Infection in Neurosurgery

Infection in Neurosurgery
Beverly C. Walters, Stephen J. Haines

Infection remains a cause of serious morbidity and mortality in neurosurgical disease and practice. From the potentially devastating consequences of brain abscess to the loss of 6 to 10% of all cerebrospinal fluid shunts placed, infections complicate the neurosurgical treatment of many diseases. Infection has been a subject of study since before neurological surgery arose as a distinct specialty, and neurosurgeons have learned a great deal and left many questions unanswered. This chapter highlights both the limitations of existing knowledge and areas in which evidence of high quality can guide practice.

Patient Assessment

Infectious diseases usually present in a classical manner, with the hall-marks of the inflammatory response. These include fever, pain, swelling, redness, and loss of function. Cerebrospinal fluid (CSF) shunt infection may present in this same manner, with pyrexia, abdominal pain, induration of the shunt tract, or swelling around the valve chamber or reservoir. When the infection is not just localized to the tissue in immediate contact with the shunt, the patient may present with meningitis, peritonitis, or septicemia. If the shunt is draining into the systemic circulation, the patient may present with an immune-complex glomerulonephritis.1^–³

The symptom complex associated with the inflammatory response can usually be depended upon to signal a shunt infection, particularly in a child with a ventriculoperitoneal shunt infected with coagulase-positive staphylococci or gram-negative organisms. These symptoms, therefore, usually lead to investigations and management appropriate for shunt infections.4 Articles written in the early days of shunt therapy were concerned with ventriculoatrial shunts, and the clinical presentations most commonly mentioned were those detailed above.⁵ However, even at that time authors pointed out the indolent, clinically inapparent nature of some shunt infections, particularly those infected with staphylococci.⁶^,⁷

As peritoneal shunting became more prevalent, shunt malfunction as a manifestation of infection was noted, and again mention was made of the “occult” infection diagnosed only on routine culture of the shunt apparatus.⁸^,⁹ Even so, shunt malfunction as a clinical presentation in infection is mentioned in only four of 20 articles found on shunt infection.⁹^–¹² No interobserver or intraobserver variability studies have been done on clinical assessments in shunt infection or on any other infections of the central nervous system.

Establishing the Diagnosis

In the general category of central nervous system (CNS) infections, meningitis is verified by lumbar puncture and microbiological examination of spinal fluid, postoperative wound infections are established by their classical (rubor, calor, dolor) signs and symptoms, and brain abscess is visualized on imaging studies, most typically as a ring-enhancing lesion. However, establishing the diagnosis of CSF shunt infection is far more difficult, especially if the shunt is colonized with coagulase negative staphylococcus and the patient presents with a malfunctioning shunt rather than a more overt infection.

An appropriate diagnostic test evaluation needs to have four different types of patients in it (e.g., true-positives, true-negatives, false-positives, and false-negatives) (Table 2–1). These data help to answer four questions, each reflecting one aspect of the test’s ability [sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV)] to accurately diagnose a given condition. The medical literature was searched from 1966 to the present in an attempt to find the literature available on the diagnosis of shunt infection. Information was found on shunt reservoir aspiration, antistaphylococcal antibody titer, and C-reactive protein (CRP) in serum. Each of these was examined for its ability to meet the required criteria for evaluation of a diagnostic test.

Aspiration of CSF from the shunt reservoir and sending the aspirate for microbiological examination is the standard approach to making the diagnosis of CSF shunt infection. However, the positive and negative predictive value of this diagnostic test has never been examined. Only two studies on this test were found in a search of the literature from 1966 to the present.¹³^,¹⁴ Neither study meets the criteria for an appropriate evaluation of a diagnostic test; only true-positives (infected, shunted) were used in the study.

