Evidence in practice.

Clinical Question: How does evidence on the diagnostic accuracy of the vertebral artery test influence teaching of the test in a professional physical therapist education program?

The purpose of "Evidence in Practice" is to illustrate the literature search process to obtain evidence that can guide clinical decision making. This article is not a case report. The examination, evaluation, and intervention sections are purposely abbreviated.

Although faculty members have always had the obligation to remain current in their teaching, (1) they are now being called on to provide evidence to support classroom teaching. (2) We believe evidence-based teaching leads to several positive outcomes. First, the importance of using evidence to make decisions is modeled to students. Second, students are shown that, for some topics, the evidence is contradictory or sparse. Addressing this ambiguity in the classroom also models ways to address similarly ambiguous clinical situations. Third, evidence-based teaching provides faculty with one standard for deciding what should or should not be included in a curriculum.

When teaching tests and measures to physical therapist students, we believe faculty should teach not only psychomotor competency, but also the appropriate use of test information in the clinical reasoning process. In order to draw confident conclusions from a clinical test, clinicians must understand the diagnostic characteristics of that test. Faculty should present the available evidence on diagnostic accuracy--including sensitivity, specificity, positive and negative predictive values, and likelihood ratios--particularly when the test itself puts the patient at risk. In the case of vertebral artery (VA) testing, the information gained from the test must outweigh the risk to the patient.

Vertebral artery testing is a traditional component of cervical spine examination for the assessment of vertebrobasilar insufficiency (VBI), (3-7) a syndrome of brainstem ischemia. (8) This syndrome can result from pathologies such as atherosclerosis of the subclavian, vertebral, and basilar arteries; postural hypotension; Stokes-Adams attacks; and mechanical compression from cervical spondylosis. (9) It also has been attributed to cervical spine examination and intervention techniques, including manipulation (6,10) and passive mobilization. (10) The risk of VBI resulting from cervical manipulation has not been definitively established, (11) although Vickers and Zollman (12) report that the risk of adverse events ranges from 1 in 20,000 to 1 in 1 million manipulations. Common symptoms associated with VBI include dizziness, visual changes, slurred speech, difficulty swallowing, lightheadedness, syncope, nausea, or sensory changes. (4,6,10,13) Symptoms of VBI will vary depending on the balance between factors that compromise circulation and compensatory factors such as bilateral circulation and the circulation through the circle of Willis. (6)

Assessment for VBI typically includes questioning the patient about the history of VBI symptoms and performing VA testing, which is believed to compromise the vertebrobasilar system. (3,4,14) The specifics of VA testing regarding patient position, hold times, and sequence of cervical movements vary among authors. (3,5,7,14-16) A common feature of VA testing, however, involves cervical spine motion to an end-range position of rotation, extension, or combined rotation and extension and then assessing for VBI symptoms. Reproduction of VBI symptoms during the test is considered a positive result and a contraindication for cervical manipulation, (3,4,14) cervical mobilization, (3,5) and end-range rotation. (14) Magee (3) describes the VA test as a "clearing" test, which implies that a negative VA test "clears" the patient of VBI and allows cervical mobilization or manipulation to be used, if indicated. We believe this to be a common interpretation of a negative VA test result.

We suggest that there are 2 major issues to consider when using the VA test. First, clinicians use the result of the VA test when making critical intervention decisions. We were concerned, however, that without knowledge of the diagnostic accuracy of the test, clinicians may be making intervention decisions based on incomplete information. Second, some authors indicate that the VA test may actually put patients at risk for VBI. (11,15,17) With this in mind, we posed the question "How does evidence on the diagnostic accuracy of the vertebral artery test influence teaching of the test in a professional physical therapist education program?"

* Database used for initial search: PubMed

Before we began our search, we knew of one study that described the diagnostic accuracy of VA testing. (18) Because we believed that a successful search strategy would retrieve this study along with other studies reporting on the diagnostic accuracy of the VA test, we used it as a marker for a successful search.

We searched MEDLINE using PubMed, which can be accessed at www.PubMed.gov. PubMed takes advantage of the MeSH (Medical Subject Heading) vocabulary for indexing bibliographic data in MEDLINE. MeSH terms are assigned to a bibliographic record based on the content of the record. We conducted the search on December 30, 2004.

* Initial search: We started by searching for a MeSH term related specifically to our topic. We used the MeSH database, which can be accessed through a link on the left sidebar of the main PubMed screen and through a dropdown list at the top of the main screen. We entered the term vertebr * into the MeSH database query box. The asterisk is a truncation symbol that instructs PubMed to search for alternative endings. Of the 466 terms that were retrieved, we decided to use the MeSH term Vertebrobasilar Insufficiency, after reading its definition. Because we were interested in studies reporting the diagnostic accuracy of VA testing, we decided to use the Clinical Queries feature available in PubMed. The Clinical Queries feature is accessed from the sidebar on the main screen. The interface of the Clinical Queries feature, previously described by Scalzitti and Sternisha, (19) has changed, and an additional search option (Medical Genetics Searches) has been added (Fig. 1). Scalzitti and Sternisha used the Clinical Queries feature to search for citations of studies examining a treatment effect; because we were asking a diagnostic accuracy question, we used this feature to search for citations of articles related to diagnostic testing.

Three options are available from Clinical Queries: Search by Clinical Study Category, Find Systematic Reviews, and Medical Genetics Searches. Because we were interested in clinical studies, we chose the Search by Clinical Study Category option. Clinical Queries offers 4 category filters and 2 scope filters. The 4 available clinical categories, which have not changed since publication of the previous article, (19) are etiology, diagnosis, therapy, and prognosis. The scope of the search (previously called "emphasis") can be set by choosing broad, sensitive search or narrow, specific search. Broad searches retrieve more records, but at the expense of retrieving records that are less relevant, whereas narrow searches retrieve fewer but more relevant records. The combination of the diagnosis and broad, sensitive search filters includes MeSH terms related to diagnostic accuracy (eg, sensitivity and specificity[MeSH Terms] ([dagger])) and terms related to diagnostics (eg, diagnos*[Title/Abstract]). A broad search is achieved by linking the filter terms with the Boolean operator OR. The combination of diagnosis and narrow, specific search filters has a single term: specificity[Title/Abstract].

