The literature review for diagnostic accuracy studies identified 4722 citations with 20 additional citations identified from the bibliographies of review articles and the authors’ libraries (eFigure); 118 full-text articles were retrieved for full assessment. Four articles evaluating 741 patients were included in the final review.35–38
One population was studied in 2 separate reports: the first involved history and physical examination findings, and the second, questionnaire items ().35,36
Prevalence of the Clinical Syndrome of LSS
The prevalence of the clinical syndrome of LSS in the eligible diagnostic accuracy studies varied from 44% to 49%. The highest quality study included approximately one-third of patients recruited directly from primary care clinics and reported a prevalence of the clinical syndrome of LSS of 47% in adults with symptoms of pain or numbness in the lower extremities.35
These patients had a mean (SD) age of 65 (14) years and 54% were women.
Accuracy of Historical Features and Symptoms
The performance characteristics of all clinically relevant tests with LR point estimates of 2.0 and higher or 0.50 or less are listed in (eTable 1 and eTable 2 include complete data of all findings and are available at www.jama.com
). These thresholds were defined by some authors as producing small but meaningful changes in posttest probability.39
Diagnostic Accuracy of History and Physical Examination
Age and Comorbidities
The likelihood of the clinical syndrome of LSS increases with age, especially for individuals older than 70 years (LR, 2.0; 95% CI, 1.6–2.5). Patients younger than 60 years are less likely to have the clinical syndrome of LSS (LR, 0.40; 95% CI, 0.29–0.57). Concurrent orthopedic problems such as osteoarthritis, inflammatory arthritis, and fractures increase the likelihood of the clinical syndrome of LSS (LR, 2.0; 95% CI, 1.2–3.5).
Among the most useful symptoms when examined in isolation were those that described pain location and provocative associations. The most useful symptoms for increasing the likelihood of the clinical syndrome of LSS were having no pain when seated (LR, 7.4; 95% CI, 1.9–30), having an unexplained urinary disturbance (LR, 6.9; 95% CI, 2.7–17), improvement of symptoms when bending forward (LR, 6.4; 95% CI, 4.1–9.9), the presence of bilateral buttock or leg pain (LR, 6.3; 95% CI, 3.1–13), or neurogenic claudication (LR, 3.7; 95% CI, 2.9–4.8). The presence of symptoms thought to be related to cauda equina syndrome, including a burning sensation around the buttocks, intermittent priapism associated with walking, or both increased the likelihood of the clinical syndrome of LSS (LR, 7.2; 95% CI, 1.6–32). However, these symptoms were insensitive and present in only 6% of patients. The absence of neurogenic claudication (LR, 0.23; 95% CI, 0.17–0.31) was the most useful symptom for decreasing the likelihood of the clinical syndrome of LSS when test results were negative.
Accuracy of the Physical Examination
Physical examination tests taken in isolation were not as useful as symptoms (). A wide-based gait (LR, 13; 95% CI, 1.9–95) and an abnormal Romberg test result (LR, 4.2; 95% CI, 1.4–13) increased the likelihood of the clinical syndrome of LSS. Two separate studies examined the effects of lumbar flexion on pain and reached similar conclusions: one study found pain with lumbar flexion to have an LR of 0.48 (95% CI, 0.24–0.96),38
and a second study found symptoms induced by bending forward to have an LR of 0.48 (95% CI, 0.34–0.66).35
Accuracy of the Clinical Examination in Multivariate Analyses
Certain individual tests may be highly intercorrelated. In these instances, the independent value of the second test is diminished once the value of the first test is accounted for. Multivariate analyses can overcome this problem and identify the independent, incremental value of diagnostic tests in the presence of other tests. Three studies examined tests in multivariate analyses.35,36,38
One study found that increased age (in years), having no pain when seated, a wide based gait, and having thigh pain with 30 seconds of lumbar extension were independently associated with the degree of expert physician confidence in the diagnosis of the clinical syndrome of LSS.38
Two studies examined multivariate predictors of the clinical syndrome of LSS in the same sample of patients and used predictor variables that were independently associated with the diagnosis to create risk scores for diagnosing the clinical syndrome of LSS35,36
(). A score of 7 or higher on a clinical diagnostic support tool including history and examination findings increased the likelihood of the clinical syndrome of LSS35
(LR, 3.3; 95% CI, 2.7–4.0; eTable 3 available at www.jama.com
). Therefore, although sensitivity was optimized by the combination of history and examination findings, a lower overall specificity contributed to a lower positive LR than what was seen with some individual tests. A score of less than 7 on this diagnostic tool made the clinical syndrome of LSS much less likely (LR, 0.10; 95% CI, 0.06–0.16). A score of 5 or higher on a diagnostic tool including only questionnaire-based items produced small increases in the likelihood of the clinical syndrome of LSS36
(LR, 1.9; 95% CI, 1.6–2.3; eTable 3). A score of less than 5 on this tool made the clinical syndrome of LSS less likely (LR, 0.33; 95% CI, 0.23–0.45). On testing in a validation sample, this questionnaire-based diagnostic tool yielded smaller magnitude LRs when positive (LR, 1.5; 95% CI, 1.1–2.1) and when negative (LR, 0.50; 95% CI, 0.28–0.88).