**Table 2–1 The Components of Diagnostic Test Evaluation**
		Patient Has Infection	Patient Has No Infection
		Gold Standard
Test Result (e.g., shunt aspiration, C-reactive protein, antistaphylococcal antibody titre	Positive: Appears to have infection	True Positive (a)	False Positive (b)	(a) + (b)
	Negative: Appears to have no infection	False Negative (c)	True Negative (d)	(c) + (d)
		(a) + (c)	(b) + (d)	(a) + (b) + (c) + (d)

Using the above table, the components of accuracy can be expressed and calculated as follows:

Sensitivity	a/(a+c)	If a patient has a positive test, how likely is (s)he to have an infection?
Specificity	d/(b+d)	If a patient has a negative test, how likely is (s)he to not have an infection?
Positive predictive value	a/(a+b)	If a patient has an infection, how likely is (s)he to have a positive test?
Negative predictive value	d/(c+d)	If a patient does not have an infection, how likely is (s)hehe to have a negative test?
Accuracy	(a+d)/(a+b+c+d)	How likely is this test to accurately identify those patients with and without an infection?

CRP, an acute-phase reactant in blood, was also suggested as a marker for shunt infection.15 This study did include both true-positives and true-negatives but did not include a variety of clinical presentations, nor did it include patients with other conditions that might have an elevated CRP. Like aspiration of the shunt, this test is certainly potentially useful, but we cannot know how useful, or under what circumstances, without more study.

Antistaphylococcal antibody titer was the first test ever suggested for the diagnosis of shunt infection, partly because shunts were—and are still—overwhelmingly infected with staphylococcus species.¹⁶^,¹⁷ One of these studies is the only one found that successfully meets all the required criteria of diagnostic test evaluation.¹⁷ This study demonstrated a high degree of accuracy (100% sensitivity, specificity, PPV, NPV) with a titer greater than 1/320. Although this study is almost 30 years old, it is the only one meeting the criteria needed for determining accuracy of a diagnostic test in a shunt infection (Table 2–2). There is no evidence that this has had any impact upon neurosurgical practice, and there is no doubt that the least evidence-based test we have—shunt aspiration—is the one universally used in this clinical circumstance.

Determination of Prognosis

Brain Abscess

To summarize the literature regarding the prognosis of brain abscess, we began by searching the national Library of Medicine’s MEDLINE data-base. The primary search term was brain abscess. The search was restricted to human subjects and the English language. Publication year was limited to 1980 or later to avoid the problem of diagnosis before the availability of computed tomographic (CT) scans. Subheadings of “mortality,” “prevention and control,” and “epidemiology” were used to narrow the search. Review of titles was used to exclude studies with a defined specific focus (such as those limited to childhood), a clear selection bias (patients with multiple brain abscesses, for example), subtypes (such as abscess related to congenital heart disease), or those with a sample size of fewer than 35 patients or no mention of the number of patients.

This search identified 13 articles of potential value. One was immediately discarded as it was a review presenting no new data.18 Two were discarded because they were restricted to abscesses of a specific origin.¹⁹^,²⁰ The remaining nine are characterized below (Table 2–3). The two that meet most criteria for valid prognostic studies both include substantial numbers of patients diagnosed before the availability of CT scanning.21^,²² Therefore, the search was extended back to 1966, the earliest publication date available online in MEDLINE. This search added one article that included some patients treated after the advent of CT scanning.²³

**Table 2–2 Antistaphylococcal Antibody Titre as a Diagnostic Test in Shunt Infection**
		Gold Standard
		Patient Has Infection	Patient Has No Infection
Test Result (e.g., antistaphylococcal antibody titre)	Positive: >1/320 Appears to have infection	9	1	10	Positive predictive value = 90%
Test Result (e.g., antistaphylococcal antibody titre)	Negative: <1/320 Appears to have no infection	0	23	23	Negative predictive value = 100%
		9	24	33
		Sensitivity = 100%	Specificity = 96%

Source: From Bayston R, Spitz L. Infective and cystic causes of malfunction of ventriculoperitoneal shunts for hydrocephalus. Z Kinderchir Grenzgeb 1977;22(4):419–421. Reprinted by permission.