From the MeSH database screen, we chose Vertebrobasilar Insufficiency and clicked the Links option at the right of the term. A dialog box appeared, offering 3 linking options: PubMed, Clinical Queries, and NLM MeSH Browser. We clicked on Clinical Queries to instruct PubMed to enter the MeSH term into the Clinical Queries screen. We then selected the diagnosis and narrow, specific filters and clicked on Go. We retrieved 23 citations. Using the Limits option, we limited retrieval to English-language studies that used human subjects and included an abstract. We did not limit the search by years. With these limits in place, our search results dropped to 15 citations. After reviewing the tides and abstracts, the only citation to a diagnostic accuracy study of VA testing was Cote et al (18)--the article we were using as a marker of a successful search. The other citations described studies of the diagnostic accuracy of imaging techniques such as magnetic resonance angiography, computerized tomography angiography, and transcranial Doppler sonography for cerebral and limb circulation.

At this point, we decided to use the Related Articles link to search for records related to Cote et al. (18) This feature is useful when you know of a study related to your topic and want to determine whether other related studies are in the database. To find related articles, PubMed uses a word-weighting algorithm to compare records. Only the rifle, abstract, and MeSH headings are compared. Records that are most closely related to the original record are retrieved. Results are displayed from the most to least relevant, with the original record displayed first. When this feature is used, limits are not in effect. Using this option, we retrieved 114 citations. We reviewed the rides and abstracts to determine if the citation referred to an article that examined the effect of cervical motion on VA blood flow and or VBI symptoms. Based on our review of the titles and abstracts, we decided to read 21 articles. (20-40)

* Database used for second search: Science Citation Index Expanded

Although the 21 articles we planned to review were related to VA blood flow and cervical motion, it was not dear whether they examined diagnostic testing of the VA as performed clinically. Up to this point, the only study of the diagnostic accuracy of VA testing that we had found was the study by Cote et al. (18) Consequently, we decided to search the literature further with Science Citation Index Expanded. A useful feature in Science Citation Index Expanded is the Cited Reference Search (Fig. 2). With this feature we were able to search forward, from 1995 to 2004, for articles that cited Cote et al. (18) We believed that other studies of the diagnostic accuracy of VA testing published after Cote et al (18) would cite this article. The ability to search forward from a given citation is not available in PubMed.

Science Citation Index Expanded is a database that provides bibliographic information from approximately 5,900 journals, including American Journal of Physical Medicine & Rehabilitation, Australian Journal of Physiotherapy, Journal of Manipulative and Physiological Therapeutics, Journal of Orthopaedic & Sports Physical Therapy, Manual Therapy, Physical Therapy, Stroke, and Spine. We accessed Science Citation Index Expanded through the Web of Science, produced by Thomson ISI ([double dagger]) (www.isinet.com/). Web of Science is available by subscription and offers 2 other databases: Social Science Citation Index and Arts & Humanities Citation Index. We had access through our university's health science center library. Our institutional subscription provided access to the records of articles published from 1995 to 2004. Our search of Science Citation Index Expanded was conducted on December 30, 2004.

* Second search: We clicked on the Cited Reference Search button (Fig. 2). On the next screen, we entered, in the appropriate query boxes, Cote P for the last name and initial for cited author, J Manip Physiol Ther (the journal abbreviation used by the database publisher) for the cited work, and 1996 for the cited year. We then clicked on Search. The next screen displayed the results of our search including the number of times the article was cited by articles published in journals included in the database (Fig. 2). In our case, Cote et al (l8) had been cited 25 times. We limited retrieval of these 25 articles to those published in English and clicked on the Finish Search button to view the references. Next, we printed the tides and abstracts of the 25 records that cited Cote et al. (18) Fifteen of the 25 records in the database were not found when we searched PubMed. Although none of these 15 articles were diagnostic studies of VA testing, we chose to read 2 articles (41,42) that appeared to be commentaries related to VA testing.

* Database used for third search: Cumulative Index to Nursing and Allied Health Literature

In order to make sure that our search was reasonably comprehensive, our last search for studies of the diagnostic accuracy of tests for vertebrobasilar insufficiency used the CINAHL (Cumulative Index to Nursing and Allied Health Literature) database. We chose CINAHL because we believed that relevant citations might be present that were not in MEDLINE. CINAHL is a smaller database than MEDLINE; however, 55% of the journals in CINAHL are not indexed in MEDLINE. (43) For example, a magazine of the Orthopaedic Section of the American Physical Therapy Association, Orthopaedic Physical Therapy Practice, is indexed in CINAHL but is not indexed in MEDLINE. The search was conducted on December 30, 2004.

* Third search: Much of what we had learned from our previous searches we incorporated into the CINAHL search strategy. We combined subject headings (CINAHL's equivalent of MeSH terms) related to VBI ("vertebral artery" and "vertebral artery dissections") and related to diagnostic accuracy ("sensitivity and specificity," "reliability and validity" [which we exploded to capture narrower terms], and "ROC curve"). Our final search string was (vertebral artery OR vertebral artery dissections) AND ([sensitivity and specificity] OR [reliability and validity] OR [ROC curve]). After limiting our results to English we ended with a set of 7 records, which included Cote et al. (18) Of these 7 records, 3 were not in MEDLINE or Science Citation Index Expanded. After reviewing the tides and abstracts, we decided not to read any of the articles because they did not appear to pertain to our question.

