Version of Record: https://www.sciencedirect.

com/science/article/pii/S2468781221001119
Manuscript_f93af0ee0eef47d3763d3d4cc4cc0bc5

Diagnostic accuracy of upper limb neurodynamic tests in the diagnosis of cervical

radiculopathy

Francis Grondin1,2, PhD-candidate, PT, Chad Cook3, PhD, PT, FAPTA, Toby Hall4, PhD, PT,
Olivier Maillard5, MSc, PharmD, Yannick Perdrix6, MSc, PT, 1Sebastien Freppel1, PhD, MD.

1
Neurosurgey Department, Centre Hospitalier Universitaire de La Réunion, Reunion, France.
2
School of Physiotherapy (IFMK), Centre Hospitalier Universitaire de La Réunion, Reunion,
France.
3
Professor. Duke Department of Orthopaedics, Duke Clinical Research Institute, Duke
University, 311 Trent Drive, Durham, NC, USA. chad.cook@duke.edu
4
Adjunct Associate Professor, School of Physiotherapy and Exercise Science, Curtin
University, Kent Street, Bentley, Perth, Australia. halltm@netspace.net.au
5
Centre Hospitalier Universitaire de La Réunion, INSERM, CIC1410, 97410, Saint Pierre,
France. olivier.maillard@chu-reunion.fr
6
Educational Manager, School of Physiotherapy (IFMK), Centre Hospitalier Universitaire de
la Réunion, Reunion, France. yannick.perdrix@ies-reunion.fr

Declaration of Conflict of Interests

The authors declare no conflict of interest for this article.

Corresponding Author
grondin.fran@yahoo.fr

© 2021 published by Elsevier. This manuscript is made available under the Elsevier user license
https://www.elsevier.com/open-access/userlicense/1.0/
 1 Introduction

2 Cervical radiculopathy (CR) is a relatively common neurological disorder caused by

3 mechanical compression from a disc or other space-occupying lesion or from inflammation to

4 the nerve root (Anekstein et al, 2012) The annual incidence of CR is 107.3 per 100,000 for

5 men and 63.5 per 100,000 for women (Radhakrishnan et al, 1990). The clinical manifestations

6 of cervical radiculopathy may include pain, sensory deficits, motor deficits, diminished

7 reflexes, or combinations of these. Cervical radiculopathy typically is self-limiting with 75%–

8 90% of patients achieving symptomatic improvement or resolution within a year with

9 conservative care (Woods et al, 2015)

10 Although no gold standard exists as a reference standard for cervical radiculopathy,

11 magnetic resonance imaging (MRI) is the preferred diagnostic method (Mink et al, 2003),

12 since it can differentiate tumors, inflammation, visualize trauma, and the extensiveness of

13 disc, arthritic, neural, and vascular cervical pathologies. Electrodiagnostic tests are capable of

14 detecting clinically significant problems in many patients as well, although they are operator

15 dependent and variable methods and normative values are used in practice (Reza Soltani et al,

16 2014). Furthermore it may be negative if performed before denervation has occurred or when

17 re-innervation is complete (Ashkan et al, 2002). Cervical radiculopathy is considered a

18 ‘clinical diagnosis with imaging confirmation’, and it is important to match valid clinical

19 signs with MRI findings and/or electrodiagnostic test results (Carette and Fehlings, 2005 ;

20 Kuijper et al, 2009).

21 There are numerous clinical tests used to diagnose cervical radiculopathy. Upper Limb

22 neurodynamic tests ((ULNT) 1, 2a, 2b and 3), or also called upper limb tension test (ULTT),

23 initially described by Elvey (Elvey, 1986), Butler (Butler, 2000) and Shacklock (Shacklock

24 1996), involve targeted sequences of movement that provoke mechanosensitivity of the nerve.

25 The tests are performed by placing and releasing progressively more tension on the proposed
26 component of the nervous system that is being tested. A recent systematic review (Thoomes

27 et al, 2017) concluded that “limited evidence for accuracy of physical examination tests for

28 the diagnosis of CR” exists.

29 Moreover, neurodynamic test procedures in studies that populated the aforementioned

30 systematic review used a variety of testing methods and results to determine a positive

31 finding. Three criteria has been advocated when testing: 1) reproduction of neurogenic pain-

32 burning or lightning-like pain, tingling sensation in the neck and arm (Apelby-Albrecht et al,

33 2012), the patient’s symptoms reproduced (Wainner et al, 2003), or reproduction of pain

34 (Ghasemi et al, 2013); 2) side to side range of motion difference (Wainner et al, 2003) or

35 side-to-side difference in painful radiation (Apelby-Albrecht et al, 2012); and 3)

36 increased/decreased symptoms with structural differentiation (Apelby-Albrecht et al, 2012) or

37 cervical structural differentiation with cervical spine movement alone (Wainner et al, 2003).

