Professional Documents
Culture Documents
Abstract
Objective. The study aims to create a diagnostic support tool to indicate the likelihood of the presence of lumbar
spinal stenosis (LSS) using a cluster of elements from the patient history and observational findings. Design. The
study is case based and case controlled. Setting. The study was performed in the tertiary care of a medical
center. Subjects. There were a total of 1,448 patients who presented with a primary complaint of back pain with
or without leg pain. Methods. All patients underwent a standardized clinical examination. The diagnosis of LSS
was made by one of two experienced orthopaedic surgeons based on clinical findings and imaging. Data from the
patient history and observational findings were then statistically analysed using bivariate analysis and contingency
tables. Results. The most diagnostic combination included a cluster of: 1) bilateral symptoms; 2) leg pain more
than back pain; 3) pain during walking/standing; 4) pain relief upon sitting; and 5) age >48 years. Failure to meet
the condition of any one of five positive examination findings demonstrated a high sensitivity of 0.96 (95%
CI = 0.940.97) and a low negative likelihood ratio (LR) of 0.19 (95% CI = 0.120.29). Meeting the condition
of four of five examination findings yielded a LR+ of 4.6 (95% CI = 2.48.9) and a post-test probability of
76%. Conclusion. The high sensitivity of the diagnostic support tool provides the potential to reduce the incidence
of unnecessary imaging when the diagnosis of LSS is statistically unlikely. In patients where the condition of four
of the five findings was present, the post-test probability of 76% suggests that imaging and further workup are
indicated. This is an inexpensive but powerful tool, with a potential to increase diagnostic efficiency and reduce
cost by narrowing the indications for imaging. Copyright 2010 John Wiley & Sons, Ltd.
Keywords
diagnostic support tool; lumbar spinal stenosis; patient history and observational findings
*Correspondence
Eric Hegedus DPT, MHSc, OCS, 2200 West Main Street-Wing B, Box 104002 DUMC, Durham, NC 27708, USA.
Email: eric.hegedus@duke.edu
Published online 12 November 2010 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/pri.500
170 Physiother. Res. Int. 16 (2011) 170178 2010 John Wiley & Sons, Ltd.
C. Cook et al. Patient History and Findings for LSS
Physiother. Res. Int. 16 (2011) 170178 2010 John Wiley & Sons, Ltd. 171
Patient History and Findings for LSS C. Cook et al.
and conclusions of findings. The study protocol was detecting change in sample populations (White and
approved by the Institutional Review Board of the Duke Velozo, 2002). The ODI was tabulated to identify the
University Health System. severity of disability for this sample.
The observational examination included assessment
Study participants of gait, which was characterized as normal or abnormal
(ataxic, wide based or poor coordination). Physical
The study included 1,448 consecutive patients seen at a
examination tests included were a battery of motor,
spine surgery center at Duke University from 2005 to
sensory and reflex tests; use of the straight leg raise; and
2009. All patients with suspicion of a condition associ-
range of motion assessment of the spine and upper
ated with origin at the lumbar spine were eligible for
motor neuron assessment. Physical examination find-
the study if a clinical diagnosis was made and if imaging
ings were not included in the clustered patient history
was performed or available for review.
and observational assessment. All clinical examination
findings were inputed into a structured Excel compat-
Standardized clinical examination
ible database that allows real-time Internet-based
All patients received a standardized clinical exam- interface.
