A modified score to identify and discriminate neuropathic pain: a study on the German version of the neuropathic pain symptom inventory (NPSI)

We translated the NPSI into German using a standard forward-backward translation and administered it in a case-control design to patients with neuropathic (n = 68) and non-neuropathic pain (headache and osteoarthritis, n = 169) to validate it and to analyze its discriminant properties, its sensitivity to change, and to detect neuropathic pain subgroups with distinct profiles.Using a sum score (the NPSI-G score), we found sensitivity to change (r between 0.37 and 0.5 for pain items of the graded chronic pain scale) and could distinguish between neuropathic and other pain on a group basis, but not for individual patients. Post hoc development of a discriminant score with optimized diagnostic properties to distinguish neuropathic pain from non-neuropathic pain resulted in an instrument with high sensitivity (91%) and acceptable specificity (70%). We detected six different pain profiles in the patient group with neuropathic pain; three profiles were found to be distinct.The NPSI-G potentially combines the properties of a diagnostic tool and an instrument to identify subtypes of neuropathic pain.Neuropathic pain must be correctly diagnosed for optimal treatment. Several screening tools have been developed to differentiate neuropathic from non-neuropathic pain [1]. These include the Leeds Assessment of Neuropathic Symptoms and Signs (LANSS [2]), the Neuropathic Pain Questionnaire [3], the DN4 [4], the painDETECT questionnaire [5], and the ID pain [6]. Another goal of pain questionnaires is the use of effective pain descriptors to identify subgroups of patients that may benefit from specific therapies. Such questionnaires are the Neuropathic Pain Scale (NPS) [7] and the Neuropathic Pain Symptom Inventory (NPSI) [8]. The NPSI has psychometric properties which suggest that it may be used to characterize subgroups of neuropathic pain patients and verify whether they respond differentially to treatment [8]. It has good construct validity, high test-retest reliability, and is s