The lack of information regarding referral patterns and state of disease at entry for all the studies limits the conclusions that can be drawn. In essence, it is possible only to draw conclusions regarding the mortality of brain abscesses treated at neurosurgical centers and gather some information on the demographics of the treated patients and the etiology and bacteriology of their abscesses. These data are summarized in Table 2–4. This careful examination of the quality of evidence regarding the prognosis of brain abscess leaves the reader with greater prognostic uncertainty than probably expected. However, a review of Table 2–4 demonstrates a substantial difference in the demographics, etiologies, microbiology, and outcome of these studies. Under these circumstances, one can only speculate about the cause of the differences and make only carefully qualified statements about the likelihood of survival (reported range: 10 to 43%).

**Table 2–3 Prognosis of Brain Abscess**
Reference	Inception Cohort?	Referral Description	Follow-Up Complete?	Outcome Objective?	Outcome Blind?	Cofactor Adjustment?
Rosenblum et al, 1978²³	Unclear—some single, some multiple, differing diagnostic methods	None	Unclear—duration and completeness not stated	Mortality—yes, otherwise no	No	No—informal analysis of etiology and multiplicity
Alderson, 1981⁴⁵	Unclear—insufficient patient description	Regional neurosurgical service	Unclear—duration and completeness not stated	Mortality—yes, otherwise no	No	No—informal analysis of etiology and multiplicity
Bradley and Shaw, 1983²¹	Unclear—insufficient patient description	Regional neurosurgical service	No—records incomplete	Mortality—yes, otherwise no	No	No—informal analysis of etiology and location
Svanteson, 1988⁴⁶	Unclear—insufficient patient description	Only neurosurgical department in region	Group 1: Unclear— duration and completeness not stated Group 2: all autopsied	Mortality—yes, other outcomes rated on category 4 scale	No	No—informal analysis of etiology
Richards, 1990⁴⁷	Unclear—insufficient patient description	Regional neurosurgical service	Unclear—duration and completeness not stated	Mortality—yes, otherwise no	No	No—informal analysis of etiology
Chalstrey, 1991⁴⁸	Unclear—insufficient patient description	Regional neurosurgical service	Unclear—duration and completeness not stated	Mortality—yes, otherwise no	No	No—informal analysis of etiology
Nicolosi et al, 1991²² (Abstract)	All ascertained cases in Olmsted County, MN	Population-based	Not stated in abstract	Mortality—yes, otherwise no	No	Unclear from abstract
Seydoux, 1992⁴⁹	Unclear—insufficient patient description	None	Two patients (5%) lost to follow-up	Mortality—yes, other outcomes	No	No—informal analysis of duration of symptoms, fever, etiology
O’Donoghue, 1992⁵⁰	Unclear—insufficient patient description	Regional neurosurgical service	Unclear—duration and completeness not stated	Mortality—yes, other outcomes reported	No	No—informal analysis of etiology and location

