|Home | About | Journals | Submit | Contact Us | Français|
The qualities of chronic neuropathic pain (NeP) may be informative about the different mechanisms of pain. We previously developed a 2-factor model of NeP that described an underlying structure among sensory descriptors on the Short-Form McGill Pain Questionnaire (SF-MPQ). The goal of this study was to confirm the correlated 2-factor model of NeP. Individual descriptive scores from the SF-MPQ were analyzed. Confirmatory factor analysis (CFA) was used to test a correlated 2-factor model. Factor 1 (stabbing pain) was characterized by high loadings on stabbing, sharp, and shooting sensory items; factor 2 (heavy pain) was characterized by high loadings on heavy, gnawing, and aching items. Results of the CFA strongly supported the correlated 2-factor model.
While chronic neuropathic pain (NeP) is a multidimensional experience, the primary focus of most clinical therapeutic trials for NeP is pain intensity (magnitude or quantity). However, other dimensions of pain such as affect (or unpleasantness), location, timing, quality, and interference with activity may be as important as pain intensity in therapeutic trials and in the characterization of chronic pain states. These pain dimensions may directly inform researchers about pain mechanisms and treatment responsiveness to the active agent or placebo. Unfortunately, these other dimensions of pain are rarely the focus of therapeutic trials for chronic NeP.
Pain qualities (eg, hot, stabbing, aching), in particular, have become the focus of recent research in chronic pain, as preliminary evidence indicates that different pain sensations may reflect different pain mechanisms.10 Improved understanding of the relationship between the sensations of qualities of pain and their underlying mechanisms would advance our knowledge of the causes of different chronic pain conditions. Furthermore, if certain pain qualities are preferentially affected by a particular therapy, these qualities may provide clinicians a basis for choosing between a multitude of available treatment options. These features of pain qualities have led investigators to call for more research into their characterization for specific pain treatments and perhaps for distinguishing pain mechanisms.18
We recently demonstrated that a correlated 2-factor model of NeP described an underlying structure among sensory descriptors on the Short-Form McGill Pain Questionnaire (SF-MPQ).9 The first factor, stabbing pain, was characterized by a high percentage of patients using the words stabbing, sharp, and shooting to describe their pain experiences. The second factor, heavy pain, was characterized by high use of the terms heavy, gnawing, and aching to describe pain. Specific and differential analgesic responses to lidocaine were identified from the pain factors. Compared with patients with low levels of heavy pain, patients with high levels of heavy pain factor had significantly greater reduction in pain intensity, as measured by a visual analog scale (VAS), in response to intravenous lidocaine but not in response to placebo.9 This study was limited by the small sample size (n = 71) and heterogeneity of the suspected NeP conditions. We sought to confirm this 2-factor model of NeP in this study of a large cohort of patients with 1 of 2 common NeP conditions, diabetic peripheral neuropathy (DPN) and postherpetic neuralgia (PHN), who were treated with pregabalin in randomized controlled clinical trials. NeP associated with DPN is present in approximately 11 to 26% of patients with diabetes.8,11,13,24 PHN occurs in 7 to 27% of patients who develop symptomatic reactivation of latent Varicella zoster virus, with the elderly at especially high risk. As many as 50% of patients aged 70 or older will have persistent pain 6 months after healing of the acute rash.16,22,27,35
Patient data from 6 DPN,1,21,25,28,29,32 4 PHN,12,30,31,33 and 1 combined DPN/PHN14 randomized, placebo-controlled clinical trials were selected for this study. These data were limited to fixed doses of pregabalin 150, 300, or 600 mg/d or placebo (Table 1).
The SF-MPQ was administered to patients at the baseline visit in each study. The questionnaire sections included in this analysis were: (a) the pain descriptors and (b) the VAS. To maintain consistency across the studies, SF-MPQ individual descriptor scores were analyzed for English-speaking patients only. For SF-MPQ pain descriptors, the patient rated the intensity of each type of pain that he/she had experienced over the past week on a 4-point intensity scale (0 = none, 1 = mild, 2 = moderate, 3 = severe). The VAS provided a rating of overall pain during the past week; the patient placed a hash mark across a 100-mm line to indicate the intensity of the pain from left (0 mm = “no pain”) to right (100 mm = “worst possible pain”).