ination that consisted of self-report, observational,
physical examination, and imaging assessment. A stan-
Diagnosis
dardized clinical examination was implemented in 2005
to improve outcome reporting and data capture at the The two board-certified orthopaedic surgeons were
institution. Both physicians involved in the study were responsible for the diagnosis of each subject. At present,
board-certified orthopaedic surgeons, one with 22 years there is no consensus regarding the criterion standard
of orthopaedic surgical experience and the other with for the diagnosis of LSS (Kalcihman et al., 2009). This
4 years of orthopaedic surgical experience. lack of consensus is confirmed by past studies and the
The standard patient history included standard guidelines set forth by the North American Spine
demographics capture and questions on the location, Society (2007) and others (Wiesel et al., 1984; Jensen
constancy, characteristics, and severity of pain in and Schmidt-Olsen, 1989; Kent et al., 1992; Mann et al.,
general and by location. In addition, questions regard- 1993; Katz et al., 1995; Fritz et al., 1997; deGraaf et al.,
ing provocative movements (e.g. walking, sitting or 2006). In our study, a diagnosis of spinal stenosis
standing), alleviating movements (e.g. walking, sitting, required a convincing history and description of symp-
rest or standing), exercise frequency and participation, toms, including pertinent clinical findings such as lower
and previous treatment were captured. Standardized extremity weakness, gait impairment, or reflex changes
outcome measures such as the Short Form 12 (SF12) and patient complaints of leg pain generally greater
(split by the Physical Component Summary [PCS] and than back pain, symptoms that were worse with stand-
Mental Component Summary [MCS]) and the Oswes- ing/walking, which were generally relieved with sitting
try Disability Index (ODI) were also captured. (often more relieved still with supine) (deGraaf et al.,
The SF12 is a generic measure and does not target a 2006). Clinical differential diagnosis included vascular
specific age or disease group. The SF12 is a shortened claudication versus neurogenic claudication.
version of the Short Form 36 and is weighted and The imaging method most commonly used by our
summed to provide an interpretable measure of quality surgeons was the T2 MRI (aa combination of axial and
of life (Ware and Kosinski, 2001). The ODI is a multi- sagittal images) (Kent et al., 1992). In addition, radio-
dimensional scale and has been used to document graphic evidence of degenerative listhesis on standing
changes in muscle activity, pain, psychological factors lateral lumbar film is often positively correlated with
and work status (Taylor et al., 1999). The ODI has been foraminal stenosis on MRI, and was used as further
used to evaluate pre- and post-surgical outcomes, as confirmation in some cases. An MRI was obtained on
well as a benchmark for the determination of treatment all patients (a radiograph was not) and was always
effectiveness. Four versions of the ODI are available in reviewed by each surgeon, and in most cases there were
English and in nine other languages. The data for the also reviews included by radiology.
ODI provide both validation and standards for other In cases where imaging findings did not support the
users and indicate the power of the instrument for clinical representation, yet the clinical findings sup-
172 Physiother. Res. Int. 16 (2011) 170178 2010 John Wiley & Sons, Ltd.
C. Cook et al. Patient History and Findings for LSS
ported the clinical diagnosis of presumed LSS, these during walking as a sensitive and specific measure, this
patients were diagnosed with LSS and were typically variable was included as well in the backward stepwise
referred for conservative care methods (physical therapy, regression. A backward stepwise logistic regression was
medications or injections). In our study, imaging served used to select variables, with p values of 0.15 to exit the
as confirmation only, consequently, when clinical find- model and 0.10 to enter it. Variables retained by the
ings did not support the diagnosis of LSS yet the MRI regression model were used to cluster findings and were
or radiograph did; a competing diagnosis (e.g. spondy- then inputed into 2 2 contingency tables that involved
lolisthesis only) was provided. In most cases, a clinical the conditions of 1 of 5, 2 of 5, 3 of 5, 4 of 5 and 5 of
diagnosis of LSS was not exclusive. A concomitant 5 positive findings. For each condition, sensitivity, spec-
primary or secondary diagnosis of acquired spondylo- ificity, and likelihood ratios were analysed. In addition,
listhesis, congenital spondylolisthesis and degenerative in each condition, post-test probability measures were
scoliosis was made in over 65% of the cases. All forms calculated using a pre-test probability of 40.3% (the
of LSS were merged for this study, including the location prevalence of LSS in this sample).