* Selection of articles for review: Based on our searches we decided to read a total of 24 articles: 21 articles from PubMed, 2 articles from Science Citation Index Expanded, and the article by Cote et al. (18) After retrieving the articles from our institution's library or through the interlibrary loan department, we found that 2 were letters, (23,24) 1 was a reliability study of the validity of Doppler ultrasound velocimetry, (39) and 1 examined the effect of head position on blood flow velocity in the internal carotid arteries. (40) Because these items were not appropriate to our question, they will not be discussed. The remaining 20 studies were classified into 4 groups (Tab. 1).

The 2 commentaries (41,42) reached different conclusions. In one of several invited commentaries on the APA guidelines for pre-manipulation testing, (14) Dunne (42) maintains that testing does not determine whether manipulation will be safe. Conversely, Barker et al (41) recommend that a recognized premanipulation test be performed before cervical manipulation. They also state, however, that "several pre-manipulation tests are used by clinicians to screen for VBI, but none have been adequately validated." (41)(p39) In a review article, Haynes (27) concluded that tests of vertebral artery patency using positional changes of the head and neck lack validity and recommended Doppler ultrasound velocimetry as an alternative.

We retrieved 3 case reports (21,22,36) that described 4 cases involving changes in blood flow with cervical motion and the presence or absence of VBI symptoms. Results from the cases did not consistently support the relationship among cervical motion, blood flow or artery compromise, and VBI signs and symptoms. Two cases were false negatives, (21,36) one case was a false positive, (22) and one case was a true positive. (22)

The third group included 13 studies that examined the effect of head and neck movements on blood flow in the cerebral vasculature. Six observational studies of people who were apparently healthy individuals described the effect of head position on blood flow, including extracranial VA blood flow, (20,33,37) intracranial VA blood flow, (29,31,38) and blood flow in the internal carotid arteries. (33) Four other observational studies investigated the relationship between cervical motion and VA blood flow in participants who had signs and symptoms that suggested VBI. (25,26,28,30) Three studies used a group design comparing cases with controls to examine the relationship between cervical motion and VA blood flow. (32,34,35) The findings of these studies did not consistently support a relationship among cervical motion, VA blood flow, and VBI signs and symptoms. The lack of consistency in findings may be the result of differences in the characteristics of the participants and in the tools and procedures used to measure blood flow.

Although not a diagnostic accuracy study, 1 of the 13 studies in the third group reported data that we could use to calculate diagnostic accuracy statistics. (35) The abstract for this study follows.

   Sakaguchi M, Kitagawa K, Hougaku H, Hashimoto H,
   Nagai Y, Yamagami H, Ohtsuki T, Oku N, Hashikawa K,
   Matsushita K, Matsumoto M, Hori M. Mechanical compression
   of the extracranial vertebral artery during neck
   rotation. Neurology. 2003 Sep 23;61(6):845-7.

   Using duplex ultrasonography (US), the authors showed
   compression of the extracranial vertebral artery (ECVA)
   during neck rotation in 5.0% of 1,108 patients. Age (per
   10-year increase, OR 0.80, 95% CI 0.67 to 0.96), vessel
   diameters (per 0.5-mm diameter increase, OR 0.63, 95%
   CI 0.51 to 0.79), and symptoms upon neck rotation (OR
   4.01, 95% CI 1.35 to 11.9) were associated with vessel
   compression. In one case, SPECT ([section]) revealed decreased
   cerebral perfusion of the hindbrain during rotation. ECVA
   US is useful in identifying vessel compression, especially in
   patients with symptoms on neck rotation.

   [[c] 2003 AAN Enterprises Inc. Abstract reproduced with
   permission of Lippincott Williams & Wilkins.]

Using duplex Doppler ultrasonography, Sakaguchi et al (35) examined blood flow in the vertebral arteries and measured the diameter of the vertebral arteries. Participants included 1,108 people who came to their department for neurovascular evaluation. One hundred thirty-six participants (12.3%) had a history of unexplained vertebrobasilar distribution symptoms. Doppler ultrasound testing was performed on participants with their cervical spine in an extended and rotated position. The position was held for at least 10 seconds. Of the 136 participants with a history of unexplained vertebrobasilar distribution symptoms, 28 had symptoms with ultrasound testing and 108 did not. Of the 28 participants with symptoms during ultrasound testing, 5 had VA compression. Of the remaining 108 participants who had no symptoms during ultrasound testing, 7 had VA compression. Of the 972 participants who had no history of vertebrobasilar distribution symptoms, 42 had VA compression. None of these 972 participants had symptoms with ultrasound testing.

Although Sakaguchi (35) did not present diagnostic accuracy statistics, we entered their data into a 2 x 2 table (Tab. 2) and calculated sensitivity, specificity, and likelihood ratios (LRs), along with their 95% confidence intervals (CI). Sensitivity is the proportion of people who actually have the disease and are correctly identified as "positive" by the screening test.44 Specificity is the proportion of people who do not have the disease and are correctly identified as "negative" by screening test. (44) Likelihood ratios combine sensitivity and specificity and indicate the change in odds favoring the disease given the screening test result. (45) In practice, LRs are used to determine the change from pretest probability to posttest probability of a disorder given a particular test result. (46) Likelihood ratios are referred to as positive (LR+) when the test results are positive; the greater the LR+, the more likely the condition is present given a positive test result.47 Likelihood ratios are referred to as negative (LR-) when the test results are negative; the smaller the LR-, the more likely the condition is absent given a negative test. (47) An LR of 1.0 does not change the odds for or against the disease. (45) Jaeschke et al (48) provide further recommendations for interpreting LRs.

When calculating these diagnostic accuracy statistics, we considered ultrasound testing to be the reference test (gold standard) and the cervical position of extension-rotation as the index (screening) test. The frequency of VA compression with ultrasound testing as reported by Sakaguchi et al (35) was the number of true positives. A finding of no VA compression with ultrasound testing was considered a true negative finding. Participants who reported signs or symptoms while their cervical spine was in an extended and rotated position as reported by Sakaguchi et al (35) were considered "positive" on the index test. Based on our interpretation of the data presented by Sakaguchi et al, (35) cervical spine extension-rotation has a sensitivity of 9.3% (95% CI=4%-19.9%), a specificity of 97.8% (95% CI=96.7%-98.5%), an LR+ of 4.243 (95% CI=1.678-10.729), and an LR- of 0.928 (95% CI=0.851-1.011). ([parallel]) The study by Cote et al (18) was the only study we found that reported diagnostic accuracy statistics for cervical spine movements used in pre-manipulation testing. The abstract for this study follows.

