UNIVERSITY OF CALGARY
The Prognostic Accuracy of Suggested Predictors of Failure of Medical Management for Patients with
Spinal Epidural Abscess
by
Alexandra Stratton
A THESIS
SUBMITTED TO THE FACULTY OF GRADUATE STUDIES
IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE
DEGREE OF MASTER OF SCIENCE
GRADUATE PROGRAM IN COMMUNITY HEALTH SCIENCES
CALGARY, ALBERTA
MAY, 2016
© Alexandra Stratton 2016
Abstract
Controversy exists as to the first-line treatment in patients with spinal epidural abscess (SEA) who
present with an intact neurologic exam—medical treatment or surgery. The aim of this thesis was to
clarify treatment decision making in such patients. A systematic review and meta-analysis were first
undertaken to define the risk of failure of medical treatment and to identify predictors of failure.
Although the incidence of failure of medical management of SEA was found to be relatively common in
published reports, estimates were highly heterogeneous between studies. Two studies were identified
in the literature with prediction criteria for failure of medical management. These prediction criteria
were tested in a cohort of patients from a tertiary care centre and both overestimated medical
treatment failure. Once the prediction models were calibrated to our data it was Patel’s model that was
superior and thus more useful for clinical decision making.
ii
Acknowledgements
My supervisor, Ken Thomas, co-supervisor, Elijah Dixon, and committee members Peter Faris and Christy
Tomkins-Lane for making themselves available when needed (sometimes with very little notice), their
guidance and wisdom. Ish Baines for her help with obtaining the patient list and charts, Leslie Truong for
her dedication to the chart review and data extraction. My husband and children for their understanding
and commitment to the ‘big picture.’
iii
Table of Contents
Abstract……………………………………………………………………………………………………………………………………………………ii
Acknowledgements…………………………………………………………………………………………………………………………………iii
Table of Contents……………………………………………………………………………………………………………………………………iv
List of Tables……………………………………………………………………………………………………………………………………………v
List of Figures………………………………………………………………………………………………………………………………………….vi
Chapter 1: Introduction and Review of the Literature………………………………………………………….………………….1
Chapter 2: Incidence and Risk Factors for Failure of Medical Management of Spinal Epidural Abscess: A
Systematic Review and Meta-analysis……………………………………………………………………………………………………..7
Chapter 3: The Prognostic Accuracy of Suggested Predictors of Failure of Medical Management for
Patients with Spinal Epidural Abscess…………………………………………………………………………………………………….28
Chapter 4: Conclusions...............................................................................................................................45
Appendix……………………………………………………………………………………………………………………………………………….46
iv
List of Tables
Chapter 2……………………………………………………………………………………………..…………………………………………..19
Table 1. Characteristics of studies included in systematic review.
Table 2. Study Quality Assessment.
Table 3. Stratified analyses. FMT=failure of medical treatment. *=Total number of patients in study.
Table 4. Published predictors for failure of medical management, with associated odds ratios (OR).
Appendix Table 1. Abscess location by study.
Appendix Table 2. Immunosuppression definition and proportion of patients by study.
Chapter 3……………………………………………………………………………………………..…………………………………………..37
Table 1. Proposed predictors of failure of medical management.
Table 2. ICD-10 diagnostic codes used for identification of patients with spinal epidural abscess.
Table 3. Characteristics of patients intentionally treated medically.
Table 4. Medical treatment failures.
Table 5. Odds ratios associated with proposed predictors of failure of medical management.
Table 6. Assessment of calibration and discrimination of the two models.
Table 7. Probability of failure of medical management based on Kim’s original and calibrated model, and
the observed failure rate in our cohort.
Table 8. Probability of failure of medical management based on Patel’s original and calibrated model.
v
List of Figures
Chapter 2……………………………………………………………………………………………..…………………………………………..19
Figure 1. Study Selection.
Figure 2. Forest plot of proportion of patients who failed medical management based on any definition
of failure reported by a study.
Figure 3. Forest plot of proportion of patients that failed based on failure defined as requiring surgery.
Figure 4. Funnel plot with pseudo-95% confidence limits.
Chapter 3……………………………………………………………………………………………..…………………………………………..37
Figure 1. Patient flow diagram.
Figure 2. Probability of failure based on Kim’s and Patel’s models versus actual failure of medical
management. 0=no failure, 1=failure.
Figure 3. ROC curves for Kim and Patel models.
vi
Chapter 1: Introduction & Review of the Literature
Spinal epidural abscess (SEA) is a serious infection of the spinal canal. The neurologic function of
patients at the time of diagnosis ranges from neurologically intact to complete absence of neurologic
function. Controversy exists surrounding the treatment of neurologically intact patients with SEA.
Medical management of such patients, while becoming increasingly common, carries a risk of failure
with neurologic deterioration necessitating surgery.
The first manuscript in this thesis (Chapter 2) is a systematic review and meta-analysis that was
performed with two objectives: 1) to define the incidence of failure of medical management of SEA and
2) to identify risk factors associated with failure.
The first comparative study of surgical to medical management of SEA with more than ten
medically managed patients was published in 1996(1). Khanna et al. reported the outcome of all 41
patients with nontuberculous SEA treated at a single centre, including 16 initially treated medically. The
authors found increasing age (by 10 year increments) and severe spinal canal encroachment (>50%) to
be the only patient factors independently associated with a poor outcome among patients in both
surgical and medical groups. Nineteen-and-a-half percent of the medically treated SEA patients’
neurologic status worsened despite appropriate antibiotics(1). The authors recommended surgery in all
patients except in the specific case of a neurologically stable patient, with absent or minimal deficits,
and a positive culture with the infecting organism identified and susceptible antibiotics. In the event a
patient is treated medically, the authors stressed the importance of close clinical follow up, recognizing
that “quick and progressive neurological decline is not uncommon.”
Three additional studies also concluded that surgery is the preferred treatment for SEA. Curry et
al. found patients treated medically were significantly more likely to deteriorate clinically and have
poorer outcomes than surgically treated SEA patients(2). A staggering 50% of medically treated patients
1
in the cohort required conversion to surgical management for new neurologic deficit while on
antibiotics. The authors cautioned against medical management, and suggested improved outcomes
with urgent surgery. Similarly, Connor et al., despite a lower failure rate of medical management (15%)
compared to the previous two studies, reached a similar conclusion, particularly in the setting of focal
neurologic deficit(3). The final and most recent study favouring early surgery, by Patel et al., is discussed
below(4).
Several authors have reported equivocal results when comparing surgical to medical groups for
the treatment of SEA. However, many have pointed out that prior to initiating treatment the groups
were likely quite different in terms of neurologic status and other baseline characteristics (5-7). In 2002,
Tang et al. retrospectively reviewed the clinical course of all 46 patients treated at their centre over a 10
year period(6). The authors reported three factors associated with a poor prognosis irrespective of
treatment type: abscess involving the cervical spine, an erythrocyte sedimentation rate (ESR) higher
than 110mm/h, and low platelet counts (<100 x 109/L). Twenty-four percent of the patients in the
medical group had a poor outcome, defined as “no improvement in neurological impairments, death, or
relapse of disease.” The authors suggested that medical management may be considered in patients
who: are poor surgical candidates, have a long abscess precluding adequate surgical debridement
without an extensive approach, and lack any neurological deficit or present with complete paralysis for
greater than three days(6). Karikari et al. were the first to examine the relationship between anatomic
abscess location and patient prognosis(5). Overall no difference was found in outcomes between the
surgical and medical groups, however it was suggested that patients with a ventrally located abscess (in
front of the spinal cord/cauda equina) were more likely to be successful with medical management than
those with a dorsally located abscess (behind the spinal cord/cauda equina). The authors concluded that
patients with a ventral abscess are more likely to have fever and present at an earlier stage of disease
than those with a dorsal abscess, thereby enabling eradication of the infection with antibiotics alone.
2
However, the outcomes did not differ significantly between the two anatomic groups. The authors
recommended consideration of anatomic location when deciding on medical versus surgical
treatment(5).
Four recent comparative studies provide evidence to support medical management of SEA(811). Siddiq et al. documented the clinical course and outcomes of all 57 patients treated at an academic
centre over 14 years ending in 2002. In the cohort, only 6% of the medically managed group required
surgery, which is one of the lowest reported rates in the literature. The authors concluded that patients
with SEA, regardless of the presence of neurologic deficit, may be safely treated with antibiotics alone or
combined with percutaneous drainage, with the caveat that close clinical follow-up is necessary(11). The
results of the study by Savage et al. were also favorable toward medical management(10). In the cohort
of 29 SEA patients who were initially treated medically, only three (11%) required conversion to surgical
management. Chen et al., demonstrated a 38% failure rate of medical management in a cohort of
exclusively Staph. aureus bacterial SEA, which is the most common infecting organism(8). The authors
concluded that certain patient groups may benefit from medical management, such as those with
minimal neurologic deficit, renal failure, or malignancy.
In summary, of the 11 studies included in the systematic review, four recommended surgical
management of SEA(1-4), four supported medical management in selected patients with close followup(8-11) and in three, the results were conflicted(5-7). Clearly, there is no consensus with respect to the
management of SEA. While a topic of debate for many years, recent literature has shown an overall
trend toward medical management and to assist in clinical decision making there have been efforts to
predict which patients are expected to fail this treatment. In 2014 the two largest studies on SEA to date
were published with 128(4) and 355(9) patients, of which 51 and 142 patients, respectively, were
treated medically. Both addressed a gap in the literature regarding the variables that predict failure of
medical management of SEA. These studies were the first to publish criteria for predicting failure, based
3
on regression analyses of multiple possible contributory patient characteristics. Aside from the presence
of diabetes, none of the criteria from these studies were shared.
The prediction models identified in the systematic review and meta-analysis were then applied
to a cohort of patients at our tertiary care centre to assess the external validity of the two models. To
our knowledge this is the first external cohort to which these criteria have been applied. The second
manuscript in this thesis (Chapter 3) pertains to this study.
