Improving the recognition of cervical arterial dissection in clinical practice: investigation of a five criteria diagnostic support tool

ABSTRACT Background Cervical arterial dissection (CeAD) is a serious condition that can mimic a musculoskeletal condition. A diagnostic tool using five key criteria could help prompt early medical referral, but these criteria may occur in healthy people or benign neck pain/headache. Objective To determine the frequency of CeAD criteria in healthy individuals and those with neck pain/headache, and identify refinements needed to improve specificity. Methods An interview and neurological screen to identify the presence of the five criteria was conducted. Definitions were refined and the frequency of the modified criteria in each individual was determined. The criteria were re-administered using data from 37 CeAD cases of the derivation cohort, to examine how the modifications impact sensitivity of the tool. Results One hundred healthy and 20 participants with neck pain/headache were interviewed. Most participants had ≤ 2 criteria, mainly age or trauma, 3% had 3 criteria, but had migraine or resolving symptoms. None had >3. Modifications to definitions were needed to improve potential specificity of the tool (96.7%). Changes did not impact sensitivity of the tool (81%). Further refinements may be required. Conclusions Strictly defined CeAD criteria may assist in identifying when to refer, when to wait and monitor, or when management can proceed. Trialing the tool in those with migraine and in emergency departments to calculate risk scores is recommended.


Introduction
Cervical arterial dissection (CeAD) is a rare adverse outcome of cervical manipulation but may be catastrophic (Albuquerque et al., 2011;Arnold and Bousser, 2005;Paciaroni and Bogousslavsky, 2009). It is also a common cause of stroke in people under 55 years who are usually otherwise healthy, although it can occur at any age with an estimated incidence of 2.6-3/100,000 (Debette, 2014;Metso et al., 2012). The early symptoms, typically headache and neck pain, may mimic migraine or a musculoskeletal problem before obvious neurological signs or symptoms arise, which can be up to 5 weeks later (Biller et al., 2014;Thomas, Rivett, Attia, and Levi, 2015). Subsequently, patients may present for inappropriate musculoskeletal treatment. The mortality rates are generally low <5%, but may be higher if CeAD progresses to ischemic stroke (Debette and Leys, 2009). Early diagnosis is crucial, but may often be missed due to the lack of obvious stroke symptoms (Gottesman et al., 2012;Grond-Ginsbach et al., 2013;Morris, Merkler, Gialdini, and Kamel, 2017), delaying medical evaluation and putting the patient at risk of a poorer outcome (Thomas, 2016).
There is a need for a succinct protocol, which can be conducted within a normal clinical encounter to quickly identify patients of concern, specifically those presenting with neck pain or headache. Existing frameworks in relation to CeAD encompass general vascular conditions and factors, such as cardiovascular risk factors, that are primarily directed to screen for risk of CeAD prior to cervical manipulative therapy, rather than specific features for recognition of CeAD in progress (Rushton et al., 2014(Rushton et al., , 2020. These are likely contributing to overrecognition of irrelevant factors and may lack sufficiently reliable (de Best et al., 2021) and efficient means to recognize an acute pathology such as a CeAD in progress, and the need for urgent medical evaluation. A simple evidencebased diagnostic support tool could help identify those in the early stages of CeAD and direct patients of concern on to further investigation, avoiding inappropriate and potentially dangerous musculoskeletal management.
Such a tool could be useful in both emergency and primary care settings and used by several different professionals. The tool could include weighting of more important features or combinations of features that would be helpful to define a cutoff level beyond which further investigation such as computed tomography (CT) or magnetic resonance (MR) angiography would be indicated and a lower cutoff score where follow-up would be recommended. A brief physical screen for neurological signs relevant to CeAD could confirm reported symptoms and strengthen findings.
A derivation cohort was used to develop a diagnostic tool, utilizing key clinical features and risk factors from both retrospective and prospective case-control studies of patients with radiographically diagnosed CeAD (Thomas, Rivett, Attia, and Levi, 2015;Thomas, Rivett, and Levi, 2011). Identified risk factors and the most common presenting clinical features with a p-value less than 0. 2 in univariate analyses, were examined in multivariate analyses, compared with individuals with stroke from some other cause, and all significant factors were included in the tool. This analysis found the following five factors to be important and these were developed into the diagnostic tool as key criteria for CeAD: 1) Age under 55 years; 2) Recent minor trauma or infection; 3) Acute onset of neck pain or headache; 4) Recent dizziness; and 5) Focal neurological signs in last 5 weeks.
However, diagnostic tools must have both good sensitivity and specificity. In this regard, the criteria above might have good sensitivity, which is critical, but because several of these features can commonly occur in healthy people or those with benign neck pain or headache disorders, this could potentially lead to a false diagnosis of CeAD or poor specificity. This is particularly relevant in this case where the condition is rare and good specificity is important to avoid both the initiation of unnecessary medical investigations on people who don't need them, and also withholding of effective manual therapy treatments. Further, if the tool demonstrates good specificity, this could reassure clinicians when it is safe to exclude CeAD in progress and proceed with the assessment.
Before the criteria can be tested on a clinical population, they need to be trialed on people without CeAD to help tighten criteria definitions to identify CeAD more accurately. Therefore, the study aimed to test the five criteria and briefly screen for focal neurological signs in healthy people and those with neck pain and/or headache to identify the need for refinement of definitions, and then retest the modified criteria. Specific research questions were: 1) How many and which combinations of the above criteria are present in healthy people and in those with neck pain and/or headache; 2) What modifications are needed for any of criteria to improve specificity of the tool; 3) How many healthy people and those with neck pain/headache have abnormal neurological signs related to CeAD; 4) How many and which combinations of the modified criteria are present in healthy and people with neck pain and/or headache; and 5) Do the changes made to improve specificity affect the sensitivity of the modified tool?

