Child's Nerv Syst (1992) 8:253-257 9 Springer-Verlag 1992

A randomized controlled trial of perioperative rifampin]trimethoprim in cerebrospinal fluid shunt surgery * Beverly C. Waiters 1,2, Liliana Goumnerova 3, Harold J. Hoffman 4, E. Bruce Hendrick 4, Robin P. Humphreys 4, and Carey Levinton 2 1 Division of Neurosurgery and a Clinical Epidemiology Unit, Sunnybrook Health Science Centre, Toronto, Ontario, Canada 3 Department of Neurosurgery, The Children's Hospital, Boston, Massachusetts, USA 4 Division of Neurosurgery, Hospital for Sick Children, Toronto, Ontario, Canada Received February 23, 1992

Abstract. A randomized, double-blind, placebo-controlled trial of perioperative rifampin-trimethoprim was undertaken at the Hospital for Sick Children from March 1984 to October 1987, in which 243 patients undergoing 300 cerebrospinal fluid (CSF) shunting procedures were randomized into groups including treatment with rifampin/trimethoprim and placebo, and then followed for a minimum of 2 years. Patients were stratified prior to randomization into those with and those without meningomyeloceles having first insertions of their shunts, and those having revisions. Patients could be entered into the study more than once, but always received the same treatment regimen once allocation had taken place. Among the patients receiving antibiotics there was an infection rate of 12%, versus 19% among patients receiving placebo. Among the surgical procedures, the rates were 9% and 15%, respectively. Because these rates of infection were a substantial increase over the rate of 7.5% overall for the few years prior to implementation of the study, and well over any acceptable rate of infection, the study was stopped before statistical significance was reached. However, had the study continued, and the proportions of patients becoming infected remained constant, we would have been able to achieve a statistically significant difference in rates of infection, and therefore demonstrate a benefit of rifampin/trimethoprim as prophylaxis against shunt infection. Methodological problems encountered in this and other studies of prevention of CSF shunt infection will be discussed. Key words: Antibiotics - Controlled trial - CSF shunt Infection - Prevention - Randomization

strating the unquestioned benefit of shunts to patients with hydrocephalus [16], these prosthetic devices have become the mainstay of treatment. Within a short time of their being introduced, however, the complications of malfunction and infection were recognized [6, 8, 17, 24, 25, 31]. CSF shunt infections are difficult to treat because of the combined need to treat the infection (which may involve removal of the device) while continuing to treat the hydrocephalus. This therapeutic challenge has led to much attention being directed toward prevention. Since most infections are a product of the surgical procedure itself [3, 7], the use of various methods of intervention during surgery has arisen. The most notable of these is the use of "antibiotic cover", or perioperative antibiotics as prophylaxis against shunt infection. Several studies have undertaken to demonstrate the efficacy of prophylactic antibiotics, but all have suffered from methodological flaws. Many studies were case series or other uncontrolled designs open to numerous biases [1, 20, 23, 27, 30]. However, even the randomized controlled trials were open to criticism, the most common error being that of small numbers of patients insufficient to establish the appropriate power of the study [2, 9, 12, 22, 26, 29]. Of all the trials carried out, only one had sufficient power to demonstrate a statistically significant difference [4]. In an attempt to find efficacious therapy for preventing shunt infection, we instituted a randomized controlled trial of a combination of trimethoprim and rifampin [Rifamprim, Merrell Dow Pharmaceuticals (Canada) Inc., Richmond Hill, Ontario, Canada] to assess its ability to decrease the rate of infection in treated patients. This trial was conducted at the Hospital for Sick Children in Toronto from March, 1984, through October, 1987.

Since the introduction of CSF shunts in the 1950s and Lorber's definitive randomized controlled trial demon* Presented at the 1989 Meeting of the American Association of Neurological Surgeons, Washington, D.C. Correspondence to: B. C. Walters, Division of Neurosurgery, Sunnybrook Health Science Centre, 2075 Bayview Avenue, North York, Ontario M4N 3M5, Canada

Methods and materials Eligibility criteria

Patients were eligible for the study if they were undergoing a CSF shunting procedure, were not known to have a shunt infection or

254 any concomitant systemic infection, had no allergies to either rifampin or trimethoprim, and informed consent was given either by the patient (if aged 16 or over) or a parent/guardian. Patients were excluded if unable to meet these criteria.

