ABSTRACT

Background. Chronic low back pain patients have a high rate of psychopathology, comprised mainly of depression, anxiety, and high levels of neuroticism. We previously found that psychopathology is associated with increased placebo analgesia in this patient group.

Objective. To better understand this finding in the context of other known predictors of placebo response (such as expectations for relief), we performed a detailed analysis of expectations and other possible covariates.

Design. We conducted a double-blind, placebo-controlled, randomized, crossover-designed trial of intravenous morphine vs. placebo in 60 subjects with chronic low back pain and discogenic abnormalities. Patients were stratified into three groups of psychiatric symptom severity (Low, Moderate, and High), based on composite scores on depression, anxiety for pain, and neuroticism scales. Subjects were given intravenous morphine and placebo in random order on separate visits, and completed serial pain ratings over 3 h at each session.

Results. With 20 subjects per group, there were small, but significant differences between groups in baseline pain ratings. No differences were found between groups in age, gender, and radicular pain. Patients with low psychological symptomatology reported 7.7% total pain relief with placebo compared with 23.4% in the Moderate group and 23.5% in the High group (P < 0.05). Expectations were not significantly different between groups, but in the High group expectation levels predicted placebo analgesia (P < 0.001). Neuropathic pain quality was also a predictor (P < 0.05).

Conclusions. This study indicates that high and moderate levels of psychopathology are associated with heightened placebo analgesia in chronic low back pain patients. Expectations were only an influence in the high psychopathology group, and neuropathic pain affects placebo responses. These findings have implications for future research characterizing placebo responders.

Introduction

When placebo responses were first detected in the placebo arm of randomized controlled trials, much effort was immediately undertaken to characterize placebo responders [1,2]. Trialists wanted to eliminate responders and clinicians wanted to learn how to harness placebo effects, resulting in extensive research efforts [3]. The most common method of investigating such placebo responses has been to use a “piggy-back,” retrospective strategy in which baseline psychological and demographic characteristics of participants in medication randomized controlled trials were correlated with responses to placebo. These studies were usually not designed to examine placebo effects in details and were often subject to confounding due to: 1) retrospective designs and poor power; 2) lengthy periods of follow-up with concomitant effects of regression to the mean and spontaneous improvement; 3) poorly defined inclusion criteria that studied identical complaints with heterogeneous underlying pathology; and 4) absence of within-subject controls to assess responsivity to change. The conclusion of a review of these efforts to find a responder is, “many variables were identified as being associated with placebo effects, but there was little or no agreement [between trials] about which variables contributed.”[4] This conclusion is similar to four earlier reviews [5–8]. Nonetheless, two related psychological factors relatively consistently emerge from these data, which showed a tendency to predict a heightened placebo response, and deserve further investigation—high levels of anxiety and neuroticism [4,7,9].

Psychiatric comorbidity is one of the most important predictors of poor treatment outcome in patients with chronic pain, particularly in chronic low back pain [10–15]. The primary psychopathologies in pain patients are major depression, anxiety, and high levels of neuroticism, which afflict at least 20% of low back pain sufferers in the general population [16] and between 50% and 80% of those in clinic settings [17–20]. As such, we predicted that those with psychiatric comorbidity would report less opioid analgesia and greater placebo analgesia, compared with a group with little psychopathology. In testing these hypotheses, we reported that those with high levels of psychopathology had significantly less analgesia when given intravenous (IV) morphine, compared with a group with low levels of psychopathology [21]. We also observed that the low back pain patients with high psychopathology had a greater placebo response. This finding, however, has not been thoroughly analyzed nor discussed.

This report is a detailed analysis of the placebo response in a randomized, double-blinded, controlled study, designed to separately examine placebo analgesia. We prospectively evaluated the role of psychopathology in opioid and placebo analgesia in patients with discogenic low back pain, using within-subject controls and an active drug with a short duration of action.

Methods

Subjects

Patients with chronic low back pain were recruited from the pain management and arthritis clinics of an urban teaching hospital, as well as through an e-mail announcement distributed to employees of the hospital. Subjects provided written informed consent prior to being enrolled, in accordance with and approved by the hospital's Institutional Review Board. The consent form describes placebo as “an inactive liquid that contains no medication.” Subjects were told that they will be receiving morphine and placebo at separate sessions in randomized, blinded order. Subjects were included if they: 1) were between the ages of 21–70 years; 2) were averaged at least 4 on a 0–10 scale of pain intensity; 3) had no back surgery within the past year; 4) were not having pain management procedures during the study period; 5) were not taking long-acting oral or intrathecal opioids; 6) had no history of opioid abuse; and 7) had discogenic low back pain of at least 6-month duration, either with or without radicular pain. Two physicians had to agree that the low back pain was mainly discogenic using the results of a thorough history and physical examination, and either computed tomography or magnetic resonance imaging (MRI) findings, confirming disc disease.

Study Procedures

These are detailed at length in a previous report of the morphine analgesia findings from this experiment, including sample size calculations and the model for classifying psychopathology [21]. This study was designed with sufficient power to separately examine morphine and placebo findings. In sum, this was a double-blind, placebo-controlled, randomized, crossover trial of IV morphine vs. placebo in 60 patients with discogenic low back pain. Subjects were divided into three groups of 20 patients each of psychological symptom severity (Low, Moderate, or High), using cutoff scores to specify the degree of psychopathology. To be in the High group subjects had to have high scores on two out of three of the psychological symptom measures. The Low group had low scores on two out of three of these scales, and the Moderate group was all others in between.

At the initial visit, subjects completed a medical and psychiatric evaluation to determine suitability for study inclusion. The subjects then completed a battery of pain and functional measures, as well as psychological symptom scales. These included:

  • Numerical Rating Scale of Pain (NRS): This is an 11-point measure of pain intensity, from 0 (no pain) to 10 (worst possible pain) [22].

  • Neuropathic Pain Questionnaire, Short Form (NPQ): This recently validated scale describes the presence or absence of neuropathic pain qualities, using self-rated descriptive terms for neuropathic pain symptoms such as burning or numbness [23]. It has a predictive accuracy for neuropathic pain of 73%.