**Table 2–4 Differences Among Studies of the Prognosis of Brain Abscess**
Reference	N	Demographics	Microbiology	Etiology	Survivor Outcome	Mortality
Rosenblum et al, 1978²³	38	Median age 26 years Gender M:F 2:1 Median duration symptoms 12 days	Streptococcus 21% Staphylococcus 16% Anaerobes 53%	Congenital heart disease 24% Pulmonary 16% Postoperative 16% Trauma 8%	Not sufficiently specified	21%
Alderson, 1981⁴⁵	90	Not stated	Streptococcus 9% Staphylococcus 6% Anaerobes 59%	Otogenic 50% Sinugenic 4% Metastatic 17% Trauma 11%	Not stated	24%
Bradley and Shaw, 1983²¹	342	Median age <30 years Gender M:F 2:1 Median duration symptoms <1 month	Not stated	Congenital heart disease 4% Pulmonary 16% Otogenic 41% Sinugenic 17% Trauma 5%	48% survivors had “disability”	43%
Svanteson, 1988⁴⁶	87	Median age 31–40 years Gender M:F not given Median duration symptoms not stated	Streptococcus 9% Staphylococcus 14% Anaerobes 20%	Congenital heart disease 10% Metastatic 5% Otogenic 22% Sinugenic 10%	Good—67% Moderate disease—15% Severe disease—1% Dead—17%	17%
Richards, 1990⁴⁷	71	Not stated	Streptococcus Aerobic 30% Anaerobic 26%	Otogenic 34% Sinugenic 14%	Not stated	10%
Chalstrey, 1991⁴⁸	74	Not stated	Not stated	Sinugenic 23% Otogenic 18% Trauma 15%	Not stated	15%
Nicolosi et al, 1991²² (Abstract)	38	Median age not stated in abstract Gender M:F 3.2:1 Median duration symptoms not stated	Streptococcus “most frequent”	Not stated in abstract	Neurologic sequelae in 44%	38%
Seydoux, 1992⁴⁹	39	Mean age 44 years Gender M:F 2:1 Mean duration symptoms 15 days	Streptococcus 49% Anaerobes 74% (of 59 isolates)	Sinugenic or Otogenic 38% Congenital heart 21% (24 patients reported)	“Severe sequelae” in 22%	13%
O’Donoghue, 1992⁵⁰	62	Mean age 44 years Gender M:F 2:1 Mean duration symptoms 15 days	Anaerobes 34% Streptococcus 27% Staphylococcus 13% (of 110 isolates)	Otogenic 29% Trauma 23% Sinugenic 10%	“Persistent neurologic deficit” in 25%	24%

Evaluation of Intervention

Prevention

POSTOPERATIVE INFECTION IN CLEAN ELECTIVE NEUROSURGERY AND IMPLANTATION OF CEREBROSPINAL FLUID SHUNTS

Both the prevention of infection following clean (wound class I) elective neurosurgical operations and implantation of CSF shunts have been the subject of intense secondary analysis including meta-analysis of primary prevention trials. Therefore, the search strategy for this section consists first in identifying existing meta-analyses and then conducting searches of MEDLINE to identify comparable studies published after the completion of the meta-analyses.

Unfortunately, we found that MEDLINE searches for cerebrospinal fluid shunts or wound infection combined with meta-analysis produced no useful results. Indeed, a search for antibiotics, infection control, or surgical wound infection and meta-analysis, although producing many meta-analyses of surgical wound infection prophylaxis, failed to identify any related to neurosurgery. We relied, therefore, entirely on our own knowledge of the literature to identify these summaries of knowledge.

The prevention of infection following CSF shunt operations is the subject of two meta-analyses.24^,²⁵ The most recent randomized trial cited in these articles was published in 1991. A MEDLINE search for cerebrospinal fluid shunts and clinical trials or controlled clinical trials produced no more recent articles.

The prevention of infection following clean elective neurosurgical operations was the subject of a single meta-analysis by Barker.²⁶ The most recent randomized trial included in this analysis was also published in 1991. A MEDLINE search for wound infection, surgical wound infection, or surgical site infection and prevention or antibiotic prophylaxis or prophylaxis produced only one randomized trial published later than the trials cited by Barker in his work.²⁷ That study found a significant protective effect attributed to antibiotic prophylaxis, consistent with the findings reported by Barker.²⁶ We assume that the authors have appropriately accessed the best available evidence; therefore, conclusions are drawn from these data as outlined below.

SHUNT INFECTION

In the case of CSF shunt infection, two groups of investigators working contemporaneously reviewed the published literature for randomized clinical trials of antibiotic prophylaxis for CSF shunt infection. Eight of 12 identified trials were included in both analyses. Three of the four studies not included in both analyses were published only in abstract form. Presumably, a difference between the investigators in their willingness to accept the limited information available in the abstracts was responsible for the difference of opinion on whether or not to include the studies. The assurance that the combined research of the two investigative teams has identified all the useful published studies is increased by the fact that additional studies have not been revealed to or by the authors since publication.

