (9.9) 11 (5.0) 36 (16.2) 35 (15.8) 38 (17.1)
Hypertension
COPD
(64.4) (35.6) (22.5) (10.4) (2.7)
resection
(16.7) (29.7) (36.0) (17.6)
of the prostate
of comorbidities (>2 vs <2 diagnoses), race (white or nonwhite), and duration of the TURP operation(a continuousvariable). A third model was developed using logistic regressionto test the effect of postTURP prophylactic antibiotic therapy on the outcome of interest, incidence of in1595
CLINICAL
THERAPEUTICS’
Table III. Prophylactic regimens used against urinary tract infection in 222 patients undergoing TURP. No. (%) of Patients
Regimen Pre-TURP No. of doses One (before surgery) Two (started the night before surgery) Parenteral antibiotics Ampicillin + gentamicin Vancomycin + gentamicin Gentamicin Cefazolin Ampicillin + ceftizoxime Post-TURP (oral antibiotics) Ofloxacin or ciprofloxacin Co-trimoxazole Cephalexin None TURP
= transurethral
resection
87 (39.2) 154 17 4 23 24
(69.4) (7.7)
121 63 30 8
(54.5) (28.4) (13.5) (3.6)
(1.8) ( 10.4) (10.8)
of the prostate.
fection. Because all infections occurred 26 days after discharge, this model did not include parenteral antibiotics given before TURP, asit is unlikely their effects would have lasted long enough to intluencethe post-dischargeinfection rate. The independent variables included in this model were age, types of oral antibiotics, duration of oral antibiotics, comorbidities, race, operation time, and year (to control for possible selection bias). To control for any selection bias introduced by developments in surgical technique or in methods of administering the antibiotics, we included year dummy variables for 1995 (control group), 1996, 1997, and 1998 in all models. RESULTS Characteristicsof the study population are summarized in Table II. The mean age of 1596
135 (60.8)
the subjects was 69 years, and the mean length of hospital stay was 3 days (range. 2 to 8 days); the mean (k SD) total medical cost was $6000 + $1990 per patient. At the time of admission, 17% of the patients had positive results on urinalysis, and 24% had negative results; the rest had no documentedurinalysis before TURP. It is possible that surgeonsat this hospital do not seethe need to order urinalysis or urine culture before surgery becausethey routinely give prophylaxis to every patient. Therefore, preoperative urine culture was not included
in the regression
analysis.The mean operating time was 69 minutes. All patients received prophylactic therapy. but there were several variations in the regimens used (Table III). Before TURP the patients in this study received either ampicillin and gentamicin or ampicillin and ceftizoxime; thosewho were al-
G.G. LIU ET AL.
Table IV. Incidence of the outcomesof interest during the 4 weeks after TURP (N = 222). Event
No. (%) of Patients
Infectionrequiringrehospitalization Infectionrequiringextra antibiotics/clinicvisits Infectiouseventsin patientsprescribed Cephalexin(n = 30) Co-trimoxazole(n = 63) Quinolones(n = 121) No antibiotic(n = 8) Infectiouseventsin patients With positiveurinalysis/culture(n = 37) With negativeurinalysis/culture(n = 53) With no urinalysis/culture(n = 132) Receivingantibioticsat admission (n = 45) Infectiousevents,by race Black (n = 50) Hispanic(n = 23) White (n = 143) TURP
= transurethral
resection
3 (1.4) 42 (18.9) 6 17 20 2
(20.0) (27.0) (16.5) (25.0)
11(29.7) 16(30.2) 18(13.6) 9 (20.0)
18(36.0) 2 (8.7) 25 ( 17.5)
of the prostate.
lergic to ampicillin were given vancomycin and gentamicin, cefazolin, or gentamicin alone. Becauseonly 4 patients received gentamicin alone, this group was added to the ampicillin-and-gentamicin group for the purposesof analysis. The pre-TURP antibiotic doses received by patients varied from 1 doseimmediately before surgery to 2 doses,1 the night before and another immediately before surgery. All patients continued to receive parenteral antibiotics for -1 to 2 days after the surgery and were discharged on 1 of 4 regimens: ofloxacin, cotrimoxazole, cephalexin, or no antibiotics. The duration of prophylactic oral antibiotic therapy varied from 1 to 21 days (mean, 9 days). Forty-four percent of patients were prescribed prophylactic oral antibiotics for >7 days. The outcomes identified included successful procedure without infectious com-
plications, infection requiring rehospitalization, and infection requiring additional doses of antibiotics or additional clinic visits (Table IV). A total of 45 infectious events were recorded in the 222 patients in the study: 3 patients developed infections (sepsisor pyelonephritis) requiring rehospitalization for another 3 days, and 42 patients developed infections requiring additional doses of antibiotics or clinic visits during the 4 weeks after TURP After controlling for comorbidities, race, age, indication for TURP, and catheterization before admission, linear regression analysis suggestedthat there was no difference in length of hospital stay for TURP between patients who received only 1 doseof preoperative antibiotic and thosewho received 2 doses(Table V). Moreover. there was no difference in length of stay for patients who received different parenteral antibiotic regimens 1597
CLINICAL
THERAPEUTICS@
Table V. Length of hospital stay model.* Variable
Intercept 4s Indication
(BPH)
Amp + gentor gentonly Van + gent
Amp + ceft Dose (1 dose) Comorbidities (>2) White male Operation time Year 1996 Year 1997 Year 1998
Estimate
SE
1.922 0.008 -0.123 0.294 -0.152 0.289 0.100 -0.166 0.232 0.005 0.143 -0.404 -0.548
0.654 0.007 0.165 0.241 0.326 0.304 0.141 0.135 0.139 0.001 0.208 0.196 0.282
BPH = benignprostatichyperplasia:amp= ampicillin:
P 0.0037 0.2496 0.4540 0.2244 0.6410 0.3425 0.4773 0.2 I87 0.0975.’ 0.0070~ 0.4939 0.0412” 0.0536:
gent = gentamicin; van = vancomycin; ceft
= ceftizoxime. ‘P = 0.0007:
r2 = 0.1467:
adjusted
r2 = 0.0977.
‘Significant at P < 0.10. ‘Significant at P < 0.01, $Significant at P < 0.0.5.
before TURP. nor did comorbidities greatly affect length of stay. However, length of stay was affected by certain other variables. In particular. patients who had longer operation times had significantly longer lengths of stay than did patients who had shorter operation times (P < 0.01). Patients who underwent TURP in 1997 and 1998 had significantly shorter lengths of stay (0.4 day and 0.5 day, respectively) than did those who underwent TURP in 1995 (P 5 0.05). In addition, white males seemed to have a longer length of stay than nonwhite males (0.2 day) (P = 0.10). In the evaluation of pre-TURP antibiotic prophylaxis, after controlling for the year of operation, our linear regression analysis of total medical cost showedthat patients who received the ampicillin-and1598
ceftizoxime regimen incurred a 14% higher total medical cost than patientswho received cefazolin. a result significant only at the 10% level (P = 0. IO) (Table VI). In the evaluation of post-TURP antibiotic prophylaxis, patients who received quinolonesincurred a significantly higher total medical cost (7%) than patients who received co-trimoxazole (P = 0.06). Again, patients who had a longer operation time also incurred a higher total cost than patients who had a shorter operation time (P < 0.01). Other variables, including age, indication for TURP, comorbidities, and year of operation, showed no significant direct relationship to total medical cost. The logistic regressionanalysis for the likelihood of infection (Table VII) found no evidence that the useof different types
G.G. LIU ET AL.