Methods
This was an observational study conducted among staff, students and the local community at the University of Queensland, Brisbane, Australia. Inclusion criteria were healthy individuals between 18 and 65 years not currently undergoing treatment for neck pain or headache. Another group of people with neck pain and/or headache were also included. Exclusion criteria were recent lower limb surgery/trauma which might impact balance testing, and preexisting neurological conditions. The study was approved by the institutional ethics committee and all participants gave their informed consent.

Protocol
The processes involved in the study are shown in Figure 1. In brief, the participants were first screened for the five criteria and any relevant physical features. The interview provided more detail and clarification in regard to any reported criteria. The results were then reviewed, and criteria definitions refined to capture the specific symptoms of greater concern. The modified criteria were re-administered on the results and then compared to the original derivation cohort of CeAD cases.
Initially, to determine the presence of any of the criteria for CeAD, all participants were assessed by trained researchers in a single session consisting of a structured interview and brief physical screen for Horner's syndrome, balance and gait or coordination disturbances. The interview was conducted by initially simply asking the participant about the presence of the five criteria exactly as stated above: 1) Age under 55 years; 2) Recent minor trauma or infection; 3) Acute onset of neck pain or headache; 4) Recent dizziness; and 5) Focal neurological signs in last 5 weeks, specifically visual, speech or balance disturbance, limb/ facial paresthesia or weakness.
For criterion 2, participants were asked about specific types of trauma, such as jerky head movements, heavy lifting, intensive exercise, and sustained end range positions of the neck that have been reported in association with CeAD. Any history of recent illness or infection particularly involving prolonged coughing (> 3 weeks) was also recorded as this can involve trauma. Although criterion 5 related to neurological signs, reports of neurological symptoms were also collected as these may be precursors to signs.
The examiners were instructed to ask follow-up questions as required to further clarify the presence of the criteria and determine possible benign causes. Details were recorded to enable possible modification of the criteria to improve specificity of the tool. Examiners were qualified physiotherapists undertaking specialty musculoskeletal training (2) and final year physiotherapy students (2) who had completed a full semester of cervical spine and neurological courses. They first underwent a formal training session of 5 hours and had the opportunity to practice over the next two weeks and were then checked by the lead researcher, an experienced musculoskeletal physiotherapist before commencing any data collection.

Participant characteristics
Participant characteristics such as age, gender, height and weight were collected from the interview and BMI calculated to identify any individuals outside healthy ranges. In addition to CeAD criteria, participants were also questioned about any potential neurological symptoms specifically any features relevant for CeAD (e.g., visual disturbance, speech, swallowing or hoarse voice, balance and coordination, and limb weakness or paresthesia) as these might preempt the onset of neurological signs (Debette et al., 2011).