   Cote P, Kreitz BG, Cassidy JD, Thiel H. The validity of the
   extension-rotation test as a clinical screening procedure
   before neck manipulation: a secondary analysis. J Manipulative
   Physiol Ther. 1996 Mar-Apr;19(3):159-64.

   OBJECTIVE: To determine the validity of the neck extension-rotation
   test as a clinical screening procedure to
   detect decreased vertebrobasilar blood flow that might be
   associated with dizziness. DESIGN: Secondary analysis
   of a clinical screening test. METHODS: Twelve subjects
   with dizziness reproduced by the extension-rotation test
   and 30 healthy control subjects had Doppler ultrasonography
   examination of their vertebral arteries with the neck
   extended and rotated. Vascular impedance to blood flow
   was measured and the presence of signs and symptoms
   of vertebrobasilar ischemia was recorded. RESULTS: Cutoff
   points for validity estimates were derived through the
   percentile and Gaussian methods using impedance to
   blood flow as the standard. The sensitivity of the extension-rotation
   test for increased impedance to blood flow was
   0%, regardless of the selected cut-off point. The specificity
   rates for the left vertebral artery were 71% and 67% for
   the percentile and Gaussian methods, respectively. The
   extension-rotation test was more specific on the right side,
   with a rate varying from 90% with the percentile method
   to 86% with the Gaussian technique. The positive predictive
   value of the test was 0% and its negative predictive
   value ranged from 63% to 97%. CONCLUSION: We were
   unable to demonstrate that the extension-rotation test is
   a valid clinical screening procedure to detect decreased
   blood flow in the vertebral artery. The value of this test for
   screening patients at risk of stroke after cervical manipulation
   is questionable.

   [[c] 1996 National University of Health Sciences. Abstract
   reproduced with permission of National University of
   Health Sciences.]

This study is a secondary analysis of data collected in a previous study by Thiel et al. (49) We found the study by Thiel et al (49) after reading Cote et al. (18) We used the study by Thiel et al (49) to clarify methodological issues reported by Cote et al. (18) Participants were recruited from chiropractic clinics and identified as either experimental (n = 12) or control (n = 30) subjects. Participants in the experimental group had a history of symptoms related to head and neck movement. In addition, these participants had a positive result on the Wallenberg test (a specific VA test in which the head and neck are held in extension-rotation for 30 seconds); 4 people who had symptoms related to head and neck movement but a negative Wallenberg test were excluded from the study. Radiographic films of the neck and examination by a neurologist were used to rule out nonvascular causes of the symptoms related to head and neck movements. Participants in the control group were healthy; they did not have a history of cardiovascular disease, symptoms related to head and neck movements, or neck pain or stiffness; and they were not pregnant.

The index test for the study was cervical extension-rotation, which was held for at least 30 seconds. The reference test was duplex Doppler ultrasound at C3 to C5. Ultrasound testing was conducted with the subject supine and the head and neck in 3 positions: neutral, head over the edge of the table with the neck fully extended, and head hanging over the edge of the table with the head and neck fully extended and rotated. (49) Both VAs were examined. The unit of measurement for the ultrasound test was the ratio of systolic peak velocity to end diastolic minimum velocity, which is an estimate of vascular impedance. (49) A positive index test result occurred when the participant experienced symptoms of vertigo, nausea, tinnitus, light-headedness, visual problems, numbness of the face or one side of the body, nystagmus, vomiting, or loss of consciousness during cervical extension-rotation.

Using data from the control group, Cote et al (18) established 2 cut points to determine whether the ultrasound test was positive. The first cut point was based on the Gaussian (normal) distribution of vascular impedance values and was set at the upper limit of the 95% CI. The second cut point was the 95th percentile of the vascular impedance values. Cote et al (18) calculated sensitivity, specificity, and predictive values for each VA using each cut point, but did not calculate LRs. The reported sensitivity and positive predictive values were 0% for both VAs and for both cut points. (18) Specificity ranged from 67% to 90%, and negative predictive values ranged from 63% to 97%. (18)

We used the data from 2 x 2 tables of the diagnostic test Results (18) to calculate 95% CIs for sensitivity (#) and LRs and their 95% CIs for the neck extension-rotation test. Because there were no title positives for the left and right VAs at each cut point, we added 0.5 to the value of each cell in the 2 x 2 table to permit calculation of LRs and their 95% CIs. (50) The results from our calculations and values from the data in Cote et al (18) are presented in Table 3. Regardless of the cut point, sensitivity for the extension-rotation test was 0%. The 95% CIs for sensitivity were 0% to 21% for the right VA using the Gaussian method and 0% to 79% for the left VA using the 95th percentile method. Specificity ranged from 67% (95% CI=53%-81%) for the left VA using the Gaussian method to 90% (95% CI=80%-99%) for the right VA using the 95th percentile method. Likelihood ratios and their 95% CIs were calculated using the log method described by Simel et al. (45) Positive LRs ranged from 0.211 (95% CI=0.014-3.174) for the left VA using the Gaussian method to 0.867 (95% CI=0.054-13.824) for the right VA using the 95th percentile method. Negative LRs ranged from 1.017 (95% CI=0.744-1.392) for the right VA using the 95th percentile method to 1.402 (95% CI=1.030-1.909) for the left VA using the Gaussian method.