Table VI. Total medical cost model.* Variable
Estimate
SE
8.323 0.001 -0.021 0.092 -0.023 0.137 0.03x
0.183 0.002 0.046 0.067 0.091 0.085 0.041
Intercept Age Indication (BPH) Amp + gent or gent only Van + gent Amp + ceft IDose ,CI dose I (Cephakxjn Qtimhmes Duration of prophylaxis No oral antibiotics ([3nmoJtidjtjes White male (Dpemilm Year 1996 Year1997
DD9ZJ D-D73 -0.003 -0.096
ADD4 0.044
D.DD)
time
0.048 0.031 0.048
Year 1998
D.Dh2 D.D37 0.004
0.111 D.D-38 0.039 D.DD5 0.058 0.057 0.079
P 0.0001 0.6608 0.6518 0.1764 0.8007
0.1079+ 0.3479 D.II3D DsDf&' 0.3929 0.3879 1).9DB 0.2689 DDDI$ 0.4089 0.5850 0.5450
‘P = 0.0628: r? = 0.1130; adjusted ? = 0.0438. +Significant at P < 0.10. ‘Significant at P < 0.01. CD? Xh%~DiXS
M
tifik&l
bWai~DIJS
05 PID-
phylactic therapy affected the likelihood of infection among TURP patients. Patients with BPH, however, had a significantly lower rate of infection than did those with prostate cancer or other diseases.In addition, ptien~s wi-lh a p-y dria-itic diagnosis (eg, depression.bipolar disorder, or schizophrenia)andpatientswith coronary artery diseasewere significantly more likely to develop infection. There was no correlation between the rate of infection and the year surgery was performed.
DISCUSSION AND CONCLUSIONS The results of this srudy indicate that there is no difference in length of stay
6~ rale of in?ec$ion among TWX? patients receiving different parenteral and oral prophylactic antibiotic regimens for UTI. However, patients who received ampicillin and ceftizoxime incurred a higher total medical cost than patients receiving cefa2di-n at the 11% significance level. Patients who received oral quinolones also incurred a significantly higher total medical cost than did patients who received oral cotrimoxazole at the 6% significance level. The overall incidence of infection in this study was 20%, which seemsrather high compared with the findings of clinical trials evaluating prophylactic antimicrobial agents.10,‘1’22 However, because 1599
CLINICAL
THERAPEUTICS”
Table VII. Logistic regressionanalysisof infection rate. Variable
P > x2
Odds Ratio
Intercept Age Indication (BPH) Cephalexin Quinolones Duration of prophylaxis No antibiotic Coronary artery disease Chronic obstructive pulmonary disease Cancer Musculoskeletal diagnosis Psychiatric diagnosis Dementia Diabetes mellitus White male Operation time Year 1996 Year 1997 Year 1998
0.9995 0.658 1 0.0067” 0.7197 0.4758 0.7192 0.8018 0.0142.’ 0.9778 0.2435 0.2165 0.0151’ 0.1320 0.4946 0.3821 0.7619 0.2676 0.7595 0.1115
0.990 0.302 0.795 0.756 1.014 0.769 2.840 0.984 I.970 1.838 3.228 0.167 1.367 0.718 1.002 0.55 I 0.852 0.329
BPH = benign prostatic hyperplasia. *Significant at P < 0.01. +Significant at P < 0.05.
this study covered the 4-week period after TURP, it may have captured more infectious events than did other studies.(For example, if the study had covered only the 5 days after TURP, the incidence of infection would have been O%.) Another reason for the high rate of infection may be our inclusion of patients with both positive and negative results on urinalysis and of patients with catheters before surgery. The majority of studieshave excluded thesepatients. All infections occurred 26 days after TURP; therefore, it is unlikely that the parenteral antibiotics given before TURP were responsiblefor the treatment failure. In this study, 13.6% of patients with no documented urinalysis developed infec1600
tion, and 30% of patients who had undergone urinalysis/culture (whether the result was positive or negative) before the procedure developed infection. Reviews of the literature show that patients with preoperative bacteriuria have a higher risk of developing postoperative UT1 than do patients with sterile urine.’ In this study. because all patients were given prophylactic antibiotics before and after surgery and because there was no difference in the rate of infection among patients prescribed different antibiotic regimens, other factors must have contributed to the occurrence of postoperative infection. In fact, it appears that a history of depression, schizophrenia, bipolar disorder, or coronary artery diseasepredisposespa-
G.G.
LIU
ET AL.
tients to infection. TURP patients with prostate cancer also had a higher rate of infection than patients without prostate cancer, perhaps due to the high-risk profile of their disease. Because all patients were given prophylactic antibiotics and because the rate of infection was similar regardless of whether the results of pre-TURP urinalysis were positive or negative, urinalysis or culture and sensitivity testing may be unnecessary in patients undergoing this procedure unless the physician has reason to believe that an organism resistant to the usual antibiotics is present. Because there was no significant difference in the rate of UTIs in TURP patients receiving different types of parenteral or oral antibiotic prophylactic therapy, there is no justification for using more expensive antibiotic regimens (eg, ampicillin and ceftizoxime or quinolones). In addition, choosing a powerful antibiotic such as a quinolone for prophylaxis may predispose patients to infection with resistant organisms and result in superinfection. Based on the acquisition cost of antibiotics at our VA hospital, cefazolin would be the most cost-effective drug for pre-TURP prophylaxis, whereas oral cotrimoxazole would be the most costeffective agent for post-TURP prophylaxis. Because duration of therapy did not show any influence on the rate of infection, 24 hours of prophylaxis seems to be more cost-effective than longer periods of prophylaxis. Some limitations of this study are worth noting. First, because we used a retrospective database, we were not able to ensure that patients were randomized to the varous treatments. As a result, although we attempted to control for some observed confounding factors (eg, age, race, comor-
bidities, and year) while examining the treatment effects through a multivariate regression model. we were unable to control for any unobserved confounding factors. One such unobserved factor may be the unidentified health conditions that determined physicians’ choice of treatment. If unidentified conditions also contributed to the clinical outcomes under investigation, our results could have been influenced by a selection bias. To control for such bias, future studies might consider using the instrumental variables approach.23s2d Another limitation was the use of reimbursement data as a measure of cost. Because VA hospitals are nonprofit organizations, we assumed that the difference between reimbursement and cost was not significant. However, this assumption may not be appropriate to other types of institutions in which reimbursement differs substantially from actual cost. Given the retrospective nature of the data, our study was also limited by its relatively small sample size. Hypothetically, the insignificant between-treatment differences in outcome measures observed here could be partly attributed to the size of the sample. Thus future studies might consider improving the robustness of the findings by including more patients. In addition, the inclusion of more patient factors (ie, whether patients resided at home or in a nursing home. and whether they were compliant with the oral antibiotics) would better explain variations in the cost model.