Authorship:
In the systematic review and meta-analysis presented in Chapter 2, the student was the primary author
and was responsible for: developing the topic and research question, writing the proposal, title/abstract
review, full text review, analysis of the results, writing the manuscript and submitting for publication.
The 2nd author, Karla Gustafson, assisted with the proposal writing, title/abstract review, full text review,
statistical analysis and writing of the manuscript. The 3rd author, Ken Thomas, was integral to the
development of the research question, oversaw the project from proposal to submission for publication,
provided clinical insights, and reviewed and edited the final manuscript. The 4th author, Matthew James,
oversaw the project from proposal to submission for publication, was particularly integral in developing
and carrying out the methodology, and reviewed and edited the final manuscript.
In the paper presented in Chapter 3, the student was the primary author and was responsible
for: developing the research question, writing the research proposal, carrying out the chart review,
performing the statistical analysis, writing the manuscript, presenting the results and submitting for
publication. The 2nd author, Peter Faris, provided assistance with respect to the statistical analysis at the
proposal stage, as well as throughout the analysis stage of the project. The 3rd author, Ken Thomas,
oversaw all stages of the project, and was involved in reviewing and editing the manuscript.
4
References
1.
Reihsaus E, Waldbaur H, Seeling W. Spinal epidural abscess: a meta-analysis of 915 patients.
Neurosurg Rev. 2000;23(4):175-204; discussion 5.
2.
Darouiche RO. Current concepts: spinal epidural abscess. New England Journal of Medicine.
2006;355(19):2012.
3.
Nussbaum ES, Rigamonti D, Standiford H, Numaguchi Y, Wolf AL, Robinson WL. Spinal epidural
abscess: a report of 40 cases and review. Surgical neurology. 1992;38(3):225-31.
4.
Rigamonti D, Liem L, Sampath P, Knoller N, Namaguchi Y, Schreibman DL, et al. Spinal epidural
abscess: contemporary trends in etiology, evaluation, and management. Surgical neurology.
1999;52(2):189-96; discussion 97.
5.
Savage K, Holtom PD, Zalavras CG. Spinal epidural abscess: early clinical outcome in patients
treated medically. Clin Orthop. 2005;439:56-60.
6.
Heusner AP. Nontuberculous spinal epidural infections. The New England journal of medicine.
1948;239(23):845-54.
7.
Davis DP, Wold RM, Patel RJ, Tran AJ, Tokhi RN, Chan TC, et al. The clinical presentation and
impact of diagnostic delays on emergency department patients with spinal epidural abscess. The Journal
of emergency medicine. 2004;26(3):285-91.
8.
Arko Lt, Quach E, Nguyen V, Chang D, Sukul V, Kim BS. Medical and surgical management of
spinal epidural abscess: a systematic review. Neurosurgical focus. 2014;37(2):E4.
9.
Tuchman A, Pham M, Hsieh PC. The indications and timing for operative management of spinal
epidural abscess: literature review and treatment algorithm. Neurosurgical focus. 2014;37(2):E8.
10.
Hlavin ML, Kaminski HJ, Ross JS, Ganz E. Spinal epidural abscess: a ten-year perspective.
Neurosurgery. 1990;27(2):177-84.
11.
Connor DE, Jr., Chittiboina P, Caldito G, Nanda A. Comparison of operative and nonoperative
management of spinal epidural abscess: a retrospective review of clinical and laboratory predictors of
neurological outcome. Journal of Neurosurgery Spine. 2013;19(1):119-27.
12.
Curry Jr WT, Hoh BL, Amin-Hanjani S, Eskandar EN. Spinal epidural abscess: Clinical presentation,
management, and outcome. Surgical neurology. 2005;63(4):364-71.
13.
Khanna RK, Malik GM, Rock JP, Rosenblum ML. Spinal epidural abscess: evaluation of factors
influencing outcome. Neurosurgery. 1996;39(5):958-64.
14.
Patel AR, Alton TB, Bransford RJ, Lee MJ, Bellabarba CB, Chapman JR. Spinal epidural abscesses:
Risk factors, medical versus surgical management, a retrospective review of 128 cases. Spine Journal.
2014;14(2):326-30.
15.
Chen WC, Wang JL, Wang JT, Chen YC, Chang SC. Spinal epidural abscess due to Staphylococcus
aureus: clinical manifestations and outcomes. Journal of Microbiology, Immunology & Infection.
2008;41(3):215-21.
16.
Kim SD, Melikian R, Ju KL, Zurakowski D, Wood KB, Bono CM, et al. Independent predictors of
failure of nonoperative management of spinal epidural abscesses. Spine Journal: Official Journal of the
North American Spine Society. 2014;14(8):1673-9.
17.
Siddiq F, Chowfin A, Tight R, Sahmoun AE, Smego Jr RA. Medical vs surgical management of
spinal epidural abscess. Archives of internal medicine. 2004;164(22):2409-12.
18.
Chen SH, Chang WN, Lu CH, Chuang YC, Lui CC, Chen SF, et al. The clinical characteristics,
therapeutic outcome, and prognostic factors of non-tuberculous bacterial spinal epidural abscess in
adults: a hospital-based study. Acta neurol. 2011;20(2):107-13.
19.
Karikari IO, Powers CJ, Reynolds RM, Mehta AI, Isaacs RE. Management of a spontaneous spinal
epidural abscess: a single-center 10-year experience. Neurosurgery. 2009;65(5):919-23; discussion 23-4.
5
20.
Tang HJ, Lin HJ, Liu YC, Li CM. Spinal epidural abscess--experience with 46 patients and
evaluation of prognostic factors. Journal of Infection. 2002;45(2):76-81.
6
Chapter 2
Manuscript Title:
Incidence and Risk Factors for Failure of Medical Management of Spinal Epidural Abscess: A Systematic
Review and Meta-analysis
Authors:
Alexandra Stratton, MD 1,2
Karla Gustafson, MD 1
Kenneth Thomas, MD 1, 2
Matthew T James, MD 1,3
1. Department of Community Health Sciences, University of Calgary, Alberta, Canada
2. Department of Surgery, University of Calgary, Alberta, Canada
3. Department of Medicine, University of Calgary, Alberta, Canada
7
Abstract
BACKGROUND CONTEXT: Spinal epidural abscess (SEA) is a life threatening infection. There is
uncertainty whether medical versus surgical treatment is the ideal initial management approach for
neurologically intact patients with SEA. Recent evidence demonstrates that initial medical management
is increasingly common, however patients who ultimately require surgery after failed medical
management may have a worse prognosis than those whose treatment was initially surgical.
PURPOSE: The primary objective of this study was to establish the current incidence of failure of medical
management for SEA. The secondary aim was to identify risk factors associated with failure of medical
management for SEA.
STUDY DESIGN/SETTING: Systematic review and meta-analysis.
METHODS: We searched electronic databases (MEDLINE, EMBASE, CINAHL, and PubMed), recent
conference proceedings, and reference lists of relevant articles. We included studies that reported
original data on consecutive adult patients with SEA treated medically.
RESULTS: We identified 12 studies that met the inclusion criteria, which included a total of 489 medically
treated patients with SEA. Agreement on articles for inclusion was very high between the reviewers
(kappa 0.86). In meta-analysis, the overall pooled risk of failure of medical management was 29.3% (95%
CI: 21.4-37.2%) and when medical to surgical crossover was used to define failure the rate was 26.3%
(95% CI: 13.0-39.7%). Only six studies provided data for analysis by intended treatment, with a pooled
estimate of failure of medical management of 35.1% (95% CI: 15.7-54.4%). Two studies reported
predictors of failure of medical management.
CONCLUSIONS: Although we found that the incidence of failure of medical management of SEA was
relatively common in published reports, estimates were highly heterogeneous between studies, thereby
8
introducing uncertainty about the frequency of this risk. A consensus definition of failure is required to
facilitate comparison of failure rates across studies.
Background
Spinal epidural abscess (SEA) is a life-threatening infection involving the epidural space. Current
mortality rates are estimated at 5-16% worldwide (12, 13), and fewer than half of patients who survive
this infection fully recover (12). While SEA is a relatively rare diagnosis, its incidence has nearly doubled
in the last fifty years possibly owing to increased IV drug abuse, long-term vascular access in conjunction
with spinal instrumentation, the aging population, and availability of magnetic resonance imaging (MRI)
that has increased sensitivity of detection (10, 13-15).
Once a diagnosis of SEA has been made, treatment is largely based upon the presence or
absence of acute or progressive neurologic deficits (16). In the latter group, there is little debate that
urgent surgical debridement is required (16). There is disagreement, however, as to the optimum initial
treatment approach for patients without neurologic deficit (17). A recent systematic review revealed
that since 1999 there has been an overall trend toward more frequent medical management for
neurologically intact patients with SEA, although practices remain highly variable(16). Nonsurgical
management consists of several weeks of IV antibiotics and close clinical follow-up. Patients who
ultimately require surgery after failed medical management may have a worse prognosis than those
whose treatment was initially surgical (18). Information on the risk of failure of medical management for
SEA is important to inform clinical decision making and enable physicians to communicate to patients
their risk of failure when discussing treatment options.
We performed a systematic review and meta-analysis to establish the current incidence of
failure of medical management for SEA. The secondary aim of our systematic review was to identify risk
9
factors associated with failure of medical management for SEA to aid clinicians in predicting who can
safely be treated medically.
Methods
We followed a pre-specified protocol for study selection and analysis of results, and followed the
reporting recommendations outlined in the MOOSE guidelines (19).
Data Sources and Searches
We searched the electronic databases of MEDLINE, EMBASE, CINAHL, and PubMed from inception to
October 27, 2014. Our search in MEDLINE included Medical Subject Headings (MeSH) for the study
population theme “central nervous system infections,” or “bacterial central nervous system infections,”
or “epidural abscess.” Our search terms for the exposure theme included the exploded MeSH “operative
surgical procedure,” or “anti-bacterial agents,” and title and abstract keyword terms “treatment,” or
“management.” We combined the results of the study population search theme with that of the
exposure search theme using the Boolean operator “AND.” We conducted an identical search, using
EMTREE headings and title and abstract keyword terms, of EMBASE restricted to adults. We also
searched CINAHL for the term “epidural abscess,” as well as PubMed using “CNS infections” and
“antibacterial agents” or “operative surgical procedure” for 2014 to identify new articles not yet indexed
in MEDLINE. All citations identified through database searching were imported into EndNote X7 citation
software and duplicates were removed.