  • Oswestry Disability Index (ODI): The ODI is an extensively used 10-item scale to describe the level of disability in patients with chronic low back pain [24].

  • Beck Depression Inventory II (BDI): The 21-item BDI has been shown to have good sensitivity and specificity for major depression in chronic pain patients, provided that a higher cutoff score for major depression is used to account for somatic symptoms due to pain [25,26]. Cutoff scores of >20 (high), and <14 (low) were adopted for depression symptoms [25].

  • Pain Anxiety Symptoms Scale (PASS): This 40-item scale has been reliably used to assess the level of anxiety in pain patients [27,28]. Based on published scores among low back pain patients, a score above the 75th percentile was considered high and less than the 25th percentile was low for anxiety symptoms [29].

  • NEO Personality Inventory-Short Form (NEO-FFI), Neuroticism Subscale: The 60-item NEO-FFI is a widely used personality measure [30] that assesses five domains: neuroticism, extraversion, openness, agreeableness, and conscientiousness [31]. It has been used extensively with pain patients and demonstrates good psychometric properties [32–35]. In reference to general population norms, NEO-FFI “T” scores greater than the 75th percentile for neuroticism was considered high, and less than the 50th percentile was low for neuroticism.

The results of the above scales as well as demographic information are summarized in Table 1. Compared with the Low psychopathology group, the High group tended to be unmarried, had less formal education, and were more likely to have an active worker's compensation claim or personal injury litigation. A significantly higher percentage of the Low group was working. There were small, but significantly different baseline pain ratings among groups (mean for all subjects 6.1/10; P < 0.05), but only the Low vs. the Moderate groups were significantly different from each other (P < 0.05).

Table 1

Demographics and pain history

Variable Psychopathology Group (N = 20/group)
 
Significance 
Low Moderate High 
Age (years), mean 43.4 ± 11.4 42.65 ± 12.1 44.2 ± 8.6 NS 
Sex, % female 65 75 70 NS 
Work status, % working 75 40 25 χ2 = 10.6** 
WC. or Lit.,% yes  0 10 20 NS 
Maritial status, % married 45 25 20 NS 
Education 
  % grade school  0  5 15 NS 
  % high school 30 30 45  
  % college 50 60 35  
  % graduate or professional 20  5  5  
Pain duration (years), mean 10.7 ± 7.8 10.8 ± 7.8  9.2 ± 6.8 NS 
Baseline pain (0–10)  5.5 ± 1.5  6.6 ± 1.5  6.2 ± 1.2   * Low vs. Moderate* 
Radicular pain, % yes 70 75 70 NS 
Neuropathic pain, % positive 40.0 55.0 65.0 NS 
Disability index, % 31.1 ± 10.1 43.0 ± 13.5 47.7 ± 10.3   ** 
Variable Psychopathology Group (N = 20/group)
 
Significance 
Low Moderate High 
Age (years), mean 43.4 ± 11.4 42.65 ± 12.1 44.2 ± 8.6 NS 
Sex, % female 65 75 70 NS 
Work status, % working 75 40 25 χ2 = 10.6** 
WC. or Lit.,% yes  0 10 20 NS 
Maritial status, % married 45 25 20 NS 
Education 
  % grade school  0  5 15 NS 
  % high school 30 30 45  
  % college 50 60 35  
  % graduate or professional 20  5  5  
Pain duration (years), mean 10.7 ± 7.8 10.8 ± 7.8  9.2 ± 6.8 NS 
Baseline pain (0–10)  5.5 ± 1.5  6.6 ± 1.5  6.2 ± 1.2   * Low vs. Moderate* 
Radicular pain, % yes 70 75 70 NS 
Neuropathic pain, % positive 40.0 55.0 65.0 NS 
Disability index, % 31.1 ± 10.1 43.0 ± 13.5 47.7 ± 10.3   ** 
*

P < 0.05;

**

P < 0.01.

Chi-square, analysis of variance (anova), and pairwise comparisons.

Active worker's compensation or personal litigation case.

NS = nonsignificant.

For those subjects on short-acting opioids at the time of enrollment, they were given a weaning schedule and had to be completely off of the medication for 2 weeks prior to the first drug infusion session. This was confirmed by a negative urine toxicology screen for opioids. Subjects were allowed to continue their other medications, including on the days of drug infusion. The breakdown of medications is illustrated in Table 2. Eleven of 60 subjects were taking short-acting opioids at the time of enrollment; the distribution of these subjects was not significantly different between groups; and the average daily dose of opioids in morphine equivalents was approximately 30 mg. In terms of other pain medications, a significantly greater number of subjects in the High group were taking anticonvulsants (P < 0.0001).

Table 2

Pain and psychiatric medications

Medication Psychiatric Group (N = 20/group)
 
Significance 
Low Moderate High 
Opioids, No. of patients (%)  2 (10)  6 (30)  3 (15) NS 
Average daily dose, mg MS04 22.5 30 30 NS 
NSAIDS, % 50 50 60 NS 
Tramadol, % 15 15 25 NS 
Muscle relaxants, %  5 10 25 NS 
Antidepressants, all, % 30 35 55 NS 
TCAs, % 10 10 10 NS 
Anticonvulsants, % 10 10 60 χ2 = 18.4* 
Benzodiazepines, %  5 20 25 NS 
Antipsychotics, %  0 10 15 NS 
Medication Psychiatric Group (N = 20/group)
 
Significance 
Low Moderate High 
Opioids, No. of patients (%)  2 (10)  6 (30)  3 (15) NS 
Average daily dose, mg MS04 22.5 30 30 NS 
NSAIDS, % 50 50 60 NS 
Tramadol, % 15 15 25 NS 
Muscle relaxants, %  5 10 25 NS 
Antidepressants, all, % 30 35 55 NS 
TCAs, % 10 10 10 NS 
Anticonvulsants, % 10 10 60 χ2 = 18.4* 
Benzodiazepines, %  5 20 25 NS 
Antipsychotics, %  0 10 15 NS 
*

P < 0.0001.