Address correspondence to: Gordon Liu, PhD, Assistant Professor, Department of Pharmaceutical Economics and Policy, University of Southern California, 1540 E Alcazar Street, Room 140, Los Angeles, CA 90089. 1601
CLINICAL
institutions evaluating 3,885 patients. / Ural. 1989:141:243. Abstract.
REFERENCES 1. Benign Prostatic and Treatment.
Hyperplasia:
Diagnosis
Silver Spring, Md: Agency for Health Care Policy and Research. AHCPR Guideline 94-0582, February 1994.
2. Christensen MM, Madsen PO. Antimicrobial prophylaxis in transurethral surgery. Urology. 1990;35(Suppl): I l-14. 3. Grabe M. Antimicrobial agents in transurethral prostatic resection. J Ural. 1987; 138:245-252. 4. Hargreave TB, Botto H, Rikken GH. et al. European Collaborative Study of Antibiotic Prophylaxis for Transurethral Resection of the Prostate. Eur Ural. 1993;23: 437443. 5. Murphy DM, Falkiner FR, Carr M, et al. Septicemia after transurethral prostatectomy. Urology. 1993;12:133-135. 6. Chodak GW, Plaut ME. Systemic antibiotics for prophylaxis in urologic surgery: A critical review. J Ural. 1979;121:695. Abstract. 7. Madsen PO, Graversen PH. Antimicrobial prophylaxis in transurethral surgery. Infection. 1986:14:201. Abstract. 8. Home AM, rioperative transurethral justifiable?
Mokhless I. Sarto N, et al. Peantibiotic prophylaxis for resection of the prostate: Is it J Ural. 1989; 142:3 17-3 19.
9. Ho11 WH, Rous SN. Is antibiotic prophylaxis worthwhile in patients with transurethral resection of prostate? Urology. 1982; 19:43346. 10. Mebust WK, Holtgrewe HL, Cockett ATK, et al. Transurethral prostatectomy: Immediate and postoperative complications. A cooperative study of 13 participating
1602
THERAPEUTICS”
11. Te Slaa E, De Wildt MJ. Debruyne FM. Urinary tract infections following laser prostatectomy: Is there a need for antibiotic prophylaxis? Br J Ural. 1996~77: 228-232. 12. Williams M, Hole DJ, Murdoch RW. 48.Hour cephradine and post-prostatectomy bacteriuria. Br J Ural. 1980:52: 311-315. 13. Grabe M. Forsgren A, Bjork T. et al. Controlled trial of a short and a prolonged course with ciprofloxacin in patients undergoing transurethral resection of the prostate. Eur J C/in Microbial infect Dis. 1987:6: 1l-17. 14. Hall JC, Christiansen KJ, England P, et al. Antibiotic prophylaxis for patients undergoing transurethral resection of the prostate. Adult Ural. 1996;47:852-856. 15. Goldwasser B. Bogokowsky B. Nativ 0. Prophylaxis antimicrobial treatment in transurethral prostatectomy. How long should it be instituted? Urology. 1983;22: 136-138. 16. Gasser TC, Wisarad M. Frei F. Oral fleroxacin prophylaxis in transurethral surgery. J Ural.
1996:156:146-148.
17. Crawford ED, Berger NS, Davis MA. et al. Prevention of urinary tract infection and bacteremia following transurethral surgery: Oral lomefloxacin compared to parenteral cefotaxime. J Ural. 1992; 147: 1053-1055. 18. Finkelstein LH. Arsht al. Ceftriaxone in the operative infection in transurethral resection J Surg. 1994;19-21.
DB, Manfrey SJ, et prevention of postpatients undergoing of the prostate. Am
G.G. LIU ET AL.
19. Hargreave TB, Hindmarsh JR, Chisholm GD, et al. Short-term prophylaxis with cefotaxime for prostatic surgery. Br Med J. 1982;284:1008-1010. 20. Iversen P, Madsen PO. Short-term cephalosporin prophylaxis in transurethral surgery. C/in Ther. 1982;5(Suppl A): 58-66. 21. Gorbach SL, Condon RE, Conte JE, et al. General guidelines for the evaluation of new anti-infective drugs for prophylaxis of surgical infections. Clin Infect Dis. 1992;15(Suppl l):S313-S338.
22. Shapiro M, Townsend TR, Rosner B, et al. Use of antimicrobial drugs in general hospitals. N Engl J Med. 1979;361: 351-35s. 23. McClellan M, Newhouse J, McNeil B. Does more intensive treatment of acute myocardial infarction in the elderly reduce mortality? JAMA. 1994;11:859-866. 24. Harris K, Remler D. Who is the marginal patient? Understanding instrumental variables estimates of treatment effects. Health Serv Res. 1998;33:1337-1360.
1603
21, NO. 9. 1999
An Economic Analysis of Antimicrobial Prophylaxis Against Urinary Tract Infection in Patients Undergoing Transurethral Resection of the Prostate
Gordon G. Liu, PhD,’ Trisha Nguyen, MS, PharmD,2 and Michael B. Nichol, PhD’ ‘University of Southern California, Los Angel& California
and 2Department of Veterans Affairs,
ABSTRACT Despite the high level of safety and low incidence of mortality associated with transurethral resection of the prostate (TURP), urinary tract infections (UTIs)the most common complication associated with this procedure+ontinue to be an important source of postoperative morbidity and costs. However, there is controversy about whether antimicrobial agents should be usedasUT1 prophylaxis in patients undergoing TURP and, if so, which agentsshouldbe usedand for what duration. This retrospective study used multivariate regressionanalysis to evaluate the different types and durations of antibiotic prophylaxis in 222 patientswho underwent TURP at a Veterans Affairs hospital between January 1. 1995, and March 30, 1998. The primary outcome measures were total medical costs (ie, Accepted for publication Printed in the USA. Reproduction in whole
0149-2918/99/$19.00
March
26,
1999.