38 Interestingly, there is inconsistency in what is advocated to measure including conflicting

39 evidence for reproduction of any pain or discomfort (Apelby-Albrecht et al, 2012 ; Ghasemi

40 et al, 2013); and side to side range of motion comparisons (Nee et al, 2012). Most studies

41 advocate the use of structural differentiation, which involves directional movement of a

42 defined body region (e.g., neck side flexion) away from the area assessed to evaluate the

43 effect of mechanical force on the nervous system and its impact on the patient’s symptoms.

44 In clinical practice, many clinicians have assessed the accuracy of neurodynamic tests

45 on two criteria as recommended by Nee et al (Nee et al, 2012) : 1) familiar patient’s

46 pain/symptoms reproduced and 2) increased/decreased symptoms with structural

47 differentiation. Since we are unfamiliar with any studies that have included both findings in

48 the assessment of CR, we investigated the accuracy of four ULNTs in comparison against a

49 reference standard of medical history and MRI confirmation in patients with and without CR.

50 We hypothesized that the findings may provide insight on the role of ULNTs (e.g., screening
51 or confirmation) and that the more rigid definition of a positive test should improve the

52 specificity of the test findings. Further, combinations of test findings should result in more

53 diagnostic accuracy than individual tests alone.

54

55 Materials and Methods

56 The study was a diagnostic accuracy study (prospective) design in which clinical

57 testing occurred in a state of diagnostic uncertainty. The study followed the updated 2015

58 STARD reporting standards (Bossuyt et al, 2015). Patients were informed about the study and

59 they gave their consent for participation before inclusion. The study was conducted in

60 accordance with the Ethical principles and the Helsinki Declaration on research involving

61 human subjects and was approved the French regulatory and ethics rules (n°2212189v0).

62 Participants were recruited from consecutive patients referred to a Neurosurgery

63 Department by a general practitioner or specialist from September 2017 to September 2019.

64 Each patient had a suspected neck disorder. Referred patients provided information and

65 questionnaires about pain intensity and neck disability. To be included patients had to be aged

66 18 to 65 years, reporting arm pain with or without neck pain of at least 3-months in duration.

67 In addition, they were required to have a self-reported pain score of at least 30mm and less

68 than 80 on a 100 mm visual analogue scale (VAS) (Horn et al, 2016) during the previous 24

69 hours, and had a self-reported score of at least 20% on the Neck Disability Index

70 questionnaire (NDI) (Masaracchio et al, 2013).

71 Subjects were excluded if they were unable to understand French, had suffered from a

72 significant neck trauma at the time of the study (i.e., recent cranio-cervical trauma including

73 cervical spine fracture), had a history of neck or arm surgery, inflammatory joint

74 condition/arthritis, fibromyalgia, diabetes, pregnancy, cardiovascular, neurological, neoplastic

75 or psychiatric pathology, cervical myelopathy, pyramidal or extrapyramidal pathology.

 76

77 Reference Standard: The diagnosis of CR or a competing diagnosis was made by a

78 single neurosurgeon with 15 years of experience from the consecutive patients included.

79 Cervical radiculopathy is a clinical diagnosis that is confirmed through imaging verification,

80 thus the diagnosis of CR was based on the following criteria: 1) history and presence of

81 dermatomal radicular pain and/or symptoms (dysesthesia, muscle weakness or altered

82 reflexes) attributable to a CR and 2) presence of MRI findings. MRI findings were specific

83 in their confirmation of nerve root compression or irritation by disc herniation or stenosis in

84 pre- or intra-foraminal space narrowing on the ipsilateral side and at the same or adjacent

85 level of radicular pain (Kuijper et al, 2008). The reference standard results were interpreted

86 without knowledge of the results of ULNT.

87 Index tests : Approximately one hour after the reference standard was provided by the

88 neurosurgeon, a single physiotherapist with 10 years of experience in neck pain management,

89 and advanced certification for orthopedic assessment evaluated the ULNT on each participant.