Allowing for some differences in the inclusion criteria of the investigative teams, the basic conclusion of both meta-analyses is virtually identical: when pooled, the data indicate that the prophylactic use of antibiotics for CSF shunt operations reduces infection risk by ~50% (Langely et al,24 Mantel–Haenzel weighted relative risk of 0.52, 95% confidence interval of 0.37 to 0.73; Haines and Walters,²⁵ Mantel–Haenzel pooled odds ratio of 0.48, 95% confidence interval of 0.31 to 0.73).

Sensitivity analyses demonstrated that the result was not impacted by removing lower quality studies and that no single study was responsible for the result. Langely et al²⁴ calculated that 41 trials demonstrating no effect would be required to negate the results of the reported trials. They point out that infection rates in the treated groups averaged 6.8%, demonstrating that antibiotic prophylaxis could reduce but not eliminate CSF shunt infection. Haines and Walters, in a post hoc analysis, noted that the protective effect was greatest in those studies with high-baseline-infection rates and might disappear if the baseline-infection rate approached 5%.²⁵

These two analyses of the best available evidence provide us with a high degree of clinical certainty (class I evidence) that the use of antibiotic prophylaxis for CSF shunt placement operations is beneficial. This is sufficient to define a standard of practice. Arguing backward from a 6.8% average infection rate with the use of antibiotic prophylaxis, the estimated rate without prophylaxis would be 13.6%. An absolute risk reduction of 6.8% converts to a “number needed to treat” of 1/0.068 or 15; in other words, for every 15 shunt operations done with antibiotic prophylaxis, one infection is prevented.

NEUROSURGICAL WOUND INFECTION

A single meta-analysis has reviewed the published data on the use of antibiotic prophylaxis in elective clean neurosurgical procedures. The analysis was restricted to reports of craniotomy not involving the placement of a foreign body (such as a shunt). The direction of treatment effect favored antibiotic prophylaxis in each study, although only three of the eight included studies showed a statistically significant treatment effect by the author’s reanalysis. The pooled conclusion was that there was a highly significant effect favoring infection reduction with the use of antibiotic prophylaxis. The estimated rate of infection without prophylaxis was 8%, with prophylaxis 2%. An absolute risk reduction of 6% gives a “number needed to treat” of 17.

Once again, the consistency of results, the robustness of the conclusion when subjected to sensitivity analysis, and the quality of the evidence on which it is based lead to a confident conclusion that antibiotic prophylaxis is helpful in reducing craniotomy infection rates. Class I evidence supports this conclusion as a standard of practice.

Treatment Efficacy

CEREBROSPINAL FLUID SHUNT INFECTION

The difficulties in treating CSF shunt infection revolve around the fact that it is a secondary disease that results from the therapy of the primary disease, hydrocephalus. A therapy that enabled the neurosurgeon to treat the infection without interfering with the functioning shunt would be ideal. To this end, people have tried to use antibiotics alone to treat the infection, leaving the shunt in situ to continue to treat the hydrocephalus. This is contrary to the principles of surgery in which a foreign body must be removed when it is the nidus of infection to effect cure of the infection. Despite this, claims for cures using this technique have appeared in the literature. A detailed critical appraisal of this literature is undertaken in Table 2–5.