of the stenosis (central, lateral, recess or foraminal ste- Most diagnostic accuracy studies are powered on
nosis), which was treated as one group. tabulated probabilities of values being lower than
acceptable confidence interval (CI) estimates (lower
levels) for sensitivity values (Flahault et al., 2005),
Variables
whereas clusters involve combining variables which
The targeted variables for the study included predictive often drive up specificity values at the expense of sen-
demographic, patient history and observational vari- sitivity values. Consequently, the study was powered
ables. Descriptive variables included the self-reported using the regression values and the 10 predictor vari-
findings of age; bilateral nature of symptoms; leg pain ables. Using Monte Carlo simulations, Peduzzi et al.
worse than back pain; pain with walking or standing; (1995) reported that an n of 1020 per predictor is
alleviation of pain during sitting; gait abnormality; appropriate for a boundary level per variable for regres-
report of severity of back, buttock, and leg pain and sion analyses. This finding suggests that 100150
pain constancy. patients would provide adequate values for the regres-
sion analysis, whereas larger numbers may be necessary
to further reduce CIs for the diagnostic accuracy
Data analysis
statistics.
Data downloaded from the Internet-based Excel system
were transferred to SPSS 13.0 for Windows (SPSS Inc.,
Chicago, IL; 606066307). Descriptive statistics were
Results
captured and sequestered into groups of patients with The study captured 584 (40.3%) patients with a clinical
and without a diagnosis of LSS. Bivariate analyses diagnosis of LSS and 864 with a competing diagnosis.
between patients with and without a clinical diagnosis Significant differences in age (p < 0.01), marital status
of LSS were performed. A p value of 0.05 was consid- (p < 0.01), Body mass index, employment status and
ered significant. SF12 MCS were found (Table 1).
All 10 predictor variables were individually exam- Age, self-report of bilateral symptoms, self-report of
ined for diagnostic accuracy. Contingency tables (2 2) moderate or greater leg pain, report of leg pain greater
were used to calculate sensitivity and specificity, and than back pain, and pain that is alleviated by sitting had
likelihood ratios (positive likelihood ratio [LR+]; nega- LR+ above 1.5 or LR 0.50 (Table 2). Of the single
tive likelihood ratio [LR]) for each predictive test item. patient history of observational findings, bilateral
Receiver operator characteristic curves were used to symptoms demonstrated the highest LR+ (2.3; 95%
determine all possible cut-off values for age and report CI = 1.14.8), whereas age >48 demonstrated the lowest
of back, buttock and leg pain severity. LR (0.25; 95% CI = 0.210.32).
The conditionally independent variables from the The most diagnostic combination included a cluster
individual 2 2 analyses that yielded LR+ above 1.5 or of: 1) bilateral symptoms; 2) leg pain more than back
LR below 0.5 were retained for the regression analysis. pain; 3) pain during walking/standing; 4) pain relief
In addition, because past studies have identified pain upon sitting; and 5) age >48 years (Table 3). Of the
Physiother. Res. Int. 16 (2011) 170178 2010 John Wiley & Sons, Ltd. 173
Patient History and Findings for LSS C. Cook et al.
BMI = body mass index; MCS = Mental Component Summary; PCS = Physical Component Summary; SF12 = Short Form 12.
p Values of <0.01 in bold were considered significant.