90 No intervention was allowed between the index test(s) and reference standard. The

91 physiotherapist was blind to the patient history, clinical/MRI findings and the diagnosis. The

92 index test results were interpreted without knowledge of the results of the reference standard

93 and the presence of CR. Before the tests, patients were instructed to communicate the onset

94 of any sensation such as stretch, tingling or pain anywhere in the arm or neck (Schmid et al,

95 2009). The patient was positioned supine without a pillow (Walsh 2005). The examiner

96 performed the ULNTs for that are purported for the median (ULNT1 and ULNT2a), radial

97 (ULNT2b) and ulnar (ULNT3) nerves (Figure 1) in randomized order using randomization

98 software. Upper limb neurodynamic testing was operated according to the standardized

99 sequence previously described (Butler, 2000; Nee et al, 2012 ; Schmid et al, 2009), with a 5-

100 minute break between each test to avoid any pain sensitization by repeating tests (Walsh
101 2005). Passive movements were achieved to the end of range or until symptoms were

102 produced (Schmid et al, 2009). The non-symptomatic side was tested first for each ULNT for

103 familiarization with sensation/pain induced by tests. A ULNT was considered as positive if

104 both of the two following criteria were met:

105 - Reproduction of a familiar symptomatic complaint of arm pain and/or neck pain at

106 least partially (pain or dysesthesia including burning, or lightning-like pain, or

107 tingling sensation) (Nee et al, 2012);

108 - Structural differentiation: Once such a familiar complaint was provoked, structural

109 differentiation between neurogenic and non-neurogenic sources was performed by

110 the addition of sensitizing movements at a site distant to the pain: ipsilateral- or

111 contralateral cervical lateral flexion, elbow or wrist extension/flexion, or shoulder

112 girdle elevation (Appendix A) (Nee et al, 2012);

113 Tests were considered negative if each failed to meet the positive criteria identified above or

114 indeterminate if the patient was unable to tolerate the test of position to allow complete

115 execution of the test.

116

117

118

119

120

121

122

123 Figure 1. Flow chart

124

125

126 Results

127

128 Statistical analysis were carried using SPSS (IBM SPSS, Version 26.0. IBM Corp.

129 Armonk, NY). Variables normality was tested with Shapiro-Wilk test for continuous data.

130 Nonparametric continuous variables were described as medians and interquartile ranges and

131 differences were tested using the Wilcoxon Mann Whitney test. Gaussian variables were

132 described with means and standard deviations and differences were tested using the Student’s

133 t-test. Categorical variables were described as numbers and percentages. They were compared

134 using Chi square test or Fisher’s exact test, as appropriate. P-value significance level was set

135 at .05 and all tests were bilateral. Two by two tables were created for each ULNT measure.

136 ULNT test performances were analyzed using calculations for sensitivity, specificity, positive

137 likelihood ratios (LR+), and negative likelihood ratios (LR-). Sensitivity is the percentage of

138 people whose test is positive for a specific disease among a group of people who have the

139 disease (Cook et al, 2020). Specificity is the percentage of people whose test is negative for a
140 specific disease among a group of people who do not have the disease [22]. LR+ are the

141 probability of a person with the disease testing positive divided by the probability of a person

142 without the disease testing positive (Cook et al, 2020). LR- are the probability of a patient

143 who has the condition of testing negative divided by the probability of a patient without the

144 disease, testing negative (Cook et al, 2020).

145 We also calculated pretest probability, which is the probability of the condition being

146 present before the diagnostic result is known and is sample specific for those enrolled in our

147 study, and post-test probability with a positive and a negative finding on the ULNTs. Post-test

148 probability is the percentage chance of the condition being present after a positive or negative

149 finding for a ULNT. Generally, a positive test will increase the post-test probability of

150 diagnosing the condition (otherwise known as ruling in the diagnosis). In contrast, a negative

151 finding will generally decrease the post-test probability of diagnosing the condition

152 (otherwise known as ruling out the condition) (Cook et al, 2020).

153 We calculated sensitivity/specificity, LR+/LR-, and post-test probabilities with a

154 positive and a negative finding for the four individual ULNT tests and combinations of these

155 tests. When calculating combinations of findings, the clusters of tests were placed in

156 “conditions” (e.g., 1 of 4 is positive, 2 of 4 is positive, etc.) and evaluated for their abilities to

157 influence post-test probability change with each defined condition. For all analyses, we also

158 evaluated post-test probability change, which is the difference between the pre-test prevalence

159 and the post-test finding with a positive or a negative result (Cook et al, 2020). Since the

160 purpose of a test is to change the post-test probability of an accurate diagnosis, larger post-test

161 probability changes were considered to have the highest clinical utility. 95% confidence

162 intervals (95%CI) were calculated for all of these features.