or part is not permitted
medication use, clinic office visits, and hospital care in the 4 weeks after the procedure), length of hospital stay (total days in hospital due to the procedure), and probability of UT1 (incidence of infection in the 4 weeks after the procedure). Results showedthat there was no difference in the length of hospital stay regardlessof the regimen or duration of pre-TURP antibiotic therapy. Patients who received pre-TURP ampicillin plus ceftizoxime incurred moderately higher total medical costs than did patients who received the least costly drug, cefazolin (P = 0.10). Similarly, patients who received postTURP quinolones incurred a significantly higher total medical cost than did patients who received co-trimoxazole (P = 0.06). We found no evidence of a relationship between use of specific parenteral or oral antibiotic prophylaxis for UT1 in patients undergoing TURP and the rate of UT1 in such patients. Thus there is no justification for the use of more expensive antibiotic regimens.At our institution, the preferred pre-TURP prophylactic regimen 1589
CLINICAL THERAPEUTICS”
would be cefazolin, whereas co-trimoxazole would be the most cost-effective post-TURP prophylactic regimen. Because duration of post-TURP prophylaxis does not appear to influence the rate of UTI, 24 hours would seem adequate. Key words: antibiotic therapy, prophylaxis, transurethral resection of the prostate, urinary tract infection. INTRODUCTION Transurethral resection of the prostate (TURP) is a procedure for relieving bladder outlet obstruction caused by benign prostatic hyperplasia (BPH) or cancer of the prostate. Among the most commonly performed surgical procedures, TURP is the gold standard for treating BPH and an accepted therapy for prostate cancer.’ Despite the relative safety and low incidence of mortality with TURP compared with other high-risk procedures such as coronary bypass surgery, urinary tract infections (UTIs)-the most common complication associated with the procedurecontinue to be an important source of postoperative morbidity and cost.? However, there continues to be controversy about whether to use antimicrobial prophylaxis against UT1 in patients undergoing TURP. We conducted a MEDLINE@ search of the literature for articles relating to the prophylactic use of antimicrobial agents for UT1 in association with TURP. In previous studies in TURP patients,sm5 antibiotics significantly decreased the number of infectious complications and eliminated bacteriuria in approximately 50% of cases. According to the findings of the European Collaborative Study of Antibiotic Prophylaxis for Transurethral Resection of the Prostate, “patients, even those with cul1590
ture-negative urine preoperatively, benefit from antibiotic prophylaxis and...antibiotic prophylaxis should be considered for all men undergoing prostatic surgery as a routine procedure.“” This randomized, controlled study in 157 patients with no bacteriuria before undergoing TURP found that patients who received no prophylaxis had a UT1 rate of 33.9%. compared with a rate of 11.6% in patients who received intravenous (IV) ceftazidime until removal of the urethral catheter (<5 days). In their 1979 literature review, Chodak and Plaut6 found no proof that prophylactic use of antimicrobial agents for UT1 prophylaxis in patients undergoing TURP was beneficial; nevertheless, a later study’ reported that prophylactic antimicrobial agents were administered to 160% of urologic patients in the United States and were given for >7 days in 49% of patients. In an open-label, prospective. randomized, controlled study in 1IO patients with sterile urine preoperatively who received either placebo or cefoperazone every 8 hours for 24 hours after surgery, only 1 patient in the control group developed infection (ie, positive culture with signs and symptoms of infection), compared with none in the cefoperazone group.” Even if clinicians were to agree on the need to administer prophylactic antibiotics. the ideal duration of such treatment is debatable. Some investigators have administered a single preoperative dose of antibiotic. others recommend 5 days of antibiotics after surgery, and still others believe in starting antibiotic therapy a day or so before surgery and continuing it for 2 to 3 weeks after surgery.9-tZ Mebust and colleagueslo reported on the use of antibiotics (unknown types) for UT1 prophylaxis in 61% of 3885 patients undergoing TURP in 13 institutions; 49% of the anti-
G.G. LIU ET AL.
biotic prophylaxis group received therapy for >8 days. The authors did not assess the correlation between infection rate and duration of prophylaxis. In a randomized study in urologic surgery patients, Grabe et alI3 compared the efficacy of short-term ciprofloxacin (3 to 4 days), long-term ciprofloxacin (8 to 9 days), and no antibiotic in 222 patients. At discharge, no patient in the short- or long-term prophylaxis group had bacteriuria, whereas 38.5% of the patients receiving no prophylaxis had bacteriuria at discharge. At -14 days after discharge, bacteriuria was detected in 20% of the short-term prophylaxis patients, 4.3% of the long-term prophylaxis patients, and 44.9% of those who had received no prophylaxis. In a prospective, randomized study of 171 patients, Hall et alI4 evaluated the efficacy of a single oral dose of fleroxacin, a single IV dose of fleroxacin, and an extended regimen consisting of an initial IV dose of fleroxacin followed by oral therapy until removal of the urinary catheter (~6 days); there were no differences in the rates of infection between the 3 groups. In another study, Goldwasser et alI5 administered 3 different regimens to 81 TURP patients who had sterile urine preoperatively: group A received parenteral co-trimoxazole, 1 dose preoperatively and 1 dose postoperatively, and then continued on oral antibiotics for 10 days; group B received only the parenteral cotrimoxazole regimen; and group C received nothing. The investigators found that by day 7, the rate of infection in groups A and B was 3.8%, compared with 32% in group C. Therefore, these investigators recommended a short course of antimicrobial prophylaxis. The choice of antibiotic is another area of controversy. Some clinicians use an-
tibiotics known to penetrate prostate tissue well (ie, the quinolones), whereas others prefer to use antibiotics that are known to be effective against the organisms likely to cause UTIs (eg, Escherichia coli. Proteus spp).“,t6 Prophylactic ampicillin, sulfonamides, tetracycline, nitrofurantoin, cephalosporins, and quinolones have all been studied in TURP patients.4~6*‘6~19 Most of these studies were placebo controlled. Crawford et al,17 however, randomized patients to a single dose of oral lomefloxacin or IV cefotaxime and found no significant difference in the rate of infection between the 2 groups. A prospective, randomized, double-masked study by Iversen and Madsen’O compared 3 regimens of cephalosporin in 116 men who had sterile urine preoperatively. One group received 1 g cefotaxime IV 30 to 90 minutes before surgery, every 60 to 90 minutes during surgery, and immediately after surgery; of the other 2 groups, 1 received 1 g cefotaxime and the other received 1 g cefazolin, following the same regimen used in the first group and every 8 hours for 24 hours postoperatively. The 3 groups had a similar incidence of infection 7 days postoperatively. In summary, our literature review found little consensus on the effectiveness of antibiotic prophylaxis for UT1 in patients undergoing TURP. The weaknesses of some of these studies included lack of a double-masked design, small number of study subjects, lack of placebo control, and variations in definition of UTI. There are currently few data on the comparative effectiveness of different antibiotics for the prophylaxis of UT1 in TURP patients and even fewer data defining the appropriate duration of such therapy. The objective of the present study was to evaluate the costs and effectiveness of 1591
CLINICAL
