A preliminary fMRI study of analgesic treatment in chronic back pain and knee osteoarthritis

Eleven CBP and eight OA patients were recruited. Complete fMRI data were collected only in eight CBP and five OA patients and only these data were analyzed. CBP patients were included if they fulfilled the IASP criteria for CBP [26], and were diagnosed in accordance with recent guidelines [27]. OA patients were included if they fulfilled the criteria of the American College of Rheumatology for the classification of OA [28]. All the patients reported chronic pain for a duration longer than 3 months with a pain magnitude of at least 30/100 on a visual analog scale (VAS). Clinical properties are summarized in Table 1 and Table 2 respectively. Participants signed written informed consent to the experimental procedure, which was approved by the Northwestern University Institutional Review Board.

Pain characteristics were determined using validated questionnaires. At every visit, CBP and OA patients filled the short form of McGill Pain Questionnaire (sf-MPQ) [29]. Additionally, CBP patients filled the Neuropathic Pain Scale (NPS) [30], and OA patients filled the Western Ontario and McMaster Osteoarthritis Index (WOMAC) [31].

During each session patients were scanned under two different conditions: 1) pain rating (up to three scans per session) where patients were asked to report the intensity of their pain using a finger-span logging device. Prior to scanning, patients learned how to use the device to continuously report changes in their pain intensity while observing their rating projected on a computer screen as a moving bar (size of the bar is a scale from 0 to 100; 0 = no pain; 100 = worst imaginable pain); 2) one visual control scan where patients used the finger-span device to rate the length of a bar moving at a rate approximating the variability of their pain (for more details see, [32, 33]. The visual task served as a motor-attention-cognitive control for non-specific brain activations due to finger motion and attention. During the pain rating task CBP patients reported the fluctuations of their spontaneous pain intensity in the absence of an overt experimental stimulus. On the other hand, OA patients rated pain in response to pressure applied to the most sensitive part of their bad knee. Pressure was delivered using a custom-made fMRI compatible device equipped with a feedback signal regarding applied pressure. The finger device and pressure signal were synchronized and time locked with the fMRI image acquisition sequence. During each scan, OA patients were presented with up to six stimulation episodes with variable intensities (range: 2–13 s.u., mean ± standard deviation : 5.5 ± 4.8 s.u., s.u. = standardized units; absolute pressure values were not used due to their strong dependence on the physical location of the stimulus on the knee), durations (range: 20 – 45 seconds, mean: 32.2 ± 15 seconds) and inter-stimulus intervals (range: 30 – 60 seconds, mean 42.8 ± 16 seconds). Examples of one CBP and one OA pain-rating task are shown in Figure 1. The applied pressure and the patient's rating of related pain were collected during scanning and used to relate brain activity to perceived pain.

Functional MRI data was acquired with a 3T Siemens Trio whole-body scanner with echo-planar imaging (EPI) capability using the standard radio-frequency head coil. Multi-slice T2*-weighted echo-planar images were obtained with the following parameters: repetition time TR = 2.5 s, echo time TE = 30 ms, flip angle = 90°, slice thickness = 3 mm, and in-plane resolution = 3.475 × 3.475 mm². The 36 slices covered the whole brain from cerebellum to vertex. Two hundred and forty volumes were acquired per scan. A T1-weighted anatomical MRI image was also acquired for each subject using the following parameters: TR = 2.1 s, TE = 4.38 ms, flip angle = 8°, FOV = 220 mm, slice thickness = 1 mm, in-plane resolution = 0.86 × 0.86 mm², and number of sagittal slices = 160.

Image analysis to reveal significant brain activity based on changes in BOLD signal was performed on each subject's data using FMRIB Expert Analysis Tool (FEAT, [34], http://​www.​fmrib.​ox.​ac.​uk/​fsl). The preprocessing of each subject's time-series of fMRI volumes encompassed: slice time correction; motion correction using MCFLIRT; spatial smoothing using a Gaussian kernel of full-width-half-maximum 5 mm; nonlinear high-pass temporal filtering (Gaussian-weighted least-squares straight line fitting, filter cutoff of 100 seconds) and subtraction of the mean of each voxel time-course from that time-course (i.e. intensity normalization). The fMRI signal was then linearly modeled on a voxel by voxel basis using FMRIB's Improved Linear Model (FILM) with local autocorrelation correction [35, 36].