In addition, we searched the reference lists of relevant articles, and the 2012-2014 conference
proceedings from the North American Spine Society and International Society for the Study of the
Lumbar Spine. Non-English articles included for full text review were translated as required. All citations
were screened and full text reviews conducted by two independent physician reviewers. We contacted
10
authors for additional information when necessary and for data from identified abstracts not published
as full manuscripts.
Study Selection
Articles were selected for inclusion using a two-step process. In the first step, we screened abstracts for
inclusion if a study reported original data on the consecutive selection of adult patients with SEA and
included patients treated medically. In the second step we conducted full text review of articles. Our
primary outcome of failure of medical management was taken broadly acknowledging the variability of
clinical definitions for this outcome. We included studies that reported outcomes of persistent severe
neurologic deficits, poor clinical outcome, worse clinical outcome, death, or requirement for surgery in
patients who had received antibiotic therapy within the definition of failure of medical management.
Articles were excluded after full text review if there were: no specific information on the outcome of
medically treated patients, ten or fewer medically treated patients reported, non-consecutive patients
included, specific etiologies such as post-operative infections or no original data provided.
Data Extraction
During full text review we extracted data on the criteria for medical management and the number of
patients treated medically and surgically. Patients who were treated with percutaneous abscess
drainage only were considered to have received medical management. We collected data from studies
for secondary analysis of the primary outcome in which failure of medical management was based solely
on the need for a medically managed patient to cross over to surgical management due to persistent or
deteriorating neurologic or medical condition. Additional information was collected on demographics,
patient comorbidities and immune status, diagnostic procedures, and outcome information of patients
where available. All data on predictors of failure of medical management was extracted, when provided,
for analysis of the secondary outcome.
11
Quality Assessment
We evaluated the quality of each included study based on a six-item checklist adapted from Altman’s
framework for assessing the internal validity of prognostic studies (20). This checklist was comprised of
the following parameters: patient inclusion criteria defined, diagnostic criteria described, patient
characteristics described, treatment described, outcome known, and follow-up described. Two
reviewers independently graded the six items as “full” or “poor.” We considered a study to fully describe
patient inclusion criteria if all patients with SEA over a given time period were included or the inclusion
and exclusion criteria were specified. We considered a full description of diagnostic criteria to explicitly
state how patients were diagnosed (for example, clinical, lab, culture/biopsy or imaging results). Full
reporting of patient characteristics meant demographic information was provided as well as patient
comorbidities. Treatment was rated as fully described for studies that outlined intended treatment
course, crossover if any, length of time to operation, type of operation, duration of antibiotic treatment,
number of patients undergoing percutaneous drainage and how patients were classified in terms of
treatment group in the analysis. Outcome was fully described if outcome by treatment group including
number of deaths in each group was presented. Studies that reported the length of follow up time to be
at least six weeks, and provided explanation of outcome assessment were rated “full” for follow-up
description.
Data Synthesis and Analysis
We determined the pooled proportion of patients who failed medical management using a random
effects model. A random effects model was chosen given the variability in preferred treatment of SEA at
different centres, the spectrum of symptoms on presentation, and inconsistent treatment regimens.
Heterogeneity was quantified using the I2 statistic. For studies with overlapping patients, only the larger
study was included in pooled analysis. Stratified and sensitivity analyses were performed to investigate
12
potential causes of heterogeniety. Stratified analysis was conducted on inclusion of patients with
tuberculosis (TB), precentage of included patients with methicillin-resistant Staphylococcus aureus
(MRSA) and diabetes, and the quality of study treatment description. Sensitivity analyses were
performed to replace the larger of the duplicate cohorts with the smaller cohort and to exclude studies
that included pediatric patients. All analyses were performed using STATA version 13 software with
alpha set at 0.05.
Results
Identification of Studies
Our electronic bibliographic database search identified 4029 unique citations (Fig. 1). Screening of title
and abstracts by both reviewers yielded 59 articles for full text review (Fig. 1). After full text review, 12
observational studies were included for analysis. Agreement between reviewers was very high (Cohen’s
kappa coefficient of 0.86 for inter-rater agreement)(21).
Study Characteristics
All included studies reported the number of patients considered to have failed medical management,
with this outcome variably defined across studies (Table 1). Two studies had an apparent overlap of
patients older than 50 years old between 1999 and 2007 (5, 22). All studies reported the proportion of
patients with positive cultures, either from blood or abscess aspirate, as well as the most common
pathogens. The proportion of patients with a positive culture ranged from 69-100%, with a mean of 86%
across studies (Table 1). All but one study did not, however, detail the pathogen by treatment group. Six
studies specifically reported those patients in the medically treated group who crossed over to the
surgery group as a result of worsening symptoms (1, 2, 4, 9-11). In five of the six studies that reported
the number of patients who crossed over to surgery, it was possible to determine the proportion who
13
failed treatment based on a composite of requiring surgery, worsening clinical status, or death to
enhance consistency across studies for comparison (Table 1)(1, 2, 9-11).
Study Quality
Reporting of inclusion criteria and patient characteristics was well done across studies (Table 2). Only
one study reported exclusion of patients presenting with neurologic compromise of greater than 48
hours, which is most applicable to the clinical practice of offering medical therapy to patients without
early indications for surgery (9). The description of treatment was poor in most studies with only four
reporting the intended treatment plan (2, 4, 9, 10). Eight studies reported results based on the final
treatment received without specifying if the final treatment was the intended one (1, 3, 5-8, 11, 22). Of
these eight studies, three specified time from diagnosis to surgery which ranged from zero to 42 days (3,
7, 8). Only five studies reported the number of patients who died in the two treatment groups (5, 8-10,
22). Description of follow-up was poor with ten studies failing to report minimum patient follow-up of
six weeks or higher to ensure resolution of infection (2-8, 10, 11, 22).
Pooled Proportion of Failure of Medical Therapy
We identified 11 studies that reported data on the proportion of patients who failed medical
management, with failure defined as including: poor outcomes, worse neurologic status, and death, in
addition to patients requiring surgery. The pooled proportion of failure of medical management was
29.3% (95% CI: 21.4-37.2%)(Fig. 2). The I2 value of 69.5% indicated moderate heterogeneity. We
performed a secondary analysis including six studies that reported medical to surgical crossover as a
definition of failure (1, 2, 4, 9-11); the pooled proportion of failure of medical management by this
definition was 26.3% (95% CI: 13.0-39.7%), with an I2 value of 87% indicating high heterogeneity (Fig. 3).
Sensitivity Analysis
14
We conducted sensitivity analyses to test the effect on the pooled estimate of removing the single study
that included pediatric patients (5), the study that was restricted to patients over age 50yrs. (22), and
the study that included only S. aureus infections (8). Removing each of these studies individually did not
significantly change the pooled risk estimate of 29.3% (95% CI: 21.4-37.2%). When the larger study by
Karikari (5) that included duplicate patients to the Adogwa study was excluded and replaced by the
study by Adogwa (22), the pooled result remained similar; 29.8% (95%CI: 21.9-38.7%).
Stratified Analysis
To investigate potential sources of heterogeneity, we performed stratified analyses and metaregression. When studies were stratified by those with greater versus less than 20% MRSA infection
rates (among all treatment groups), we found no difference in the risk of failure of medical management
between the two strata (Table 3), with the proportion of MRSA (high or low) accounting for only 3.4% of
the heterogeneity observed. Stratification of studies by inclusion or exclusion of TB infection showed no
difference in risks of medical treatment failure, and resulted in an increase in heterogeneity between
studies within strata (Table 3). With stratification according to adequacy of treatment description, the
difference in the estimated failure of medical management between those studies with full versus poor
treatment descriptions was not statistically significant (Table 3), with 10.0% of the heterogeneity
accounted by this variable. When studies were stratified by those with greater versus less than 20% of
patients with diabetes (among all treatment groups), we did not find a significant difference in the risk
of failure of medical management between the two strata (Table 3). The proportion of patients with
diabetes accounted for only 2.9% of the heterogeneity observed.
Publication Bias
15
A funnel plot and Begg's test (p=0.27) provided no evidence of statistically significant small study effects,
however, we noted more studies with fewer patients reporting higher proportions of failure of medical
management (Fig. 4).
Predictors of Medical Treatment Failure
Two studies reported predictors of failure of medical management (Table 4). Kim et al. reported that
“age greater than 65, diabetes, MRSA and neurologic impairment” were significant risk factors for
medical treatment failure in multivariable models (9), while “positive blood cultures, diabetes, WBC >
12.5, and c-reactive protein >115” were reported as significant univariable predictors by Patel et al. (4).
The odds ratios associated with each criterion are reported in Table 4.
Discussion
Our systematic review of the incidence of failure of medical management for SEA identified twelve
observational studies with at least ten patients initially treated medically. In pooled analysis of 11
studies reporting independent patients, we found that the proportion of SEA patients who failed
medical management was 29% with moderate heterogeneity between studies. When we defined
treatment failure based on six studies that reported on patients requiring surgery after initial attempts
at medical management, the resulting estimate of proportion of failure was 26% with high
heterogeneity between studies. In stratified analyses to explore potential sources of heterogeneity, we
found that studies that reported outcomes according to intended treatment strategy reported higher
risks of treatment failure than those that reported based on the final treatment received (35% vs. 25%).
Our findings highlight the limitation of studies which do not report outcomes based on the intended
treatment strategy, which likely underestimate the failure rate of initial medical management.