163

164 Between September 2016 to December 2018, 85 participants, from 109 individuals
165 who were screened, were enrolled in the study. Of the 85 participants, 27 (31.7%) were

166 diagnosed with CR, 42 with neck and non-radiculopathic arm pain, 12 with peripheral nerve

167 entrapment, and 4 with diffuse shoulder pain (Table 1). All participants received the same

168 reference standard and were included for analysis (Figure 1). Diagnostic accuracy of the

169 four individual ULNTs are presented in Table 2. All four of the tests were more specific,

170 than sensitive, with the ULNT3 demonstrating the highest specificity. None of the four tests

171 markedly influenced post-test probability with a positive or a negative finding, with post-test

172 probability changes from baseline prevalence ranging from 41.58% with a positive for

173 ULNT3 to 15.72% with a negative for ULNT 2a.

174

175 Table 1: Baseline characteristics of the subjects (n=85)

Cervical Non-radiculopathic

radiculopathy arm pain

Age* 43.96 (8.94) 45.27 (9.74) p = 0.61

Height (m) 1.66 (0.09) 1.67 (0.08) p = 0.54

Body Mass Index * 24.73 (3.91) 24.88 (4.97) p = 0.74

Duration (self-report) in months * 93.25 (98.41) 70.51 (62.31) p = 0.69

Visual Analogue Scale for Pain* 5.14 (1.58) 5.03 (1.53) p = 0.73

Neck Disability Index (%) 38.16 (14.14) 43.07 (13.90) p= 0.19

176
177
178
179 Table 2: Diagnostic Accuracy of Individual Upper Limb Neurodynamic Tests. Pre-test
180 Prevalence = 31.7%.
Sensitivity Specificity LR+ (95% LR- (95% Post-test Post-test
CI) CI) Probability Probability with
 with a Positive a Negative
Finding (95% Finding (95%
CI) CI)
ULNT 1 59.26 (38.80, 75.86 (62.83, 2.46 (1.41- 0.54 (0.33- 53.30 (39.55- 20.04 (13.28-
77.61) 86.13) 4.27) 0.87) 66.46) 28.76)

ULNT 2a 70.37 (49.82, 72.41 (59.10, 2.55 (1.57- 0.41 (0.22- 54.20 (42.15- 15.98 (9.26-
86.25) 83.34) 4.14) 0.75) 65.77) 25.82)

ULNT 2b 55.56 (35.33, 75.86 (62.83, 2.30 (1.30- 0.59 (0.38- 51.63 (37.63- 21.49 (14.99-
74.52) 86.13) 4.06) 0.92) 65.33) 29.92)

ULNT 3 40.74 (22.39, 93.10 (83.27, 5.91 (2.07, 0.64 (0.46- 73.28 (48.99- 22.95 (17.59-
61.20) 98.09) 16.87) 0.88) 88.65) 28.99)
181
182
183 Diagnostic accuracy of test conditions for combinations of ULNTs are presented in

184 Table 3. Characteristically, with lower conditions (e.g., 1 of 4 is positive) values exhibit

185 high sensitivity and low specificity, whereas higher conditions (4 of 4 are positive) values

186 exhibit low sensitivity and high specificity. As expected, the condition of 1 out of 4 ULNT

187 tests positive was the most sensitive combination whereas the condition of 4 out of 4 ULNT

188 tests was the most specific. The condition of 1 out of 4 tests positive has the ability to “rule

189 out” CR (LR-=0.08), exhibiting a post-test probability change of 28.12% with a negative

190 finding. The condition of 4 of 4 tests positive had an infinite LR+ but there were only 3

191 cases in which all four tests were positive. The condition of 3 of 4 tests positive occurred in

192 12 of the 27 patients with CR and provided a LR+ of 12.89 and a post-test probability of

193 85.71 (post-test probability change of 54.01%). No adverse events from performing the

194 index test or the reference standard were observed.