* Interpretation of the evidence: After searching PubMed, Science Citation Index Expanded, and CINAHL, we found one study that examined the diagnostic accuracy of VA testing, (18) and we were able to calculate sensitivity, specificity, LRs, and 95% CIs for another study. (35) None of the remaining studies examined the diagnostic accuracy of VA testing.

Although Cote et al (18) reported sensitivity, specificity, and predictive values, possible methodological problems limit the usefulness of the findings. For example, in an empirical study on the influence of bias in studies of diagnostic tests, Lijmer et al (51) found that diagnostic accuracy studies using separate groups of participants with and without the disorder (eg, a case-control design) significantly overestimated diagnostic accuracy (relative diagnostic odds ratio=3.0, 95% CI=2.0-4.5) when compared with studies that used consecutive patients. Cote et al (18) used a case-control design. Lijmer et al (15) also found that there was bias in favor of the apparent accuracy of the index test when interpretation of the reference standard was done by people who had knowledge of the results of the index test.

Likewise, in a systematic review of bias in studies of diagnostic accuracy, Whiting et al (52) found that diagnostic accuracy was biased (sensitivity was increased) when clinical information, such as age, sex, and symptoms, was available during the interpretation of test results. Whether knowledge of clinical history should be available to the person interpreting the index test depends on whether the information would be available in the clinical setting. (52) In Cote et al, (18) it is not clear if the clinicians who were asking participants about possible symptoms of VBI during the testing were aware of the clinical history of the participant being tested. This knowledge may change how the questions were asked and influence the diagnostic accuracy of the test; that is, participants who were cases may have been questioned more closely than participants who were controls. Finally, Johnson et al (53) criticized the method used to measure VA blood flow as reported by Thiel et al, (49) calling its use as a reference standard into question.

Putting aside the methodological concerns related to the Cote et al (18) study, the sensitivity values reported for the extension-rotation test and the CIs we calculated do not support the use of this test as screening test for VBI. The sensitivity of the test was 0%. (18) The CIs we calculated for the sensitivity were wide (Tab. 3), indicating a lack of precision, which may have been a result of the small number of participants. (54) In almost all instances, the LRs and their 95% CIs that we calculated (Tab. 3) were close to 1.0, indicating that the extension-rotation test as performed by Cote et al (18) would only slightly change the posttest probability of VBI. (45,48) The exception was the LRs calculated from data using the 95th percentile method as the cut point. In this instance, the upper limits of the positive LRs were large; however, the CI was wide (95% CI=0.073-9.686), indicating a lack of precision and small sample size. (54) Studies with larger sample sizes could provide more precise estimates of the diagnostic accuracy of the extension-rotation test.

Although the study by Sakaguchi et al (35) was not designed as a diagnostic accuracy study, we did review the methods for potential biases known to affect diagnostic accuracy, (51,52) and we calculated diagnostic accuracy statistics based on the data reported by the authors. The participants were from a consecutive series of patients referred for neurovascular examination, and the spectrum of problems seen was broad. (35) Both of these features strengthen the methodological quality of the paper. (51) The authors did not state whether personnel examining the participants were blinded to the participants' clinical history. (35) As previously noted, knowledge of the clinical history may bias the results and increase sensitivity. (52)

The sensitivity values and 95% CIs that we calculated from the data in the study by Sakaguchi et al (35) were larger than those reported by Cote et al (18) (Tab. 3); however, they were still too low for the extension-rotation test to be considered useful as a test to rule out VBI. Likewise, the LR- and 95% CI was close to 1.0. The specificity value and its 95% CI were close to 100% and larger than the specificity values reported by Cote et al (18) (Tab. 3). Such a large specificity value could be interpreted as indicating that the extension-rotation test is useful for ruling in VBI55; however, LRs that are calculated from sensitivity and specificity are better indicators of the usefulness of a diagnostic test. (46,56) The LR+ was 4.243 and the 95% CI was 1.678 to 10.729. Positive LRs between 2 and 5 indicate that a positive test result would lead to a small increase from the pretest to posttest probability. (48) Whether the change would be important would depend on the clinical situation. Although the upper limit of the 95% CI indicates that the extension-rotation test may lead to a large increase from the pretest to posttest probability, the lower limit of the CI indicates that the increase from the pretest to posttest probability would be small and may not be important. (48) In this case, the data are not definitive. (57)

* Teaching decision: Our teaching decision has a psychomotor component and a cognitive component. We believe that, to gain the psychomotor skill needed to perform the VA test, students can review videos or still pictures of the test maneuver; they do not need to actually perform the test. Two aspects of the VA test led us to this conclusion. First, an individual may be at risk for VBI with the VA test. (10,11,15,17,42) Second, the VA test using end-range extension-rotation is an uncomplicated movement that does not require advanced psychomotor skills.

Our teaching decision regarding the cognitive aspects of the VA test has 3 components that we believe are crucial in the development of clinical reasoning: (1) interpreting the test result, (2) using VA test results in making intervention decisions, and (3) understanding the strength of the evidence for using the test.

Whether reported by the author (18) or calculated using reported data,(18,35) the sensitivity, LR-, and associated 95% CIs do not support the interpretation of a negative VA test result as a way to rule out VBI. In fact, these data indicate that a negative VA test does not add anything to the clinical decision-making process. The specificity values and LR+ values indicate that a positive VA test may suggest the presence of VBI and require that the clinician consider referring the patient for further diagnostic testing.

