METHODS

Between January 1994 and mid-July (15th) 2001, 126 patients underwent shunt insertions and revision procedures in the Department of Neurosurgery at University Hospitals Coventry & Warwickshire NHS Trust, Walsgrave. A total of 176 shunt procedures were performed in 126 patients with hydrocephalus. Although the majority of procedures involved ventriculoperitoneal shunts, a variety of other shunts were also inserted, including lumboperitoneal, cystoperitoneal, ventriculopleural, syringopleural, and ventriculoatrial shunts. Primary shunt procedures had been performed previously at other centres in 33 patients; these patients underwent shunt revision for various reasons, and have been included in this study. All shunt material removed during revision was sent for bacteriological analysis.

Operative technique

All operations were carried out in a dedicated neurosurgical operating theatre, which was seldom, if ever, used for general surgical cases. The operating theatre personnel were all neurosurgically trained. Most operations were carried out by single surgeons. All operating theatre staff were reminded that a shunt procedure was in progress, and strict aseptic sterile technique was observed throughout. Lapses in theatre discipline were not tolerated, and this attitude was inculcated into all present; we term this “zero tolerance”. All ingress and egress from the operating theatre was prohibited except for genuine emergencies. The number of operating theatre personnel varied between five and seven. All personnel wore full operating theatre dress, including facemasks. A wide area was cleared around the operating table itself, so that the diathermy, suction, and anaesthetic machines were kept well away from the sterile area.

Intravenous antibiotics were given with the induction of general anaesthesia to all patients, in most cases benzylpenicillin 1.2 gm and flucloxacillin 1 gm intravenously. Those who were allergic to penicillin were given rifampicin 600 mg intravenously with induction. The patient was positioned on the operating table in the anaesthetic room (details of the operating theatre and the preparation solutions used are given in the Appendix). The head, trunk, and abdomen were lathered with Betadine® surgical scrub, and shaved widely with a sterile cut throat razor. MSC’s preference was to use a right frontal burr hole whenever possible, and therefore the pinna of the ear was carefully taped forwards.

The skin was prepared using a solution of chlorhexidine and cetrimide, followed by alcoholic Betadine®. The incisions were marked with a sterile pen and then draping was carried out, using sterile adhesive drapes (Klinidrapes®, by Molnlycke), to outline the operative area, followed by a transparent sterile drape (Ioban®) impregnated in Betadine®. Two further layers of standard towels were then used to drape the patient right down to floor level, so that no part of the operating theatre table was exposed. Finally, the sterile drape was daubed with a further layer of alcoholic Betadine®, which was allowed to dry while the surgeon left the operating room to put on sterile dress.

The skin incisions were then carried out and a swab soaked in aqueous Betadine® placed in the incisions, with particular care to soak the exposed skin edges. The burr hole was made, the abdominal cavity opened by a transverse right upper quadrant incision, and a small intervening stab incision was added just above the ligamentum nuchae about 5 cm behind the ear. Two separate shunt tunnellers were daubed with aqueous Betadine® and positioned subcutaneously. The interior of each tunneller was then irrigated with aqueous Betadine® solution. Having completed all the preparation for shunt placement, including opening of the dura and diathermy of the cortical surface, the operating surgeon’s gloves were washed with aqueous Betadine® and covered with a further pair of sterile gloves. Only then was the shunt system opened and placed upon a separate trolley, using sterile instruments that had not been in contact with the skin. In the majority of cases Delta valves (PS Medical) were chosen, with a Medtronic barium-impregnated catheter, without biocide.

The shunt system was irrigated with 5 mg of intrathecal gentamicin, made up in theatre with saline to 5 ml. The ventricle was tapped using a Dandy needle, and a further 5 mg of intrathecal gentamicin, made up to 5 ml with saline, was injected slowly into the ventricles, with frequent back and forth aspiration and injection, to ensure thorough mixing with ventricular CSF. The shunt system was tunnelled down to the peritoneal cavity, with care to keep it away from the skin on a separate sterile towel.

The ventricular catheter was inserted and cut to length with new, sterile scissors. The connection between the ventricular catheter and the shunt was made as quickly as possible, handling the shunt with fingers once again soaked in Betadine®. The ventricular catheter was tied to the shunt with a heavy silk tie (2/0 or larger), to avoid the risk of cutting through the shunt catheter. The shunt was then pulled down, and checked to ensure that it was free of kinks and dripping well. The burr hole was lined with surgicel, and bone wax used to cover the burr hole defect and reduce the risk of CSF leakage. The distal end of the shunt was placed in the peritoneal cavity. Haemostasis was ensured. The wounds were then closed using Vicryl®, sprayed with Betadine® dry powder spray, followed by sterile Opsite® dressing spray, and covered with Mepore® dressings. Strict instructions were given that the dressings should not be disturbed until the staples were due for removal at seven days (14 days for infants).