Patient allocation Once patient eligibility was established and consent obtained, patients were stratified into the following groups:

Group A: Patients with meningomyelocele undergoing first insertion

Group B. Patients without meningomyelocele undergoing first insertion

Group C: Patients undergoing shunt revision, regardless of the underlying cause of hydrocephalus, not previously entered into the study

Group D: Patients undergoing shunt revision, previously entered into group A, B, or C The patients thus stratified were then randomized into treatment or control groups by the research pharmacist at the Hospital for Sick Children using a block randomization schema of four patients at a time for a given stratum. Patients could be entered into the study more than once, but they were always treated by the same regimen as dictated by randomization in the initial entry in the study. If a patient with a malfunctioning shunt was recruked into the study, randomized to one of the treatment regimens, and then found at operation to have colonized CSF or shunt tubing, that patient was withdrawn at the time of analysis.

Drug administration Once a patient was recruited into the study, the Doctor's Orders sheets indicated that this patient was to be randomized according to the above grouping, and the patient's weight was included. The randomized drug was then supplied in oral formulation by body weight at a rate of 20 mg/kg rifampin and 5 mg/kg trimethoprim per day in divided doses. Each patient received one dose 2 h preoperatively, one dose 8 h postoperatively, and one dose every 8 h for a further 48 h. Those patients over 16 years of age were given a capsule with a sip of water, and infants were given liquid via a nasogastric tube. Surgeons and patients (and their parents/ guardians) were blinded to the treatment received.

Sample size For our significance level, we chose 0.05, one-tailed, and our/~ was 0.20 (giving the study a power of 0.80, or an 80% chance of finding a clinically significant difference, if it existed.) We determined that the predicted rate of infection in the control group would be 10%, and that in the treatment group could be reduced by 50%, to 5%. Using these parameters, we determined that we needed a sample size of 690 patients, divided equally amongst the two treatment modalities. With a large shunted population requiring 400-500 procedures per year, we estimated a 2-year accrual period, with a 2-year follow-up: a total of 4 years to complete. However, accrual rates were well below the estimate, with a total of 294 patients treated according to the protocol for 336 surgical procedures over a 3-year period. In addition, overall infection among the patients was higher than had been previously the case at the Hospital for Sick Children. At this point, the study was terminated. Patients taken into the study prior to termination were then followed for a minimum of 2 years.

OHtcome measures Shunted hydrocephalic patients were followed routinely in the neurosurgical outpatient clinic, regardless of whether they were in the study or not, and irrespective of treatment allocation. Patients were declared free of infection if, at further surgery for shunt obstruction, they were found to have negative cultures from CSF and/or shunt apparatus, or they passed a 2-year period of clinical well-being as judged by scrupulous follow-up in the clinic if no further surgery was required. Conversely, those patients who required additional surgery for malfunction and were found at operation to have positive microbiological cultures from shunt tubing and/or CSF were called infected. Additionally, those patients who presented with frank infection in the form of a reddened and oedematous shunt tract, pus from the wound, meningitis, or other systemic forms of infection were also categorized as infected if the clinical findings were verified by positive cultures. Both clinicians assessing patients and microbiologists examining specimens were blinded to the treatment received.

Data management and statistical analysis Patients recruited into the study were entered into a computerized database for tracking, and pertinent demographic and other data were collected prospectively at the time of any hospital admission and at every follow-up clinic appointment. At the termination of the study, patient charts were examined to verify the quality of the data and the database corrected as needed. Data from the pharmacy were added indicating the treatment received by patients entered into the study. The data were then analyzed using the Kaplan-Meier product limit procedure to take into account patient progress over time.