different antibiotics and durations of antibiotic prophylaxis against UT1 in patients undergoing TURP. Two specific issues were addressed:(1) whether the rate of infection differs significantly between patients prescribed different prophylactic regimens before and after TURP, and (2) whether and how the length of hospital stay and total medical cost differ between patients prescribed different prophylactic antibiotics for different durations. Our perspective was that of a Veterans Affairs (VA) hospital, and our results may help similar hospitalsinvestigating ways of providing TURP in the most cost-effective manner. MATERIALS
AND METHODS
This retrospective study was conducted at a VA hospital in West Los Angeles. California. It was designedto analyze the experience of patients during 1 episode of hospitalization for TURP and for 4 weeks after discharge, a sufficient period for the development of possible infectious complications of TURP and for assessingthe benefits of antimicrobial prophylaxis.” The fact that a patient may have undergone >I TURP procedure during the study should not affect the validity of the study results, since each TURP event was assessedindependently. Our analysis applies to all patients being considered for TURP, whetherfor BPH or prostatecancer. From the hospital database.we identified 222 patients who had undergone TURP between January 1, 1995, and March 30, 1998. Patients were excluded if they had severe infection causing their transfer to the intensive care unit before the procedure. Data recorded from patients’ charts included date of surgery; date of discharge; birth date and age; in1592
THERAPEUTICS@
dication for TURP; comorbidities; presence of an indwelling Foley catheter; presence of bacteriuria on urinalysis before surgery; antibiotics received for an existing infection at the time of admission; prophylactic antibiotic regimen and duration of therapy before and after TURP; operative time (defined asthe time in minutes from operating table to recovery room); and postsurgicalinfection (defined as any additional post-TURP intervention for which the diagnosiswas UTI, including additional antibiotic prescriptions, genitourinary clinic or emergency department visits, or hospitalizations for UTI). The number of postsurgical infections and number of clinic visits were calculated for the 4-week period after the procedure. According to general guidelines for the evaluation of new anti-infective therapies for the prophylaxis of surgical infections, any unexplained use of anti-infective agentsin the 4-week period after surgery should be regarded as failure of prophylaxis, as should development of postoperative bacteriuria. with or without systemic signsof infection. I5 However, our review of patients’ charts found that due to the possible transient nature of these events, physicians had not treated some patients having a single increase in white blood cell count, temperature, or bacteriuria without symptoms.In accordancewith the physicians’judgment, we did not consider these events infections. Patient comorbidities included in our chart analysis were hypertension, coronary artery disease, chronic obstructive pulmonary disease,cancer, dementia,diabetes mellitus, degenerative joint disorder, psychiatric illness (eg, depression, bipolar disorder, and schizophrenia), and Parkinson’s disease.These comorbidities
G.G. LIU ET AL.
Table I. Description
of variables
used in the study.
Variable
Age
Type
Operation time Race Use of urethral catheter before TURP Positive bacteriuria test or culture Receiving antibiotics at time of admission No. of patients undergoing TURP Indication for TURP Pre-TURP antibiotic regimen Amp + gent Van + gent Amp + ceft Cefazolin (control option) Dose Post-TURP antibiotic regimen Cephalexin Quinolones Co-trimoxazole (control option) None Duration of prophylaxis Comorbidities (diagnoses recorded in medical history) Hypertension Coronary artery disease Chronic obstructive pulmonary disease Cancer Parkinson’s disease Dementia Diabetes mellitus Degenerative joint disorder Psychiatric illness (depression or schizophrenia) Year of TURP 1995 1996 1997 1998 = transurethral resection of the prostate; BPH gentamicin; van = vancomycin; ceft = ceftizoxime.
TURP
are rarely investigated in relation to TURP2 but were included in the present study for completeness and to control for extraneous variables. Other comorbidities
Continuous Continuous 1 = white; 0 = other 1 = yes; 0 = no I = yes; 0 = no 1 = yes; 0 = no Continuous 1 = BPH; 0 = prostate cancer or other 1= 1= 1= 1= 1=
amp + gent; 0 = other van + gent; 0 = other amp + ceft; 0 = other cefazolin; 0 = other 1 dose; 0 = >l dose
1 = yes; 0 = other 1 = yes; 0 = other 1 = yes: 0 = other Continuous (days) 1= 1= 1= 1= 1= 1= 1= 1= 1=
yes; 0 yes; 0 yes; 0 yes; 0 yes; 0 yes; 0 yes; 0 yes; 0 yes; 0
= = = = = = = = =
no no no no no no no no no
1 =95;O=other 1 = 96; 0 = other 1 = 97; 0 = other 1 = 98; 0 = other
= benign prostatic hyperplasia; amp = ampicillin; gent =
occurring in ~1% of patients pothyroidism, gout, and anemia) controlled for in our analysis. Length of stay was calculated
(eg, hywere not from the 1593
CLINICAL
day the patient underwent TURP to the day of discharge. If the patient was kept in the hospital for a reason unrelated to TURP (eg, admission to a psychiatric ward or delay in nursing home admission), the day the patient was discharged from the genitourinary service was considered the dischargedate, since this was the date at which the patient was considered stable enough to be discharged.The same criterion was applied to patients transferred to a nursing home or intermediate facility. However, if a patient was transferred to the intensive care unit, the period spent in the unit was included in the length of stay. The costs evaluated included acquisition costs of the prophylactic antibiotics and costs of additional clinic visits and hospitalizations associated with TURP. These costs were obtained from the VA reimbursement databasefor surgical patients and adjusted for inflation using the medical care component of the Consumer Price Index. Total medical costs were expressedin 1998 dollars. Antibiotic Regimensand Outcome Measures We considered 4 pre-TURP antibiotic regimens:ampicillin and gentamicin, vancomycin and gentamicin, ampicillin and ceftizoxime. and cefazolin alone. Prophylactic antibiotic administration started either immediately before surgery or the night before surgery.All patientswere continued on parenteral antibiotics for 1 to 2 days after surgery and before discharge. Four post-TURP antibiotic regimenswere considered: co-trimoxazole, quinolones, cephalexin, and none. These oral medications were given to patients at discharge and continued for various durations. 1594
THERAPEUTICF’
Three outcome measureswere used: meanlength of hospital stay for the TURP procedure. total direct cost, and incidence of post-TURP UTI. Indirect costssuch as lost productivity at home and lost earnings at work were not included, mainly becauseindirect costs are not significant from a payer’s perspective. Statistical Analysis Linear regressionanalysis was used to assessthe effects of the various pre-TURP and post-TURP antibiotic regimenson the 3 outcome measures:length of stay, total medical costs, and incidence of infection. Ordinary least squaresmethodology was employed to estimatethe 3 outcome models using SAS software (SAS Institute, Cary, North Carolina). The variables assessedin the study are defined in Table I. For hospital length of stay, we used a multivariate model that measuredthe effect of pre-TURP antibiotics, with cefazolin as a control. while controlling for certain other factors that may influence this variable. In particular, the model was designedto include type of parenteral antibiotics given before TURP (ampicillin and gentamicin, vancomycin and gentamicin, ampicillin and ceftizoxime, or cefazolin). age, indication for surgery (BPH or prostate cancer), number of doses of pre-TURP antibiotic (1 vs 2), number of comorbidities (~2 vs 52 diagnoses),race (white or nonwhite), and duration of the TURP operation (a continuous variable). Post-TURP antibiotics were not included in this model becausethese medications were generally given just before discharge and thus had no influence on length of stay. Becauseboth could contribute to total medical cost, the effects of pre- and post-
G.G. LIU ET AL.