Chi-square and anova.

NSAIDS = nonsteroid anti-inflammatory drug; TCA = tricarboxylic acid; NS = nonsignificant.

Expectations Assessment

At the end of the first session patients were asked to rate their anticipated pain level from 0 to 10 after being given morphine (Expectations for Relief Scale [ERS]). A lower rating indicates a positive expectation for relief. At visits 2 and 3 prior to medication infusion, they were asked to rate their anticipated pain level from 0 to 10 after being given the medication at the day's visit. Thus, expectations could be examined separately for whether they predicted morphine or placebo analgesia.

Primary Outcome Measure

Repeated administration of a Numerical Rating Scale of Pain Relief (NRSR, 0 = no pain relief, 10 = total relief) was used to calculate percent total pain relief (TOTPAR) for each infusion session [36], the primary analgesia outcome.

Infusion Protocol

Approximately 1 week after the initial screening visit, subjects returned and completed baseline pain rating (NRS) and ERS. They then received an IV bolus injection of either morphine 0.075 mg/kg ideal body weight (approximately 5 mg) or saline placebo. Previous studies of IV opioids have found IV saline to be an effective placebo control [37]. An active placebo, such as diphenhydramine or diazepam, was not chosen because these medications are anxiolytics. In the groups with moderate or high psychopathology, they may blunt the affective components of pain to a greater degree than in those with low psychopathology.

Three study nurses, blinded to which group each subject belonged, administered the medications. Both the investigators and study nurses were also blinded to treatment order. A 10-mL syringe containing either morphine or saline was prepared by the investigational drug pharmacy of our institution, which randomized treatment order in a block design. Each nurse administered the medication through a peripheral IV and monitored appropriate vital signs. The medication sessions were performed in the General Clinical Research Center (GCRC) of our institution, in a private hospital room. Subjects completed the NRSR every 20 min for 3 h. Patients had to remain in the hospital room for the entire session, and were permitted to walk, sit, or lie down. However, they could not lie down in the hospital bed for greater than 5 min of each 20-min interval (monitored by a study nurse).

Approximately 1 week after the first infusion session, subjects returned for the second infusion session, crossed over to either morphine or placebo, and identical pain rating procedures were performed.

Statistical Analysis

All data were analyzed with SPSS (Statistical Package for the Social Sciences v.11.5, SPSS Inc., Chicago, IL, USA). Relations among demographic data, interview items, and questionnaire data were analyzed using Pearson Product Moment correlations, chi-square, and analysis of variance (anova). anova was used to examine the relationship between psychological group and TOTPAR for placebo sessions. To determine which groups differed from each other, we used either Tukey (in the case of equal variances between groups) or Games-Howell (in the case of unequal variances between groups) corrections for multiple comparisons. Post hoc analyses included an analysis of covariance (ancova) to test other main predictors and covariates.

Results

Placebo Analgesia Results

No differences were found among groups on ideal body weight (average 66.4 kg) and morphine dose (average 5.0 mg). Table 3 and Figure 1 display the analgesia results for the primary outcome variable (TOTPAR) for the placebo infusion. Both the High and Moderate groups had significantly greater placebo analgesia than the Low group (23.5%, 23.4%, and 7.7% TOTPAR, respectively; F2,57 = 4.10, P = 0.022).

Table 3

Placebo analgesia results: percent total pain relief (TOTPAR)

Variable Psychopathology Group (N = 20/group)*
 
Group Significance (PPairwise Comparisons 
Low Moderate High 
Placebo 7.7 ± 12.6 23.4 ± 22.6 23.5 ± 23.2 0.022 High vs. Low, P = 0.03; Moderate vs. Low, P = 0.03 
Variable Psychopathology Group (N = 20/group)*
 
Group Significance (PPairwise Comparisons 
Low Moderate High 
Placebo 7.7 ± 12.6 23.4 ± 22.6 23.5 ± 23.2 0.022 High vs. Low, P = 0.03; Moderate vs. Low, P = 0.03 
*

Means ± SD (the higher the number, the greater the analgesic response).

anova group comparison.

anova pairwise comparisons.

Figure 1

TOTPAR results. a, b = significant pairwise comparisons: a = 0.03, b = 0.03.

Figure 1

TOTPAR results. a, b = significant pairwise comparisons: a = 0.03, b = 0.03.

Expectations

Table 4 displays the expectations outcomes, and Table 5 displays correlations between expectations and analgesia. Because expectations were rated at each visit, expectations could be compared for differences between visits, differences between groups, and for the effect of randomization order (morphine or placebo at the first session), using repeated measures anova and Pearson correlations. Predrug expectations were collected at the first visit, prior to any medication infusion and reflect a subject's estimate of how morphine will lower their pain level on a 0–10 scale. A lower estimate of pain after medication infusion indicates an expectation for some relief. Thus, lower numerical values on Table 4 represent greater expectations for relief, and the negative coefficient values of expectations to morphine and placebo analgesia in Table 5 represent a positive correlation to pain relief. To examine the role of randomization order, we compared the placebo expectations between those who received placebo the first vs. the second session.

Table 4

Expectations outcomes

Variable Psychopathology Group*
 
Group Significance (P) 
Low Moderate High 
Predrug expectations 1.4 ± 1.3 (20) 2.2 ± 2.6 (20) 2.6 ± 1.9 (20) NS 
Placebo expectations§ 2.0 ± 1.3 (20) 2.5 ± 2.8 (20) 2.8 ± 1.6 (20) NS 
Variable Psychopathology Group*
 
Group Significance (P) 
Low Moderate High 
Predrug expectations 1.4 ± 1.3 (20) 2.2 ± 2.6 (20) 2.6 ± 1.9 (20) NS 
Placebo expectations§ 2.0 ± 1.3 (20) 2.5 ± 2.8 (20) 2.8 ± 1.6 (20) NS 
*

Means ± SD (N).

anova.

Expectations of pain level (0–10) after being given morphine (rated at initial visit).

§

Expectations of pain level after being given medication, rated at the session they received placebo.