As we described earlier, the VA test has been called a clearing test, (3) implying that a negative test "clears" the patient of VBI and allows cervical mobilization or manipulation if indicated. The diagnostic accuracy data we found do not support this interpretation. Furthermore, because the VA test does not expose the patient to a high-velocity, low-amplitude thrust, a negative VA test cannot be used to determine the safety of cervical manipulation. (17,42)

We believe the strength of the evidence for VA testing is weak. In spite of the common clinical recommendation to use the VA test for VBI, (3-5,14) we found only one study that reported diagnostic accuracy statistics for the VA test. (18) This study appeared to have methodological flaws shown to bias results. (51,52) Many of the studies we reviewed (20,25,26,28-30,32-35,37,38) indicate that our understanding of the usefulness of VA testing is at an early phase of development. (58)

Sackett and Haynes (58) describe 4 phases in the development of a diagnostic test. The first phase is supposed to demonstrate that the test results differ between those people known to have the disorder and those who do not. Results from these studies can add insight into the biological basis of the disorder but cannot be used to demonstrate the clinical usefulness of the test. (58) During the second phase of test development, testing may be done under ideal conditions, comparing people clearly with the disease to those without the disease. (58) At this time, cut points, which clearly differentiate the groups, can be selected. Because participants known to have or not have the disorder are tested in a phase 2 study, the results cannot be used to infer how the test would perform in a clinical situation. (58) In the third phase of test development, testing is done under typical clinical conditions. (58) The pool of participants include people with varying levels of the disorder and people with conditions that could be confused with the disorder of interest. The spectrum of the severity of the disorder should reflect what would be seen in clinical situations where the test will be used. (58) Studies in the fourth phase are designed to determine whether people who undergo testing have better health outcomes than those who do not undergo testing. (58)

Using these criteria, the Cote et al (18) study represents a phase 2 study. Although not designed as a diagnostic study, we consider Sakaguchi et al (35) to represent a phase 2 study as well. Excluding the literature

review, commentaries, and case reports, the remaining group studies (20,25,26,28-34,37,38) represent phase 1 studies at best. We believe that, before the VA test can be used to indicate the presence or absence of VBI, phase 3 studies using consecutive patients in clinical settings and that support the diagnostic accuracy of the test are needed.

The purpose of this article was to examine the diagnostic accuracy of the VA test to inform our teaching of the test. Although we will still teach the VA test to the students, this search has substantially changed how we teach the psychomotor skill and the clinical decisions based on test results. The decision of whether to perform the VA test for a given patient is beyond the scope of this article; however, we believe that any patient seen for cervical conditions should be questioned for symptoms related to VBI. We recognize that diagnostic accuracy data for these questions do not exist; however, they do not put the patient's health at risk. Based on the patient's response to these questions and the clinical judgment of the practitioner, performance of the VA test and referral for diagnostic imaging of VA patency may be appropriate. If the VA test is performed, the practitioner should carefully interpret the results in light of the diagnostic accuracy evidence for the test.

A presentation based on an earlier draft of this paper was given to the Eastern District, Missouri Chapter of the American Physical Therapy Association.

The authors thank Mary Krieger, MLIS, RN, for her informative discussions about PubMed and CINAHL and for her comments on an earlier draft of this manuscript.

References

(1) Commission on Accreditation in Physical Therapy Education. Accreditation Handbook: Evaluative Criteria for the Accreditation of Education Programs for the Preparation of Physical Therapists. Alexandria, Va: Commission on Accreditation in Physical Therapy Education; 1998.

(2) Rothstein JM. Graduation day for education programs [Editor's Note]. Phys Ther. 2003;83:518-519.

(3) Magee DJ. Orthopedic Physical Assessment. 4th ed. Philadelphia, Pa: Saunders; 2002.

(4) Maitland GD. Vertebral Manipulation. 5th ed. London, United Kingdom: Butterworths; 1986.

(5) Hertling D, Kessler RM. Management of Common Musculoskeletal Disorders: Physical Therapy Principles and Methods. 3rd ed. Philadelphia, Pa: JB Lippincott; 1996.

(6) Grant R. Vertebral artery concerns: premanipulative testing of the cervical spine. In: Grant R, ed. Physical Therapy of the Cervical and Thoracic Spine. 2nd ed. New York, NY: Churchill Livingstone Inc; 1994:145-166.

(7) Donatelli R, Wooden MJ. Orthopaedic Physical Therapy. 3rd ed. New York, NY: Churchill Livingstone Inc; 2001.

(8) Noble J. Textbook of Primary Care Medicine. 3rd ed. St Louis, Mo: Mosby; 2001.

(9) Cummings CW, ed. Otolaryngology: Head and Neck Surgery. 3rd ed. St Louis, Mo: Mosby; 1998.

(10) Magarey ME, Rebbeck T, Coughlan B, et al. K. Pre-manipulative testing of the cervical spine review, revision, and new clinical guidelines. Man Ther. 2004;9:95-108.

(11) Di Fabio RP. Manipulation of the cervical spine: risks and benefits. Phys Ther. 1999;79:50-65.

(12) Vickers A, Zollman C. ABC of complementary medicine: the manipulative therapies: osteopathy and chiropractic. BMJ. 1999;319:1176-1179.

(13) Saeed AB, Shuaib A, Al-Sulaiti G, Emery D. Vertebral artery dissection: warning symptoms, clinical features and prognosis in 26 patients. Can J Neurol Sci. 2000;27:292-296.

(14) Australian Physiotherapy Association. Clinical Guidelines for Pre-manipulative Procedures for the Cervical Spine. Melbourne, Victoria, Australia: Australian Physiotherapy Association; 2000.

(15) Baxter RE. Pocket Guide to Musculoskeletal Assessment. 2nd ed. Philadelphia, Pa: Saunders; 2003.

(16) Saunders HD, Saunders R. Evaluation, Treatment and Prevention of Musculoskeletal Disorders. Vol I: Spine. 3rd ed. Chaska, Minn: The Saunders Group; 1993.

(17) Mann T, Refshauge KM. Causes of complications from cervical spine manipulation. Aust J Physiother. 2001;47:255-266.

(18) Cote P, Kreitz BG, Cassidy JD, Thiel H. The validity of the extension-rotation test as a clinical screening procedure before neck manipulation: a secondary analysis. J Manipulative Physiol Ther. 1996; 19:159-164.

(19) Scalzitti D, Strernisha C. Does exercise during hospitalization after stem cell transplantation decrease reports of fatigue and reduce the duration of the hospital stay? Phys Ther. 2002;82:716-721.