Table II. Characteristicsof the study population at baseline(N = 222). No. (‘3%) of Patients
Characteristic Age (y) <65 26.5 Unknown Race White Nonwhite Black Hispanic Asian Use of urethral catheter before TURP None or unknown Yes Bacteriuria test or culture Positive Negative Undocumented urinalysis
81 (36.5) 119 (53.6) 22 (9.9) 143 79 50 23 6
187 (84.2) 35 (15.8) 37 53 132 45
Receivingantibioticsat time of admission Comorbidities Coronaryartery disease COPD Cancer Dementia Diabetes mellitus Degenerative joint disorder Psychiatric illness (depression or schizophrenia) Year of TURP 1995 1996 1997 1998 = chronic
obstructive
pulmonary
37 66 80 39
disease; TURP = transurethral
TURP antibiotic regimens on total cost were modeledin a secondmultivariate regression.Independentvariables controlled for in this model were age. indication for TURP (BPH or prostate cancer), number of dosesof pre-TURP antibiotic (1 vs 2), duration of post-TURP antibiotic, number
(16.7) (23.9) (59.5) (20.3)
110 (49.5) 49 (22.0) 26 (11.7) 22 (9.9) 11 (5.0) 36 (16.2) 35 (15.8) 38 (17.1)
Hypertension
COPD
(64.4) (35.6) (22.5) (10.4) (2.7)
resection
(16.7) (29.7) (36.0) (17.6)
of the prostate
of comorbidities (>2 vs <2 diagnoses), race (white or nonwhite), and duration of the TURP operation(a continuousvariable). A third model was developed using logistic regressionto test the effect of postTURP prophylactic antibiotic therapy on the outcome of interest, incidence of in1595
CLINICAL
THERAPEUTICS’
Table III. Prophylactic regimens used against urinary tract infection in 222 patients undergoing TURP. No. (%) of Patients
Regimen Pre-TURP No. of doses One (before surgery) Two (started the night before surgery) Parenteral antibiotics Ampicillin + gentamicin Vancomycin + gentamicin Gentamicin Cefazolin Ampicillin + ceftizoxime Post-TURP (oral antibiotics) Ofloxacin or ciprofloxacin Co-trimoxazole Cephalexin None TURP
= transurethral
resection
87 (39.2) 154 17 4 23 24
(69.4) (7.7)
121 63 30 8
(54.5) (28.4) (13.5) (3.6)
(1.8) ( 10.4) (10.8)
of the prostate.
fection. Because all infections occurred 26 days after discharge, this model did not include parenteral antibiotics given before TURP, asit is unlikely their effects would have lasted long enough to intluencethe post-dischargeinfection rate. The independent variables included in this model were age, types of oral antibiotics, duration of oral antibiotics, comorbidities, race, operation time, and year (to control for possible selection bias). To control for any selection bias introduced by developments in surgical technique or in methods of administering the antibiotics, we included year dummy variables for 1995 (control group), 1996, 1997, and 1998 in all models. RESULTS Characteristicsof the study population are summarized in Table II. The mean age of 1596
135 (60.8)
the subjects was 69 years, and the mean length of hospital stay was 3 days (range. 2 to 8 days); the mean (k SD) total medical cost was $6000 + $1990 per patient. At the time of admission, 17% of the patients had positive results on urinalysis, and 24% had negative results; the rest had no documentedurinalysis before TURP. It is possible that surgeonsat this hospital do not seethe need to order urinalysis or urine culture before surgery becausethey routinely give prophylaxis to every patient. Therefore, preoperative urine culture was not included
in the regression
analysis.The mean operating time was 69 minutes. All patients received prophylactic therapy. but there were several variations in the regimens used (Table III). Before TURP the patients in this study received either ampicillin and gentamicin or ampicillin and ceftizoxime; thosewho were al-
G.G. LIU ET AL.
Table IV. Incidence of the outcomesof interest during the 4 weeks after TURP (N = 222). Event
No. (%) of Patients
Infectionrequiringrehospitalization Infectionrequiringextra antibiotics/clinicvisits Infectiouseventsin patientsprescribed Cephalexin(n = 30) Co-trimoxazole(n = 63) Quinolones(n = 121) No antibiotic(n = 8) Infectiouseventsin patients With positiveurinalysis/culture(n = 37) With negativeurinalysis/culture(n = 53) With no urinalysis/culture(n = 132) Receivingantibioticsat admission (n = 45) Infectiousevents,by race Black (n = 50) Hispanic(n = 23) White (n = 143) TURP
= transurethral
resection
3 (1.4) 42 (18.9) 6 17 20 2
(20.0) (27.0) (16.5) (25.0)
11(29.7) 16(30.2) 18(13.6) 9 (20.0)
18(36.0) 2 (8.7) 25 ( 17.5)
of the prostate.
lergic to ampicillin were given vancomycin and gentamicin, cefazolin, or gentamicin alone. Becauseonly 4 patients received gentamicin alone, this group was added to the ampicillin-and-gentamicin group for the purposesof analysis. The pre-TURP antibiotic doses received by patients varied from 1 doseimmediately before surgery to 2 doses,1 the night before and another immediately before surgery. All patients continued to receive parenteral antibiotics for -1 to 2 days after the surgery and were discharged on 1 of 4 regimens: ofloxacin, cotrimoxazole, cephalexin, or no antibiotics. The duration of prophylactic oral antibiotic therapy varied from 1 to 21 days (mean, 9 days). Forty-four percent of patients were prescribed prophylactic oral antibiotics for >7 days. The outcomes identified included successful procedure without infectious com-
plications, infection requiring rehospitalization, and infection requiring additional doses of antibiotics or additional clinic visits (Table IV). A total of 45 infectious events were recorded in the 222 patients in the study: 3 patients developed infections (sepsisor pyelonephritis) requiring rehospitalization for another 3 days, and 42 patients developed infections requiring additional doses of antibiotics or clinic visits during the 4 weeks after TURP After controlling for comorbidities, race, age, indication for TURP, and catheterization before admission, linear regression analysis suggestedthat there was no difference in length of hospital stay for TURP between patients who received only 1 doseof preoperative antibiotic and thosewho received 2 doses(Table V). Moreover. there was no difference in length of stay for patients who received different parenteral antibiotic regimens 1597
CLINICAL
THERAPEUTICS@
Table V. Length of hospital stay model.* Variable
Intercept 4s Indication
(BPH)
Amp + gentor gentonly Van + gent
Amp + ceft Dose (1 dose) Comorbidities (>2) White male Operation time Year 1996 Year 1997 Year 1998
Estimate
SE
1.922 0.008 -0.123 0.294 -0.152 0.289 0.100 -0.166 0.232 0.005 0.143 -0.404 -0.548
0.654 0.007 0.165 0.241 0.326 0.304 0.141 0.135 0.139 0.001 0.208 0.196 0.282
BPH = benignprostatichyperplasia:amp= ampicillin:
P 0.0037 0.2496 0.4540 0.2244 0.6410 0.3425 0.4773 0.2 I87 0.0975.’ 0.0070~ 0.4939 0.0412” 0.0536:
gent = gentamicin; van = vancomycin; ceft
= ceftizoxime. ‘P = 0.0007:
r2 = 0.1467:
adjusted
r2 = 0.0977.
