Background Upper limb neurodynamic tests (ULNT) are used to diagnose neuropathic conditions such as cervical radiculopathy (CR). Within the literature, a positive ULNT is defined in markedly variable ways, which is likely why the diagnostic accuracy of these tests lacks consistency across studies. Objectives To determine the diagnostic accuracy of single and combined upper limb neurodynamic tests ((ULNT)1,2a, 2b and 3) for cervical radiculopathy using test findings that are similar to those used in practice. Design Diagnostic accuracy study (prospective) design following the updated STARD 2015 reporting guideline. Method From 109 consecutively enrolled individuals with suspected CR. Of the 85 participants included, 27 (31.7%) were diagnosed with CR (mean age, 43.9years; Neck Disability Index 38,16%). ULNTs test were performed by a blind examiner to a CR reference standard of clinical diagnosis and magnetic resonance imaging verification provided by a neurosurgeon. Results In general, the single tests were better at ruling in CR versus ruling out. Of the single ULNT, the ULNT3 demonstrated the strongest post-test probability change with a positive finding (73.28%). Three of four test combinations demonstrated the highest clinical utility for changing the post-test probability with a positive finding at 83.29% and with LR+ = 12.89 (95%CI: 3.10–53.62). Having none of the test's positive was able to rule out CR with LR- = 0.08 (95%CI: 0.01–0.56). Conclusion ULNTs fail to significantly alter post-test probability when used singularly for diagnosis of CR. However, combinations of ULNT (3 out of 4 positive) can rule in CR, and rule out CR when all ULNT are negative.