Each condition described above (rating of spontaneous pain in CBP, rating of stimulus pain in OA, and rating of visual stimuli in both populations) was considered to generate a hemodynamic response described by the convolution of the corresponding vector with a canonical hemodynamic response function (gamma function: lag = 6 seconds, standard deviation = 3 seconds). Head motion vector (derived from motion correction) was used at this level as a covariate of no interest to further remove any residual variance due to head motion. The significance of the model fit to each voxel time-series was calculated, yielding statistical parametric maps for the pain and visual rating conditions in each subject.

Average group activity pain maps for OA and CBP for in session 1 and session 2 were generated by first subtracting the visual task related activity from the pain related activity map for each subject using FEAT in a higher level fixed effects analysis following the co-registration of individual scans to standard space (152 subject average MNI space, http://​www.​bic.​mni.​mcgill.​ca/​cgi/​). This results in a contrast of parametric estimate (COPE) map of statistically significant pain minus visual activity for each patient. Group average pain minus visual activity maps were then generated by averaging COPE maps for CBP (n = 7) and OA (n = 5) in a third higher level fixed effects analysis. The resultant group average pain minus visual maps were thresholded at z-scores > 2.3, and a cluster probability threshold of p = 0.01 was applied to compute the significance of each cluster, thereby correcting for multiple comparisons [37]. Group average pain minus visual brain activity for CBP and OA are shown in Figure 2.

Brain regions exhibiting a significant decrease in activity in response to lidoderm treatment in CBP and OA were identified using a paired t-test (fixed effects analysis, z-score > 2.3, cluster threshold p = 0.01) between the higher-level (pain – visual) COPEs statistical images of session 1 and session 2. The resultant z-map for session 1 minus session 2 for each patient population was then masked with its respective group average activity of the pain minus visual of session 1 to eliminate false positive results. These maps thus depict the brain areas showing pain-specific activity that significantly decreased with treatment.

To determine those brain regions encoding pain intensity, we performed a whole brain covariate analysis using FSL (fixed effects analysis, z-score > 2.3, cluster threshold p = 0.01) across CBP and OA patients. For CBP, the first level pain activity maps were pooled from session 1 and 2 (n = 42) and regressed with the average spontaneous pain during the scan. Similarly, the first level OA pain activity maps were pooled from sessions 1 and 2 (n = 37) and regressed with average pain divided by applied pressure, thereby normalizing perception to unit intensity of applied stimulus. We used standardized stimulus units rather than absolute values of pressure to account for the variability of pressure intensities necessary to induce pain across the knee. The z-maps for the covariate analysis for CBP and OA were then masked with their respective group average activity for pain minus visual contrast, to restrict the analysis to pain specific brain regions. To show the relationship of brain activity and pain intensity in more detail, we also performed a region of interest analysis (ROI). For each scan, we extracted the z-values of a 1 cc ROI centered on the peaks obtained in the covariate maps and examined their relationship to perceived pain (Figure 3).

To delineate the brain regions encoding pain intensity that were also modulated by treatment, we performed a whole brain voxel-wise conjunction analysis between the thresholded z-map for treatment and the pain intensity covariate maps for CBP and OA. For each patient group we then identified the ROIs from the conjunction map and extracted their z-values from a 1 cc volume and used them to perform an unpaired t-test between session 1 and session 2.

Surface-based mapping (figures 2 and 3) was constructed using the Population-Average, Landmark- and Surface-based (PALS) average fiducial surface from 12 individual subjects as the atlas target [38].

CBP patients showed a robust decrease in pain intensity as well as other clinical pain parameters in response to a two-week topical Lidocaine treatment applied to the back. Pain intensity was assessed from VAS scores obtained in sf-MPQ, and by averaging the spontaneous pain reported by the subject during the scan. Both of these measurements exhibited significant reduction after treatment (Figure 1B, Table 3). In addition, CBP patients showed a significant decrease in the surface score (p < 0.05) and borderline significant decreases for the deep, unpleasantness, and overall neuropathic pain score (Table 3).

In contrast to CBP, OA patients did not show improvement after Lidocaine treatment in their knee pain as measured by VAS in sf-MPQ, or by intensity of mean pain ratings for knee stimulation, and other OA related pain characteristics on either sf-MPQ or on WOMAC. They did exhibit however a borderline decrease in ratings of knee pain (p = 0.06) after normalization with applied pressure (Figure 1B, Table 3).

Before treatment (session 1), spontaneous pain in CBP gave rise to increased brain activity mainly in the frontal cortex, including the medial prefrontal cortex (mPFC), the rostral anterior cingulate cortex (rACC), bilateral superior frontal gyrus (SFG), and the nucleus accumbens (NAc). Increased activity was also observed in the inferior temporal gyrus (ITG), and the left posterior parietal cortex (PP). Therefore, CBP is represented mainly in emotion/reward (mPFC and NAc) mediating areas as reported previously [39]. After treatment (session 2), only a small cluster in the left MT area showed significant increased activity for spontaneous pain (Figure 2, Table 4).