All randomized, intent-to-treat patients with nonmissing SF-MPQ scores at baseline were analyzed. SF-MPQ individual descriptor scores were summarized by disease and treatment groups.
Confirmatory factor analysis (CFA)19 was used to test the correlated 2-factor measurement structure of pain originally reported in the literature.9 Accordingly, the first factor, stabbing pain, was characterized by a set of three descriptors: stabbing, sharp, and shooting. The second factor, heavy pain, was characterized by a different set of three descriptors: heavy, gnawing, and aching. The covariance between the stabbing and heavy pain factors were also estimated. The diagonally weighted least squares estimator was used to take account of ordinal item distributions23; missing-at-random was assumed. All available data were used to estimate the model.
The pooled population (N = 2,516) was modified owing to clarification of an SF-MPQ descriptor (at least one non-missing English-only SF-MPQ descriptor). In order to maintain consistency across the analysis, SF-MPQ individual descriptive scores are described and analyzed only for English-speaking patients. Thus, the final analyzed cohort (n = 2,267) included patients with DPN (n = 1,250) and PHN (n = 1,017, Table 2). Baseline demographics were comparable across the treatment groups in the pooled population. The baseline mean VAS scores were similar in the pregabalin (68.98) and placebo (70.00) groups.
The percentage of patients who completed the SF-MPQ and selected each of the pain descriptors at the time of randomization within the studies, separately for treatment group and disease group, is given in Figure 1. At least 50% of patients with either DPN or PHN chose most of the pain descriptors. The aching, hot-burning, sharp, throbbing, shooting, stabbing, and tiring-exhausting descriptors were chosen by at least 75% of all patients at randomization, which suggests that these pain descriptors are characteristic of painful peripheral neuropathy in general. Among the patients with PHN only, the tiring-exhausting, tender, aching, hot-burning, and sharp descriptors were chosen by at least 75% of patients in each treatment group (placebo and pregabalin) at randomization (Figure 1).
The correlated 2–factor model of pain was sufficient to explain the observed covariance structure among SF-MPQ sensory items.9 Factor 1, labeled as “stabbing pain quality,” was characterized by high factor loadings on stabbing, sharp, and shooting sensory descriptors. Factor 2, labeled as “heavy pain quality,” was characterized by high factor loadings on heavy, gnawing, and aching sensory descriptors. “Throbbing” and “splitting” qualities appeared to load equally on both factors; thus, these 2 items lacked the trait discriminant validity and were not used to measure the 2 factors.
A CFA on the DPN, PHN, and DPN/PHN pooled studies was performed separately to test and confirm the 2-factor structure (sharp pain, heavy pain, and the pain qualities that grouped with each) previously reported.9 As a competing model, a unidimensional pain factor model (1-factor model) was also tested. The CFA independently and strongly supported the correlated 2-factor structure in terms of absolute fit and relative fit. Absolute fit statistics were within the acceptable range (ie, root mean square error of approximation ≤0.07, comparative fit index ≥0.99, and Tucker-Lewis Index ≥0.99; Table 3).17 In latent variable modeling, it is customary to inspect multiple indices of fit because each index has its own strengths and limitations.4
The correlation between the heavy and stabbing factors was moderate (0.67/0.70/0.68 for the DPN/PHN/combined data). In terms of relative and absolute model fit, the correlated 2-factor model held in both samples (Table 4).
Our results confirm that in a large cohort of patients with either DPN or PHN, factor analysis of the 11 baseline SF-MPQ sensory pain qualities reveals a 2-factor model of pain: (1) stabbing pain quality (utilizing stabbing, sharp, and shooting pain descriptors) and (2) heavy pain quality (utilizing heavy, gnawing, and aching pain descriptors). Interestingly, the quality of ‘throbbing’ did not discriminate between the 2 factors. As an independent and large sample, these results confirm our prior work that was performed in a much smaller and more heterogeneous group of patients with suspected NeP.9
Similar pain factors were described for patients with localized osteoarthritis (OA).34 “Stabbing” was used as a descriptor of pain by 18% of patients with localized OA who completed the MPQ, as opposed to 4% of patients with generalized OA or rheumatoid arthritis (RA). While heavy pain was not explicitly measured in this study, terms such as “cramping” and “pressing” were used by patients with localized OA more frequently than by patients with RA or generalized OA. The authors suggested that specific pain descriptors may be sufficiently sensitive to discriminate between related painful conditions, but their sample size was too small to substantiate this claim.34
The 2-factor pain quality structure is important because it may inform us about mechanisms of neuropathic pain in general. It is notable that the 2-factor structure coincides with distinctive pain sensations carried by Aδ and C fibers. In particular, Aδ fibers have been associated with the pain sensation described as acute or “first pain” akin to the stabbing pain factor in this study, whereas C fiber stimulation has been described as possessing pain qualities such as dull, slow, “second pain,” 5,26 similar to the heavy pain factor described in this report. While we did not attempt to characterize the mechanisms associated with the stabbing and heavy factors in this study, and any association is admittedly speculative, future research should characterize the pain factor quality–mechanism relationships.