‘Significant at P < 0.10. ‘Significant at P < 0.01, $Significant at P < 0.0.5.
before TURP. nor did comorbidities greatly affect length of stay. However, length of stay was affected by certain other variables. In particular. patients who had longer operation times had significantly longer lengths of stay than did patients who had shorter operation times (P < 0.01). Patients who underwent TURP in 1997 and 1998 had significantly shorter lengths of stay (0.4 day and 0.5 day, respectively) than did those who underwent TURP in 1995 (P 5 0.05). In addition, white males seemed to have a longer length of stay than nonwhite males (0.2 day) (P = 0.10). In the evaluation of pre-TURP antibiotic prophylaxis, after controlling for the year of operation, our linear regression analysis of total medical cost showedthat patients who received the ampicillin-and1598
ceftizoxime regimen incurred a 14% higher total medical cost than patientswho received cefazolin. a result significant only at the 10% level (P = 0. IO) (Table VI). In the evaluation of post-TURP antibiotic prophylaxis, patients who received quinolonesincurred a significantly higher total medical cost (7%) than patients who received co-trimoxazole (P = 0.06). Again, patients who had a longer operation time also incurred a higher total cost than patients who had a shorter operation time (P < 0.01). Other variables, including age, indication for TURP, comorbidities, and year of operation, showed no significant direct relationship to total medical cost. The logistic regressionanalysis for the likelihood of infection (Table VII) found no evidence that the useof different types
G.G. LIU ET AL.
Table VI. Total medical cost model.* Variable
Estimate
SE
8.323 0.001 -0.021 0.092 -0.023 0.137 0.03x
0.183 0.002 0.046 0.067 0.091 0.085 0.041
Intercept Age Indication (BPH) Amp + gent or gent only Van + gent Amp + ceft IDose ,CI dose I (Cephakxjn Qtimhmes Duration of prophylaxis No oral antibiotics ([3nmoJtidjtjes White male (Dpemilm Year 1996 Year1997
DD9ZJ D-D73 -0.003 -0.096
ADD4 0.044
D.DD)
time
0.048 0.031 0.048
Year 1998
D.Dh2 D.D37 0.004
0.111 D.D-38 0.039 D.DD5 0.058 0.057 0.079
P 0.0001 0.6608 0.6518 0.1764 0.8007
0.1079+ 0.3479 D.II3D DsDf&' 0.3929 0.3879 1).9DB 0.2689 DDDI$ 0.4089 0.5850 0.5450
‘P = 0.0628: r? = 0.1130; adjusted ? = 0.0438. +Significant at P < 0.10. ‘Significant at P < 0.01. CD? Xh%~DiXS
M
tifik&l
bWai~DIJS
05 PID-
phylactic therapy affected the likelihood of infection among TURP patients. Patients with BPH, however, had a significantly lower rate of infection than did those with prostate cancer or other diseases.In addition, ptien~s wi-lh a p-y dria-itic diagnosis (eg, depression.bipolar disorder, or schizophrenia)andpatientswith coronary artery diseasewere significantly more likely to develop infection. There was no correlation between the rate of infection and the year surgery was performed.
DISCUSSION AND CONCLUSIONS The results of this srudy indicate that there is no difference in length of stay
6~ rale of in?ec$ion among TWX? patients receiving different parenteral and oral prophylactic antibiotic regimens for UTI. However, patients who received ampicillin and ceftizoxime incurred a higher total medical cost than patients receiving cefa2di-n at the 11% significance level. Patients who received oral quinolones also incurred a significantly higher total medical cost than did patients who received oral cotrimoxazole at the 6% significance level. The overall incidence of infection in this study was 20%, which seemsrather high compared with the findings of clinical trials evaluating prophylactic antimicrobial agents.10,‘1’22 However, because 1599
CLINICAL
THERAPEUTICS”
Table VII. Logistic regressionanalysisof infection rate. Variable
P > x2
Odds Ratio
Intercept Age Indication (BPH) Cephalexin Quinolones Duration of prophylaxis No antibiotic Coronary artery disease Chronic obstructive pulmonary disease Cancer Musculoskeletal diagnosis Psychiatric diagnosis Dementia Diabetes mellitus White male Operation time Year 1996 Year 1997 Year 1998
0.9995 0.658 1 0.0067” 0.7197 0.4758 0.7192 0.8018 0.0142.’ 0.9778 0.2435 0.2165 0.0151’ 0.1320 0.4946 0.3821 0.7619 0.2676 0.7595 0.1115
0.990 0.302 0.795 0.756 1.014 0.769 2.840 0.984 I.970 1.838 3.228 0.167 1.367 0.718 1.002 0.55 I 0.852 0.329
BPH = benign prostatic hyperplasia. *Significant at P < 0.01. +Significant at P < 0.05.
this study covered the 4-week period after TURP, it may have captured more infectious events than did other studies.(For example, if the study had covered only the 5 days after TURP, the incidence of infection would have been O%.) Another reason for the high rate of infection may be our inclusion of patients with both positive and negative results on urinalysis and of patients with catheters before surgery. The majority of studieshave excluded thesepatients. All infections occurred 26 days after TURP; therefore, it is unlikely that the parenteral antibiotics given before TURP were responsiblefor the treatment failure. In this study, 13.6% of patients with no documented urinalysis developed infec1600
tion, and 30% of patients who had undergone urinalysis/culture (whether the result was positive or negative) before the procedure developed infection. Reviews of the literature show that patients with preoperative bacteriuria have a higher risk of developing postoperative UT1 than do patients with sterile urine.’ In this study. because all patients were given prophylactic antibiotics before and after surgery and because there was no difference in the rate of infection among patients prescribed different antibiotic regimens, other factors must have contributed to the occurrence of postoperative infection. In fact, it appears that a history of depression, schizophrenia, bipolar disorder, or coronary artery diseasepredisposespa-
G.G.