We found significant heterogeneity between studies that could not be fully explained in
stratified or sensitivity analyses. Further sources of clinical heterogeneity between studies that we were
16
not able to test for include: delay between onset of symptoms and diagnosis of SEA, anatomic location
of abscess(es), timing of first dose of antibiotics (before or after cultures), use of appropriate pathogenspecific antibiotics, type and virulence of organism (other than MRSA and TB), patient immune status,
use of percutaneous drainage/aspiration, criteria for selecting treatment strategy, rationale for crossing
over to surgery, time to surgery in the surgical group, outcome definitions, and length of follow-up. Two
reviews have recently published on the treatment of SEA (16, 17); however, neither was specifically
designed to study the failure of medical management in this disease. Arko et al. reported large
variability in the proportion of medical treatment failure (defined as requiring surgery) ranging from 649% across studies (16). This is consistent with the estimate we identified using the same definition of
treatment failure (26%). Tuchman et al. did not report the proportion of failure (17).
The strengths of our systematic review and meta-analysis include a comprehensive literature
search including non-English language articles and high inter-rater agreement for study selection. Our
systematic review is, however, limited by the quality of the published literature. Studies on the
management of SEA have all been retrospective in nature as the incidence of SEA is small. The definition
of treatment failure has varied considerably across studies, suggesting a need for a consensus on the
definition for this outcome to facilitate outcomes research. Furthermore, it is critical for authors to
specify whether patients were treated medically because of a lack of surgical indication(s), or rather
comorbidities precluding surgical intervention. Another important limitation that we identified was the
small proportion of studies reporting the outcome by intended treatment course. In the studies that did
not report outcome based on intended treatment, there was a lengthy delay from the time of diagnosis
to surgery (up to 42 days) in the three studies that reported this information (3, 7, 8). This delay
suggests that the intended treatment course in an unspecified number of patients was not likely
surgical, despite the fact that these patients were included in the surgical group for outcome reporting.
17
This is the first meta-analysis to report the proportion of failure of medical management of
spinal epidural abscess by intended treatment course, and by failure definition. Our secondary objective
was to identify risk factors associated with failure of medical management for SEA to aid clinicians in
predicting which patients can safely be treated medically. We found that in 2014 the two largest studies
on SEA both addressed a gap in the literature regarding the variables that predict failure of medical
management of SEA. These studies were the first to publish criteria for predicting medical treatment
failure, based on regression analyses of multiple possible contributory patient characteristics. However,
the predictor variables studied differed, resulting in the identification of distinct prediction criteria
(Table 3) which do not overlap aside from the presence of diabetes. Further research will be needed to
synthesize this information to derive and validate more robust multivariable prediction models.
The results of this meta-analysis are generalizable to all patients with SEA treated medically;
however, a more clinically useful estimate would be based on a population of patients excluding those
who are not surgical candidates due to medical comorbidities and those with prolonged neurologic
deficit (greater than 36-48 hours). Inclusion of these patients in studies comparing medical to surgical
management is problematic because such patients have a poor prognosis regardless of treatment. By
excluding patients with symptoms of a complete spinal cord injury present for greater than 48 hours on
presentation, the study by Kim et al. is the most relevant to initial treatment decisions for early surgical
versus medical management among patients eligible for either approach (9). This study reported the
proportion of failure of medical management to be 30% (requiring surgery) and 38% (including deaths).
This rate of failure of medical management is extremely high and reinforces the need for further
research to help physicians identify patients who may be appropriately treated medically. Future
research should focus on this subset of SEA patients without prolonged neurologic deficit and without
contraindications to surgery and employ standardized outcome reporting using validated functional
outcome questionnaires. These standards would make results comparable across studies and assist in
18
clinical decision making. Further research to validate the prognostic accuracy of the prediction criteria
suggested by Kim et al. and Patel et al. is warranted.
In conclusion, we found high variability across studies with respect to rationale for chosen
treatment, definition of treatment failure and outcome reporting for SEA. Although we found that the
incidence of failure of medical management of SEA was relatively common in published reports,
estimates were highly heterogeneous between studies, thereby introducing uncertainty about the
frequency of this risk. Future research in this area should clearly document: the intended treatment
course for all patients, the time delay between initiation of antibiotic treatment and surgery when
required, and outcomes based on intended treatment. Furthermore, it is critical for a consensus
definition of failure to be developed to facilitate comparison of failure rates across studies.
Figures and Tables
Citations from
Literature Search
(n = 5323)
MEDLINE n = 2217
EMBASE n = 2471
PubMed n = 235
CINAHL n = 398
Conference Abstracts
n=2
Citations Remaining
After Duplicates
Removed n = 4029
Citations from
Reference Lists
n=4
Articles Identified for
Full Text Review
n = 59
Citations Excluded
after Title/Abstract
Screen n = 3974
Articles Excluded (n = 47)
Inadequate Outcome Information n = 14
Ten or Fewer Medical Patients n = 27
No Original Data n = 3
Non-Consecutive Patient Inclusion n = 3
Articles Included in
Meta-Analysis
n = 12
Figure 1. Study Selection.
19
%
Study
ES (95% CI)
Weight
Khanna (1996)
0.56 (0.33, 0.77)
6.06
Tang (2002)
0.24 (0.11, 0.45)
8.07
Siddiq (2004)
0.18 (0.08, 0.34)
10.29
Savage (2005)
0.14 (0.05, 0.31)
10.40
Curry (2005)
0.52 (0.33, 0.71)
7.28
Chen WC (2008)
0.20 (0.08, 0.42)
8.33
Karikari (2009)
0.23 (0.14, 0.34)
11.33
Chen SH (2011)
0.37 (0.19, 0.59)
6.85
Connor (2003)
0.15 (0.05, 0.36)
9.09
Patel (2014)
0.41 (0.29, 0.55)
9.99
Kim (2014)
0.38 (0.30, 0.46)
12.32
Overall (I^2 = 69.5%, p = 0.000)
0.29 (0.21, 0.37)
100.00
0
.1
.8
Proportion (95% CI)
ES = effect size (proportion)
Figure 2. Forest plot of proportion of patients who failed medical management based on any definition
of failure reported by a study.
%
Study
ES (95% CI)
Weight
Khanna (1996)
0.31 (0.14, 0.56)
12.87
Siddiq (2004)
0.06 (0.02, 0.19)
19.12
Savage (2005)
0.10 (0.04, 0.26)
17.96
Curry (2005)
0.48 (0.29, 0.67)
13.86
Patel (2014)
0.41 (0.29, 0.55)
16.94
Kim (2014)
0.30 (0.23, 0.38)
19.25
Overall (I^2 = 87.0%, p = 0.000)
0.26 (0.13, 0.40)
100.00
0
.1
.8
Proportion (95% CI)
ES = effect size (proportion)
Figure 3. Forest plot of proportion of patients that failed based on failure defined as requiring surgery.
20
0
Standard Error
.05
.1
.15
0
.2
.4
Proportion
Figure 4. Funnel plot with pseudo-95% confidence limits.
21
.6
Author
Year
Exclusion Criteria
Adogwa1
2014
<50 yrs, postoperative infections
<18 yrs, postoperative infections, complete
SCI from SEA >48 hrs,
spondylodiscitis, osteomyelitis
<18 yrs, post-intervention infections, TB,
treatment initiated at outside facility, discitis
or osteomyelitis, negative intraoperative
findings and culture
<17 yrs, concurrent intracranial pathology,
no data for SEA diagnosis
Kim12
2014
Patel14
2014
Connor6
2013
Chen SH4
2011
Karikari10
2009
Chen WC5
2008
Curry7
2005
Savage18
2005
Siddiq19
2004
<17 yrs, TB
postoperative infections,
note: <18 yrs included
<18 yrs, any organism except Staph. aureus,
polymicrobial infections
TB, spinal infections without epidural
extension
Tang21
2002
TB, infective spondylitis, paraspinal abscess
without epidural involvement
Khanna11
1996
TB
Mean
Age
(yrs)
Total
Pts
Pts with
Positive
Culture (%)
65
82
88
59.9
355
Most Common Organisms (%
total pts growing each)
MSSA (28), MRSA (28), coag.
negative Staph. (11)
MSSA (51),
MRSA
(17),
Strep. sp. (12)
92
Management
Failure Definition
Failure
Pts
Percent
Failure
52
poor*/death
14
27%
required surgery
42
30%
142
+death
+12
38%
51
required surgery
21
41%
20
Poor~/death
3
15%
19
poorβ
7
37%
62
poor*/death
14
23%
20
poorα/death
required surgery
+poor*
required surgery
+death
required surgery
4
11
+1
3
+1
2
20%
48%
52%
11%
14%
6%
34
+poor#/death
+4
18%
21
poorµ/death
required surgery
5
5
24%
31%
16
+poorπ/death
+4
56%
MSSA (40),
52.9
51.4
(med)
128
90
77
79
62
45
96
58
55
(med)
61
(med)
100
88
31
100
48
81
45.1
52
69
57
100
60
46
70
54
41
93
MRSA
(30),
Streptococcus (6)
MRSA (31), MSSA (30),
Staph. epidermidis (3)
MSSA (38), MRSA (27), gram
negatives (29)
MSSA (33), MRSA (30), coag.
negative Staph. (10)
MSSA (68),
MRSA
(32)
MSSA (54), MRSA (6), Strep.
sp. (8)
MSSA (42), MRSA (12), Strep.
sp. (12)
Staph. aureus (60), Strep.
sp. (16),
coag.
negative. Staph. (9)
MSSA (24), MRSA (15), Strep.
agalactiae (9)
Staph. aureus (61),
Strep.
viridans (10), Strep.
pyogenes (5)
Initial Medical
Treatment
23
29
Table 1. Characteristics of studies included in systematic review. For studies in which it was possible to separate out outcome measures, a second line was
added to display poor outcomes and deaths. The definitions of “poor” were variable and are listed here: *=worse neurologic status at follow-up compared to
presentation, ~=no improvement or worsening, β=Barthel index score <12 or death, α= recurrence or worsening clinical symptoms after discharge, #=treatment
failure or relapse, µ=no improvement in neurologic impairments or relapse, π=severe paresis or plegia +/- bowel/bladder dysfunction; pts=patients,
med=median, TB=tuberculosis, SEA=spinal epidural abscess, SCI=spinal cord injury, MSSA=Methicillin/oxacillin sensitive Staphylococcus aureus, MRSA=
Methicillin/oxacillin resistant Staphylococcus aureus, Strep=Streptococcus, sp=species, Staph=Staphylococcus, coag=coagulase.