(20) Arnold C, Bourassa R, Langer T, Stoneham G. Doppler studies evaluating the effect of a physical therapy screening protocol on vertebral artery blood flow. Man Ther. 2004;9:13-21.

(21) Bolton PS, Stick PE, Lord RS. Failure of clinical tests to predict cerebral ischemia before neck manipulation. J Manipulative Physiol Ther. 1989;12:304-307.

(22) Combs SB, Triano JJ. Symptoms of neck artery compromise: case presentations of risk estimate for treatment. J Manipulative Physiol Ther. 1997;20:274-278.

(23) Cote P. Rotation: A valid premanipulative dizziness test? Does it predict safe manipulation? [letter]. J Manipulative Physiol Ther. 1994;17:413-414.

(24) Giles LG. The validity of the extension-rotation test as a clinical screening procedure before neck manipulation: a secondary analysis [letter]. J Manipulative Physiol Ther. 1997;20:65.

(25) Haynes M, Milne N. Color duplex sonographic findings in human vertebral arteries during cervical rotation. J Clin Ultrasound. 2001;29:14-24.

(26) Haynes MJ. Doppler studies comparing the effects of cervical rotation and lateral flexion on vertebral artery blood flow. J Manipulative Physiol Ther. 1996;19:378-384.

(27) Haynes MJ. Vertebral arteries and cervical movement: Doppler ultrasound velocimetvy for screening before manipulation. J Manipulative Physiol Ther. 2002;25:556-567.

(28) Jargiello T, Pietura R, Rakowski P, et al. Power Doppler imaging in the evaluation of extracranial vertebral artery compression in patients with vertebrobasilar insufficiency. Eur J Ultrasound. 1998;8:149-156.

(29) Li YK, Zhang YK, Lu CM, Zhong SZ. Changes and implications of blood flow velocity of the vertebral artery during rotation and extension of the head. J Manipulative Physiol Ther. 1999;22:91-95.

(30) Licht PB, Christensen HW, Hoilund-Carlsen PE Is there a r01e for premanipulative testing before cervical manipulation? J Manipulative Physiol Ther. 2000;23:175-179.

(31) Mitchell J, Keene D, Dyson C, et al. Is cervical spine rotation, as used in the standard vertebrobasilar insufficiency test, associated with a measureable change in intracranial vertebral artery blood flow? Man Ther. 2004;9:220-227.

(32) Petersen B, von Maravic M, Zeller JA, et al. Basilar artery blood flow during head rotation in vertebrobasilar ischemia. Acta Neurol Scand. 1996;94:294-301.

(33) Refshauge KM. Rotation: a valid premanipulative dizziness test? Does it predict safe manipulation? J Manipulative Physiol Ther. 1994;17:15-19.

(34) Rivett DA, Sharpies KJ, Milburn PD. Effect of premanipulative tests on vertebral artery and internal carotid artery blood flow: a pilot study. J Manipulative Physiol Ther. 1999;22:368-375.

(35) Sakaguchi M, Kitagawa K, Hougaku H, et al. Mechanical compression of the extracranial vertebral artery during neck rotation. Neurology. 2003;61:845-847.

(36) Westaway MD, Stratford P, Symons B. False-negative extension/rotation pre-manipulative screening test on a patient with an atretic and hypoplastic vertebral artery. Man Ther. 2003;8:120-127.

(37) Zaina C, Grant R, Johnson C, et al. The effect of cervical rotation on blood flow in the contralateral vertebral artery. Man Ther. 2003;8:103-109.

(38) Mitchell JA. Changes in vertebral artery blood flow following normal rotation of the cervical spine. J Manipulative Physiol Ther. 2003;26:347-351.

(39) Haynes MJ. Vertebral arteries and neck rotation: Doppler velocimeter and duplex results compared. Ultrasound Med Biol. 2000;26:57-62.

(40) Licht PB, Christensen HW, Hoilund-Carlsen PE Carotid artery blood flow during premanipulative testing. J Manipulative Physiol Ther. 2002;25:568-572.

(41) Barker S, Kesson M, Ashmore J, et al. Professional issue: guidance for pre-manipulative testing of the cervical spine. Man Ther. 2000;5:37-40.

(42) Dunne J. Pre-manipulative testing: predicting risk or pretending to? Aust J Physiother. 2001;47:165.

(43) CINAHL-MEDLINE comparison. CINAHL news. 2003;22(3-4):1-4. Available at http://www.cinahl.com/library/library.htm. Accessed March 24, 2005.

(44) Altman DG, Bland JM. Diagnostic tests, 1: sensitivity and specificity. BMJ. 1994;308:1552.

(45) Simel DL, Sama GP, Matchar DB. Likelihood ratios with confidence: sample size estimation for diagnostic studies. J Clin Epidemiol. 1991;44:763-770.

(46) Deeks JJ, Altman DG. Diagnostic tests, 4: likelihood ratios. BMJ. 2004;329:168-169.

(47) Davidson M. The interpretation of diagnostic test: a primer for physiotherapists. AustJPhysiother. 2002;48:227-332.

(48) Jaeschke R, Guyatt GH, Sackett DL, the Evidence-Based Medicine Working Group. Users' guides to the medical literature, III: how to use an article about a diagnostic test, B: what are the results and will they help me in caring for my patients? JAMA. 1994;271:703-707.

(49) Thiel H, Wallace K, Donat J, Yong-Hing K. Effect of various head and neck positions on vertebral artery blood flow. Clin Biomech. 1994;9:105-110.

(50) Altman D. Diagnostic tests. In: Altman D, Machin D, Bryant T, Gardner M, eds. Statistics With Confidence: Confidence Intervals and Statistical Guidelines. 2nd ed. London, United Kingdom: BMJ Books; 2000:105-119.

(51) Lijmer JG, Mol BW, Heisterkamp S, et al. Empirical evidence of design-related bias in studies of diagnostic tests. JAMA. 1999;282:1061-1066.