Results

Patient population F r o m M a r c h 1984 to O c t o b e r 1987, 294 p a t i e n t s were r e c r u i t e d into the study. O f these, 15 p a t i e n t s were lost to f o l l o w - u p a n d 36 p a t i e n t s were f o u n d to b e infected at the time o f initial s u r g e r y after being r a n d o m i z e d a n d given their s t u d y drug. This left a t o t a l o f 243 p a t i e n t s for analysis, 113 receiving p l a c e b o a n d 130 receiving Rif a p r i m . O f the 300 surgical p r o c e d u r e s involved, 145 were c o v e r e d w i t h p l a c e b o a n d 155 w i t h R i f a p r i m . A m o n g the p a t i e n t g r o u p ( n = 243), there w a s a n infection rate o f 12% a m o n g t h o s e receiving R i f a p r i m , versus a r a t e o f 19% a m o n g t h o s e receiving p l a c e b o . L o o k i n g a t the r a t e o f infection for the surgical p r o c e d u r e s , we f o u n d a rate o f 9 % a m o n g t h o s e receiving R i f a p r i m as o p p o s e d to 15% a m o n g those receiving p l a c e b o . I n o r d e r to e x a m i n e the effect o f m u l t i p l e entries into the study, we e x a m i n e d the s u b s e t o f 216 p a t i e n t s w h o were o n l y in the s t u d y once. T h e r e was a r a t e o f 1 0 % infections in the R i f a p r i m g r o u p , a rate o f 19% in the p l a c e b o g r o u p (Fig. 1). N o n e o f these differences is statistically significant. We p a i d p a r t i c u l a r a t t e n t i o n to the p a t i e n t g r o u p inc l u d i n g infants u n d e r 1 y e a r o f age, since t h e y have been s h o w n to have a g r e a t e r l i k e l i h o o d o f infection. T h e r e were 95 p a t i e n t s in this c a t e g o r y , 50 o f w h o m received p l a c e b o a n d 45 o f w h o m received R i f a p r i m . T h e infection rates were 22% a n d 18% respectively. I n revisions a m o n g s t this s u b g r o u p , t h e r e was a h i g h l y statistically

L

1.00

(%) [ ] Rifaprim

[ ] Placebo 03

0.20 ///////i,1

~////////i

YZ21~/I

///19

0.15

*~ 0.95 UA

//////2/J

03 03 c

o

7/,227"/,I

0.90

.~-, 03

iiiiii!~;iii~il

0.10 -

looy/A ili}};);ii}!}iii

0.05

"6 0.85

I

Rifaprim

25 ~J Placebo

0.80

I

n

i:i:i:i:i:in=243"///A ,:-:.:.:-:-.-...-.-zv///////A

All patients

255

0.75

i

I

All events

Patients in study only once

Fig. 1. Proportions of infections among all patients, all surgical procedures, and a subset of patients only entered into the study once

i

i

i

i

50 Time (months) Fig. 4. Survival curves for patients entered into the study only once (n=216) 0

10

20

30

40

1.0-

1.00 03 03

0.95 03

03

o

I

0.90

Rifaprim

0.9

g "5 03

45 03 0.85 "E' ~6

Placebo

[

I

I__

0.80 /D O [3_

Rifaprim

"b 0.8 m

1

25

I

n

o 13._

0.75 0

i

i

i

i

r

10

20

30

40

50

0.7

Fig. 2. Survival curves for all patients entered into the study (n = 243)

o.)

0.95 03 d)

[

0.90 -

g

Rifaprim

03

I

25

0.8-

LI

Rifaprim

0.6

I

/3

0.80 J 0.75 / 0

0.9.

