Evoked Pressure Pain Sensitivity Is Associated with Differential Analgesic Response to Verum and Sham Acupuncture in Fibromyalgia

Abstract

Objective. Fibromyalgia is a chronic pain condition with few effective treatments. Many fibromyalgia patients seek acupuncture for analgesia; however, its efficacy is limited and not fully understood. This may be due to heterogeneous pathologies among participants in acupuncture clinical trials. We hypothesized that pressure pain tenderness would differentially classify treatment response to verum and sham acupuncture in fibromyalgia patients.

Design. Baseline pressure pain sensitivity at the thumbnail at baseline was used in linear mixed models as a modifier of differential treatment response to sham versus verum acupuncture. Similarly, needle-induced sensation was also analyzed to determine its differential effect of treatment on clinical pain.

Methods and Patients. A cohort of 114 fibromyalgia patients received baseline pressure pain testing and were randomized to either verum (N = 59) or sham (N = 55) acupuncture. Participants received treatments from once a week to three times a week, increasing in three-week blocks for a total of 18 treatments. Clinical pain was measured on a 101-point visual analog scale, and needle sensation was measured by questionnaire throughout the trial.

Results. Participants who had higher pain pressure thresholds had greater reduction in clinical pain following verum acupuncture while participants who had lower pain pressure thresholds showed better analgesic response to sham acupuncture. Moreover, patients with lower pressure pain thresholds had exacerbated clinical pain following verum acupuncture. Similar relationships were observed for sensitivity to acupuncture needling.

Conclusions. These findings suggest that acupuncture efficacy in fibromyalgia may be underestimated and a more personalized treatment for fibromyalgia may also be possible.

 Acupuncture, Fibromyalgia, Quantitative Sensory Testing, Needling Sensation, Pressure Pain Threshold

Introduction

Acupuncture has been used as a therapeutic intervention for over 2,500 years in China [1] and remains widely used in Asian medical systems. Worldwide acceptance of acupuncture as a complementary or alternative medical intervention has recently increased, with estimates of its use ranging from 0.3% (Saudi Arabia) to 2.8% (Australia), with an estimated 0.6–1.4% of Americans utilizing acupuncture each year [2]. A commonly cited reason for seeking acupuncture treatment is the presence of chronic pain [3].

Fibromyalgia (FM) is a chronic widespread pain condition that affects approximately 2–5% of the population in industrialized countries [4,5]. There are few interventions that effectively treat pain in FM, and consequently an estimated 15–20% of FM patients worldwide seek alternative medical care such as acupuncture within two years of diagnosis [5,6]. While some trials have shown efficacy of acupuncture, particularly electroacupuncture, in FM patients, other trials indicate that acupuncture has minimal or no efficacy when compared with sham (placebo) acupuncture treatment [7–9].

This controversy centers on the lack of a well-accepted protocol for sham acupuncture. Indeed, randomized controlled clinical trials of acupuncture for other chronic pain conditions indicate that analgesic effects attributed to needle insertion and manipulation are relatively small when compared with nonspecific effects of the treatment (effect size = 0.26 for sham-controlled trial vs 0.55 for nonacupuncture controls) [10]. However, these trials often lack an assessment of needling sensations and sensitivity to experimental pain as outcomes. This may be critical as the level of sensation patients experience during needling may factor into treatment response. Lundberg and Lund proposed that more sensitive fibromyalgia patients might receive less benefit, or even a worsening of symptoms, from strong needle manipulation [11].

Recent work suggests that quantitative sensory testing (QST) can be used to predict which chronic pain patients are likely to respond to pharmacologic [12–14] as well as nonpharmacologic [15,16] interventions. Interestingly, we have previously demonstrated that QST using evoked experimental pressure pain is associated with a patient’s response to sham acupuncture [17], wherein FM patients that were more sensitive to pressure pain at baseline were more likely to respond to sham, but not verum, acupuncture. Given the uncertainty in the literature surrounding the efficacy of both verum and sham acupuncture, a QST marker could be used to stratify patients into different treatment methods and uncover a difference in verum and sham acupuncture, thus providing a validation of this therapy. Moreover, QST could allow providers to practice personalized medicine in the setting of acupuncture, delivering the most directed and efficacious treatment to their patients.

To explore the veracity of these results, we sought to validate them in a separate cohort of FM patients from our previously reported clinical trial of acupuncture in FM [18]. Our 2005 study determined that both verum and sham acupuncture led to analgesic affects in FM: Needle placement and stimulation did not seem to be critical factors in analgesia [18]. However, while pretreatment QST data was collected in that study, it was not analyzed for its association with treatment response. Moreover, we also collected specific needling sensation outcomes as another metric of sensitivity, and those have not been reported. Given the preliminary findings from our previous trial, we hypothesized that FM patients who displayed greater pain tenderness at baseline would respond preferentially to sham as opposed to verum acupuncture.