Neurological screening
Physical screening was performed to assess for presence of any neurological signs that could be related to CeAD. Horner's syndrome; unilateral ptosis (drooping eyelid) or miosis (constricted pupil) was assessed by visual inspection of the participant's eyes. Balance was assessed in quiet tandem stance in bare feet with eyes closed. A positive test was identified as the inability to maintain this position for 30 seconds but was repeated with the eyes open if the participant could not achieve this. Coordination was evaluated by tandem gait over two meters. As gait assessment might not be possible in a patient group, it was also assessed by observable dysdiadochokinesia in a rhythmical alternating forearm pronation and supination test and observable overshoot or undershoot in the finger to nose test (Blumenfeld, 2014).

Reported initial CeAD criteria
Numbers of participants who reported any of the 5 key CeAD criteria below were recorded. Any clarifications regarding the presence of these criteria were also documented. Criteria were numbered in order of clinical relevance; from those likely to be commonly present to those potentially indicative of more serious pathology: 1) age under 55 years; 2) recent minor trauma or strain to head or neck or recent infection; 3) unusual or acute onset of headache or neck pain; 4) recent dizziness; and 5) focal neurological signs in the last 5 weeks The frequency of participants reporting combinations of criteria from less serious (e.g. 1 + 2 + 3) to increasingly more serious (1 + 2 + 3 + 4 or 5) was recorded to determine which combinations might indicate the need for urgent referral in a patient population.

Statistical analysis
Demographic details and number of initial CeAD criteria per participant were reported with descriptive statistics including frequencies and percentages. Combinations of criteria for both (1 + 2 + 3; 1 + 2 + 3 + 4; 1 + 2 + 3 + 5; 1 + 2 + 3 + 4 + 5) were also recorded with frequencies and percentages and the specificity of the tool was calculated by dividing the number of true negatives by the number of non-CeAD cases x 100. These values were inspected to determine whether certain criteria were overrepresented, which would result in false-positive findings for the tool.

Participant characteristics
One hundred healthy participants, 58 females, with a mean (SD) age of 34 (± 13) years, and 20 participants with neck pain, 11 females with a mean age 27 years (± 6.1) years, entered the study. All had BMI within normal ranges.

Reported initial CeAD criteria
The number, n (%) of participants (healthy, neck pain) who reported each of the initial criteria for CEAD and those reporting combinations of criteria are shown in Table 1. There were several suggested modifications to certain criteria after reviewing the clarifications obtained during the interview.

Criteria 1 and 2: age and minor trauma or infection
Age under 55 years was common due to the inclusion criteria. Reports of recent minor trauma or infection were reasonably common particularly in the neck pain group. Clarifications did not suggest any changes to these criteria.

Criterion 3: acute onset neck pain or headache
Despite not requiring treatment, there were still reports of neck pain (48%) or headache (58%) in the healthy group, but on clarification, all pain was infrequent, mild, and familiar. In one participant, there was an acute onset. In the neck pain group, four participants reported acute pain; for one this was only mild, but three reported moderate-severe pain. For two participants, the pain was also new and unusual, but the pain was familiar for the other. Thus criterion 3 was modified to 'recent acute onset of new and unusual moderate to severe neck pain and headache.'

Criterion 5: focal neurological signs
Three healthy participants and one with neck pain reported mild balance/gait disturbance; one had a vestibular problem, one participant had postural hypotension, the other had a two-year history of gait disturbance. There were no reports of speech deficits or hoarse voice apart from that associated with colds, and one participant reported loss of strength in hands on waking. All were long standing signs. Criterion 5 was therefore modified to 'new focal neurological signs in the last five weeks.' On physical examination all participants were able to perform tandem stance with eyes open, 13 had difficulty with eyes closed; four of whom were over 55 years and one also had difficulty with tandem walk. Four of these participants reported problems with balance due to known conditions as stated above. There were no other coordination difficulties observed. No participants had a Horner's Syndrome.

Neurological symptoms
While focal signs were rare in either group, participants did report some neurological symptoms; Nine had visual disturbance (7 of these had diagnosed migraine, 1 possible migraine) and three participants with neck pain reported intermittent paresthesia in the arms/hands, feet, or face.