LIU
ET AL.
tients to infection. TURP patients with prostate cancer also had a higher rate of infection than patients without prostate cancer, perhaps due to the high-risk profile of their disease. Because all patients were given prophylactic antibiotics and because the rate of infection was similar regardless of whether the results of pre-TURP urinalysis were positive or negative, urinalysis or culture and sensitivity testing may be unnecessary in patients undergoing this procedure unless the physician has reason to believe that an organism resistant to the usual antibiotics is present. Because there was no significant difference in the rate of UTIs in TURP patients receiving different types of parenteral or oral antibiotic prophylactic therapy, there is no justification for using more expensive antibiotic regimens (eg, ampicillin and ceftizoxime or quinolones). In addition, choosing a powerful antibiotic such as a quinolone for prophylaxis may predispose patients to infection with resistant organisms and result in superinfection. Based on the acquisition cost of antibiotics at our VA hospital, cefazolin would be the most cost-effective drug for pre-TURP prophylaxis, whereas oral cotrimoxazole would be the most costeffective agent for post-TURP prophylaxis. Because duration of therapy did not show any influence on the rate of infection, 24 hours of prophylaxis seems to be more cost-effective than longer periods of prophylaxis. Some limitations of this study are worth noting. First, because we used a retrospective database, we were not able to ensure that patients were randomized to the varous treatments. As a result, although we attempted to control for some observed confounding factors (eg, age, race, comor-
bidities, and year) while examining the treatment effects through a multivariate regression model. we were unable to control for any unobserved confounding factors. One such unobserved factor may be the unidentified health conditions that determined physicians’ choice of treatment. If unidentified conditions also contributed to the clinical outcomes under investigation, our results could have been influenced by a selection bias. To control for such bias, future studies might consider using the instrumental variables approach.23s2d Another limitation was the use of reimbursement data as a measure of cost. Because VA hospitals are nonprofit organizations, we assumed that the difference between reimbursement and cost was not significant. However, this assumption may not be appropriate to other types of institutions in which reimbursement differs substantially from actual cost. Given the retrospective nature of the data, our study was also limited by its relatively small sample size. Hypothetically, the insignificant between-treatment differences in outcome measures observed here could be partly attributed to the size of the sample. Thus future studies might consider improving the robustness of the findings by including more patients. In addition, the inclusion of more patient factors (ie, whether patients resided at home or in a nursing home. and whether they were compliant with the oral antibiotics) would better explain variations in the cost model.
Address correspondence to: Gordon Liu, PhD, Assistant Professor, Department of Pharmaceutical Economics and Policy, University of Southern California, 1540 E Alcazar Street, Room 140, Los Angeles, CA 90089. 1601
CLINICAL
institutions evaluating 3,885 patients. / Ural. 1989:141:243. Abstract.
REFERENCES 1. Benign Prostatic and Treatment.
Hyperplasia:
Diagnosis
Silver Spring, Md: Agency for Health Care Policy and Research. AHCPR Guideline 94-0582, February 1994.
2. Christensen MM, Madsen PO. Antimicrobial prophylaxis in transurethral surgery. Urology. 1990;35(Suppl): I l-14. 3. Grabe M. Antimicrobial agents in transurethral prostatic resection. J Ural. 1987; 138:245-252. 4. Hargreave TB, Botto H, Rikken GH. et al. European Collaborative Study of Antibiotic Prophylaxis for Transurethral Resection of the Prostate. Eur Ural. 1993;23: 437443. 5. Murphy DM, Falkiner FR, Carr M, et al. Septicemia after transurethral prostatectomy. Urology. 1993;12:133-135. 6. Chodak GW, Plaut ME. Systemic antibiotics for prophylaxis in urologic surgery: A critical review. J Ural. 1979;121:695. Abstract. 7. Madsen PO, Graversen PH. Antimicrobial prophylaxis in transurethral surgery. Infection. 1986:14:201. Abstract. 8. Home AM, rioperative transurethral justifiable?
Mokhless I. Sarto N, et al. Peantibiotic prophylaxis for resection of the prostate: Is it J Ural. 1989; 142:3 17-3 19.
9. Ho11 WH, Rous SN. Is antibiotic prophylaxis worthwhile in patients with transurethral resection of prostate? Urology. 1982; 19:43346. 10. Mebust WK, Holtgrewe HL, Cockett ATK, et al. Transurethral prostatectomy: Immediate and postoperative complications. A cooperative study of 13 participating
1602
THERAPEUTICS”
11. Te Slaa E, De Wildt MJ. Debruyne FM. Urinary tract infections following laser prostatectomy: Is there a need for antibiotic prophylaxis? Br J Ural. 1996~77: 228-232. 12. Williams M, Hole DJ, Murdoch RW. 48.Hour cephradine and post-prostatectomy bacteriuria. Br J Ural. 1980:52: 311-315. 13. Grabe M. Forsgren A, Bjork T. et al. Controlled trial of a short and a prolonged course with ciprofloxacin in patients undergoing transurethral resection of the prostate. Eur J C/in Microbial infect Dis. 1987:6: 1l-17. 14. Hall JC, Christiansen KJ, England P, et al. Antibiotic prophylaxis for patients undergoing transurethral resection of the prostate. Adult Ural. 1996;47:852-856. 15. Goldwasser B. Bogokowsky B. Nativ 0. Prophylaxis antimicrobial treatment in transurethral prostatectomy. How long should it be instituted? Urology. 1983;22: 136-138. 16. Gasser TC, Wisarad M. Frei F. Oral fleroxacin prophylaxis in transurethral surgery. J Ural.
1996:156:146-148.
17. Crawford ED, Berger NS, Davis MA. et al. Prevention of urinary tract infection and bacteremia following transurethral surgery: Oral lomefloxacin compared to parenteral cefotaxime. J Ural. 1992; 147: 1053-1055. 18. Finkelstein LH. Arsht al. Ceftriaxone in the operative infection in transurethral resection J Surg. 1994;19-21.
DB, Manfrey SJ, et prevention of postpatients undergoing of the prostate. Am
G.G. LIU ET AL.
19. Hargreave TB, Hindmarsh JR, Chisholm GD, et al. Short-term prophylaxis with cefotaxime for prostatic surgery. Br Med J. 1982;284:1008-1010. 20. Iversen P, Madsen PO. Short-term cephalosporin prophylaxis in transurethral surgery. C/in Ther. 1982;5(Suppl A): 58-66. 21. Gorbach SL, Condon RE, Conte JE, et al. General guidelines for the evaluation of new anti-infective drugs for prophylaxis of surgical infections. Clin Infect Dis. 1992;15(Suppl l):S313-S338.
22. Shapiro M, Townsend TR, Rosner B, et al. Use of antimicrobial drugs in general hospitals. N Engl J Med. 1979;361: 351-35s. 23. McClellan M, Newhouse J, McNeil B. Does more intensive treatment of acute myocardial infarction in the elderly reduce mortality? JAMA. 1994;11:859-866. 24. Harris K, Remler D. Who is the marginal patient? Understanding instrumental variables estimates of treatment effects. Health Serv Res. 1998;33:1337-1360.
1603