In addition, as our study focuses on patients with FM who have been proven to be more sensitive to sensory stimulation [19], we hypothesized that any analgesic effect from strong needling may be masked as the needling sensation itself may counteract the analgesic effects of acupuncture. This theory has been previously been proposed by Lundeberg and Lund [11], and therefore we also hypothesized that FM patients who were more sensitive to pressure pain might have an adverse response to verum acupuncture, particularly those individuals who experienced strong needle stimulations.

Methods

Participants

All subjects in this trial were participants in a previously reported study [18]. All subjects provided written informed consent, and all study protocols were approved by the Georgetown University Institutional Review Board. The original trial analyzed data from 114 participants; however, only 89 had complete needle sensation data, which was not reported previously. The final selected group included 73 subjects with nonmissing pain score measurements. List-wise deletion of missing data was implemented.

Intervention

As previously described, our 2005 study utilized a 2 X 2 factorial design to examine the individual and synergistic effects of both needle location and needle stimulation on the efficacy of acupuncture analgesia [18]. Briefly, subjects were randomly assigned to receive acupuncture needling in one of the following locations: 1) traditional locations with stimulation, 2) traditional locations without stimulation, 3) nontraditional locations with stimulation, or 4) nontraditional locations without stimulation. As there is generally no consensus on what constitutes sham acupuncture [20], we collapsed the 2 × 2 factorial by needle location, creating two groups, traditional locations (with and without stimulation) and nontraditional locations (with and without stimulation). These two groups were designated as verum and sham (or placebo) acupuncture, respectively, in this analysis. Verum acupuncture points included nine traditional acupuncture points (Du 20, right Ear-Shenmen, left Large Intestine 11, right Large Intestine 4, left Gall Bladder 34, left Stomach 36, left Spleen 6, and right Liver 3) used in classical Traditional Chinese Medicine treatment of FM, while sham points did not fall on any acupuncture meridians or points (see previous 2005 study for a diagram of point locations for sham needling). Characteristic de qi needle sensation was sought in the participants receiving manual manipulation by the acupuncturist. Needle manipulation involved lifting and thrusting with even rotation (approximately 12 times at 180 degrees clockwise and counterclockwise). Needles were retained for 20 minutes. Participants were treated with three different doses each for three weeks, such that treatments were performed once per week, followed by twice per week and finally three times per week. A washout period of two weeks was inserted between treatment doses. The flow of participants through the study can be found in Figure 1.

Figure 1

Flow of participants in the study. Only patients with complete sensation data were analyzed.

Outcomes

The main outcome in this study was clinical pain intensity as assessed by a visual analog scale (VAS) ranging from 0 “no pain” to 100 “worst pain imaginable.” Clinical pain was measured longitudinally over time through the three different treatment doses. Needling sensations were also assessed via a questionnaire completed by each participant after every treatment. Participants were asked whether they felt any sensation with needling and if so, to describe the sensation (coded as dull, sharp, or mixed) and to rate the intensity of the sensation (coded as mild, moderate, or severe). Acupuncture needling sensations are typically not rated as painful.

Quantitative sensory testing was conducted using a multiple random staircase method to determine pain threshold as reported previously [17]. Briefly, a pressure inducer was used to increase pressure on the participant’s thumbnail in discreet pressure and time intervals. Pain intensity ratings for each pressure were noted on a 21-point scale [21]. First, the participant’s pain sensitivity range was obtained by inducing stimuli of five-second duration in increasing order with an interstimulus interval of 20 seconds. The initial stimulus was 0.25 kg/cm² and increased by 0.25–0.50 kg/cm² increments each time until either the participant’s pain tolerance or 10 kg/cm² was reached. Next, software titrated three independent series of increasing stimuli based on the method of multiple random staircases [22]. Each staircase was titrated to elicit pain sensations that were rated between 0 and 1 (faint pain), between 7 and 8 (mild to moderate pain), and between 13 and 14 (strong to slightly intense pain) on the 21-box numerical descriptor scale as previously described. Each staircase delivered 12 five-second duration stimuli at 20-second intervals. Only stimulus intensities (in kg/cm²) obtained from the middle staircase (mild to moderate pain) were used to group patients into pressure pain threshold tertiles (low threshold [most painful], medium threshold, and high threshold [least painful]) for further analysis, as described below. This middle staircase has shown to be more associated with clinical pain in this population [23].

Statistical Analysis

Data were analyzed using R 3.3.1. The statistician was blinded with respect to patient management and the conduct of the trial. Descriptive univariate analysis was based on a variety of models and tests specific to the scale of the specific variable being analyzed. Between-group comparisons of continuous variables were performed using a t test (descriptive statistics only), a chi-square test was used for binary variables, a Wilcoxon test was used for ordered categorical variables, and a nominal multinomial logistic model was used to study categorical variables that were not associated with any specific order. In descriptive analyses, the baseline QST pressure pain threshold (tenderness) variable was divided into three separate tertiles (built using one-third and two-third percentiles; low threshold, medium threshold, high threshold), with lower thresholds indicating more tenderness/pain. In the multivariate models, tenderness was treated as a continuous variable and also as a binary variable dichotomized at the median (as a secondary analysis similar to analysis in our 2013 manuscript). Stimulation of acupuncture needle (yes vs no), needling sensation score measured at each of the three treatment dose levels, age, gender (male vs female), and race (white, black, or other) were also included in the analysis.