(52) Whiting P, Rutjes AWS, Reitsma JB, et al. Sources of variation and bias in studies of diagnostic accuracy: a systematic review. Ann Intern Med. 2004;140:189-202.

(53) Johnson C, Grant R, Dansie B, et al. Measurement of blood flow in the vertebral artery using colour duplex Doppler ultrasound: establishment of the reliability of selected parameters. Man Ther. 2000;5:21-29.

(54) Sim J, Reid N. Statistical inference by confidence intervals: issues of interpretation and utilization. Phys Ther. 1999;79:186-195.

(55) Sackett DL, Straus SE, Richardson WS, et al. Evidence-Based Medicine: How to Practice and Teach EBM. 2nd ed. New York, NY: Churchill Livingstone Inc; 2000.

(56) Pewsner D, Battaglia M, Minder C, et al. Ruling a diagnosis in or out with "SpPIn" and "SnNOut": a note of caution. BMJ. 2004;329:209-213.

(57) Montori VM, Kleinbart J, Newman TB, et al. Tips for learners of evidence-based medicine, 2: measures of precision (confidence intervals). CMAJ. 2004;171:611-615.

(58) Sackett DL, Haynes RB. Evidence base of clinical diagnosis: the architecture of diagnostic research. BMJ. 2002;324:539-541.

([section]) SPECT = single photon emission computerized tomography.

([parallel]) Values were calculated using the software program Confidence Interval Analysis, version 2.1.1, build 48. The program is included with Altman D, Machin D, Bryant T, Gardner M, eds. Statistics With Confidence: Confidence Intervals and Statistical Guidelines, 2nd ed. London, United Kingdom: BMJ Books; 2000. 95% CIs for sensitivity and specificity were calculated using the Wilson method. The log method was used to calculate the 95% CIs for the likelihood ratios.

(#) 95% CIs for sensitivity were calculated using the Wilson method available through the software program Confidence Interval Analysis, version 2.1.1, build 48.

([dagger]) Items in brackets are the portions of the MEDLINE record that PubMed will search. In this case, the asterisk instructs PubMed to search for terms with alternate endings such as "diagnosis," "diagnose," or "diagnostic."

([double dagger]) Thomson Scientific, 3501 Market St, Philadelphia, PA 19104.

Randy R Richter, PT, PhD, is Associate Professor, Department of Physical Therapy, Edward and Margaret Doisy School of Allied Health Professions, Saint Louis University, St Louis, Mo.

Mark F Reinking, PT, PhD, SCS, ATC, is Assistant Professor, Department of Physical Therapy, Edward and Margaret Doisy School of Allied Health Professions, Saint Louis University.

Table 1. Studies Reviewed: Grouped by Type (a)

A. Commentaries and a review of vertebral artery testing

Barker et al (41)
Dunne (42)
Haynes (27)

B. Case reports

Bolton et al (21)
Combs and Triano (22)
Westaway et al (36)

C. Studies of blood flow in participants who were healthy
or in participants with signs though to be indicative
of vertebrobasilar insufficiency while their neck was
placed in positions believed to affect the vertebral or
internal carotid arteries.

Arnold et al (20)
Haynes (26)
Haynes and Milne (25)
Jargiello et al (28)
Li et al (29)
Licht et al (30)
Mitchell (38)
Mitchell et al (31)
Petersen et al (32)
Refshauge (33)
Rivett et al (34)
Sakaguchi et al (35)
Zaina et al (37)

(a) Studies in red were found by the search of Science Citation
Index Expanded. The remaining studies were found by the search
of PubMed.

Table 2. A 2 x 2 Table Based on Our Interpretation of Data
From Sakaguchi et al (35)

Signs/Symptoms
With Cervical    Vertebral Artery
Position (b)     Compression (a)

                 Yes     No

Yes               5       23
No               49    1,031

(a) Based on ultrasound testing.

(b) Extension-rotation.

Table 3. Sensitivity (%), Specificity (%), Likelihood Ratios (LR) and
Their 95% Confidence Intervals (CI) for the Cervical Extension-Rotation
Test Calculated From Data From Cote et al (18,a)

                            Sensitivity    Specificity
Cut Point (b)                (95% CI)       (95% CI)

Gaussian        Right       0% (0%-21%)   86% (76%-96%)
method          vertebral
                artery

                Left        0% (0%-39%)   67% (53%-81%)
                vertebral
                artery

95th            Right       0% (0%-49%)   90% (80%-99%)
percentile      vertebral
method          artery

                Left        0% (0%-79%)   71% (57%-84%)
                vertebral
                artery

                                Positive LR            Negative LR
Cut Point (b)                     (95% CI)              (95% CI)

Gaussian        Right       0.215 (0.012-3.73)     1.144 (0.954-1.371)
method          vertebral
                artery

                Left        0.211 (0.014-3.174)    1.402 (1.030-1.909)
                vertebral
                artery

95th            Right       0.867 (0.054-13.824)   1.017 (0.744-1.392)
percentile      vertebral
method          artery

                Left        0.840 (0.073-9.686)    1.068 (0.468-2.434)
                vertebral
                artery

(a) The 95% Cls for sensitivity were calculated using the software
program Confidence Interval Analysis, version 2.1.1, build 48. The
program is included as part of Altman D, Machin D, Bryant T, Gardner M,
eds. Statistics With Confidence: Confidence Intervals and Statistical
Guidelines. 2nd ed. London, United Kingdom: BMJ Books; 2000. Likelihood
ratios and their 95% Cls were calculated using procedures recommended
by Altman et a150 and the log method described by Simel et al. (45)

(b) Using data from the control group, 2 cut points were established to
determine whether the ultrasound test was positive. The first cut point
was based on the Gaussian (normal) distribution of vascular impedance
values and was set at the upper limit of the 95% confidence interval
(CI). The second cut point was the 95th percentile of the vascular.
impedance value.