NS = nonsignificant.

Table 5

Analgesia and expectations correlations

Pearson Correlations (N = 60) Mor TOT Plac TOT P-D Exp Mor Exp Plac Exp 
Morphine TOTPAR (Mor TOT)      
Placebo TOTPAR (Plac TOT)  0.10     
Predrug expectation (P-D Exp) −0.29*  0.05    
Morphine expectation (Mor Exp) −0.24  0.05 0.54**   
Placebo expectation§ (Plac Exp) −0.24 −0.15 0.39** 0.69**  
Random order −0.01  0.17 0.31* 0.05 0.04 
Pearson Correlations (N = 60) Mor TOT Plac TOT P-D Exp Mor Exp Plac Exp 
Morphine TOTPAR (Mor TOT)      
Placebo TOTPAR (Plac TOT)  0.10     
Predrug expectation (P-D Exp) −0.29*  0.05    
Morphine expectation (Mor Exp) −0.24  0.05 0.54**   
Placebo expectation§ (Plac Exp) −0.24 −0.15 0.39** 0.69**  
Random order −0.01  0.17 0.31* 0.05 0.04 
*

Correlation is significant at the 0.05 level (two-tailed).

**

** Correlation is significant at the 0.01 level (two-tailed).

Expectations of pain level (0–10) after being given morphine (rated at initial visit).

Expectations of pain level after being given the day's medication, rated at the morphine session.

§

Expectations of pain level after being given the day's medication, rated at the placebo session.

Randomization order for receiving morphine (either the second or third visit).

The primary finding is that the High and Moderate groups did have lower expectations for pain relief prior to both the morphine and placebo injections, but these differences were not significant. Importantly, there were no significant differences found in expectations between visits (within-subject change over time), between groups, or in those who received placebo the first vs. the second session. Table 5 indicates that predrug expectations were moderately correlated to morphine TOTPAR (P < 0.05) and not correlated to placebo TOTPAR. Predrug expectations were strongly correlated to morphine expectations (P < 0.01) and moderately correlated to placebo expectations (P < 0.01). The expectations at the morphine and placebo sessions were also strongly correlated to each other (P < 0.01). Morphine analgesia had a poor relationship to placebo analgesia. Randomization order was only correlated to predrug expectations, which we consider to be a spurious finding as randomization order was double-blinded. In sum, an aggregate analysis of all three groups reveals that expectations did not predict placebo analgesia. Each subject's rating of expectations at the initial visit was moderately correlated to their rating at either the morphine or placebo session, regardless of randomization order.

Sensitivity Analysis. Testing the Relationship Between Psychopathology and Placebo Analgesia

To test the strength of the relationships among baseline pain level, the level of psychopathology, and placebo analgesia, potential confounders were included in exploratory correlational, anova, and ancova analyses. Advantages of these analyses are that potential confounders can be examined as independent predictors of placebo response, for their interaction effects within different subgroups, and can be controlled for in testing the primary predictor variable. Pearson correlations were used to test the relationship between baseline pain level and scores on the depression, anxiety, and neuroticism scales. Since neuropathic pain was measured as a dichotomous variable along a continuum (present or absent), a biserial correlation was used to test the relationship between baseline pain level and neuropathic pain qualities. Baseline pain level was significantly correlated with the level of depression, anxiety, or the neuropathic quality of the pain (correlation coefficients = 0.3–0.36, P < 0.05), but not neuroticism. A greater score on the depression or anxiety scale or the presence of neuropathic pain was associated with a higher baseline pain rating.

In terms of the anova, for the dichotomous or ordinal variables: opioid use, anticonvulsant use, education level, work status, workers' compensation/personal injury litigation involvement, neuropathic pain rating, radicular pain, randomization order, marital status, race, and gender were tested as main effects in addition to psychopathology group using anova. For the continuous variables: pain duration, baseline pain rating, age, and expectations of relief were tested as covariates using ancova.

Of these main effects and covariates, work status, neuropathic pain, and predrug expectations proved to be important (Table 6). Of note, despite differences in baseline pain ratings between groups, this was not a significant covariate. It was neither necessary to control for baseline pain rating in testing the effect of psychopathology level on placebo analgesia in this patient cohort, nor did this alter the strength of the relationship between psychopathology group and placebo analgesia. In each of the alternative models described in Table 6, psychopathology group was still a significant predictor of placebo analgesia (P= 0.001–0.05). Work status was a confounder such that regardless of group assignment, those who were working had greater placebo analgesia. Controlling for work status improves the model (R2 = 0.19, P < 0.01), which is reflected in greater differences in estimated TOTPAR means between the Low and High groups (P < 0.05). Similarly, neuropathic pain confounds the relationship between psychopathology group and placebo analgesia such that those with neuropathic pain had greater placebo responses. Even though the differences in estimated TOTPAR means between the Low and High groups are more narrow, it is still significant (P < 0.05). The proportion of variability in analgesia between subjects explained by both the work and neuropathic pain models was significantly different from the original model including psychopathology group only as a predictor (P < 0.05, for the difference in R2 between models).

Table 6

Alternative anova and ancova models

Psychopathology Group Model (TOTPAR Estimated Means ± SE)
 
Psych Group Working Neuropathic Pain§ Expectation 
Low (20)  7.7* (4.5)  4.8* (4.6)  8.8* (4.4)  7.0* (4.6) 
Moderate (20) 23.4* (4.5) 24.5* (4.4) 22.8* (4.4) 24.5* (4.2) 
High (20) 23.5* (4.5) 26.4* (4.6) 21.8* (4.4) 21.0* (4.1) 
P (model)  0.022  0.008  0.008  0.0001 
  R2  0.13  0.19  0.19  0.35 
P (R2 change from the Crude model)  <0.05 <0.05 <0.001 
Psychopathology Group Model (TOTPAR Estimated Means ± SE)
 
Psych Group Working Neuropathic Pain§ Expectation 
Low (20)  7.7* (4.5)  4.8* (4.6)  8.8* (4.4)  7.0* (4.6) 
Moderate (20) 23.4* (4.5) 24.5* (4.4) 22.8* (4.4) 24.5* (4.2) 
High (20) 23.5* (4.5) 26.4* (4.6) 21.8* (4.4) 21.0* (4.1) 
P (model)  0.022  0.008  0.008  0.0001 
  R2  0.13  0.19  0.19  0.35 
P (R2 change from the Crude model)  <0.05 <0.05 <0.001 
*

Significance pairwise comparisons, High vs. Low and Moderate vs. Low, P < 0.05.