Test Item Sensitivity Specificity Positive likelihood ratio Negative likelihood ratio
(95% CI) (95% CI) (95% CI) (95% CI)
Age > 48 0.88 (0.850.89) 0.49 (0.470.50) 1.7 (1.61.8) 0.25 (0.210.32)
Bilateral symptoms 0.03 (0.020.04) 0.98 (0.980.99) 2.3 (1.14.8) 0.98 (0.970.99)
Leg or back pain worse 0.16 (0.140.18) 0.92 (0.910.93) 2.1 (1.52.8) 0.91 (0.870.94)
Pain with walking/standing 0.67 (0.640.69) 0.44 (0.420.46) 1.2 (1.11.3) 0.75 (0.660.86)
Sitting relieves pain 0.26 (0.240.29) 0.86 (0.840.88) 1.9 (1.52.3) 0.86 (0.820.91)
Gait abnormality 0.29 (0.270.32) 0.81 (0.790.83) 1.6 (1.21.9) 0.87 (0.820.92)
Moderate back pain 0.95 (0.940.96) 0.02 (0.010.03) 0.97 (0.960.99) 2.1 (1.23.9)
Moderate buttock pain 0.81 (0.770.84) 0.33 (0.310.35) 1.2 (1.11.3) 0.60 (0.460.72)
Moderate leg pain 0.90 (0.880.92) 0.24 (0.220.25) 1.2 (1.11.2) 0.43 (0.320.58)
Pain constancy 0.23 (0.200.26) 0.78 (0.760.80) 1.1 (.861.3) 0.98 (0.921.04)
CI = confidence interval.
Useful likelihood ratios appear in bold.
174 Physiother. Res. Int. 16 (2011) 170178 2010 John Wiley & Sons, Ltd.
C. Cook et al. Patient History and Findings for LSS
1 of 5 positive findings 0.96 (0.940.97) 0.20 (0.190.21) 1.2 (1.11.2) 0.19 (0.120.29) 44 11.4
2 of 5 positive findings 0.68 (0.650.71) 0.62 (0.600.64) 1.8 (1.62.0) 0.51 (0.450.58) 55 25.6
3 of 5 positive findings 0.29 (0.270.31) 0.88 (0.870.90) 2.5 (2.03.1) 0.80 (0.760.85) 63 35.1
4 of 5 positive findings 0.06 (0.050.07) 0.98 (0.980.99) 4.6 (2.48.9) 0.95 (0.940.97) 76 39.1
5 of 5 positive findings <0.01 (0.0010.003) 1.0 (0.991.0) infinite (0.77infinite) 0.99 (0.991.0) 99+ 40.1
Note: Five findings are included in the rule: 1) bilateral symptoms; 2) leg pain more than back pain; 3) pain during walking/standing; 4) pain
relief upon sitting; and 5) age >48 years. The associated post-test probability values are based on a pre-test probability of 40.3%.
combinations, failure to meet the condition of one of presence of a gait impairment in our study (Fritz et al.,
five positive tests demonstrated high sensitivity (0.96; 1998).
95% CI = 0.940.97) and a low LR (0.19; 95% CI = Others have combined findings to develop clinical
0.120.29), providing a post-test probability of 11.4% prediction rules. Thigh pain within 30 seconds of
if this condition was negative. A combination of four of lumbar extension, age, no pain when seated and wide-
five tests yielded a LR+ of 4.6 (95% CI = 2.48.9) and based gait (Katz et al., 1995) has been previously
a post-test probability of 76%. Three of five yielded a reported as useful when combined during examination.
higher sensitivity, capturing a larger population of Konno and colleagues (2007b) advocate a scoring
patients with LSS, but had a lower LR+ 2.5 (95% CI = system that involves age, self-report symptomatic crite-
2.03.1) and a post-test probability of 63%. ria and physical examination findings, and absence of
diabetes. Scores of 7 were associated with a positive
likelihood ratio of 3.31 for the diagnosis of LSS. The
Discussion tool was recently validated in a prospective study
Our diagnostic support tool improved post-test prob- involving 118 patients (Kato et al., 2009). Another study
abilities to 6376%; thus this cluster of findings may be (Konno et al., 2008) examined a self-administered
useful to alter pre-test probability prior to physical patient history questionnaire and found small improve-
examination testing. When the condition of one of five ments in post-test probability (LR+ = 1.9). Included
variables was not met, there was a low probability that within the questionnaire were queries regarding leg
the patient actually had LSS (11.4%). This finding is pain, pain that worsens when walking, numbness in
useful when one considers ordering imaging to rule out both legs, worse pain during standing and improve-
LSS as the sensitivity and LR of our findings are com- ment of symptoms during bending over.