22
Inclusion
Criteria
Diagnostic
Criteria
Patient
Characteristics
Treatment
Outcome
Follow-up
Study
Adogwa1 (2014)
Chen SH4 (2011)
Chen WC5 (2008)
Connor6 (2003)
Curry7 (2005)
Karikari10 (2009)
Khanna11 (1996)
Kim12 (2014)
Patel14 (2014)
Savage18 (2005)
Siddiq19 (2004)
Tang21 (2002)
full
full
full
full
full
full
full
full
full
full
full
full
full
full
full
poor
poor
full
poor
full
poor
partial
poor
partial
full
full
full
full
full
full
full
full
full
full
full
full
poor
poor
poor
poor
full
poor
poor
full
full
full
poor
poor
full
partial
full
full
partial
full
poor
full
poor
full
full
poor
poor
poor
poor
poor
poor
poor
poor
poor
poor
poor
poor
poor
Table 2. Study Quality Assessment.
Variable
Number of
studies
Number of
patients*
Proportion of
FMT (%)
Meta-regression
p-value
Heterogeneity
(%)
Treatment
description:
0.29
10.0
Full
4
583
35.1 (15.7-54.4)
Poor
7
397
25.4 (13.1-37.6)
MRSA:
0.32
3.4
>20%
4
235
24.1 (4.0-44.1)
<20%
5
629
33.1 (20.2-46.1)
TB patients:
0.82
-19.0
Included
5
637
28.3 (7.9-48.8)
Excluded
6
343
30.5 (16.1-44.9)
Diabetes:
0.38
2.9
>20%
7
792
32.3 (12.1-52.5)
<20%
4
188
24.0 (7.9-40.2)
Table 3. Stratified analyses. FMT=failure of medical treatment. *=Total number of patients in study.
23
Kim et al. 201412
Multivariable OR
Patel et al. 201414
Univariable OR
Age >65
2.5
Positive blood culture
3.5
Diabetes
2.5
Diabetes mellitus
3.8
MRSA
3.8
Leukocytosis > 12.5
3.3
Neurologic impairment
15.2
CRP > 115
4.7
(incomplete or complete cord
injury)
Table 4. Published predictors for failure of medical management, with associated odds ratios (OR).
Author
Total number
of patients
Kim12
Patel14
Connor6
Chen SH4
Karikari10
Chen WC5
Curry7
Savage18
Siddiq19
Tang21
Khanna11
Totals
355
128
77
45
104
31
48
52
57
46
41
577
Number with
abscess cervical
involvement
not reported
46 (35.9%)
18 (23.4%)
7
15
7 (22.6%)
11
not reported
16 (28%)
9
13
142 (24.6%)
Number with
abscess thoracic
involvement
not reported
50 (39.1%)
39 (50.6 %)
15
28
7 (22.6%)
7
not reported
10 (18%)
14
14
187 (32.4%)
Appendix Table 1. Abscess location by study.
24
Number with
Number with
abscess lumbar abscess sacral
involvement
involvement
not reported
not reported
70 (54.7%)
30 (23.4%)
20 (26.0 %)
23
35
19 (18%)
19 (61.3%)
22 (46%)
not reported
not reported
31 (54%)
23
5
1
248 (43.0%)
50 (8.7%)
Author
Percent classified as
immunosuppressed
Definition of
immunosuppression
Kim12
Total
number of
patients
355
15%
Patel14
Connor6
128
77
Chen SH4
Karikari10
Chen WC5
45
100
31
Curry7
48
Savage18
52
Siddiq19
57
Tang21
Khanna11
46
41
Totals
984
6%
Not directly reported
4% HIV
Not reported
10%
Not directly reported
13% Malignancy
6.5% HIV
Not directly reported
(~ from chart)
~17% Diabetes
~10% Malignancy
~2.5% AIDS
~2.5% Steroids
Not directly reported
2% HIV
2% Cancer
46% total –
- 14% liver disease
- 19% diabetes
- 9% malignancy
- 9% chemotherapy or
corticosteroids
- 7% autoimmune
disease
- 5% chronic renal
failure
- 5% alcoholism
- 2% splenectomy
Not reported
Not directly reported
5% Malignancy
2% Steroids
-------
Immunosuppressive
conditions and patients on
medications (chemotherapy
or long-term corticosteroids)
Not provided
Not provided
22%
26%
Not provided
Not provided
Not provided
33%
34%
19%
“Diabetes, malignancy,
dialysis-dependent chronic
renal disease, AIDS, and
steroid use are wellcharacterized causes of
immunosuppression”
Not provided
( ~ from chart)
~17%
19%
“At least 1
19%
immunosuppressive risk
factor including: cirrhosis or
chronic liver disease,
diabetes, malignancy,
cytotoxic chemotherapy or
corticosteroids, autoimmune
disease, chronic renal
failure, alcoholism, and
splenectomy”
Not provided
Not provided
--------
Appendix Table 2. Immunosuppression definition and proportion of patients by study.
25
Percent
classified as
diabetic
26%
46%
54%
19 – 54%
References
1.
Adogwa O, Karikari IO, Carr KR, Krucoff M, Ajay D, Fatemi P, et al. Spontaneous spinal epidural
abscess in patients 50 years of age and older: a 15-year institutional perspective and review of the
literature: clinical article. Journal of Neurosurgery Spine. 2014;20(3):344-9.
2.
ltman DG. Systematic reviews of evaluations of prognostic variables. BMJ (Clinical research ed).
2001;323(7306):224-8.
3.
Arko Lt, Quach E, Nguyen V, Chang D, Sukul V, Kim BS. Medical and surgical management of
spinal epidural abscess: a systematic review. Neurosurgical focus. 2014;37(2):E4.
4.
Chen SH, Chang WN, Lu CH, Chuang YC, Lui CC, Chen SF, et al. The clinical characteristics,
therapeutic outcome, and prognostic factors of non-tuberculous bacterial spinal epidural abscess in
adults: a hospital-based study. Acta neurol. 2011;20(2):107-13.
5.
Chen WC, Wang JL, Wang JT, Chen YC, Chang SC. Spinal epidural abscess due to Staphylococcus
aureus: clinical manifestations and outcomes. Journal of Microbiology, Immunology & Infection.
2008;41(3):215-21.
6.
Connor DE, Jr., Chittiboina P, Caldito G, Nanda A. Comparison of operative and nonoperative
management of spinal epidural abscess: a retrospective review of clinical and laboratory predictors of
neurological outcome. Journal of Neurosurgery Spine. 2013;19(1):119-27.
7.
Curry Jr WT, Hoh BL, Amin-Hanjani S, Eskandar EN. Spinal epidural abscess: Clinical presentation,
management, and outcome. Surgical neurology. 2005;63(4):364-71.
8.
Darouiche RO. Current concepts: spinal epidural abscess. New England Journal of Medicine.
2006;355(19):2012.
9.
Hlavin ML, Kaminski HJ, Ross JS, Ganz E. Spinal epidural abscess: a ten-year perspective.
Neurosurgery. 1990;27(2):177-84.
10.
Karikari IO, Powers CJ, Reynolds RM, Mehta AI, Isaacs RE. Management of a spontaneous spinal
epidural abscess: a single-center 10-year experience. Neurosurgery. 2009;65(5):919-23; discussion 23-4.
11.
Khanna RK, Malik GM, Rock JP, Rosenblum ML. Spinal epidural abscess: evaluation of factors
influencing outcome. Neurosurgery. 1996;39(5):958-64.
12.
Kim SD, Melikian R, Ju KL, Zurakowski D, Wood KB, Bono CM, et al. Independent predictors of
failure of nonoperative management of spinal epidural abscesses. Spine Journal: Official Journal of the
North American Spine Society. 2014;14(8):1673-9.
13.
Nussbaum ES, Rigamonti D, Standiford H, Numaguchi Y, Wolf AL, Robinson WL. Spinal epidural
abscess: a report of 40 cases and review. Surgical neurology. 1992;38(3):225-31.
14.
Patel AR, Alton TB, Bransford RJ, Lee MJ, Bellabarba CB, Chapman JR. Spinal epidural abscesses:
Risk factors, medical versus surgical management, a retrospective review of 128 cases. Spine Journal.
2014;14(2):326-30.
15.
Posner KL, Sampson PD, Caplan RA, Ward RJ, Cheney FW. Measuring interrater reliability among
multiple raters: an example of methods for nominal data. Statistics in medicine. 1990;9(9):1103-15.
16.
Reihsaus E, Waldbaur H, Seeling W. Spinal epidural abscess: a meta-analysis of 915 patients.
Neurosurg Rev. 2000;23(4):175-204; discussion 5.
17.
Rigamonti D, Liem L, Sampath P, Knoller N, Namaguchi Y, Schreibman DL, et al. Spinal epidural
abscess: contemporary trends in etiology, evaluation, and management. Surgical neurology.
1999;52(2):189-96; discussion 97.
18.
Savage K, Holtom PD, Zalavras CG. Spinal epidural abscess: early clinical outcome in patients
treated medically. Clin Orthop. 2005;439:56-60.
19.
Siddiq F, Chowfin A, Tight R, Sahmoun AE, Smego Jr RA. Medical vs surgical management of
spinal epidural abscess. Archives of internal medicine. 2004;164(22):2409-12.
26
20.
Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, et al. Meta-analysis of
observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies
in Epidemiology (MOOSE) group. Jama. 2000;283(15):2008-12.
21.
Tang HJ, Lin HJ, Liu YC, Li CM. Spinal epidural abscess--experience with 46 patients and
evaluation of prognostic factors. Journal of Infection. 2002;45(2):76-81.
22.
Tuchman A, Pham M, Hsieh PC. The indications and timing for operative management of spinal
epidural abscess: literature review and treatment algorithm. Neurosurgical focus. 2014;37(2):E8.