Limitations of the Literature
In an effort to capture all possibly relevant studies, we used an initial search strategy favoring sensitivity over specificity and trusted the manual search would remove studies not relevant to the area of research. However, although the topic of the diagnosis of LSS has been widely addressed in expert commentaries, surgical case series and cohort studies of patients with LSS, and a limited number of diagnostic accuracy studies using a purely radiographic reference standard, very few studies examined the accuracy of the history and physical examination using a clearly defined and appropriate reference standard such as the clinical syndrome of LSS. Stringent criteria for quality were applied in this review.
The included studies had methodological differences that did not permit pooling of data in a true meta-analysis, and generally did not allow comparison of individual tests between studies. Three studies of 2 different patient populations excluded some patients with indeterminate findings by the reference standard.35,36,38
However, 2 of these studies excluded only 1 patient out of 469, which we thought was inconsequential.35,36
No studies permitted stratification by subtype of radiographic LSS severity.
Although the prevalence of the clinical syndrome of LSS was high in the included studies of primarily older adults, it is important to note the prevalence of the clinical syndrome of LSS in all patients presenting to a primary clinic with leg pain, back pain, or both may be substantially lower. Only 2 diagnostic accuracy studies, which used the same study sample, included a substantial proportion (one-third) of patients recruited from primary care ().35,36
Therefore, a greater severity of disease in the specialty clinic populations from which these accuracy estimates were derived may overestimate sensitivity and underestimate specificity when these tests are applied to primary care populations. In addition to this bias induced by the spectrum of disease, there is a problem with incorporation bias whereby the overall clinical findings are taken into account in establishing the diagnosis. Because a diagnosis of the clinical syndrome of LSS requires information from the clinical examination, such bias is unavoidable. Potential incorporation bias may have been mitigated in the included studies by using the consensus diagnosis of multiple expert spine clinicians,35
and patient-reported data.36,37
The clinical diagnostic support tool using combinations of history and physical examination findings by Konno et al35
was subsequently tested in a separate validation study that did not meet inclusion criteria for this review, due to the inclusion of non-adults. This study by Kato et al40
found that a positive result on the diagnostic support tool had an LR of 1.6 (95% CI, 1.3–2.0) and a negative result had an LR of 0.13 (95% CI, 0.04–0.41). Taken together, these data demonstrate that this diagnostic tool is most useful for ruling out the clinical syndrome of LSS but is of limited value for ruling in disease. This may reflect the heterogeneity of the clinical syndrome of LSS, for which anatomic stenosis at different locations and multiple lumbar spinal interspaces may interact with person-specific factors to result in a wide spectrum of possible disease presentations and severity on a population level. Estimates from the validation study of the clinical diagnostic tool should be viewed cautiously given methodological differences from the derivation study. In contrast to the clinical diagnostic tool using history and physical examination findings, a validated questionnaire-based diagnostic tool had quite modest diagnostic value that is unlikely to be clinically useful.36