Postoperatively, patients were maintained on the same intravenous antibiotics for a further 48 hours, and were mobilised as soon as they were fit. Under no circumstances were patients allowed to have the staples removed by any agency other than the nurses on the neurosurgery unit, under clean conditions.

Follow up

The follow up period ranged from 16 to 90 months (table 1). All patients who had had shunt insertions were allowed free access directly to the neurosurgical unit at any time. The follow up periods for three patients were curtailed by death. Of these, two died as a result of their high grade tumours, and one was killed in a road traffic accident. No other patients were lost to follow up.

Criteria for CSF shunt infection

We observed the criteria for CSF shunt infection defined by Odio et al.²⁵ Shunt infection was considered present when a pathogenic organism was cultured from lumbar or ventricular CSF, or blood. In addition, a shunt infection was also considered present if the patient developed a pyrexia higher than 38.5^oC, shunt malfunction, or abdominal or neurological symptoms; these findings were attributed to infection even if cultures were negative.

Statistical analysis

The statistical analysis of this series was to a degree limited by the absence of controls. However, there is sufficient literature to justify the assumption that the expected level of infection lies between 5 and 15%. At the 5% level, we would have expected 9 infections; at the 15% level, 27 infections. After discussion with our statisticians, we applied the Fisher one-sided exact test to our data.

RESULTS

Between 1994 and 2001, a total of 176 shunt procedures were performed in 126 patients with hydrocephalus. The details of the patients in this series are set out in tables 2–8. Eight patients required emergency shunt revision for imminent death or blindness. The remaining 168 procedures were performed on the next routine operating list, usually within 24 hours (table 5). The reasons for shunt revision are given in table 7; the majority were in patients with neural tube defects. The duration of the procedures varied from 45 to 95 minutes.

Statistical analysis

The observed infection rate was one patient out of 176 (0.57%). Applying the Fisher exact one-sided test, and assuming an expected infection rate of 5% (nine infections), the significance was 0.01 (p <0.01). At the 15% expected infection rate, the significance was greater (p <0.0001). These statistical tests show a highly significant reduction in shunt infection when compared with historically accepted data.

Our single infection was seen in an 18 year old patient with aqueduct stenosis. Before transfer to our unit, he had had multiple procedures performed at another institution, including external ventricular drainage for haemorrhage following a shunt operation. When his CSF cleared, we inserted a shunt. His infection occurred seven months after this last procedure, and appeared to follow appendicitis and peritoneal sepsis. His shunt and ventricular CSF both grew Streptococcus faecalis. Although included in this series, this infection may not have been directly attributable to the shunt placement.

DISCUSSION

Shunt infection remains a potent cause of mortality and morbidity. The quoted infection rate ranges from 5–27%,^{111222528293134} with most between 5% and 15%. This is one reason for the recent interest in alternative CSF diversion procedures that do not involve implanted prostheses; of these, third ventriculostomy is the most commonly carried out.

Different studies have pointed to a variety of factors that may possibly be responsible for shunt infection. Age,²⁶²⁷ skin condition,²⁸ gender, pathology, aetiology of the hydrocephalus,²¹ and immunological status of the patient¹ have all been implicated. Surgical factors that have been studied are length of preoperative hospitalisation, duration of surgery, number of revisions,¹ type of material used,²⁵²⁹³⁰ aseptic technique,⁴⁴ and prophylactic antibiotics.^14113132

The bacteria that most frequently appear in shunt infections are staphylococci (S epidermidis and aureus)^1101833; the percentage varies from 62–90%.^1101833 Infections due to Gram negative bacteria are infrequent but important, because mortality is very high at 40–90%.¹

The variation in quoted infection rates may partly be caused by different study designs. The incidence of infection is related either to the number of patients²⁹³⁵³⁶ or to the number of procedures, inclusive of revisions.¹²⁸ Clearly, infection rates per patient will be higher than infection rates per procedure, as in any series many patients will have multiple procedures.