**Table 2–5 Treated Prognosis of Shunt Infection**
Reference	# Patients	Inception Cohort	Referral Pattern	Follow-Up Complete	Objective Outcome Criteria	Blinded Outcome Assessment	Extraneous Prognostic Factor Adjustment
Perrin and McLaurin, 1967³⁹; Immediate replacement	6	No	No	No	Yes	Not stated	No
White et al, 1969⁴²; EVD	33	No	No	No	No	Not stated	No
Visconti, 1979⁴¹; Delayed replacement	2	No	No	No	No	Not stated	No
Whang et al, 1974⁴³; EVD	6	No	No	No	No	Not stated	No
Mori and Raimondi, 1975¹²; EVD	6	Yes	No	No	Yes	Not stated	No
Archer et al, 1978²⁸; Antibiotics alone	1	No	No	No	No	Not stated	No
Scarff et al, 1978⁴⁴; EVD	52	Yes	No	Yes	No	Not stated	No
Gombert et al, 1981²⁹; Antibiotics alone	1	No	No	Yes	Yes	Not stated	No
Bayston and Milner, 1981³⁰; Antibiotics alone	44	No	No	Not stated	Yes	Not stated	No
Mates et al, 1982³¹; Antibiotics alone	8	No	No	No	Not stated	Not stated	No

Abbreviations: EVD, external ventricular drain.

In the case of a malfunctioning shunt, surgery is mandatory to secure the homeostasis of the ventricular system, and thus protect the brain. Various permutations of removing the shunt and replacing it with a new, functioning device have evolved over the years, so that now, three surgical methods are commonly used. These include removing the shunt and replacing it at the same operation (immediate replacement), removing it and replacing it at a second operation a few days later (delayed replacement), and externalizing it for a period of time with replacement with a new shunt when the signs of infection have disappeared (external ventricular drainage). All of the surgical modalities include intensive antibiotic therapy as an adjunct. All of the articles found by standard search criteria are examined according to whether they represent a case series of a single treatment (prognosis studies) or compare two or more treatments (comparative treatment studies).9^,¹⁰^,¹²^,²⁸^–⁴⁴ These are examined critically for their conformation to the criteria outlined in the beginning of this book for appropriate clinical study design. These are summarized below and presented in detail in Tables 2–5 and 2-6, respectively.

**Table 2–6 Comparative Studies of Treatment of Shunt Infection**
Reference	Treatments Studied	Patients Assigned by Randomization	All Clinically Relevant Outcomes Reported	Study Patients Similar to Usual Clinical Population	Statistical and Clinical Significance Considered	All Patients Entered Accounted For
Nicholas et al, 1970⁴⁰	Immediate and delayed replacement	No; similarity not documented; no prognostic stratification	Mortality no reported; no blinding of outcomes	No exclusion criteria; tertiary care center	No; clinically important; insufficient power	Yes
Luthardt, 1970³²	Antibiotics alone; immediate replacement; EVD	No; similarity not documented; no prognostic stratification	Mortality reported; no blinding of outcomes	No exclusion criteria; tertiary care center	No; clinically important; insufficient power	No
Morrice, 1974³³	Antibiotics alone; immediate and delayed replacement	No; similarity not documented; no prognostic stratification	Mortality reported; no blinding of outcomes	No exclusion criteria; type of care environment not documented	No; clinically important; insufficient power	Yes
McLaurin, 1973³⁴	Antibiotics alone; immediate and delayed replacement	No; similarity not documented; no prognostic stratification	Mortality reported; no blinding of outcomes	No exclusion criteria; tertiary care center	No; clinically important; insufficient power	Yes
Shurtleff et al, 1974¹⁰	Antibiotics alone; immediate and delayed replacement	No; similarity not documented; no prognostic stratification	Mortality not reported; no blinding of outcomes	No exclusion criteria; tertiary care center	No; clinically important; insufficient power	Yes
McLaurin, 1975³⁵	Antibiotics alone; immediate and delayed replacement	No; similarity not documented; no prognostic stratification	Mortality reported; no blinding of outcomes	No exclusion criteria; tertiary care center	No; clinically important; insufficient power	Yes
James et al, 1980³⁶	Antibiotics alone; immediate and delayed replacement	Yes; similarity not documented; no prognostic stratification	Mortality reported; no blinding of outcomes	Exclusion criteria used; tertiary care center	No; clinically important; insufficient power	Yes
James et al, 1981³⁷	Antibiotics alone; immediate and delayed replacement	No; similarity not documented; no prognostic stratification	Mortality reported; no blinding of outcomes	No exclusion criteria; tertiary care center	No; clinically important; insufficient power	Yes
Forward et al, 1983³⁸	Antibiotics alone; immediate and delayed replacement	No; similarity not documented; no prognostic stratification	Mortality reported; no blinding of outcomes	No exclusion criteria; tertiary care center	No; clinically important; insufficient power	Yes

Abbreviations: EVD, external ventricular drain.

