A systematic review using the DiTA (Diagnostic Test Accuracy) database was performed. DiTA database is a comprehensive index, which is updated monthly using automated searches, of diagnostic test accuracy studies developed specifically for the discipline of physical therapy.16,17 It is a sister database of the PEDro (Physiotherapy Evidence Database) which includes intervention studies within the physical therapy domain. We registered our protocol on the UTS (University of Technology Sydney) data repository, and it can be requested from the corresponding author.

The search was conducted on the 22 November 2021 using the search terms: Upper Limb Tension Test and associated synonyms, like ULTT, ULNT, Brachial Plexus Tension, Elvey Test, Upper limb nerve tension test, Peripheral neuropathic pain, Cervical radiculopathy, Carpal tunnel syndrome, Cubital tunnel syndrome.

The inclusion criteria consisted of a) adult participants (18 years or over) who present with signs and symptoms that indicate upper limb pathology, b) evaluated sensitivity and/or specificity of the ULTT in diagnosing upper limb pathology by comparing the ULTT to any reference test, c) primary diagnostic test accuracy studies. We excluded studies using non-human subjects or cadavers and where the ULTT was used in combination with other special tests for diagnosis. The eligibility of each study, identified by the search or included in (systematic) reviews, was determined by two out of five independent assessors (GB, CB, CA, HD, SF). Discrepancies were resolved by consensus or by a third independent investigator (AV).

The QUADAS-2 is the recommended tool to assess risk of bias (RoB) in diagnostic test accuracy studies.18 As this tool is rather difficult to administer for clinicians/researchers, we simplified it to six criteria (Table 1).19 Each criterion can be scored yes/no/unclear. Two out of five assessors (GB, CB, CA, HD, SF) independently assessed each study and any discrepancies were discussed and resolved by consensus. If a consensus could not be reached, a third independent investigator (AV) made a final decision.

Two of five assessors (GB, CB, CA, HD, SF) extracted the following data from each of the studies: author(s) and year of publication, participant details (number, mean age and range, sex, clinical characteristics), examination details (clinical setting, examiner profession and expertise), reference standard, test procedures, and criteria for positive test result, e.g. pain, range of motion (either actively or passively performed). Discrepancies and a data check was done by two review authors (HB, AV).

We considered a specificity or sensitivity value high enough to be clinically useful in ruling in or out a condition, according to the SpPIn (specificity high and positive test rules in a condition) and SnNOut (sensitivity high and negative test rules out a condition) rules. As most diagnostic test accuracy studies are performed in highly specific populations, generalising their accuracy to the public requires sensitivity and specificity to be sufficiently high, with recommended cut-offs between 0.85 and 0.90–0.95 being used.20.21 We conducted a frequency analysis of the number of papers outlining the test protocol used for the ULTT to determine the common denominator of those procedures. Similarly, a descriptive analysis was completed to assess the positive diagnostic criteria.

From the initial search in the DiTA database, 22 original studies were retrieved, in addition to two systematic reviews10,11 and three narrative reviews.3,22,23 After reviewing the primary research papers included in these reviews and removing duplicates, we retrieved an additional 5 references, resulting in 27 papers being available for full-text review (Fig. 1). After the application of the selection criteria, we excluded three papers as these were narrative reviews,3,22,23 three papers did not evaluate the ULTT,24-26 and one was in a population with low back pain.27 Finally, we included 10 papers reporting on 9 studies.

Fig. 1.

PRISMA flow diagram.

(0.2MB).

Participants. Four studies aimed at diagnosing CR (386 participants),14,28-30 and four studies aimed at diagnosing CTS in 295 participants.31-34 One study (in two publications) included both patients with CR and CTS (Table 2).35,36 Individuals with CR were all referred to a specialized clinic or a neurosurgery department. Three out of the four studies on patients with CR specifically stated they included consecutive patients. The average age of the participants varied between 43.2 and 54.3 years and most of the participants were female (varying between 49 and 83%).

Index test. The most evaluated test was ULTT1; five studies only evaluated the ULTT1, two studies evaluated all four ULTT tests and provided accuracy data for all tests separately,14,29 one study did not specify which ULTT was evaluated,28 and one study evaluated the ULTT-A (which was like the ULTT1) and ULTT-B (an alternative to ULTT1).35

Reference test. Nerve conduction study was the most common reference test mainly in the studies investigating CTS,31-35 followed by clinical presentation and magnetic resonance imaging (MRI), in the studies investigating CR.14,29,30 The CR studies reported history taking and physical examination to be part of the reference test, and in two (out of four) studies the reference test was performed by one specialist, while none of the CTS studies did so.

Accuracy data. All studies provided data on the accuracy of the ULTT, and eight studies reported data for the ULTT1. Based on the sparsity of data related to other ULTT tests, we only report outcomes for the ULTT1. Sensitivity of the four test procedures for the ULTT1 for CR varied between 0.35 and 0.83, and the specificity varied between 0.40 and 0.76, meaning none met the lowest cut-off mentioned in literature of 0.85. For CTS studies, which reported on eight test procedures, reported sensitivity was between 0.06 and 0.93 and specificity between 0.10 and 0.93. In three procedures the accuracy data met at least the 0.85 cut-off, but not the 0.95.

Risk of bias. The RoB varied from 2 to 6 criteria scored positive (Table 3).