27
Chapter 3
Manuscript Title:
The Prognostic Accuracy of Suggested Predictors of Failure of Medical Management for Patients with
Spinal Epidural Abscess
Authors:
Alexandra Stratton, MD 1,2
Peter Faris, PhD 1
Kenneth Thomas, MD 1, 2
1. Department of Community Health Sciences, University of Calgary, Alberta, Canada
2. Department of Surgery, University of Calgary, Alberta, Canada
28
Abstract
RESEARCH QUESTION: Are the two published prediction criteria for failure of medical management for
patients with spinal epidural abscess (SEA) valid in an external cohort?
DESIGN: Retrospective cohort study.
SETTING/SUBJECTS: All inpatients over the last ten years at Foothills Medical Centre with a diagnosis of
SEA.
METHODS: Patients were identified using ICD-10 diagnostic codes; electronic and paper charts were
reviewed and relevant information extracted. The incidence of SEA at FMC was calculated, and how
often patients with SEA were treated medically. The rate of failure of medical management was
determined. The published prediction models were applied to our data to determine how predictive
they were of failure in our cohort.
RESULTS: 550 patients were identified using ICD-10 codes, 160 of whom had an MRI confirmed diagnosis
of SEA. Seventy-five patients were found to be intentionally managed and were included in the analysis.
The incidence of SEA was 16 patients per year. Thirteen patients failed medical management (17%). By
the published prediction criteria, 26% (Kim et al.) and 45% (Patel et al.) of our patients were expected to
fail. Patel’s model demonstrated better discrimination.
CONCLUSIONS: Neither model was valid in our cohort. Once calibrated to our cohort, Patel’s model
consisting of: positive blood culture, diabetes, WBC >12.5 and CRP >115 was the better model for our
data.
Background
Spinal epidural abscess (SEA) is a life-threatening infection involving the epidural space that can affect
the spine at any level. Current mortality rates are estimated at 5-16% worldwide(12, 13), and fewer than
29
half of patients who survive this infection fully recover(12). While SEA is a relatively rare diagnosis its
incidence has nearly doubled in the last fifty years, possibly owing to increased IV drug abuse, long-term
vascular access, spinal instrumentation, the aging population and magnetic resonance imaging (MRI)
enabling early and accurate detection(10, 13-15). The “classic triad” of SEA symptoms includes: fever,
back pain and neurologic deficit(23). Delay in initial diagnosis is reported in 11 to 75% of cases, as only
10% of patients present with this classic triad and early symptoms are nonspecific for this infrequent
diagnosis(13, 24). Once the elusive diagnosis has been made, treatment is largely based upon the
presence or absence of progressive neurologic deficit(16). In patients with progressive neurologic
deficits, there is little debate that urgent surgical debridement is required(16). There is disagreement,
however, as to the optimum initial treatment approach for patients without neurologic deficit(17).
A recent systematic review revealed that since 1999 there has been an overall trend toward
medical management for neurologically intact patients with SEA(16). Medical management consists of
weeks of IV antibiotics and close clinical follow-up. Patients who ultimately require surgery after failed
medical management, due to progressive neurological deficit, pain, or persistent bacteremia, have a
worse prognosis than those whose initial treatment was surgical. The failure rate, estimated to be
between 30 and 41%, for medical management requiring surgery is high(4, 9).
In 2014 two separate authors, Kim and Patel, proposed criteria to predict failure of medical
management of SEA(4, 9). The criteria published in these papers are depicted in Table 1. Kim et al.
suggested that: age over 65 years old, diabetes, methicillin-resistant Staph. aureus (MRSA) infection and
incomplete or complete spinal cord injury to be independent predictors of failure of medical
management; Patel et al. determined the following to be predictors of failure: positive blood culture,
diabetes mellitus, white blood cell count greater than 12.5 and c-reactive protein (CRP) level above 115.
Aside from the presence of diabetes, none of the criteria overlap. These criteria were derived from
retrospective observational studies and, importantly, have not been applied to patients outside of the
30
original cohorts. If the prediction criteria are to be implemented in the clinical setting, it is crucial to
determine the external validity of these two studies. The intention of our study was to validate these
prediction criteria using an external cohort and thus provide clarity to the ongoing debate regarding the
optimum treatment of SEA.
To accomplish this task our specific research objectives were to: 1) define the rate of spinal
epidural abscess at our tertiary care centre, 2) evaluate the practice patterns at this centre; specifically
the proportion of patients with SEA who are treated medically, 3) determine how well the published
predictors of failure of medical management predict failure in a cohort of patients from this centre and
4) develop our own model to predict failure of medical management in patients with SEA.
Methods
We conducted a retrospective cohort study, given the infrequent nature of this diagnosis. All patients
treated at a single tertiary care centre from January 2005-February 2015 with a diagnosis of spinal
epidural abscess confirmed by MRI/CT myelogram at any point during admission were identified using
ICD-10 diagnostic codes. The 32 codes used are shown in Table 2. Electronic and paper medical records
were reviewed. Exclusion criteria were: age <18yrs, patients presenting with a complete spinal cord
injury greater than 48 hours in duration, surgical management as the initial intended treatment course,
post-intervention SEA infections, patients with less than eight weeks follow-up, and tuberculosis
infections (Pott’s disease). For excluded patients, a rejection log was kept to document the reason for
exclusion.
Data extraction included: patient demographics (age, gender, and medical comorbidities), social
habits (IV drug use, smoking), degree of neurologic impairment on presentation, duration of symptoms
prior to diagnosis, infecting organism and susceptibilities, admission bloodwork (including WBC count,
ESR, CRP and blood culture results), spinal level of abscess, antibiotics administered (including duration),
31
whether or not surgery was later required and if so for what reason, timing of surgical intervention if
required, length of hospital stay, length of follow-up, and neurologic outcome at last follow-up.
Neurologic impairment was graded using the modification by Kim et al. of Heusner’s original description:
phase 1—subjective complaints of pain (axial and/or radicular), phase 2—radicular weakness, phase 3—
incomplete spinal cord injury (including myelopathy, conus medullaris and cauda equina syndromes),
phase 4—complete spinal cord injury(9, 23).
A patient was considered to have failed medical management if he/she required surgery despite
medical management. This definition of failure is consistent with that used in multiple studies in the SEA
literature, including Patel’s study(1, 2, 4, 9-11).
Statistical analysis
We used descriptive statistics to summarize the incidence of SEA at our centre, and how often patients
with SEA were intentionally treated medically. These descriptive statistics allowed for comparison with
the two papers with prediction criteria. The proportion of patients in the medically treated group who
required surgery (failure rate) was determined. The published prediction models were applied to our
data to determine how predictive they were of failure of medical management in our cohort.
Results
Five hundred and fifty patients were identified, using ICD-10 diagnostic codes, as potentially having a
diagnosis of SEA. Three hundred and twenty-two were excluded upon review of electronic
records/imaging for not having a SEA. Upon full paper chart review of the remaining 222 patients, a
further 62 did not have MRI or CT confirmed SEA. Of the remaining 160 patients, 56 were excluded
based on intended surgical treatment, 13 for post-intervention infection, eight for default medical
treatment (patients were deemed not surgical candidates), four for inadequate follow up, two for TB
32
infections, and two for complete spinal cord injuries for greater than 48 hours (Figure 1). Seventy-five
intentionally medically treated patients were included in the analysis.
The incidence of SEA was: 15.7 patients per year (160 patients over 10.16 years). Of the patients
in whom follow up information was complete and TB, SCI >48hrs and post-intervention were ruled out,
56 were intentionally managed surgically, eight were treated medically due to contraindications to
surgery, and 75 were intentionally managed medically (54%).
Of the two papers identified with prediction criteria, the number of patients initially treated
medically were 140 (Kim et al.) and 51 (Patel et al.). In the paper by Kim et al., patient characteristics
were presented for the definitively medically managed group (patients who failed initial medical
management and ultimately required surgery were reported in the surgical cohort), and we compared
this group to the patients in our cohort (Table 3). We could not, however, do the same for the patients
in Patel’s cohort as the patient characteristics were described together for both treatment groups.
Compared to Kim’s cohort of patients who remained in the medically managed group, our patients were
slightly younger (57 versus 62 years old) and there was a higher percentage of males in our study (73%
versus 57%). With respect to neurologic status, in our cohort 71% had no deficit and the remaining 29%
had radicular weakness (Table 3). This is in contrast to Kim’s medically treated cohort in which 73% had
no deficit, 17% had radicular weakness and 10% had an incomplete spinal cord injury. The percentages
of patients with diabetes were similar, however our cohort had far fewer patients with malignancy (4%
versus 24%). Social habits, spinal level of SEA, and microbiology were similar.
Thirteen patients in our cohort failed medical treatment, as defined by requiring surgery on a
delayed basis. This corresponds to a failure rate of 17.1%. Of those who required surgery, the mean
delay between SEA diagnosis and surgery was 20 days (Table 4). Eight of the 13 patients required
33
surgery for a deterioration in neurologic status. The various other reasons for converting to surgical
management are listed in Table 4.
The odds of failure of medical management for each of the risk factors identified by Kim and
Patel were calculated in our cohort, with the exception of incomplete SCI as we did not have any
patients with this risk factor. Our odds ratios (OR) for the published risk factors were all in the same
direction as those published, however for some variables differed in magnitude. The OR for age greater
than 65 years old was 1.8, diabetes was 5.8, MRSA was 2.1, positive blood culture was 2.3, white blood
cell count greater than 12.5 was 4.1 and CRP over 115 was 2.3 (Table 5).
Calibration
We used logistic regression to calibrate the published models, using the predicted probabilities of failure
based on each model and, as the outcome measure, actual failure of medical treatment. To assess the
calibration, observed log odds of failure were plotted against the predicted log odds of failure for each
of the two models. A slope of 1 and intercept of 0 represent a perfect prediction. The intercept for Kim’s
calibrated model was -0.45 and the slope 1.15; for Patel’s calibrated model, the intercept was -1.65 and
the slope 0.70 (Table 6).
The calibrated models were used to calculate the probability of failure of medical management
for each combination of the published risk factors. Tables 7 and 8 compare the predicted probability of
failure before and after calibration of the two published models. For Kim’s model we could not include
neurologic deficit as a risk factor because we did not have any patients in our cohort positive for this risk
factor. We calculated the observed failure rate in our cohort for each combination of risk factors in Kim’s
model. We did not perform this step for Patel’s model owing to the low number of failures and number
of categories (Table 8). When age was added as a variable to Kim’s model, there was no evidence of
improved fit.