.~ 0.7 c "b

Placebo

0,85 -

s

i

30

1.0

03 03

B_

i

20

Fig. 5. Survival curves for infants under i year of age entered into the study, including both insertions and revisions (n = 95)

1.00

to

i

10 Time (months)

Time (months)

03 03

Placebo

Placebo

~3 c',

I

s

13_

I

i

i

1

10

20

30

40

0.5 0.4

50

Time (months)

Fig. 3. Survival curves for all surgical events included in the study (n = 300)

i

i

i

i

i

2

3

4

5

6

Time (months)

Fig. 6. Survival curves for revisions in infants under 1 year of age (n = 16)

256 significant difference in favour of Rifaprim, but this can be explained by the small patient population involved (16 patients). Survival curves were generated which are shown in Figs. 2 - 6 , and all of which demonstrate a visual trend toward benefit from treatment with Rifaprim. However, none of these curves, with the exception of that depicting the revisions in patients under 1 year of age, attains statistical significance.

Discussion Every surgeon, regardless of specialty, recognizes the threat of infection in ruining otherwise satisfactory surgery. Nowhere is this more true than in surgery utilizing implantable prosthetic devices. Since most colonization of these devices occurs at surgery and from the patient's own tissues [3, 7], it makes biological sense to attempt to intervene at the time of surgery to prevent infection, via antibiotic administration. In fact, this manoeuvre has been shown to help in other specialties employing prosthetics [11]. Several studies have been carried out in an attempt to demonstrate the efficacy of antibiotic prophylaxis in CSF shunt surgery. Many of them are uncontrolled case series studies, and fail to meet the criteria for good scientific evidence [1, 20, 23, 27, 30]. Others are controlled trials which fail methodological scrutiny due primarily to small sample sizes [2, 9, 12, 22, 26, 29]. The only one to be able to demonstrate a significant difference as a result o f the use of antibiotics had an infection rate of 23% in the placebo group versus 6.5% in the treated group - almost a four-fold difference [4]. This rate of infection would be considered abnormally high in most institutions. A comparison of the various powers (and risk of/~ error) of previously published studies is shown in Table 1. The far right-hand column in the table shows how many patients would have been necessary per group to establish statistical significance at the infection levels seen in the various studies. As can be seen, it is unknown whether the lack of demonstrable benefit of perioperative antibiotics is due to true lack of benefit or to j~ error [10].

Prior to undertaking a randomized trial of our own, we endeavoured to use an alternative study design using the same design principles as a randomized controlled trial. This method, the case-control study, had been suggested as an alternative method [13, 14, 21], and using this we were able to show that patients who were not given perioperative antibiotics had at least a three times greater risk of becoming infected than those who were given antibiotics [28]. We were encouraged by this preliminary study to pursue a more formal and rigorous design. Our first task was to ensure that we had sufficient numbers of patients. In fact, we ultimately failed in this objective because of problems in accrual of patients in the first instance, and because of unacceptably high rates of shunt infection in the study patients, requiring early termination. The cause of the elevation in the rate of infection was unrelated to the study itself, since the non-randomized population had an even higher rate of infection. In terms of antibiotic choice, we leaned heavily on in vitro studies by Borges showing leukocyte failure in the presence of shunt tubing and suggesting that any effective antibiotic ought to have intraleukocytic action [5]. Such action has been proven to be a property of both trimethoprim and rifampin [19]. In addition, the drugs are synergistically effective against the micro-organisms most commonly implicated in CSF shunt infection [18]. For ease of administration, we sought oral formulations, and colleagues at the Hospital for Sick Children had had excellent experience with tolerance to the drug combination in a previous clinical trial [15]. The role of meningomyelocele in predisposing patients to CSF shunt infection has never been clearly proven, but it has been suggested in the literature [31]. For this reason we wished to distribute the patients with and without this potential risk factor among both treatment groups. Likewise, we were concerned that first insertions might be different in some inherent way from subsequent revisions, and therefore determined to stratify patients prior to randomization (Fig. 7). This is an example of a study designed to answer a clinical question keeping in mind as many potential threats to validity as possible. Had our overall infection

Table 1. Comparison of study power in published randomized controlled trials of perioperative antibiotics in CSF shunt surgery Study

Wang et al. [29] Bayston [2] Schmidt et al. [26] Rieder et al. [22] Haines and Taylor [12] Djindjian et al. [9] Blomstedt [4] Present study