Limitations of our study are predominately due to the restricted number of pain qualities represented in the SF-MPQ. The SF-MPQ uses 11 words to characterize the sensory qualities of pain: throbbing, shooting, stabbing, sharp, cramping, gnawing, hot-burning, aching, heavy, tender, and splitting. Additionally, the SF-MPQ uses 4 affective qualities: tiring-exhausting, sickening, fearful, and punishing-cruel. We did not evaluate the affective qualities in this study. Several instruments have been developed to specifically characterize pain qualities in NeP including, but not limited to, the, Neuropathic Pain Scale,15 Leeds Assessment of Neuropathic Symptoms and Signs,3 Neuropathic Pain Symptom Inventory,7 Neuropathic Pain Diagnostic Questionnaire,6 and Neuropathic Pain Questionnaire20 (and its Short Form2). While useful, several of these instruments do not address the non-traditional pain qualities found in the heavy pain factor such as aching, dull, or gnawing. However, the additional pain descriptors in these instruments may enhance the content validity of pain scores estimated from our 2-factor model. Furthermore, the additional pain quality descriptors may allow for identification of distinctive factor structures for DPN and PHN. In this study, we did not find a different structure for the 2 NeP conditions—despite the clearly different initial conditions—that supported the generalizability of the proposed 2-dimensional measure of pain in independent samples. Finally, we note that the population size of this study may be large with regards to the inherent variability of some descriptors, which could lead to overfitting of some factors in the analysis. However, the sample size for this study needed to be large enough to adequately power the CFA for interpretable results.
The specific qualities of pain may be indicative of pain mechanisms and may eventually guide physicians toward the most effective therapies for particular conditions. We previously described a 2-factor structure for the qualities of general neuropathic pain in a small study.9 In this report, a unidimensional conceptualization of pain (ranging from low to high levels) has been rejected. Instead, our results confirm the two related dimensions of pain (heavy and stabbing) in a large cohort of patients with two distinct painful neuropathies (DPN and PHN). In clinical settings, one patient may present a high level of heavy pain accompanied by a low level of stabbing pain whereas another patient may present high levels of both heavy and stabbing pain. Future studies should investigate whether this two dimensional representation of pain qualities is associated with distinctive pain mechanisms and treatment responsiveness to therapeutic agents, as well as the generalizability of two dimensional representation of pain in non-neuropathic populations.
This article validates a model that describes the qualities of neuropathic pain associated with diabetic peripheral neuropathy and postherpetic neuralgia. These data suggest that specific pain qualities may be associated with pain mechanisms or may be useful for predicting treatment response.
Medical writing and editorial support was provided by Patricia Rohrs (Stanford) and Patricia McChesney, PhD, of UBC Scientific Solutions. Patricia McChesney was funded by Pfizer Inc.
Dr Mackey is funded by NIH K24DA029262 and R01NS053961. Dr. Carroll is funded by NIH K23 DA25152. Dr Dumenci is funded by NIH R01CA140151. Drs Mackey, Carroll, and Dumenci have received no financial support from Pfizer for this work. Additional funding was provided by Pfizer Inc.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Institution where work was done: Study concept and design were conducted at Stanford University, Palo Alto, CA, USA; Statistical analyses were performed at Pfizer Global Pharmaceuticals, New York, NY, USA and Virginia Commonwealth University School of Medicine, Richmond, VA, USA.
DRS EMIR, MURPHY, AND WHALEN ARE EMPLOYEES OF PFIZER INC.