Reported modified CeAD criteria
Following analysis of the initial criteria results and information gained from the follow-up questioning, the criterion definitions were modified accordingly (bold and italic text below) and the numbers of participants who reported any of the modified criteria were recorded and presented in Table 1: 1) age under 55 years; 2) recent minor trauma or strain to head or neck or recent infection; 3) acute onset of unusual or unfamiliar, moderate to severe headache or neck pain recent onset new dizziness; 4) new focal neurological signs in the last 5 weeks; and 5) The new criteria reduced the number of positive responses to acute neck pain, recent dizziness, and focal neurological signs.
Three participants with neck pain and one healthy had three initial criteria (1 + 2 + 3) but with the modified criteria this reduced to two with neck pain and no healthy participants (Table 1). No participants in either group fulfilled the criteria for the other combinations. Using the definition of the presence of three or more criteria, the specificity of initial tool (96.67%), improved to 98.33% when modifications were made.

Sensitivity of modified criteria
As an additional step, the initial and modified criteria were then re-administered using data on 37 CeAD cases that had formed the basis of the initial criteria (Thomas, Rivett, Attia, and Levi, 2015). This was done as the criteria modifications, whilst improving specificity of the tool, may have now decreased sensitivity, perhaps requiring further modifications of the tool. The number of initial and modified criteria per participant and combinations of criteria were described for this group. Further, to determine the potential diagnostic ability of the modified criteria to predict the presence or absence of CeAD, analysis was undertaken using a cut-off score of ≥ three criteria. Analysis was performed using an online diagnostic test evaluation calculator (MedCalc 2021, Ostend, Belgium) to estimate sensitivity, specificity, and accuracy of the tool.
Using the initial criteria, 73% of those with CeAD reported acute onset of pain, though this reduced to 51% when criteria were modified as severity was either mild or unreported and unusual pain was not always recorded. Reported minor trauma was like those with neck pain. A greater proportion of the CeAD group reported recent dizziness (38%) and focal neurological signs (86%) and these were unchanged with the criteria modification. Table 2 presents the data for the initial and modified criteria for the CeAD group. *n (%) presenting with initial and modified CAD criteria (* data obtained from Thomas, Rivett, Attia, and Levi, 2015).
Considering that most participants in the healthy and/or neck pain groups had less than three CeAD modified criteria and the majority of the CeAD cases had more than three (Table 3); a cutoff of ≥ 3 criteria was then used to explore the diagnostic characteristics of the modified tool (Table 4).

Discussion
The study aimed to test a simple, quick diagnostic support tool for CeAD in healthy people and those with neck pain and/or headache to ensure the included criteria could discriminate potential features of CeAD as accurately as possible. The study identified that strict definitions are needed to reduce the potential for falsepositive findings, so that patients are not referred for unnecessary medical investigations but also not denied treatment for benign conditions. Clear definitions will also assist clinicians in being alert for specific features in people presenting for treatment of acute neck pain and/ or headache and to be confident to proceed with assessment if these are absent.

How many healthy people have criteria for CeAD?
The majority of healthy people and those with benign neck pain or headache had ≤ 2 criteria; most commonly age and recent trauma, so these criteria on their own may not be particularly helpful to discriminate CeAD. However, age under 55 years is reflective of people seeking physiotherapy care who may attend following recent trauma (Beliveau et al., 2017;Carter and Rizzo, 2007).