No adjustment, psych group is the only predictor.

Controlling for work status: psych group + work.

§

Controlling for neuropathic pain: psych group + neuropathic pain.

Controlling for predrug expectations: psych group + expectation + psych group × expectation.

The addition of predrug expectations improves the model most significantly, such that 35% of the variability in analgesia between subjects can be explained by an ancova model that includes psychopathology group, predrug expectations, and a predrug expectations–psychopathology group interaction (P < 0.0001). In this model, predrug expectations are an effect modifier because the interaction was significant (P < 0.002), not expectations by itself. Expectations were only an important interaction in the High group, such that those with greater predrug expectations for relief in this group had greater placebo analgesia. Thus, while expectations were not significantly different between groups, they were a significant predictor of placebo analgesia within the High psychopathology group.

Discussion

Injections are powerful placebos [38,39], and our findings indicate that this is more so the case in those with psychopathology. Patients with discogenic low back pain and moderate or high levels of psychopathology were found to have three times the rate of analgesia to an IV saline injection compared with a matched group with low levels of psychopathology. The groups were well matched on multiple demographic, social, and pain syndrome variables. Major differences between groups were in the social and disability realms, such that the High group had fewer working and married, were less educated, and had a greater level of disability. Similar to previous studies, higher baseline pain ratings were associated with elevated levels of depression or anxiety, or a neuropathic quality to pain [11,23,40]. There is a possibility that there is an interaction between psychopathology level and neuropathic pain qualities, and future studies are needed to explore this issue. Moreover, there was not a significant difference in the distribution of neuropathic pain between groups. While it is known that these factors co-occur, it is unclear in the literature what the relationship is between them. In sum, in concordance with previous findings [41], the subgroup of patients with pain and psychopathology is characterized by a syndrome of dysfunction in psychiatric, social, and physical arenas. Hence, one can view these factors as characteristics of those with chronic pain and psychiatric comorbidity.

Shapiro and Shapiro [4] and Turner et al. [9] have previously reported that greater psychological symptoms predict an enhanced placebo response for chronic pain, but opioid analgesia was not specifically addressed. Our findings extend the conclusions of Turner et al. to quantify how the degree of psychopathology heightens the response to placebo analgesia. Studies in related populations with major depression but no chronic pain have also found strong placebo responses, averaging 30% in the medication trials for depression [42].

Another relevant issue is the within-subject relationship of morphine and placebo response. This relationship between opioid and placebo analgesia was addressed by Benedetti et al. in a double-blind study of IV buprenorphine in 57 subjects, administered for postoperative pain following a lobectomy for lung cancer [43]. Following a study design that introduced some element of conditioning to the effects of IV opioid before administering placebo, they found a high correlation between those who responded to opioid and those who responded to placebo. We found the opposite; there was a low correlation between those who responded to morphine and those who responded to placebo in our study population. However, the two results are not directly contradictory. The study populations and designs are quite different, and it is very possible that the participants in the study by Benedetti et al. did not have as high a rate of psychopathology. We were specifically examining the impact of psychopathology, which proved to be an overriding determinant of placebo analgesia in our study groups.

We can only speculate as to how psychopathology amplifies placebo analgesia for chronic pain. On a psychological level, those with negative affective symptoms have a greater awareness of and “hypervigilance” toward bodily symptoms [44], leading to symptom amplification. This, in turn, may have resulted in the Moderate and High groups having a smaller nociceptive contribution and a larger somatoform component to their pain perception. Presumably, the somatoform component is mediated by psychiatric processes involving mind and brain, which may play a greater role in the Moderate and High groups' analgesia responses compared with the Low group. For instance, it is thought that symptom amplification may be mediated by lesser descending pain inhibition in subcortical brain areas [45]. Functional MRI experiments have conclusively shown that placebo analgesia is mediated by activity in several cortical and subcortical brain regions [46–48]. Perhaps the physiological analgesia mechanisms underlying the activity in these brain areas are altered in those with psychopathology, leading to enhanced placebo responses. It is difficult to reconcile this notion, however, with the work of Levine et al. [49], Petrovic et al. [47], Finniss and Benedetti [50], and others who have found that placebo analgesia is at least partially mediated by the release of and the response to endogenous opioids. As the group in our study with heightened placebo analgesia had poor opioid analgesia, endogenous opioid response is unlikely to mediate the placebo analgesia findings in our study. This is consistent with the recent findings of Vase and colleagues in a study of placebo responses in irritable bowel syndrome patients [51].

Expectations for outcome have been noted to influence the response to active treatments [52], including those for low back pain [53,54]. But it is less clear whether expectations modulate the placebo response to chronic pain treatments [55]. Expectations for relief do influence the magnitude of placebo analgesia in experimental pain studies [56–58]. We did not, however, find such a relationship in our study of chronic low back pain patients. Expectations did not influence the degree of placebo analgesia in a uniform manner across the three groups. The ancova analyses indicated that only in the High group did increased expectations for relief predict greater analgesia, accounting for 35% of the variance in analgesia scores. There was a significant correlation between predrug expectations and morphine analgesia, but not placebo analgesia. Thus, expectations for relief may mediate the morphine, but not the placebo analgesia results in our study (except in the High group), which is consistent with the findings of Turner et al. [55].