In some series²⁸³⁷ there was no difference in infection rate between initial and revision shunt procedures. However, Odio et al showed an increased risk of CSF shunt infection in shunt revisions.²⁵ Meirovitch in contrast reported a similar rate of shunt infection in primary and shunt revision operations.³⁸ This is in accordance with our study, where there was no increase in the shunt infection rate with revisions. There are many reports demonstrating an increased risk of infection in the younger age group.¹¹²⁸³⁸ George et al found a significantly increased rate of infection in children and the elderly.¹ Kontny et al³⁹ found no significant difference in the rates of infection in different age groups, which corresponds with our results. In concordance with other studies such as that of Shoenbaum and Gardner,⁵ we did not see a significant difference in infection rates with implantation of different types of shunts for CSF diversion procedures. In parallel with Kontny et al³⁹, we did not observe any significant difference in infection rates according to different aetiologies of hydrocephalus.

The use of prophylactic antimicrobials remains controversial.^4113132 In a recent control study, the infection rate was 6.25% in patients without antibiotic prophylaxis as compared with 2.56% in patients with antibiotic prophylaxis.²⁶ This difference was not found to be statistically significant. At present, the benefit of perioperative antimicrobial prophylaxis remains unestablished. As CSF shunt infections continue to be common and life threatening for patients with hydrocephalus, we give prophylactic antibiotics to our patients. The regimen used in this series (penicillin, flucloxacillin) may be criticised for lack of Gram negative cover. However, the vast majority of shunt infections are caused by Gram positive cocci, and the senior author (MSC) has used this regimen successfully for over 12 years, including the use of rifampicin, although this may be controversial.

Regarding the duration of operation, Kontny³⁹ did not observe any significant difference in infection rates, which agrees with our study, where length of surgery had no significant effect. Several reports identified perioperative airborne contamination as the main source of infection, particularly in procedures that included the insertion of a prosthetic device.²⁸⁴⁰⁴¹ In one series there was a twofold difference in infection complications between different surgical teams (6–11.7%).²⁸ Thus, aseptic technique appeared to be a significant factor influencing the shunt infection rate.

We agree with Welch¹² that shunt infection is a potentially preventable complication. Odio et al²⁵ found a median time for postoperative infection of 8 days, whereas Mazza et al³⁶ found the highest rate of infection at 15–28 days after shunt operation. Hence, both of these studies confirm that infection is temporally closely related to surgery. A similar report was presented by George et al,¹ who found that the experience of the surgeon was the single most important factor in the reduction of shunt infection rates, which is in accordance with our study. Venes,⁴² and McCarthy and Wenzel⁴³ also placed particular emphasis on surgical technique in the control of infection.

Choux et al⁴⁴ claimed different results in two different series conducted over different time periods. In the first series (1978 to 1982), they demonstrated a rate of shunt infection of 15.56%. In 1983 they introduced a protocol with stringent technical perioperative guidelines, including vigorous asepsis and limited personnel in the operation theatre. This resulted in an infection rate of 0.33% in the second series. The only difference in the two series was a change in operative practice. All remaining parameters such as age, gender, pathology, type of shunt, shunt material, surgeon, and number of revisions caused no significant differences. Choux et al stated that it was possible to reach a nearly zero shunt infection rate.

We have demonstrated that, with careful aseptic and antiseptic surgical technique and prophylactic antimicrobials, it is possible virtually to eliminate shunt infections. It is impossible to say which component of this approach is most important: the rigid aseptic technique, the liberal use of Betadine®, the avoidance of haematomas, or the antimicrobial prophylaxis. If infection is multifactorial, then perhaps all components of the approach act in concert.

CONCLUSION

Shunt infection rates can be reduced or virtually eliminated by strict adherence to the following simple surgical principles: asepsis, antisepsis, antimicrobial therapy, and the avoidance of haematomas.

APPENDIX

The operating theatre was a standard theatre as found in most NHS hospitals, measuring 6.5 metres square and not equipped with a laminar flow hood. The air was subject to 20 changes per hour. The theatre was part of a suite of four, which was monitored continuously at a central console. Air flow and filter alarms were in place, and the filters and plant were checked every day in accordance with standard operating theatre practice in the United Kingdom.

The solutions used were as follows:

• Betadine® antiseptic solution (Seton Healthcare Group) containing povidone-iodine 10% w/v

• Betadine® alcoholic solution (Seton Healthcare Group) containing povidone-iodine 10% w/v

• Betadine® Surgical Scrub (Seton Healthcare Group) containing povidone-iodine 7.5% w/v

• Betadine® dry powder spray (Seton Healthcare Group) containing povidone-iodine USP 2.5% w/v

• Chlorhexidine and cetrimide solution (MIZA Pharmaceuticals UK) containing 0.25% chlorhexidine gluconate solution 20% BP and 1.25% strong cetrimide solution 40%BP in purified water

• Ioban® iodine impregnated transparent dressings (3M).