Evoked knee OA pressure-pain, on the other hand, showed more extensive activity pattern, with similarities to activity for acute pain in normal subjects [40]. Before treatment (session 1), painful knee stimulation evoked activity in: bilateral thalamus, secondary somatosensory cortex (SII), insula, supplementary motor area (SMA), anterior cingulate (ACC) and medial frontal gyrus; and unilaterally in right putamen, and left amygdala. After two weeks of treatment (session 2), the brain activity map was similar with some exceptions. No significant activity was observed in the left amygdala, right putamen, and left anterior insula. Additionally a new, large activation was observed in right ventrolateral orbitofrontal (VLOFC), as well as more activations in mid temporal and visual cortices (Figure 2, Table 4).

As we noted earlier, treatment of CBP with lidocaine led to a significant decrease in pain intensity rating. This was accompanied by a significant decrease in brain activity in mPFC, rACC, SFG, and PP when maps from session 1 and 2 were statistically contrasted. In OA, lidocaine treatment only showed a borderline decrease in pain per applied pressure, but the statistical contrast between session 1 and 2 indicated significantly decreased brain activity in left MFG, left anterior insula, and bilateral thalamus (Figure 2, Table 4).

Brain areas encoding pain intensity in CBP and OA were identified using a covariate analysis. In CBP the covariate was mean spontaneous pain during the scan, and in OA the covariate was the mean pain during the scan divided by the normalized applied pressure. The mPFC and parts of the rACC at the level of the genu encoded pain intensity in CBP (figure 3A). In contrast, the mid ACC, right insula, bilateral thalamus, right S2 and right MFG encoded pain/pressure in OA (figure 3B). Thus, pain intensity representation in CBP and OA (normalized with pressure in OA) pain map to different brain regions.

To show the relationship of brain activity and pain intensity in more detail, we performed a region of interest analysis. For each scan, we extracted the z-values of a 1 cc ROI centered on the peaks obtained by the covariate maps. The ROIs used include mPFC (2,42,2), left thalamus (12,-20,4), right insula (40,8,-4), right S2 (46,-18,32), mid ACC (-4,24,28) and right middle frontal gyrus (MFG) (38, 38, 36). The mPFC showed a significant correlation with pain intensity for CBP (p < 0.01) but not for OA, while the left thalamus was positively correlated with pain intensity for OA (p < 0.01) but not CBP (figure 3). The insula, S2 mid-ACC, right MFG also showed positive correlations with pain intensity for OA, with p values of 0.035, 0.042, 0.041, and 0.039 respectively, but not for CBP.

If we examine pain measures either by ratings or by questionnaire outcomes in the OA group, no treatment effects are observed. The borderline treatment effect only shows up when pain is divided with applied pressure, and this parameter relates to statistically significant brain activity changes. This is a normalization of perception per unit stimulus, it is equivalent to earlier efforts of matching pain perception by varying stimulus strength [68‐70], and might be a critical step in assessing pain treatment effects on evoked pain, especially in clinical pain conditions.

In the present study, while the number of participants in each group was small, the data was nonetheless reproducible enough to show significant changes in brain activity. Due to the small number of subjects studied we have used fixed effects statistics. Therefore, the obtained results cannot be generalized to the population at large, neither for CBP nor for OA.

Here we demonstrate that decreased brain activity with treatment may be identified prior to the patients' report of decreased pain on standard questionnaires commonly used in clinical trials. Open labeled clinical trials indicate decreases in OA pain, as measured by WOMAC or NPS [18, 71], even in small groups of patients. Therefore, it is not clear why our OA patients did not have clinical signs of decreased pain. The design of the present study was based on the idea that painful mechanical stimuli applied to the knee are relevant to knee OA, and that manipulation of this pain should also have clinical relevance. We are surprised that even though knee stimulation evoked pain is modulated with the treatment, clinical questionnaires based outcomes are not. This is in contrast to the CBP patients, at least suggesting that lidocaine therapy is more efficacious in CBP. The discrepancy between questionnaire outcomes and stimulus-evoked pain implies that in OA the clinical pain depends on factors more than just the knee pressure-pain relationship. Therefore, although we observe that lidocaine patch therapy modulates the pressure-pain relationship and thalamic activity reflects this modulation, we cannot conclude that this effect is clinically significant. Alternatively, these results in OA suggest that perhaps continued treatment for longer durations and at higher doses may result in a clinically significant outcome.