34
Discrimination
To assess the ability of each model to predict failure of medical treatment in our cohorts, we calculated
the probability of failure for each individual patient based on each model of failure and compared to
actual outcome. Both models over-estimate the observed failure rate for our patients. We calculated
the mean probability of failure based on each of the published models. Kim’s model predicted 26.3% of
our patients would fail medical treatment. Patel’s model estimated 44.6% of patients would be expected
to fail medical treatment. The actual proportion of failure was 17.1% in our cohort. The distribution of
patients’ expected probabilities of failure versus actual outcomes for both Kim’s and Patel’s models are
depicted in Figure 2. The area under the ROC curve was 0.687 for Kim’s model and 0.721 for Patel’s
(Figure 3).
Discussion
We present data on 75 medically managed patients with SEA from a single centre over a period of ten
years and two months. This study represents the second largest series of intentionally medically
managed patients in the SEA literature to date. Our intention was to test the published prediction
criteria for failure of medical management on a cohort other than that from which the criteria were
originally developed. Failure as we define it is: a patient with SEA managed medically initially requiring
surgery for any reason.
The rate of medical treatment of primary nontuberculous SEA over the study period at our
centre was 54%, which is consistent with the recent systematic review suggesting a trend toward
medical treatment for neurologically intact patients with SEA(16). The majority of our patients (71%)
were neurologically intact on presentation, and the remainder had radicular weakness. No patients had
acute incomplete or complete spinal cord injuries, which is in contrast to the cohort presented by Kim et
al. in which 10% of patients had an incomplete spinal cord injury. This difference may account for some
35
of the discrepancy between the predictions using Kim’s model and actual outcome in our patients. Kim’s
model predicted 26.3% of our patients would fail medical treatment which was an overestimation of
9.2%; Patel’s model estimated 44.6% of patients would be expected to fail medical treatment, an
overestimation of 27.5%. We cannot comment on differences in baseline characteristics between our
cohort and that of Patel because the patients in the medically treated group were not reported
separately from those in the surgical group in Patel’s paper. It is not surprising that both models
overestimated failure of medical management because the proportion of patients who failed in our
cohort (17.1%) was lower than published estimates.
Patel’s model originally had poor calibration, and therefore validity, in our cohort; however once
calibrated, it is the better model for our data. Although Patel’s model overestimated failure to a greater
extent than did Kim’s model in our cohort, Patel’s model discriminated better than did Kim’s. This is
indicated by the boxplot in Figure 2 and the area under the ROC curve (0.72 versus 0.69) (Figure 3). It is
possible that Patel’s model performed better in our cohort with respect to discrimination because we
were able to apply all four risk factors, rather than only three of four. Furthermore, our definition of
failure was the same as Patel, and did not include deaths. Kim, on the other hand, included deaths as
failures of medical management. Our concern in including deaths is that it is difficult to definitely
attribute death in a patient with SEA to the epidural abscess as many such patients have multiple
comorbidities and acute pathologies at the time of SEA diagnosis.
Limitations of our study include the apparent low volume of patients with SEA, as identified by
diagnostic codes. We made every attempt to be inclusive in our database search, using broader codes
such as those for discitis and osteomyelitis in all regions of the spine, however if a patient with SEA was
not, upon discharge, coded as having an infected epidural space, we were unable to include such a
patient. Of the patients with SEA, our subset of interest were those initially treated medically which was
an even smaller number of patients (75) and the number of failure events was particularly small at 13
36
(17%). This small number of events compared to that in the two studies from which the prediction
criteria were derived (38% Kim, 41% Patel) may indicate either a difference in baseline patient
characteristics or indications for surgery.
Conclusions
The observed failure rate for SEA treated medically at our institution was 17%. Rates in the recent
literature range from 30-41%. Analysis of the two published prediction models in our cohort of 75
medically managed patients with SEA demonstrated neither were valid in our population. However,
Patel’s model, consisting of: positive blood culture, diabetes, WBC > 12.5 and CRP > 115, showed both
better discrimination and fit, and once calibrated to our data was the better model. A future direction of
this research would be to partner with another tertiary care centre with a high volume of spine patients
and bolster the number of patients in the cohort. Increased numbers may allow for the development of
a new and more clinically useful prediction model.
Figures and Tables
Assessed for eligibility (n=550)
SEA diagnosis not confirmed
with CT/MRI (n=390)
SEA confirmed (n=160)
Included in analysis (n=75)
Figure 1. Patient flow diagram.
37
Excluded (n=85)
 Intended treatment surgical (n=56)
 Post-intervention (n=13)
 Default medical treatment (n=8)
 Inadequate follow-up (n=4)
 TB infection (n=2)
 Complete SCI >48 hours (n=2)
0
.2
.4
.6
.8
Probability of failure based on Patel's model
1
1
.9
.8
.7
.6
.5
.4
.3
.2
.1
0
0
0
1
1
0.25
0.50
Sensitivity
0.50
0.00
0.25
0.00
Sensitivity
0.75
0.75
1.00
1.00
Figure 2. Probability of failure based on Kim’s and Patel’s models versus actual failure of medical
management. 0=no failure, 1=failure.
0.00
0.25
0.50
1 - Specificity
0.75
0.00
1.00
0.25
0.50
1 - Specificity
0.75
1.00
Area under ROC curve = 0.7208
Area under ROC curve = 0.6873
Figure 3. ROC curves for Kim and Patel models.
Kim et al. 2014
Patel et al. 2014
Age >65
Positive blood culture
Diabetes
Diabetes mellitus
MRSA
Leukocytosis > 12.5
Neurologic impairment (incomplete or complete cord injury)
CRP > 115
Table 1. Proposed predictors of failure of medical management. MRSA=Methicillin-resistant S. aureus.
38
ICD-10 code
G061
M4620
M4622
M4625
M4628
M4629
M4630
M4632
M4633
M4634
M4635
M4636
M4637
M4639
M4640
M4642
M4643
M4644
M4645
M4646
M4647
M4649
M4650
M4651
M4652
M4653
M4654
M4655
M4656
M4657
M4658
M4659
Corresponding diagnosis
Intraspinal Abscess and Granuloma
Osteomyelitis of Vertebra, Multiple Sites in Spine
Osteomyelitis of Vertebra, Cervical Region
Osteomyelitis of Vertebra, Thoracolumbar Region
Osteomyelitis of Vertebra, Sacral and Sacrococcygeal region
Osteomyelitis of Vertebra, Unspecified Site
Infection Of Intervertebral Disc (Pyogenic), Multiple Sites in Spine
Infection Of Intervertebral Disc (Pyogenic), Cervical Region
Infection of Intervertebral Disc (Pyogenic), Cervicothoracic Region
Infection of Intervertebral Disc (Pyogenic),Thoracic Region
Infection of Intervertebral Disc (Pyogenic), Thoracolumbar Region
Infection of Intervertebral Disc (Pyogenic), Lumbar Region
Infection of Intervertebral Disc (Pyogenic), Lumbosacral Region
Infection of Intervertebral Disc (Pyogenic), Unspecified Site
Discitis, Unspecified, Multiple Sites in Spine
Discitis, Unspecified, Cervical Region
Discitis, Unspecified, Cervicothoracic Region
Discitis, Unspecified, Thoracic Region
Discitis, Unspecified, Thoracolumbar Region
Discitis, Unspecified, Lumbar Region
Discitis, Unspecified, Lumbosacral Region
Discitis, Unspecified, Unspecified Site
Other Infective Spondylopathies, Multiple Sites in Spine
Other Infective Spondylopathies, Occipto-atlanto-axial Region
Other Infective Spondylopathies, Cervical Region
Other Infective Spondylopathies, Cervicothoracic Region
Other Infective Spondylopathies, Thoracic Region
Other Infective Spondylopathies, Thoracolumbar Region
Other Infective Spondylopathies, Lumbar Region
Other Infective Spondylopathies, Lumbosacral Region
Other Infective Spondylopathies, Sacral and Sacrococcygeal Region
Other Infective Spondylopathies, Unspecified Site
Table 2. ICD-10 diagnostic codes used for identification of patients with spinal epidural abscess.
39
Stratton (n=75)
57.2 (+/- 14.0)
55 (73.3)
Kim (n=100)
62.4 (+/- 16.1)
57 (57)
Age (yr) with std deviation
Male gender, n (%)
Neurologic status n (%)
1-no deficit
53 (70.7)
73 (73)
2-radicular weakness
22 (29.3)
17 (17)
3-incomplete SCI
0
10 (10)
4-complete SCI
0
0
Medical hx
Diabetes
11 (14.7)
18 (18)
Malignancy
3 (4)
24 (24)
Social habits, n (%)
Smoker
34 (45)
Not reported
IVDU
10 (13)
19 (20)
Alcohol abuse
11 (15)
16 (17)
Level of SEA, n (%)
Above conus
30 (40)
48 (48)
>5 levels
8 (11)
11 (11)
Skip lesions
4 (5)
3 (3)
Microbiology, n (%)
No growth
18 (24)
15 (15)
Staphyloccocus
40 (53)
61 (61)
MRSA
7 (9)
11 (11)
MSSA
33 (44)
42 (42)
Streptococcus
8 (11)
13 (13)
Gram negative rods
1 (1)
4 (4)
>2 organisms
0
2 (2)
Other
8 (11)
7 (7)
Table 3. Characteristics of patients intentionally treated medically.
40
Time from SEA diagnosis
to OR (days)
1
1
3
3
5
6
8
9
11
22
23
70
98
Rationale for Surgery
Progression of systemic illness, sepsis
Neurologic deterioration
Neurologic deterioration
Neurologic deterioration
Persistent abscess collection with spinal cord edema
Neurologic deterioration
Persistent bacteremia despite maximal medical treatment
Neurologic deterioration
Need for tissue diagnosis, increasing pain
Neurologic deterioration
Neurologic deterioration
Neurologic deterioration
Recurrence of symptoms and abscess after completing 3 months of
antibiotics
Mean: 20
Table 4. Medical treatment failures.
