RECENT brain imaging studies have furthered our understanding of the neurobiologic mechanisms underlying placebo analgesia.1,2The majority of these studies have used experimental pain stimuli in a laboratory setting3–6and used a controversial criterion of 10–20% difference in pain rating to classify subjects as placebo responders.7In addition, they only performed a single placebo challenge, which leaves us ignorant about the persistence of the placebo effect over time. We here describe behavioral, pharmacologic, and brain imaging results of a long-term placebo response in a chronic pain patient.
A 53-yr-old chronic pain patient presented at the pain clinic with severe pain in the lower lumbar region radiating to the left lower leg. In 1999, she developed back pain problems which became progressively worse. A lumbar myelography and magnetic resonance imaging examination revealed a discus protrusion, compressing the fifth lumbar root. Two laminectomies provided only temporary pain relief. In 2003, an arthrodesis was performed between L3 and L5, resulting in aggravation of the pain. Various drug and interventional therapies had little therapeutic effect. The patient was referred to us for a trial therapy with spinal opioids. As part of the routine screening procedure of candidates for an intrathecal drug delivery system, approved by the local ethics committee, patients are tested first for their response to epidurally administered morphine and saline. The patient was randomly assigned to be first submitted to placebo testing. In short, an epidural catheter is implanted and connected to a patient-controlled analgesia (PCA) pump filled with saline. Patients are sent home and are asked to fill in a pain diary three times daily. Each second day, they are visited by a nurse to check the PCA system and to collect the pain ratings. A major difficulty when assessing placebo responses in individual patients is to separate placebo responses from factors such as natural history of the disease and regression to the mean.8To circumvent these possible confounds, we randomly selected two “placebo holidays” over the course of the placebo-evaluation period, during which the PCA pump was switched off knowingly to the patient and remained inactive for the whole day. To test for the involvement of the opioid system in the placebo response, naloxone (0.14 mg/kg) and saline (0.9% NaCl) were injected intravenously (0.1 ml/s; total infusion time < 2 min) by means of a programmable infusion pump in a double-blind manner.9The naloxone test was conducted on a day of normal use of the PCA pump (fig. 1A). Pain was measured before and 2, 5, 10, 15, 20, 30, and 60 min after the start of the infusion.
After given written informed consent, she participated in a positron emission tomography (PET) study approved by the ethics committee (week 5). The PCA pump was switched off 16 h before scanning. Changes in regional cerebral blood flow (rCBF) data were measured using a CTI 951 16/32 PET scanner (Knoxville, TN). Fifteen scans were acquired in three-dimensional mode after injections (5 ml) of 300 MBq of H215O over a period of 20 s.10The first four scans were taken with the PCA pump switched off. We next assessed the placebo effect using a stepwise titration method: The patient was told that the PCA pump would be turned on and that infusion rate would be gradually increased in 33% steps after each second scan. When maximal placebo responses had been reached (scan 12), the patient was informed that an opioid “antagonist” was going to be injected, and three further scans were obtained. In reality, a saline solution was injected. After each scan, the patient rated pain intensity and pain unpleasantness on a 10-point rating scale (0 = no pain and 10 = worst pain imaginable). PET data were preprocessed and analyzed using Statistical Parametric Mapping (SPM2) as described elsewhere.10Results were thresholded for significance at P < 0.001 (small volume corrected for multiple comparisons at P < 0.01). Coordinates (x, y, z) of peak voxels are given in standardized stereotaxic space.
The patient showed a prominent placebo response throughout the 50-day observation period (fig. 1B). Mean pain ratings decreased from 6.7 ± 0.6 (week preceding placebo) to 0.6 ± 0.7 (P < 0.0001). During placebo holidays, pain intensity ratings rapidly returned to prestudy values (fig. 1C). Naloxone did not abolish the placebo response (P > 0.05, paired t test). The patient was pain free at the moment of the naloxone test and remained pain free for 60 min after injection. During PET scanning, pain ratings gradually decreased after epidural placebo, and the patient was pain free during scans 9–12. Suggestion that an opioid antagonist was given (scans 13–15) made the pain increase to prestudy levels. PCA-related rCBF decreases were identified bilaterally in medial thalamus (x =−2, y = 18, z = 12 mm; Z value = 4.81), rostral anterior cingulate cortex (rACC; x =−14, y = 40, z = 18 mm; Z value = 4.01), and left nucleus accumbens (x =−10, y = 10, z =−8 mm; Z value = 3.81) (fig. 2A). Figure 2Billustrates the relation between PCA-related changes in thalamic rCBF and reported pain ratings. A similar rCBF time course was observed for rACC and accumbens. We observed no significant rCBF increases. Finally, placebo-induced changes in functional connectivity (i.e. , cross-correlation in rCBF, as described in Laureys et al. 11) were observed between the previously identified rACC and the right superior frontal gyrus (area 10; x = 24, y = 68, z =−4; Z value = 3.45).