In terms of other factors predicting placebo analgesia, only work status and neuropathic pain rating improved the ancova model in addition to psychopathology group, explaining 19% of the variability in individual analgesia scores. It is intriguing that those who were working had more placebo analgesia, and to our knowledge, work status has never been investigated as a factor predicting placebo analgesia. This may be a spurious finding, or there may indeed be a reason for this result, which deserves further study. Patients with major depression in medication trials who have heightened placebo responses also have a high rate of unemployment [42]. This result is seemingly at odds with our findings, which suggest, however, that employment is associated with increased placebo analgesia. But the depression medication trials did not control for the level of depression or other psychopathology as they only enrolled patients with major depression. Thus, no conclusions about the influence of work status on placebo analgesia can be inferred from those studies and those results cannot be directly compared with our findings.

In sum, the ancova analyses have identified three factors (in addition to psychopathology group) influencing placebo analgesia in our subjects, which deserve further investigation: 1) expectations for relief; 2) work status; and 3) neuropathic pain qualities. In particular, there are very few data examining the impact of different pain syndromes (e.g., nociceptive vs. neuropathic) on placebo response. While neuropathic pain quality was associated with elevated baseline pain in our subjects, there was no significant relationship between initial pain level and placebo analgesia in the ancova analyses. Thus, neuropathic pain does appear to be a unique factor influencing placebo analgesia, despite its relationship to other covariates. Even though expectations, work status, and neuropathic pain influenced our findings, they did not dilute the strength of the relationship between psychopathology and placebo analgesia. This highlights the salience of psychopathology to any study of placebo analgesia. The double-blind, placebo-controlled, random-ordered design of this study facilitated tight control of experimental conditions and confounders.

One might be tempted to conclude that the elevated placebo response in the High psychopathology group indicates that they have “psychogenic pain.” Inclusion criteria selecting for discogenic pain suggests that the High group had a nociceptive source for pain in the body that may have been amplified or further modulated by affective processes in the brain. Our findings suggest, however, that there may be some common characteristics among placebo responders to low back pain treatments, notably, their level of negative affective symptoms. This study documents associations between psychopathology and analgesia, and no cause–effect relationship can be determined. While psychopathology is associated with heightened IV placebo analgesia in patients with low back pain, we cannot assume that this effect carries over to oral medications, nerve blocks, or other pain treatment procedures.

In addition, other limitations merit discussion. First, our study design did not include a natural history control, a relevant issue in placebo studies [59]. It is possible that the subjects in the High group could have had an improvement in their chronic pain over the 3-h observation period due to natural variability in pain levels alone. Neither from this study nor reported in the literature is there evidence to suggest, however, that the variability would have been distributed differently among groups; that is, there are no data to indicate that a high psychopathology group would be more likely to experience spontaneous improvements in pain on an hourly basis than a low psychopathology group of low back pain patients. Although hourly fluctuations in pain levels are possible in any group of chronic pain patients [60], it is likely that this confounder would apply equally to all three groups and thus would not account for the differences in placebo analgesia between groups. In addition, electronic diary data from a study by one of the authors (R.N.J.), gathered from a similar population of low back pain patients at the same institution, indicate that the natural history of chronic low back pain in this patient cohort is that it is unlikely to spontaneously remit below a level of 4/10 [60]. This suggests that while a natural history control may have been useful in our study, it is unlikely to have mitigated our findings.

Second, in monitoring expectations, we did not ask subjects at the initial session whether they thought the placebo injection would affect their pain level. Our expectations data reflect primarily relief expectations from morphine. Similarly, we did not study desire for relief, a factor noted by Vase and colleagues to predict placebo responses [61]. However, in subsequent studies this same group found that desire for relief was not an independent predictor of placebo analgesia, and that it significantly interacts with expectations for relief [51]. Moreover, in experimental pain studies designed to directly compare the impact of expectancy vs. desire for relief on placebo analgesia, Price and colleagues concluded that expectations were a significant predictor of analgesia, not desire for relief [56]. Third, we do not know whether subjects with poor analgesia to a single IV saline injection would achieve better pain relief had they been given additional doses. Conversely, it is unknown whether those with placebo analgesia would experience diminishing results if given additional injections. It is also unknown whether greater anticonvulsant use in the High group affected their placebo response. Finally, we used clinical psychiatric examinations instead of structured interviews and stratified the level of psychopathology with the use of scales (which, in themselves, have their own error variances). Future studies may consider using a structured psychiatric interview (such as the primary care evaluation of mental disorder [62]).

Conclusions

Despite these limitations, the results of this study suggest that psychopathology predicts increased placebo analgesia in those with discogenic low back pain. The prospective, double-blind, crossover design provided within-subject controls and permitted a close examination of responsivity to change. Comorbid psychopathology is highly prevalent in chronic low back pain patients and is a relevant factor to any future research characterizing placebo responders among those with chronic pain. In addition to replicating these findings, future studies may also consider the modulating role of various pain conditions on placebo response.

Acknowledgments

This study was performed in the GCRC of Brigham and Women's Hospital, Boston, Massachusetts. This work was conducted while Dr. Wasan was the Lief Post-Doctoral Fellow in the Department of Anesthesiology, Perioperative, and Pain Medicine at Brigham and Women's Hospital, and the Pfizer Medical-Academic Partnership in Pain Medicine Post-Doctoral Fellow. The authors wish to thank Drs. Edgar Ross and Srdjan Nedeljkovic for their enthusiastic support of this study.

References

1
Kaptchuk
TJ.
Powerful placebo: The dark side of the randomized controlled trial
.
Lancet
 
1998
;
351
:
1722
5
.
2
Kaptchuk
TJ.
Intentional ignorance: A history of blind assessment and placebo controls in medicine
.
Bull Hist Med
 
1998
;
72
:
389
433
.
3
DeCraen
AJM
Kaptchuk
TJ
Kleijnen
J.
Placebos and placebo effects in medicine: Historical overview
.
J R Soc Med
 
1999
;
92
:
511
5
.
4
Shapiro
AK
Shapiro
E.
Predicting placebo response
. In:
Shapiro
AK
Shapiro
E
, eds.
The Powerful Placebo: From Ancient Priest to Modern Physician.
 