Kim et al. 2014
OR
Stratton Patel et al. 2014
OR Stratton
Age >65
2.4
1.8
Positive blood culture 3.5 2.3
Diabetes
2.5
5.8*
Diabetes mellitus
3.8 5.8*
MRSA
3.8
2.1
Leukocytosis > 12.5
3.3 4.1*
Neurologic impairment
15.5 ---CRP > 115
4.7 2.3
Table 5. Odds ratios associated with proposed predictors of failure of medical management.
Value
Kim
Patel
Area under ROC curve
0.687
0.721
Calibration Intercept
0.45
-1.65
Calibration Slope
1.15
0.70
Mean of Predictions*
26.3%*
44.6%*
Observed Rate of Failure
17.1%
17.1%
Table 6. Assessment of calibration and discrimination of the two models. *=prior to calibration.
41
Probability of Probability of failure, Total Failed Observed
Age >65y Diabetes MRSA failure (%)
calibrated model (%) (n)
(n)
failure rate (%)
Yes
Yes
Yes
83
80
0
0
n/a
No
Yes
Yes
67
59
1
1
100
Yes
No
Yes
66
58
1
0
0
Yes
Yes
No
56
46
3
2
67
Yes
No
No
33
22
17
3
18
No
Yes
No
34
23
7
2
29
No
No
Yes
44
33
5
1
20
No
No
No
17
9
41
4
10
Table 7. Probability of failure of medical management based on Kim’s original and calibrated model, and
the observed failure rate in our cohort.
Blood culture
Probability of Probability of failure,
positive
Diabetes WBC>12.5 CRP >115 failure (%)
calibrated model (%)
Yes
Yes
Yes
Yes
95
60
Yes
Yes
Yes
No
80
33
Yes
Yes
No
Yes
85
39
Yes
No
Yes
Yes
83
37
No
Yes
Yes
Yes
84
38
No
No
Yes
Yes
58
20
Yes
No
No
Yes
60
20
Yes
Yes
No
No
55
18
No
Yes
Yes
No
53
17
No
Yes
No
Yes
62
21
Yes
No
Yes
No
51
16
Yes
No
No
No
24
8
No
Yes
No
No
26
8
No
No
Yes
No
23
8
No
No
No
Yes
30
10
No
No
No
No
8
3
Table 8. Probability of failure of medical management based on Patel’s original and calibrated model.
42
References
1.
Khanna RK, Malik GM, Rock JP, Rosenblum ML. Spinal epidural abscess: evaluation of factors
influencing outcome. Neurosurgery. 1996;39(5):958-64.
2.
Curry Jr WT, Hoh BL, Amin-Hanjani S, Eskandar EN. Spinal epidural abscess: Clinical presentation,
management, and outcome. Surgical neurology. 2005;63(4):364-71.
3.
Connor DE, Jr., Chittiboina P, Caldito G, Nanda A. Comparison of operative and nonoperative
management of spinal epidural abscess: a retrospective review of clinical and laboratory predictors of
neurological outcome. Journal of Neurosurgery Spine. 2013;19(1):119-27.
4.
Patel AR, Alton TB, Bransford RJ, Lee MJ, Bellabarba CB, Chapman JR. Spinal epidural abscesses:
Risk factors, medical versus surgical management, a retrospective review of 128 cases. Spine Journal.
2014;14(2):326-30.
5.
Karikari IO, Powers CJ, Reynolds RM, Mehta AI, Isaacs RE. Management of a spontaneous spinal
epidural abscess: a single-center 10-year experience. Neurosurgery. 2009;65(5):919-23; discussion 23-4.
6.
Tang HJ, Lin HJ, Liu YC, Li CM. Spinal epidural abscess--experience with 46 patients and
evaluation of prognostic factors. Journal of Infection. 2002;45(2):76-81.
7.
Chen SH, Chang WN, Lu CH, Chuang YC, Lui CC, Chen SF, et al. The clinical characteristics,
therapeutic outcome, and prognostic factors of non-tuberculous bacterial spinal epidural abscess in
adults: a hospital-based study. Acta neurol. 2011;20(2):107-13.
8.
Chen WC, Wang JL, Wang JT, Chen YC, Chang SC. Spinal epidural abscess due to Staphylococcus
aureus: clinical manifestations and outcomes. Journal of Microbiology, Immunology & Infection.
2008;41(3):215-21.
9.
Kim SD, Melikian R, Ju KL, Zurakowski D, Wood KB, Bono CM, et al. Independent predictors of
failure of nonoperative management of spinal epidural abscesses. Spine Journal: Official Journal of the
North American Spine Society. 2014;14(8):1673-9.
10.
Savage K, Holtom PD, Zalavras CG. Spinal epidural abscess: early clinical outcome in patients
treated medically. Clin Orthop. 2005;439:56-60.
11.
Siddiq F, Chowfin A, Tight R, Sahmoun AE, Smego Jr RA. Medical vs surgical management of
spinal epidural abscess. Archives of internal medicine. 2004;164(22):2409-12.
12.
Reihsaus E, Waldbaur H, Seeling W. Spinal epidural abscess: a meta-analysis of 915 patients.
Neurosurg Rev. 2000;23(4):175-204; discussion 5.
13.
Darouiche RO. Current concepts: spinal epidural abscess. New England Journal of Medicine.
2006;355(19):2012.
14.
Nussbaum ES, Rigamonti D, Standiford H, Numaguchi Y, Wolf AL, Robinson WL. Spinal epidural
abscess: a report of 40 cases and review. Surgical neurology. 1992;38(3):225-31.
15.
Rigamonti D, Liem L, Sampath P, Knoller N, Namaguchi Y, Schreibman DL, et al. Spinal epidural
abscess: contemporary trends in etiology, evaluation, and management. Surgical neurology.
1999;52(2):189-96; discussion 97.
16.
Arko Lt, Quach E, Nguyen V, Chang D, Sukul V, Kim BS. Medical and surgical management of
spinal epidural abscess: a systematic review. Neurosurgical focus. 2014;37(2):E4.
17.
Tuchman A, Pham M, Hsieh PC. The indications and timing for operative management of spinal
epidural abscess: literature review and treatment algorithm. Neurosurgical focus. 2014;37(2):E8.
18.
Hlavin ML, Kaminski HJ, Ross JS, Ganz E. Spinal epidural abscess: a ten-year perspective.
Neurosurgery. 1990;27(2):177-84.
19.
Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, et al. Meta-analysis of
observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies
in Epidemiology (MOOSE) group. Jama. 2000;283(15):2008-12.
43
20.
Altman DG. Systematic reviews of evaluations of prognostic variables. BMJ (Clinical research ed).
2001;323(7306):224-8.
21.
Posner KL, Sampson PD, Caplan RA, Ward RJ, Cheney FW. Measuring interrater reliability among
multiple raters: an example of methods for nominal data. Statistics in medicine. 1990;9(9):1103-15.
22.
Adogwa O, Karikari IO, Carr KR, Krucoff M, Ajay D, Fatemi P, et al. Spontaneous spinal epidural
abscess in patients 50 years of age and older: a 15-year institutional perspective and review of the
literature: clinical article. Journal of Neurosurgery Spine. 2014;20(3):344-9.
23.
Heusner AP. Nontuberculous spinal epidural infections. The New England journal of medicine.
1948;239(23):845-54.
24.
Davis DP, Wold RM, Patel RJ, Tran AJ, Tokhi RN, Chan TC, et al. The clinical presentation and
impact of diagnostic delays on emergency department patients with spinal epidural abscess. The Journal
of emergency medicine. 2004;26(3):285-91.
44
Chapter 4: Conclusions
The aim of this thesis was to clarify appropriate treatment decision making in patients with newly
diagnosed with SEA with an intact neurologic exam. The results of the systematic review and metaanalysis were less conclusive than expected owing to high variability across studies with respect to
rationale for chosen treatment, definition of treatment failure and outcome reporting for SEA. Although
we found that the incidence of failure of medical management of SEA was relatively common in
published reports, estimates were highly heterogeneous between studies, thereby introducing
uncertainty about the frequency of this risk. Medical treatment failure estimates ranged from 26-35%.
Two studies were identified in the literature with prediction criteria for failure of medical
management, which was the second objective of the systematic review. We tested these prediction
criteria in a cohort of patients from a tertiary care centre and found the validity to be poor as both
overestimated medical treatment failure in our cohort. Once the prediction models were calibrated to
our data it was Patel’s model that was superior and thus more useful for clinical decision making.
45
Appendix
I give permission for Alexandra Stratton to reproduce in her thesis submission the following manuscript,
which I co-authored: Incidence and risk factors for failure of medical management of spinal epidural
abscess: A systematic review and meta-analysis.
Karla Gustafson
I give permission for Alexandra Stratton to reproduce in her thesis submission the following manuscript,
which I co-authored: Incidence and risk factors for failure of medical management of spinal epidural
abscess: A systematic review and meta-analysis.
Yours truly,
Matthew James, MD PhD FRCPC
Assistant Professor
Department of Medicine
Department of Community Health Sciences
Cumming School of Medicine
University of Calgary
I give permission for Alexandra Stratton to reproduce in her thesis submission the following manuscript,
which I co-authored: The Prognostic Accuracy of Suggested Predictors of Failure of Medical
Management for Patients with Spinal Epidural Abscess.
and
I give permission for Alexandra Stratton to reproduce in her thesis submission the following manuscript,
which I co-authored: Incidence and risk factors for failure of medical management of spinal epidural
abscess: A systematic review and meta-analysis.
46
Sincerely,
Ken Thomas MD,FRCSC,MHSc
Orthopaedic Surgeon
Director Combined Spine Program
Clinical Associate Professor
University of Calgary
I give permission for Alexandra Stratton to reproduce in her thesis submission the following manuscript,
which I co-authored: The Prognostic Accuracy of Suggested Predictors of Failure of Medical
Management for Patients with Spinal Epidural Abscess.
Peter Faris
47