Total no. of patients

Placebo group

Treatment group

n

Infections (%)

n

Infections (%)

120 132 152 63 74 60 122 243

65 54 73 31 35 30 60 113

8 2 6.5 10 14 20 23 19

55 78 79 32 39 30 62 130

7 3 9 6 5 3 6.5 12

a Ability of the study to detect a significant difference b In this column are given the number of patients that would have been required significant difference, for the infection rates observed

in each group

Power a

/3=0.80 b

0.04 0.04 0.08 0.09 0.27 0.58 0.80 0.30

10 879 7 003 1 926 712 158 50 61 327

to attain an 80% chance of detecting a

257

1

First insertions without spina bifida

"' First insertions with spina bifida

Revisions

I0.

11. 12.

I

I

13. 14.

Fig, 7. Allocation schema showing stratification and randomization 15. r a t e n o t s k y r o c k e t e d , we w o u l d h a v e been a b l e to c o n t i n ue a n d to h a v e a n excellent c h a n c e o f s h o w i n g b o t h clinical a n d statistical significance. A s it was, the p o w e r o f o u r s t u d y was o n l y 3 0 % , l e a v i n g us a 7 0 % risk o f m i s s i n g a clinically significant difference. A s i n d i c a t e d b y the g r a p h s , t h e r e was a n a p p a r e n t benefit f r o m the use o f R i f a p r i m , a n d we c a l c u l a t e d w h a t w o u l d h a v e h a p p e n e d if the p r o p o r t i o n s o f infections h a d s t a y e d the s a m e in a s a m p l e size o f a p p r o x i m a t e l y 650 patients. N o t surprisingly, we w o u l d h a v e r e a c h e d statistical significance. This once a g a i n p o i n t s o u t the a b s o l u t e necessity o f s a m p l e sizes large e n o u g h to w i t h s t a n d statistical scrutiny.

16. 17. 18.

19. 20.

Acknowledgements. This study was funded in part by the Hospital for Sick Children Foundation and Merrell Dow Pharmaceuticals (Canada) Inc. The authors wish to thank the many neurosurgical residents and clinical fellows at the Hospital for Sick Children for their help with the study. We also wish to express our gratitude to Irene McNeill and Trudi Schueller of the Neurosurgery Research Design Office at Sunnybrook Health Science Centre for reviewing each patient's chart to verify all data included in this study.

21. 22.

23.

References I. Ajir F, Levin AB, DuffTA (1981) Effect of prophylactic methicillin on cerebrospinal fluid shunt infections in children. Neurosurgery 9:6-8 2. Bayston R (1975) Antibiotic prophylaxis in shunt surgery. Dev Med Child Neurol 17 [Suppl 35]:99-103 3. Bayston R, Lari J (1974) A study of the sources of infection in colonised shunts. Dev Med Child Neurol 16 [Suppl 32]: 16-22 4. Blomstedt GC (1985) Results of trimethoprim-sulfamethoxazole prophylaxis in ventriculostomy and shunting procedures. J Neurosurg 62:694 5. Borges LF (1982) Cerebrospinal fluid shunts interfere with host defenses. Neurosurgery 10:55-60 6. Bruce AM, Lorber J, Shedden WIH, Zachary RB (1963) Persistent bacteraemia following ventriculo-caval shunt operations for hydrocephalus in infants. Dev Med Child Neurol 1:461-470 7. Burke JF (1963) Identification of the sources of staphylococci contaminating the surgical wound during operation. Ann Surg 158:898-904 8. Cohen SJ, Calaghan RP (1961) A syndrome due to the bacterial colonization of Spitz-Holter valves. A review of five cases. Br Med J II:677-680 9. Djindjian M, Fevrier MJ, Otterbein G, Soussy JC (1986) Oxacillin prophylaxis in cerebrospinal fluid shunt procedures:

24. 25.

26.

27. 28.

29.