We found a wide variation with a total of 13 different ULTT1 procedures (Table 4). One study did not report the procedure and just provided a reference.31,37 Three other studies provided the procedure and mentioned a reference of the origin of the procedure.29,31,34 All papers mentioned between two and seven different steps in the procedure, unfortunately, no two studies performed the exact same procedure except Vanti et al.33,34 For one study we presented the procedure for the ULTT-A, as this one is most like the ULTT1.35

In six studies, shoulder depression (which includes shoulder girdle/scapula depression/fixation/stabilisation) is listed as the starting position. One study did not mention shoulder depression and started with the shoulder in abduction,28 one study started with a contralateral flexion of the neck,31 and one study combined most procedures that are separate steps in all other studies into one starting position and added just one extra step (elbow extension).32 Six studies ended the procedure with the last step being ‘contralateral/ipsilateral cervical lateral flexion (bending)’, while one study started with this procedure and released it at the last step.31 Two studies did not mention cervical lateral flexion at all.28,32

All studies clearly stated their positive diagnostic criteria, although the criteria between studies differed (Table 5). In most studies, multiple symptoms were required for a test to be regarded positive, either in combination or as an ‘either/or’. Three studies evaluated more than one set of positive diagnostic criteria.31,33,34 ‘Reproducible patient symptoms’ and ‘symptoms decrease when relaxing the stress on the nerves’ were the most used criteria (n = 9 and n = 6, respectively). With ‘symptoms decrease when relaxing the stress on the nerves’ (or increased symptoms with increased stretching) authors mean relaxing the stretch on the arm; ‘structural differentiation’ means that symptoms were evaluated to either increase or decrease when the cervical spine was laterally flexed. Three studies used a combination of criteria (Wainner's criteria: “reproducible patient symptoms OR limited range of motion of the elbow OR decrease of symptoms when relaxing”) as defined by Wainer et al.32,34,35 The criterion ‘symptoms in the first 3 digits’ was only mentioned in three studies including patients with CTS.31,33,34

We found nine diagnostic test accuracy studies of which four included patients with CR, four studies patients with CTS, and one included both patients. The most evaluated test was the ULTT1, with in total 13 test procedures described in eight studies. We found a wide variety of test procedures and criteria for a positive test. Based on our findings, we propose a more standardised test procedure of the ULTT1 in patients with CR as well as CTS. The associated positive diagnostic criteria are reproduction of the symptoms and an increase or decrease of symptoms when stressing or relaxing the nerves.

Previously, several reviews evaluating the ULTT concluded that there is a lack of a clear definition of terms (‘investigators definition of a positive test’) and procedures for the ULTT1.3,10,23 The authors all concluded that the validity of ULTT1 is probably hampered by the diversity in the procedure and interpretation of the index test. The major online support tools also use a variety of procedures in their videos, which reflects the variability in test procedures found in the literature and hampers clear implementation.2,12,13 By evaluating the test procedures of the ULTT1 we found that the order of movements varied between studies. Based on the included studies, we suggest a more standardised test procedure and positive diagnostic criteria. Hopefully in future research the diagnostic accuracy of the ULTT1 can be more accurately assessed through a standardised procedure and positive diagnostic criteria.

Overall, all studies stabilised the shoulder in abduction, extended the elbow, supinated the forearm, extended the wrist, and performed structural differentiation by side bending of the neck, although the order of movements varied. We propose a more standardised test procedure to consist of these movements. The suggested order is like the one performed in the studies with the lowest risk of bias (score of 6): 1) stabilising the shoulder in abduction, 2) extending the wrist/fingers, 3) supinating the forearm, 4) externally rotate the shoulder, 5) extending the elbow, and 6) structural differentiation by laterally flexing (i.e., side bending) the neck. This proposed test procedure aims to assess whether the patient's symptoms are reproduced and whether symptoms increase or decrease when increasing or decreasing the tension on the nerves. When symptoms are present in the first three digits during the ULTT, this acts as an extra diagnostic criteria for patients with CTS, and can be regarded as ‘reproduction of symptoms’, same for ‘limited range of motion’.

To our knowledge, this is the first systematic review evaluating the test procedures and positive diagnostic criteria of the ULTT1 in patients with neck or arm pain. A possible limitation of our review may exist in the search strategy, which was conducted solely in the DiTA database. DiTA is updated monthly, drawing on automated optimised searches of MEDLINE, EMBASE, CINAHL, and the Cochrane Database of Systematic Reviews.38 Therefore, we consider the DiTA database up to date and when we performed a search in PubMed (January 2022) we did not find any additional or missed studies. Another limitation may be present in the study selection, as patients selected in the included studies are specific patients, referred to specialised clinics. Although this finding probably does not influence test procedures and positive diagnostic criteria, it might influence the accuracy data as these are dependent on prevalence of the condition.39 Concerning the reference test, in the CR studies it included physical examination in three of the four studies, potentially leading to risk of confirmation bias that might result in higher accuracy data. In two CR studies it was also unclear whether the index and reference test were performed independently. Our findings, however, did not show higher accuracy data in the CR studies, therefore we believe that confirmation bias might not have played an important role. Another limitation could be our risk of bias tool. We used a simpler 6 item ‘checklist’ for risk of bias assessment, to allow clinicians and students to conduct risk of bias assessments more easily. We are confident that the outcomes of the risk of bias assessment will not differ much from the QUADAS-2. We registered our protocol on the UTS data repository only, as we were unable to register the protocol in Prospero which focusses on systematic reviews with a patient related/relevant outcome. Lastly, due to the limited data, we were only able to make recommendations about the ULTT1, and not the ULTT2 (a and b) and ULTT3.

Clinical implications. We proposed a standardised test procedure and positive diagnostic criteria for use in clinical practice. A more standardised set of test procedures will help clinicians as it is a consistent message to patients. Reproduction of symptoms and a decrease of symptoms when reducing the tension are the recommended criteria for a positive test.

Research implications. We suggest that future research should validate our proposed set of test procedures as well as our proposed set of positive diagnostic criteria. This may lead to greater consistency and standardisation between studies, reducing heterogeneity regarding treatment effectiveness studies, systematic reviews, and meta-analysis of test accuracy studies.