What modifications to criteria were needed?
The main refinement needed following initial testing was to criterion 3 (acute or unusual onset of pain) to acute onset of unusual or unfamiliar, moderate to severe pain. Previous studies have reported pain descriptors in people with CeAD as 'pain worse than ever experienced before,' or 'pain like no other' (Arnold et al., 2007). While a few participants in the current study reported acute onset of pain, on further questioning this was not usually new or unfamiliar and for most was not moderate or severe. Importantly, those with acute onset of moderate to severe pain were often migraine sufferers so the pain was not new and unfamiliar. The modification to the criteria should therefore select out those with a dramatic change in symptoms. Interestingly, the modification to criterion 3 reduced the percentage of cases in the CeAD group, as not all had moderate or severe pain, although the pain was new or unfamiliar (Table 2). So, in the future separating this into two criteria: 1) acute onset moderate to severe; and 2) new and unusual pain) may be necessary to further improve the tool and improve specificity for neck pain. Recent dizziness was reported by a few participants in both non-CeAD groups (Table 1) but again was longstanding, so changing the definition to recent new dizziness removed all participants from this criterion. Similarly, focal neurological signs were reported by a few participants, but again all these were longstanding and in all, but one, could be explained by benign causes. Adding new dizziness to criterion 5 then excluded these participants. Therefore, the presence of neurological symptoms might also be important, though not strictly part of the criterion, as they may be precursors to focal signs (Morris, Merkler, Gialdini, and Kamel, 2017;Nadarajan, Perry, Johnson, and Werring, 2014). The presence of 'new' neurological symptoms might therefore also be an important factor to consider when questioning about focal neurological signs and highlight people in whom more careful monitoring and follow-up is required. As dizziness may be a neurological symptom criteria 4 and 5 could be combined into 'new neurological signs or symptoms.'

What combinations of criteria should alert clinicians to the need for medical investigation or closer follow-up?
At this stage, the presence of three or more of the modified criteria (e.g., 1 + 2 + 3) should raise suspicion and alert the clinician to investigate more closely. Specific follow-up questioning and/or testing (e.g., for neurological signs) will be required to determine if the patient then still meets the modified criteria or if benign  Preliminary exploration of the diagnostic ability of the tool using ≥ 3 of the modified criteria as the cut-off for the potential presence of CeAD appears promising, with high sensitivity and specificity values. Falsenegative findings were identified in 7 CeAD participants who had only 2 criteria; however, 5/7 had several neurological signs and 2/7 had acute pain, so some weighting of criteria is likely to be needed in the final tool to improve sensitivity. False-positive findings may also be reduced with further modification and weighting of the criteria.

Strengths and limitations
We were able to examine the frequency of the CeAD criteria in a large sample of healthy individuals or people with neck pain. The sample of people with neck pain and/or headache was modest but was sufficient to identify important modifications to criteria definitions to better distinguish a positive criterion. One limitation of the tool currently is that people with migraine have the potential to fulfill modified criteria 3, 4 and 5 particularly in the early phases of migraine presentation and also during transitional stages of the disorder where the signs and symptoms might change (Kelman, 2006). Migraine is also a risk factor for CeAD, so it will be important to determine if additional refinements to the criteria are necessary to distinguish presenting signs and symptoms from a migraine episode. The modified criteria, with the additional suggestions above to improve sensitivity of the tool, should be now tested in people with migraine in an attempt to gain the highest level of specificity of the tool. This will be important prior to the next stage of development, to test the tool in an acute setting, such as an emergency department where the potential for CeAD is higher. This is important to avoid identifying false positives who have a benign disorder but is equally critical to eliminate false negatives and ensure that high-risk cases are not missed. Testing in an emergency department setting on people with acute neck pain or headache, some of whom may present with early CeAD will enable the calculation of risk scores and weighting of individual criteria to increase the specificity of the tool.
In an acute setting, physiotherapists need to accurately identify those in whom prompt further investigation is required, but also which patients should be followed up more closely to monitor for deterioration of symptoms. Therefore, the ability of the tool to screen out those with a high likelihood of more serious pathology is paramount. Conversely, the tool should also allow physiotherapists who have adequately explored these warning signs of more serious pathology (i.e., the 5 criteria) and not identified anything of concern to be reassured that there are no evident barriers for musculoskeletal management. However, further research is needed to determine those at greater risk of adverse outcomes following cervical musculoskeletal management.

Conclusions
The modified CeAD tool appears to have potential for high diagnostic accuracy (Supplementary File). As it currently stands, people presenting with three or more of the modified CeAD criteria should be investigated more closely. Clinicians need to be accurate with definitions to avoid incorrectly identifying positive criteria to avoid unwarranted investigation of benign features. The criteria may need to be examined in people with migraine and suitable further modification made before being formally tested on people in an emergency department setting to calculate appropriately weighted risk scores. Ultimately, this could form the basis of a simple and effective screening tool for CeAD.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
The author(s) reported that there is no funding associated with the work featured in this article.