195
196
197
198
199
200 Table 3. Diagnostic Accuracy of Clustered Upper Limb Neurodynamic Test findings
201 (Conditions). Pre-test Prevalence = 31.7%.
202
Sensitivity Specificity LR + (95% LR- (95% CI) Post-test Post-test
Probability Probability with
 CI) with a Positive a Negative
Finding (95% Finding (95%
CI) CI)
1 of 4 96.30 (81.03, 46.55 1.80 (1.40, 0.08 (0.01, 45.51 (39.38- 3.58 (0.46-20.62)
Positive 99.91) (33.34, 60.13) 2.32) 0.56) 51.84)
2 of 4 85.19 (66.27, 74.14 (60.96, 3.29 (2.07, 0.20 (0.08, 60.42 (48.99- 8.49 (3.59-18.83)
Positive 95.81) 84.74) 5.23) 0.50) 70.82)
3 of 4 44.44 (25.48- 96.55 (88.09- 12.89 (3.10- 0.58 (0.41, 85,71 (59.06- 23.237 (18.57-
Positive 64.67) 99.58) 53.62) 0.81) 96.14) 28.82)
4 of 4 11.11 (2.35, 100.00 (93.84, Inf. 0.89 (0.78. 100 29.23 (26.58-
Positive 29.16) 100.00) 1.02) 32.13)
203

204

205 Discussion
206
207
208 This study sought to determine the diagnostic accuracy of four ULNTs in identifying

209 CR in comparison with a reference standard of clinical diagnosis with MRI confirmation.

210 The study was performed in a situation of diagnostic uncertainty and used a more rigid

211 definition of what constitutes a positive test compared to previous studies [12-14]; the tests

212 also more closely matched how the tests are used in clinical practice. Findings were that

213 ULNTs when used in isolation did not lead to acceptable LR-, LR+ or post-test probability.

214 However, 3 out of 4 tests positive can rule in CR with a LR+ of 12.89. One of four positive

215 tests provided a LR- of 0.08 indicating that CR can be ruled out if no tests are positive. Of

216 the four tests, the ULNT3 influenced post-test probability the most with a positive test

217 (73.28%), whereas the ULNT2a influenced post-test probability the most with a negative

218 test (15.98%).

219

220 Each ULNT provided stronger LR+ values than LR-, thus influencing post-test probability

221 with a positive finding more notability than a negative finding. Our findings are markedly

222 different than those from Wainner et al. who found very low values of LR+ (<1.3) -

223 suggesting that they did not rule in - and moderately low LR- values (>0.12) - suggesting
224 they are better for ruling out (Wainner et al, 2003). Ghasemi and colleagues [14] failed to

225 report a LR+ (or a LR-) (Ghasemi et al, 2013) and our calculations from their sensitivity and

226 specificity values yielded LR+ values similar or worse than those of Wainner and associates

227 (Wainner et al, 2003). The differences in findings compared to those of others (Wainner et

228 al, 2003 ; Ghasemi et al, 2013) are likely related to the way we defined a positive index test

229 (familiar compliant that was altered by structural differentiation). In Wainner and

230 colleagues’ study, an ULNT was defined as positive if only one of the following criteria

231 were present: reproduction of the patient's symptoms, or side to side range of motion deficit,

232 or structural differentiation using the cervical spine (Wainner et al, 2003). Ghasemi et al.,

233 reported a positive finding if ‘pain’ occurred during testing (hasemi et al, 2013). Basing the

234 test outcome on one criterion alone as identified by those authors could lead to an increase

235 in false positive findings, thus decreasing specificity, and worsening the LR+ value

236 (Schiffman et al, 2014). Apelby-Albrecht et al. defined as positive if all the three following

237 criteria were met: reproduction of neurogenic symptoms according to a dermatomal pattern,

238 increased or decreased symptoms with structural differentiation, and a difference in painful

239 radiation between sides (Apelby-Albrecht et al, 2013). Our LR+ values are very similar to

240 previous findings by a recent systematic review (Thoomes et al, 2017) calculated from data

241 reported by Apelby-Albrecht et al (Apelby-Albrecht et al, 2013). However, in our study

242 ULNT was defined as positive according to two criteria and we include a more mixed

243 control group population (58 neck or shoulder pain, thoracic outlet syndrome and carpal

244 tunnel syndrome) than Apelby-Albretch (only 18 subjects with neck pain or carpal tunnel

245 syndrome) (Apelby-Albrecht et al, 2013). These findings highlight the importance to

246 clinicians of determining a positive ULNT based on symptom reproduction together with the

247 effects of structural differentiation, at least in diagnosing cervical radiculopathy.