ANTIBIOTICS ALONE

The possibility of successfully treating a shunt infection without surgery would be a welcome addition to the therapeutic armamentarium. Aside from being able to save the patient the additional risk of surgery, there are other advantages. If the patient’s shunt is functioning and the patient is relatively well, oral antibiotics might be employed, allowing for early discharge home, with a substantial savings in health care costs and separation of patients (especially children) from their families. In addition, there would be no new mechanical trauma to the brain caused by insertion of a new shunt. This form of treatment is not without its disadvantages, however. Among these is the possibility of having to puncture the shunt reservoir percutaneously to instill intraventricular antibiotics, thereby running the risk of introducing an infection with a more virulent organism. Oral or parenteral antibiotics may not penetrate the CSF, despite being known to do so, or they may bind reversibly with the target organism, fluctuating in their effectiveness. In addition, the organisms may be resistant due to some of their own defense mechanisms. However, in a patient with a functioning shunt, the neurosurgeon is loathe to disturb it if there is any hope of success of medical therapy alone. An examination of the literature on medical therapy alone is included in Tables 2–5 and 2-6.

From this assessment it can be seen that the literature reporting success (or failure) of medical therapy alone does not provide a definitive answer to its usefulness. However, several points can be made:

Medical therapy alone may be successful in selected patients.
Because there are overriding clinical factors that may determine usefulness, randomizing patients to this treatment without stratifications may be unethical.
Failure of the therapy may be explained by many factors not controlled for in the studies examined (e.g., inadequate antibiotic levels, confounding prognostic factors).
Lack of accurate outcome criteria, may lead to erroneous conclusions that therapy was successful.

None of the studies reviewed examined the variables that may have led to treatment success or failure. The results in the one randomized controlled trial performed so far 36 were so poor in the medical therapy group that the trial was stopped and patients were subsequently treated arbitrarily by one of the more successful surgical therapies.³⁷ However, careful inspection of the report reveals that two of the patients in the antibiotics-alone group developed shunt malfunction in the course of their therapy and were thus ineligible for it. They were among those counted as treatment failures. In addition, two of the other treatment failures in the antibiotics alone group had more than one shunt in place. It has been shown that in this situation, the CSF culture may grow only one organism, whereas the shunts are colonized by many different organisms.⁴ In summary, there is no clear evidence of effectiveness of this therapy, and no delineation of the types of patients in whom it might ultimately be useful.

IMMEDIATE REPLACEMENT (SHUNT REMOVAL AND REPLACEMENT IN A SINGLE OPERATION)

If an operation is required to remove a malfunctioning and infected shunt, the most benign procedure may be that involving removal and replacement at the same operation. Not only does this treatment plan involve only one operative procedure, but it also obviates the need for ventricular taps to relieve the increase in intracranial pressure that occurs if a shunt is removed. If it is successful, it also shortens hospitalization time. There are, however, certain potential disadvantages. These include the possibility of continuity of the infection if the infected and new shunts are handled at the same operation, the inability to use this mode of therapy when there is a wound infection nearby, and its lack of effectiveness in treating virulent organisms. In an uncontrolled retrospective study, this was the least effective treatment, usually requiring some other mode of surgical therapy before cure was effected.4 An examination of the studies on this mode of treatment is included in Table 2–6, and indicates that this therapy is relatively successful. Even the randomized controlled trial and its cohort sequel show a high success rate.³⁶^,³⁷ However, the outcome criteria did not include a negative shunt culture or long-term asymptomatic follow-up. The proof of cure was taken to be a negative culture of shunt fluid aspirate 48 hours after completion of therapy and a second culture 1 to 4 months later. Follow-up ranged from 2 to 48 months, but the distribution among patients was wide (mean = 23, SD = 14 months). For all the studies, there are rival hypotheses that can be proposed for the findings.