We present a case of long-term placebo analgesic response in a clinical pain case. Our patient showed a long-lasting average pain reduction of more than 90%, which stands in sharp contrast with the moderate placebo effects observed in most experimental pain studies.3–6During placebo holidays, pain returned to prestudy levels within a couple of hours. When placebo treatment was resumed the next morning, pain ratings rapidly decreased again within the same time span. This finding is difficult to reconcile with the idea that regression to the mean or natural course of the disease are at the basis of the observed changes in pain perception.8We tested our patient with an epidural catheter for 50 days, which is much longer than in our routine clinical practice. We are aware that this unusually long period of long-term outpatient epidural administration, partly imposed by the study demands, carries the risk of meningitis.
The purported role of the endogenous opioid system in the mediation of the placebo response remains an issue of debate. Behavioral studies have shown that the placebo response can be blocked by the opioid antagonist naloxone.9,12PET studies further showed a significant overlap in the brain areas activated by an opioid and placebo.4However, some forms of placebo analgesia—mainly in clinical pain conditions or in experimental pain conditions using longer-lasting pain stimuli, which more closely resemble clinical forms of pain—are not or are only partially antagonized by naloxone.13,14The placebo response in our patient was not antagonized by naloxone. In a second placebo responder, we got similar results. Additional information regarding this patient is available on the Anesthesiology Web site at http://www.anesthesiology.org. The dose and infusion rates were the same as those used in studies that reported successful blockade of the placebo analgesic response in experimental pain.9Our data therefore provide additional evidence that some forms of placebo analgesia are not opioid mediated. One of the possible explanations for the lack of effect of naloxone is that opioids play no role in placebo effects that are not opioid conditioned.9
Placebo analgesia was associated with significantly reduced activity in the medial thalamus, rACC, and nucleus accumbens. So far, only two brain imaging studies have shown placebo-induced rCBF reductions in the pain matrix,5,15which may be explained by the low amplitude of the placebo analgesic responses in these studies. To be able to dissociate placebo effects from possible time-related changes in rCBF, we added a nocebo suggestion, which resulted in a sharp increase in pain ratings and a concomitant rCBF increase in the identified areas (fig. 2B). Placebo-related negative correlations between activity in right superior frontal gyrus area 10 and rACC suggest a top-down modulation from the prefrontal cortex on activity in part of the pain matrix.16We did not observe increased activity in periaqueductal gray, perigenual anterior cingulate cortex, or orbitofrontal cortex as shown in previous studies.3–6This might reflect a type II error due to the limited statistical power (n = 1). At lower level for significance the orbitofrontal cortex (coordinates x = 12, y = 30, z =−24 mm; voxel-level P = 0.008, cluster-level P < 0.05) did show PCA-related rCBF increases. The periaqueductal gray is often a difficult area to assess because of its small size and partial volume effect. Also, activity in these brain areas might be more transient and in direct relation to the initial instigation of the placebo treatment, making them undetectable for longer-lasting placebo responses.
In conclusion, our data further underscore the powerful contribution of placebo in clinical pain practice. Placebo analgesia was not blocked by naloxone, suggesting the involvement of nonopioidergic mechanisms, and correlated with a deactivation in parts of the pain matrix, possibly under top-down control from the prefrontal cortex.
The authors thank Henrik Kehlet, M.D., Ph.D. (Professor, Section of Surgical Pathophysiology, The Juliane Mare Center, Rigshospitalet, Copenhagen, Denmark), for his critical comments on a previous version of the manuscript.