248

249 In their recent systematic review of diagnostic tests for CR, Koulidis et al [25] concluded

250 that ULNTs could only be used as a “ruling out” strategy (Koulidis et al, 2019) based on

251 Apelby-Albrecht et al’s data (Apelby-Albrecht et al, 2013). Conversely, in our sample,

252 ULNT when used in isolation were better at ruling in CR versus ruling out, yet clustering

253 the ULNT findings produced large changes in post-test probability with either a negative

254 finding or a positive finding. The condition of one of four positive tests yields a LR- of 0.08

255 (95%CI=0.01-0.56). This means that when none of the four ULNTs are positive it can rule

256 out CR with only a 3.58% chance that the patients in this sample had CR. Moreover, using

257 multiple combination of ULNT demonstrated that the condition of 3 of 4 positive tests

258 yielded a LR+ of 12.89 (95%CI=3.10-53.62) which means it can rule in CR with a post-test

259 probability of 85.71%. We recommend the use of 3 of 4 conditions over 4 of 4, since this

260 finding was uncommon and because the confidence intervals crossed 1.0 for the LR-

261 analyses.

262 We are also the first to report post-test probability of a positive and negative finding with an

263 ULNT, an analysis omitted from past works (Apelby-Albrecht et al, 2013; Ghasemi et al,

264 2013). Post-test probability provides a better understanding of how markedly one’s decision

265 is influenced by single, or combined, positive or negative test results. This is of particular

266 importance since the reporting of individual sensitivity and specificity values is not

267 recommended (Hegedus and Stern, 2009 ; Baeyens et al, 2019) and may yield conflicting

268 results for ruling in or ruling out conditions.

269

270 Study limitations

271 Although there is notable debate on an appropriate sample size for a diagnostic

272 accuracy study (Hajian-Tilaki, 2014 ; Bujang and Adnan, 2016), we feel compelled to
273 identify our sample of 85 (including 27 CR) as a potential limitation. A smaller sample size

274 may lead to less precision (e.g., wide confidence intervals). Only one clinician was involved

275 in determining the reference standard and another was involved in determining the ULNT.

276 Although the ULNT tester was blinded to the diagnosis of the patient, the transferability of

277 their findings is unknown, since we did not test interrater agreement. Future research is

278 needed to assessed the validity of ULNT with a larger sample of patients with CR and a

279 larger control group with similar symptoms (thoracic outlet syndrome, neck/shoulder pain,

280 peripheral nerve entrapment, etc.), and with more examiner and reference standards

281 including magnetic resonance neurography and small fiber function (Schmid et al, 2013).

282

283

284 Conclusions

285 Our results support past findings that the singular use of ULNT to rule in or rule out

286 CR is not recommended. When combinations are used, findings have higher clinical utility.

287 When all ULNTs are negative, CR can be ruled out, whereas when 3 of 4 tests are positive,

288 CR can be ruled in. As such, we recommend the use of ULNT tests as combinations only.

289 Our study does not test the validity of ULNT tests for specific nerve trunks, which it is

290 hypothesized to perform.

291

292

293

294

295
296 Table 1: Baseline characteristics of the subjects (n=85)
297 * Wilcoxon rank sum test
298
299
300 Table 2: Diagnostic Accuracy of Individual Upper Limb Neurodynamic Tests. Pre-test
301 Prevalence = 31.7%.
302
303 95%CI: Confidence interval at 95%
304 LR+: Positive likelihood ratio
305 LR-: Negative likelihood ratio
306 ULNT: Upper limb neurodynamic test
307
308
309 Table 3. Diagnostic Accuracy of Clustered Upper Limb Neurodynamic Test findings
310 (Conditions). Pre-test Prevalence = 31.7%.
311
312 95% CI: Confidence interval at 95%
313 LR+: Positive likelihood ratio
314 LR-: Negative likelihood ratio

315

316

317

318

319

320

321

322

323

324

325

326

327

328
329 Appendix A. Standard sequence of joint movements and suggested structural differentiation
330 maneuvers (sensitizing movements at a site distant to the pain) for each ULNT (Nee et al,
331 2012)
ULNT 1 (median nerve) :
• Shoulder girdle stabilization
• Shoulder abduction
• Wrist/finger extension
• Forearm supination
• Shoulder external rotation
• Elbow extension
• Structural differentiation:
Cervical side bending or release wrist extension

ULNT 2a (median nerve) :

• Shoulder girdle depression
• Elbow extension
• Shoulder external rotation and forearm supination
• Wrist/finger extension
• Shoulder abduction
• Structural differentiation:
Cervical side bending, or release shoulder girdle
depression or release wrist extension

ULNT 2b (radial nerve) :

• Shoulder girdle depression
• Elbow extension
• Shoulder external rotation and forearm pronation
• Wrist/finger flexion
• Shoulder abduction
• Structural differentiation :
Release shoulder girdle depression or release wrist
flexion

ULNT 3 (ulnar nerve) :