Baltimore, MD
:
Johns Hopkins University Press
;
1997
:
217
8
.
5
Fisher
S.
The placebo reactor: Thesis, antithesis, synthesis, and hypothesis
.
Dis Nerv Syst
 
1967
;
28
:
510
5
.
6
Doongaji
DR
Vahia
VN
Bharucha
MP.
On placebos, placebo responses and placebo responders. A review of psychological, psychopharmacological and psychophysiological factors
.
J Postgrad Med
 
1978
;
24
:
147
57
.
7
Joyce
C.
Non-specific aspects of treatment from the point of view of a clinical pharmacologists
. In:
Shepard
M
Sartorium
N
, eds.
Non-Specific Aspects of Treatment.
 
Toronto
:
Han Hubers
;
1989
:
57
94
.
8
Honigfeld
G.
Non-specific factors in treatment. I. Review of placebo reactions and placebo reactors
.
Dis Nerv Syst
 
1964
;
25
:
145
56
.
9
Turner
JA
Deyo
R
Loesser
J
Von Korff
M
Fordyce
WE.
The importance of placebo effects in pain treatment and research
.
JAMA
 
1994
;
271
(
20
):
1609
14
.
10
Linton
S.
A review of psychological risk factors in back and neck pain
.
Spine
 
2000
;
25
:
1148
56
.
11
Dworkin
R
Richlin
D
Handelin
D
Brandt
L.
Predicting treatment response in depressed and non-depressed chronic pain patients
.
Pain
 
1986
;
24
:
343
53
.
12
McCreary
C.
Empirically derived MMPI profile clusters and characteristics of low back pain patients
.
J Consult Clin Psychol
 
1985
;
53
(
4
):
558
60
.
13
Andersson
G.
Epidemiological features of low back pain
.
Lancet
 
1999
;
354
:
581
5
.
14
Vendrig
A
Derksen
J
DeMey
H.
Utility of selected MMPI-2 scales in the outcome prediction for patients with chronic back pain
.
Psychol Assess
 
1999
;
11
:
381
5
.
15
Bair
M
Robinson
R
Katon
W
Kroenke
K.
Depression and pain comorbidity: A literature review
.
Arch Int Med
 
2003
;
163
:
2433
45
.
16
McWilliams
L
Goodwin
R
Cox
B.
Depression and anxiety associated with three pain conditions: Results from a nationally representative sample
.
Pain
 
2004
;
111
:
77
83
.
17
Dersh
J
Gatchel
R
Polatin
P
Mayer
T.
Prevalence of psychiatric disorders in patients with chronic work-related musculoskeletal pain and disability
.
J Occup Environ Med
 
2002
;
44
(
5
):
459
68
.
18
Fishbain
D.
Approaches to treatment decisions for psychiatric comorbidity in the management of the chronic pain patient
.
Med Clin North Am
 
1999
;
83
(
3
):
737
59
.
19
Katon
W
Egan
K
Miller
D.
Chronic pain: Lifetime psychiatric diagnosis family history
.
Am J Psych
 
1985
;
142
(
10
):
1156
60
.
20
Koenig
T
Clark
MR.
Advances in comprehensive pain management
.
Psychiatr Clin North Am
 
1996
;
19
(
3
):
589
611
.
21
Wasan
AD
Davar
G
Jamison
RN.
The association between negative affect and opioid analgesia in patients with discogenic low back pain
.
Pain
 
2005
;
117
:
450
61
.
22
Jensen
MP
Karoly
P.
Self-report scales and procedures for assessing pain in adults
. In:
Turk
DC
Melzack
R
, eds.
Handbook of Pain Assessment.
 
New York
:
Guilford Press
;
2001
:
15
34
.
23
Krause
SJ
Backonja
MM.
The development of a neuropathic pain questionnaire
.
Clin J Pain
 
2003
;
19
(
5
):
315
6
.
24
Fairbank
JCT
Pynsent
PB.
The Oswestry disability index
.
Spine
 
2000
;
25
:
2940
53
.
25
Geisser
ME
Roth
RS.
Assessing depression among persons with chronic pain using the center for epidemiological studies depression scale and the beck depression inventory: A comparative analysis
.
Clin J Pain
 
1997
;
13
:
163
70
.
26
Turner
JA
Romano
JM.
Self-report screening measures for depression in chronic pain patients
.
J Clin Psychol
 
1984
;
40
:
909
13
.
27
McCracken
L
Zayfert
C
Gross
R.
The pain anxiety symptoms scale: Development and validation of a scale to measure fear of pain
.
Pain
 
1992
;
50
:
67
73
.
28
McCracken
L
Gross
R
Aikens
J
Carnrike
C.
The assessment of anxiety and fear in persons with chronic pain: A comparison of instruments
.
Behav Res Ther
 
1996
;
34
(
11
):
927
33
.
29
McCracken
L
Gross
R
Sorg
P
Edmans
T
.
Prediction of pain in patients with chronic low back pain: Effects inaccurate prediction pain-related anxiety
.
Behav Res Ther
 
1993
;
31
(
7
):
647
52
.
30
Cox
B
Borger
S
Taylor
S
Fuentes
K
Ross
L.
Anxiety sensitivity and the five-factor model of personality
.
Behav Res Ther
 
1999
;
37
:
633
41
.
31
Costa
PT
Herbst
JH
McCrae
RR
Siegler
IC.
Personality at midlife: Stability, intrinsic maturation, and response to life events
.
Assessment
 
1992
;
7
(
4
):
365
78
.
32
Evers
A
Kraaimaat
F
Van Reil
P.
Bijlsma
J.
Cognitive, behavioral and physiological reactivity to pain as a predictor of long-term pain in rheumatoid arthritis patients
.
Pain
 
2001
;
9
:
139
46
.
33
Wade
JB
Dougherty
LM
Hart
RP
Cook
DB.
Patterns of normal personality structure among chronic pain patients
.
Pain
 
1992
;
48
:
37
45
.
34
Wade
JB
Dougherty
LM
Hart
RP
Rafii
A
Price
DD.
A cannonical correlation analysis of the influence of neuroticism and extraversion on chronic pain, suffering, and pain behavior
.
Pain
 
1992
;
51
:
67
73
.
35
Groth-Marnat
G
Fletcher
A.
Influence of neuroticism, catastrophizing, pain duration, and receipt of compensation on short-term response to nerve block treatment for chronic back pain
.
J Behav Med
 
2000
;
23
(
4
):
339
50
.
36
Max
MB
Laska
EM.
Single dose analgesic comparisons
. In:
Max
MB
Portenoy
R
Laska
EM
, eds.
The Design of Analgesic Clinical Trials.
 