DELAYED REPLACEMENT (SHUNT REMOVAL, ANTIBIOTIC INTERVAL, SHUNT REPLACEMENT IN SECOND OPERATION)

Following the principles of treating foreign body infection, the shunt may be removed completely, leaving the child’s hydrocephalus transiently untreated. In infants, ventricular taps through the anterior fontanel to relieve the pressure can overcome this. However, this is not an entirely benign procedure, and the pressure is relieved in fits and starts rather than in a continuous (and physiological) manner. The development of porencephaly along needle tracks has been reported with this form of treatment. In older children, this method of relieving pressure does not exist because there is no access through the skull as there is in the infant. In one retrospective review, this was the most effective therapy for the shunt infection, but because of the problem of leaving the patient’s hydrocephalus untreated, it was hardly ever used as the first method. A methodological review of the available literature on this method of treatment is seen in Tables 2–5 and 2-6.

EXTERNAL VENTRICULAR DRAINAGE (SHUNT REMOVAL WITH EXTERNAL DRAIN REPLACEMENT IN SINGLE OPERATION)

A method of therapy that attempts to combine the two surgical approaches already mentioned is that of removing the infected/colonized shunt and replacing it with a device that drains the CSF externally in a collecting bag. The hydrocephalus is therefore treated continuously, but the offending shunt has been removed. The criticism of this type of therapy is that the external ventricular drain (EVD) has been placed into an infected field and may then become colonized itself, thus perpetuating the infection. There has also been shown to be a tendency of these devices to become “supra-infected” with organisms far more virulent than the original organism involved in the shunt infection. This occurs because the antibiotics are being used to treat the original organisms in an open (external) system rather than in the closed (internal) system of the usual shunt. The open system is exposed to nosocomial pathogens that are free to grow in an environment in which host defenses are ineffective and antibiotic treatment destroys the relative homeostasis among microbial invaders. These more virulent organisms can make a bothersome infection a lethal one. For this reason, this method of treatment has often been reserved for situations in which a virulent organism is already involved, or in which other methods of therapy have failed.4 A critical evaluation of the literature on this method of treatment is shown in Tables 2–5 and 2-6.

TREATMENT SUMMARY

In summary, there are distinct clinical situations in which the various methods are generally used. These situations and treatments are as follows:

Medical therapy may be used in the circumstance of an infected but functioning shunt; once a shunt is shown to be malfunctioning, it must be treated surgically to correct both the malfunction and the infection.
The one combination therapy that continues to treat the hydrocephalus while observing surgical principles to remove an infected prosthesis is that of immediate shunt replacement. Unfortunately, this mode of therapy has worked infrequently; however, it can be used in the case of a highly shunt-dependent patient.
The form of surgical/medical therapy that allows time to sterilize the CSF in the absence of the foreign body is that of delayed replacement following removal. However, this method leaves the patient without treatment of the hydrocephalus; it cannot be used when the patient is highly shunt-dependent.
External ventricular drainage removes the foreign body and continues to treat the hydrocephalus but potentially places the patient at increased risk for more virulent infection. It has been most successful, however, in the studies of prognosis with treatment, including the one randomized controlled trial.

This literature on treatment stops abruptly at the time of the last attempted randomized controlled trial because of the variety of treatments available, some of which cannot be used in certain clinical circumstances. This inhibits randomization and diminishes the numbers of patients available for any comparative clinical trial. In this instance, treatment paradigms are based upon class III studies.