• Wrist/finger extension
• Forearm pronation
• Elbow flexion
• Shoulder external rotation
• Shoulder girdle depression
• Shoulder abduction
• Structural differentiation :
Cervical side bending, or release shoulder girdle
depression or release wrist extension
332

333

334

335 Abbreviations :

336 CR : Cervical radiculopathy

337 ULNT : Upper limb neurodynamic tests

338 MRI : magnetic resonance imaging

339 LR+ : Positive likelihood ratio

340 LR- : Negative likelihood ratio

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356
357 References

358 1. Anekstein Y, Blecher R, Smorgick Y, Mirovsky Y. What is the best way to apply the

359 spurling test for cervical radiculopathy? Clinical Orthopaedics and Related Research:

360 The Association of Bone and Joint Surgeons. 2012; 470(9):2566-2572. Doi:

361 10.1007/s11999-012-2492-3

362 2. Radhakrishnan K, Litchy WJ, O'Fallon WM, Kurland LT. Epidemiology of cervical

363 radiculopathy. A population-based study from Rochester, Minnesota, 1976 through

364 1990. Brain. 1994;117 ( Pt 2):325-335. doi:10.1093/brain/117.2.325

365 3. Woods, Barrett I. MD; Hilibrand, Alan S. MD Cervical Radiculopathy, Journal of

366 Spinal Disorders and Techniques: June 2015 - Volume 28 - Issue 5 - p E251-E259 doi:

367 10.1097/BSD.0000000000000284

368 4. Mink JH, Gordon RE, Deutsch AL. The cervical spine: radiologist's perspective. Phys

369 Med Rehabil Clin N Am. 2003;14(3):493-vi. doi:10.1016/s1047-9651(03)00035-4

370 5. Reza Soltani Z, Sajadi S, Tavana B. A comparison of magnetic resonance imaging

371 with electrodiagnostic findings in the evaluation of clinical radiculopathy: a cross-

372 sectional study. Eur Spine J. 2014;23(4):916-921. doi:10.1007/s00586-013-3164-z

373 6. Ashkan K, Johnston P, Moore AJ. A comparison of magnetic resonance imaging and

374 neurophysiological studies in the assessment of cervical radiculopathy. Br J

375 Neurosurg 2002;16:146–8.

376 7. Carette S, Fehlings MG. Clinical practice. Cervical radiculopathy. N Engl J Med.

377 2005;353(4):392-399. doi:10.1056/NEJMcp043887

378 8. Kuijper B, Tans JT, Schimsheimer RJ, et al. Degenerative cervical radiculopathy:

379 diagnosis and conservative treatment. A review. Eur J Neurol. 2009;16(1):15-20.

380 doi:10.1111/j.1468-1331.2008.02365.x
381 9. Elvey RL. Treatment of arm pain associated with abnormal brachial plexus tension.

382 Aust J Physiother . 1986;32(4):225-30. doi: 10.1016/S0004-9514(14)60655

383 10. Butler, DS. The Sensitive Nervous System Unley, NOI Group Publications. Adelaide,

384 Australia 2000

385 11. Shacklock, MO. (1996). Positive upper limb tension test in a case of surgically proven

386 neuropathy: analysis and validity. Manual Therapy, 1996;1(3): 154-161. Doi:

387 10.1054/math.1996.0265

388 12. Thoomes EJ, Van Gees S, Van der Windt DA, Falla D, Verhagen AP, Koes BW,

389 Thoomes-de Graaf M, Kuijper B, Scholten-Peeters WGM, Vleggeert-Lankamp CL.

390 Value of physical tests in diagnosing cervical radiculopathy: a systematic review.

391 Spine Journal. 2017; 18(1):179-189. doi: 10.1016/j.spinee.2017.08.241

392 13. Apelby-Albrecht M, Andersson L, Kleiva IW, Kvale K, Skillgate E, Josephson A.

393 Concordance of Upper Limb Neurodynamic Tests with medical examination and

394 magnetic resonance imaging in patients with cervical radiculopathy: a diagnostic

395 cohort study. Journal of Manipulative and Physiological Therapeutics.

396 2013;36(9):626-632. Doi: 10.1016/j.jmpt.2013.07.007

397 14. Wainner RS, Fritz JM, Irrgang JJ, Boninger ML, Delitto A, Allison S. Reliability and

398 Diagnostic Accuracy of the Clinical Examination and Patient Self-Report Measures

399 for Cervical Radiculopathy. Spine, 2003;28 (1):52-62.