New York
:
Raven Press
;
1991
:
55
95
.
37
Dellemijn
P
Vanneste
J.
Randomised, double-blind, active-placebo controlled crossover trial of intravenous fentanyl in neuropathic pain
.
Lancet
 
1997
;
349
:
753
8
.
38
Pengel
HM
Maher
CG
Refshauge
KM.
Systematic review of conservative interventions for subacute low back pain
.
Clin Rehabil
 
2002
;
16
(
8
):
811
20
.
39
Kaptchuk
TJ
Goldman
P
Stason
WB.
Do medical devices have enhanced placebo effects?
J Clin Epidemiol
 
2000
;
53
:
786
92
.
40
Nelson
D
Novy
D.
Self-report differentiation of anxiety and depression in chronic pain
.
J Pers Assess
 
1997
;
69
(
2
):
392
407
.
41
Gatchel
R
Dersh
J.
Psychological disorders and chronic pain: Are there cause-and-effect relationships?
In:
Turk
D
Gatchel
R
, eds.
Psychological Approaches to Pain Management: A Practitioner's Handbook
 ,
2nd edition
.
New York
:
Guilford Press
;
2002
:
30
51
.
42
De La Fuente-Fernandez
R
Schulzer
M
Stoessl
AJ.
The placebo effect in neurological disorders
.
Lancet Neurol
 
2002
;
1
:
85
91
.
43
Benedetti
F
Amanzio
M
Baldi
S
et al
The specific effects of prior opioid exposure on placebo analgesia and placebo respiratory depression
.
Pain
 
1998
;
75
:
313
9
.
44
Turk
DC
Monarch
ES.
Biopsychosocial perspective on chronic pain
. In:
Turk
DC
Gatchel
R
, eds.
Psychological Approaches to Pain Management.
 
New York
:
Guilford Press
;
2002
:
3
29
.
45
Millan
MJ.
Descending control of pain
.
Prog Neurobiol
 
2002
;
66
:
355
474
.
46
Lieberman
MD
Jarco
JM
Berman
S
et al
The neural correlates of placebo effects: A disruption account
.
Neuroimage
 
2004
;
22
:
447
55
.
47
Petrovic
P
Kalso
E
Petersson
KM
Ingvar
M.
Placebo and opioid analgesia: Imaging a shared network
.
Science
 
2002
;
295
:
1737
40
.
48
Wager
TD
Rilling
JK
Smith
EE
et al
Placebo-induced changes in fMRI in the anticipation and experience of pain
.
Science
 
2004
;
303
:
1162
6
.
49
Levine
JD
Gordon
NC
Fields
HL.
The mechanism of placebo analgesia
.
Lancet
 
1978
;
2
:
654
7
.
50
Finniss
DG
Benedetti
F.
Mechanisms of the placebo response and their impact on clinical trials and clinical practice
.
Pain
 
2005
;
114
:
3
6
.
51
Vase
L
Robinson
ME
Verne
GN
Price
DD.
Increased placebo analgesia over time in irritable Bowel syndrome (IBS) patients is associated with desire and expectation but not endogenous opioid mechanisms
.
Pain
 
2005
;
115
:
338
47
.
52
Mondloch
MV
Cole
DC
Frank
JW.
Does how you do depend on how you think you'll do? A systematic review of the evidence for a relation between patients' recovery expectations and health outcomes
.
CMAJ
 
2001
;
165
(
2
):
174
9
.
53
Atlas
SJ
Deyo
RA
Keller
RB
et al
The main lumbar spine study part II: 1-year outcomes of surgical and non-surgical management of sciatica
.
Spine
 
1996
;
21
:
1777
86
.
54
Koes
BW
Bouter
LM
Van Der Heijden
GJ.
Methodological quality of randomized clinical trials on treatment efficacy on low back pain
.
Spine
 
1995
;
20
:
228
35
.
55
Turner
JA
Jensen
M
Warms
C
Cardenas
D.
Blinding effectiveness and association of pretreatment expectations with pain improvement in a double-blind randomized controlled trial
.
Pain
 
2002
;
99
:
91
9
.
56
Price
DD
Milling
LS
Kirsch
I
et al
An analysis of factors that contribute to the magnitude of placebo analgesia in an experimental paradigm
.
Pain
 
1999
;
83
:
147
56
.
57
Jensen
MP
Karoly
P.
Motivation and expectancy factor in symptom perception: A laboratory study of the placebo effect
.
Psychosom Med
 
1991
;
53
:
144
52
.
58
Fillmore
M
Vogel-Sprott
M.
Expected effect of caffeine on motor performance predicts the type of response to placebo
.
Pharmacology
 
1992
;
106
:
209
14
.
59
Hoffman
GA
Harrington
A
Fields
HL.
Pain and the placebo: What have we learned?
Perspect Biol Med
 
2005
;
48
(
2
):
248
65
.
60
Jamison
RN
Raymond
SA
Levine
JG
et al
Electronic diaries for monitoring chronic pain: 1-year validation study
.
Pain
 
2001
;
91
:
277
85
.
61
Vase
L
Robinson
ME
Verne
GN
Price
DD.
The contributions of suggestion, desire, and expectation to placebo effects in irritable bowel syndrome patients: An empirical investigation
.
Pain
 
2003
;
105
(
1–2
):
17
25
.
62
Spitzer
RL
Williams
JBW
Kroenke
K
et al
Utility of a new procedure for diagnosing mental disorders in primary care: The Prime-MD 1000 Study
.
JAMA
 
1994
;
272
(
22
):
1749
56
.