parable with those reported using a CT and an MRI In concert with past findings, our study found low
(Lurie, 2005). A finding of four of five positive examina- sensitivity and specificity associated with pain during
tion findings provides a small to moderate improve- walking (Katz et al., 1995; Konno et al., 2008), although
ment in the LR+ (4.6; 95% CI = 2.48.9) and an Konno and colleagues (2008) did indicate that pain
improvement in post-test probability when compared during walking was a specific finding with LSS. One
with three of five positive examination findings. possible reason may be the method in which walking
Others have investigated similar patient report was captured on the patient response capture. Patients
and observational diagnostic support tools during were able to check if they had pain during the activity
assessment of LSS. Jarvik and Deyo (2002) reported of walking, but were also able to check a number of
that age >65 is modestly sensitive whereas absence of other responses in addition to walking, such as: 1)
pain when seated has low sensitivity (0.46) but is very sitting; 2) standing; 3) running; and 4) driving and
specific (0.93). The most highly specific finding involved others. Pain with standing was combined with walking,
wide-based gait (sensitivity 0.43 and specificity 0.97), which may have diluted the results. The decision to
an independent finding that was combined as the include standing was based on the recent findings from
Physiother. Res. Int. 16 (2011) 170178 2010 John Wiley & Sons, Ltd. 175
Patient History and Findings for LSS C. Cook et al.
176 Physiother. Res. Int. 16 (2011) 170178 2010 John Wiley & Sons, Ltd.
C. Cook et al. Patient History and Findings for LSS
Physiother. Res. Int. 16 (2011) 170178 2010 John Wiley & Sons, Ltd. 177
Patient History and Findings for LSS C. Cook et al.
Pheasant HC, Dyck P. Failed lumbar disc surgery: cause, Ware JE, Kosinski M. Interpreting SF-36 summary health
assessment, treatment. Clinical Orthopaedics and measures: a response. Quality of Life Research 2001; 10
Related Research 1982 Apr; (164): 93109. (5): 40513; discussion 415420.
Ray CD. Extensive lumbar decompression. In: White A, White LJ, Velozo CA. The use of Rasch measurement to
Rothman R, Ray C (eds). Lumbar Spine Surgery, CV improve the Oswestry classification scheme. Archives of
Mosby, St. Louis, 1987; 165. Physical Medicine and Medical Rehabilitation 2002; 83
Rutjes AM, Reitsma J, DiNisio M, Smidt N, van Rijn J, (6): 822831.
Bossuyt P. Evidence of bias and variation in diagnostic Wiesel SW, Tsourmas N, Feffer HL, Citrin CM, Patronas
accuracy studies. CMAJ 2006; 174: 112. N. A study of computer-assisted tomography. I. The
Stucki G, Liang MH, Lipson SJ, Fossel AH, Katz JN. Con- incidence of positive CAT scans in an asymptomatic
tribution of neuromuscular impairment to physical group of patients. Spine 1984; 9: 549551.
functional status in patients with spinal stenosis. Journal Zeifang F, Schiltenwolf M, Rainer B, Moradi B. Gait anal-
of Rheumatology 1994; 1: 13381343. ysis does not correlate with clinical and MR imaging
Sugioka T, Hayashino Y, Konno S, Kikucki S, Fukuhara S. parameters in patients with symptomatic lumbar
Predictive value of self-reported patient information for spinal stenosis. BMC Musculoskeletal Disorders 2008;
the identification of lumbar spinal stenosis. Family 9: 89.
Practice 2008; 25: 237244.
Taylor S, Taylor AE, Foy MA, Foog AJB. Responsiveness of
common outcome measures of patients with low back
pain. Spine 1999; 24: 18051812.
178 Physiother. Res. Int. 16 (2011) 170178 2010 John Wiley & Sons, Ltd.