400 15. Ghasemi M, Golabchi K, Mousavi SA, et al. The value of provocative tests in

401 diagnosis of cervical radiculopathy. J Res Med Sci . 2013 Mar;18(Suppl 1):S35-8

402 16. Schiffman, E., Ohrbach, R., Truelove, E., Look, J., Anderson, G., Goulet, JP, …

403 Dworkin, SF. Diagnostic Criteria for Temporomandibular Disorders (DC/TMD) for

404 Clinical and Research Applications: Recommendations of the International

405 RDC/TMD Consortium Network and Orofacial Pain Special Interest Group. J Oral

406 Facial Pain Headache. 2014;28(1):6-27. Doi: 10.11607/jop.1151

407 17. Nee RJ, Jull GA, Vicenzino B, Coppieters MW. The validity of upper-limb

408 neurodynamic tests for detecting peripheral neuropathic pain. The Journal of

409 Orthopaedic and Sports Physical Therapy. 2012; 42(5):413-424. Doi:

410 10.2519/jospt.2012.3988

411 18. Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig L, et al. .

412 STARD 2015: an updated list of essential items for reporting diagnostic accuracy

413 studies. BMJ. (2015) 351:h5527. 10.1148/radiol.2015151516

414 19. Horn ME, Brennan GP, George SZ, Harman JS, Bishop MD. A value proposition for

415 early physical therapist management of neck pain: a retrospective cohort analysis.

416 BMC Health Serv Res. 2016; 16: 253.

417 20. Masaracchio M, Cleland JA, Hellman M, Hagins M. Short-term combined effects of

418 thoracic spine thrust manipulation and cervical spine nonthrust manipulation in

419 individuals with mechanical neck pain: a randomized clinical trial. J Orthop Sports

420 Phys Ther 2013;43(3):118e27.

421 21. Schmid AB, Brunner F, Luomajoki H, et al. Reliability of clinical tests to evaluate

422 nerve function and mechanosensitivity of the upper limb peripheral nervous system.

423 BMC Musculoskelet Disord. 2009;10:11. Published 2009 Jan 21. doi:10.1186/1471-

424 2474-10-11

425 22. Walsh MT. (2005). Upper limb neural tension testing and mobilization. Fact, fiction,

426 and a pratical approach. J Hand Ther. 2005;18(2):241-258. DOI:

427 10.1197/j.jht.2005.02.010

428 23. Cook CJ, Cook CE, Reiman MP, Joshi AB, Richardson W, Garcia AN. Systematic

429 review of diagnostic accuracy of patient history, clinical findings, and physical tests in
430 the diagnosis of lumbar spinal stenosis. Eur Spine J. 2020;29(1):93-112.

431 doi:10.1007/s00586-019-06048-4

432 24. Koulidis K, Veremis Y, Anderson C, Heneghan NR. Diagnostic accuracy of upper

433 limb neurodynamic tests for the assessment of peripheral neuropathic pain: A

434 systematic review. Musculoskelet Sci Pract. 2019 Apr;40:21-33. doi:

435 10.1016/j.msksp.2019.01.001. Epub 2019 Jan 12. PMID: 30665045

436 25. Hegedus EJ, Stern B. Beyond SpPIN and SnNOUT: Considerations with Dichotomous

437 Tests during Assessment of Diagnostic Accuracy. J Man Manip Ther. 2009;17(1):E1-

438 E5. doi:10.1179/jmt.2009.17.1.1E

439 26. Baeyens JP, Serrien B, Goossens M, Clijsen R. Questioning the “SPIN and SNOUT”

440 rule in clinical testing. Arch Physiother 9, 4 (2019). https://doi.org/10.1186/s40945-

441 019-0056-5

442 27. Hajian-Tilaki K. Sample size estimation in diagnostic test studies of biomedical

443 informatics. J Biomed Informatics. 2014;48:193-204.

444 28. Bujang MA, Adnan TH. Requirements for Minimum Sample Size for Sensitivity and

445 Specificity Analysis. J Clin Diagn Res. 2016;10(10):YE01-YE06.

446 doi:10.7860/JCDR/2016/18129.8744

447 29. Schmid, A.B., Nee, R.J., Coppieters, M.W., 2013. Reappraising entrapment

448 neuropathies– mechanisms, diagnosis and management. Man. Ther. 18 (6), 449–457.

449

450
451

452

453
454
455
456
457
458
459 Acknowledgments : The authors thank the assistance of François-Xavier Bénard, Adeline
460 Vergé, Cécile Roesch and Audrey Tseng Qun.

461
462
463
464
465
466