30. 31.

results of a randomized open study in 60 hydrocephalic patients. Surg Neurol 25:178-180 Freiman JA, Chalmers TC, Smith H Jr, Kuebler RR (1978) The importance of beta, the type II error and sample size in the design and interpretation of the randomized controlled trial. Survey of 71 "negative" trials. N Engl J Med 299:690-694 Haines SJ (1980) Systemic antibiotic prophylaxis in neurological surgery. Neurosurgery 6:355-361 Haines SJ, Taylor F (1982) Prophylactic methicillin for shunt operations: effects on incidence of shunt malfunction and infection. Child's Brain 9:10-22 Hayden GF, Kramer MS, Horwitz RI (1982) The case-control study. A practical review for the clinician. JAMA 247: 326-331 Horwitz RI, Feinstein AR (1981) The application of therapeutic trial principles to improve the design of epidemiologic research: a case-control study suggesting that anticoagulants reduce mortality in patients with myocardial infarction. J Chron Dis 34:575-583 Jadavji T, Cheung R, Bannatyne RM, Prober CG (1986) Rifampin alone or with trimethoprim for contacts of children with Haemophilus influenzae type b infections. Can Med Assoc J 135:328-331 Lorber J, Zachary RB (1968) Primary congenital hydrocephalus. Long-term results of controlled therapeutic trial. Arch Dis Child 43:516-527 Luthardt T (1970) Bacterial infections in ventriculo-auricular shunt systems. Dev Med Child Neurol 12 [Suppl 22]: 105-109 Mandell GL, Moorman DR (1980) Treatment of experimental staphylococcal infections: effect of rifampin None and in combination on development of rifampin resistance. Antimicrob Agents Chemother 17:658-662 Mandell GL, Vest TK (1972) Killing ofintraleukocytic Staphylococcus aureus by Rifampin: in-vitro and in-vivo studies. J Infect Dis 125:486-490 McCullough DC, Kane JG, Presper JH, Wells M (1980) Antibiotic prophylaxis in ventricular shunt surgery. Child's Brain 7:182-189 Morgan PP (1981) Clinical trials on trial: must we always do a randomized trial? Can Med Assoc J 125:1309-1311 Rieder M J, Frewen TC, Del Maestro RF, Coyle A, Lovell S (1987) The effect of eephalothin prophylaxis on postoperative ventriculoperitoneal shunt infections. Can Med Assoc J 136:935-938 Savitz MH, Katz SS (1981) Rationale for prophylactic antibiotics in neurosurgery. Neurosurgery 9:142-144 Scarff JE (~ 963) Treatment of hydrocephalus: a historical and critical review of methods and results. J Neurol Neurosurg Psychiatry 26:1-26 Schimke RT, Black PH, Mark VH, Swartz MN (1961) Indolent Staphylococcus albus or aureus bacteremia after ventriculoatriostomy: role of foreign body in its initiation and perpetuation. N Engl J Med 264:264-270 Schmidt K, Gjerris F, Osgaard O, Hvidberg EF, Kristiansen JE, Dahlerup B, Kruse-Larsen C (1985) Antibiotic prophylaxis in cerebrospinal fluid shunting: a prospective randomized trial in 152 hydrocephalic patients. Neurosurgery 17:1-5 Venes JL (1976) Control of shunt infection. Report of 150 consecutive cases. J Neurosurg 45:311-314 Walters BC, Hoffman HJ, Hendrick EB, Humphreys RP (1985) Decreased risk of infection in cerebrospinal fluid shunt surgery using prophylactic antibiotics: a case-control study. Z Kinderchir 40:15-18 Wang E, Prober C, Hoffman H J, Hendrick EB, Humphreys RP (1984) Prophylactic sulfamethoxazole and trimethoprim in ventriculo-peritoneal shunt surgery. A double-blind, randomised, placebo-controlled trial. JAMA 251: 1174- i 177 Welch K (1979) Residual shunt infection in a program aimed at its prevention. Z Kinderchir 28:374-377 Yashon D, Sugar O (1964) Today's problems in hydrocephalus. Arch Dis Child 39:58-60