Because treatments varied over time, the effect of time was not used as a summary of treatment effects. Instead, for any time point, a cumulative acupuncture dose variable was constructed by summing up the number of acupuncture treatments preceding the time point since the end of the most recent washout period or start of the trial on the patient, whichever was first. The cumulative dose dropped to zero at the beginning of the next washout period. The effect of time in this context modeled the slow trend in pain responses under no treatment including the carryover effect.

Linear mixed models with a Gaussian subject-specific random intercept term were the primary tool of multivariate repeated measurements analysis. Hypotheses in the multivariate analysis were tested using Wald and likelihood ratio tests. The model-building procedure was organized in stages. First, a backward variable selection procedure was used on a model that included demographic variables (age, sex, race), cumulative number (dose) of acupuncture treatments since baseline or last washout (cdose), and the type of acupuncture (acutype, verum vs sham) included as the main effect as well as an effect modifier for the cdose. The model was trimmed one model term at a time, each time starting with the term that has the largest P values for removal. Second, effect modification by baseline QST pressure pain threshold (tenderness) was studied. The best model of the first stage was expanded to include cdose effect modifications by acutype and tenderness (up to three-way interactions). Backward term selection was applied again to the resultant model to obtain the best parsimonious model. Finally, effect modification by needling sensation was studied, adding sensation measurements as effect modifiers of the cdose to the best model of stage 2. Backward elimination procedure was then performed on the added interaction terms. Results of the analyses were reported based on the final parsimonious models. A significance level of 5% for hypothesis testing and 15% for variable retention in the elimination procedures was used throughout the analysis. The combination of staged model extension and backward elimination allowed us to maintain a stable model-building procedure by the rule of thumb of no less than approximately 10 patients per variable at a time. The pressure pain threshold variable was also evaluated for the presence of quadratic terms.

Results

The final group selected for analysis had an average age of 48.3 years (SE = 10.2, range = 24–66) and consisted of 67 females (91.8%) (Table 1). The majority of participants were white (87.7%) while the rest were black (9.6%) and other (2.7%). Participants were split into tertiles based on baseline pain pressure threshold with tertile 1 containing 27 participants, and tertiles 2 and 3 both containing 23 participants, with corresponding threshold intervals of 0–1.81, 1.81–3.12, and 3.12–6.5 kg, for the three tertiles, respectively. Table 2 contains further descriptive statistics. The tables show that, based on univariate analysis, there were no significant differences (all P > 0.05) between tertiles in terms of age, sex, race, clinical pain (baseline, week 8, week 15), acupuncture type, stimulation, or sensation.

Table 1

Demographics and between-group comparisons

Variable	Sample size	Tertile 1	Tertile 2	Tertile 3	P	P	P	P
		Mean/median (SD/IQR) or proportion	Mean/median (SD/IQR) or proportion	Mean/median (SD/IQR) or proportion	All equal	1 = 2	1 = 3	2 = 3
Age	73	46.4/49 (10.9/17)	47.3/49 (10.4/9.5)	51.6/53 (8.75/13.5)	0.17	0.74	0.07	0.15
Sex	73				0.55	0.65	0.51	0.29
Female	67	0.93	0.96	0.87
Male	6	0.07	0.04	0.13
Race	73				0.62	0.41	0.72	0.7
White	64	0.85	0.87	0.91
Black	7	0.11	0.09	0.09
Other	2	0.04	0.04	0

Groups are defined based on pain threshold tertiles (in kg: 1 = 0–1.81, 2 = 1.81–3.12, 3 = 3.12–6.5).

IQR = interquartile range.

Table 2

Descriptive statistics and between-group comparisons

Variable	Sample size	Tertile 1	Tertile 2	Tertile 3	P	P	P	P
		Mean/median (SD/IQR) or proportion	Mean/median (SD/IQR) or proportion	Mean/median (SD/IQR) or proportion	All equal	1 = 2	1 = 3	2 = 3
Baseline clinical pain	73	56.5/50	52.4/55	57.6/55	0.71	0.52	0.86	0.43
		(19.9/30)	(20.3/32.5)	(27.3/47.5)
Week 8 clinical pain	68	49.8/50	48.1/50	48.6/45	0.98	0.83	0.88	0.96
		(27.7/50)	(23.4/35)	(30.0/45.0)
Week 15 clinical pain	73	54.1/65	46.5/40	55.9/50	0.44	0.31	0.81	0.23
		(25.9/40)	(26.2/32.5)	(27.2/42.5)
Acupuncture type	73				0.78	0.98	0.52	0.55
Sham	33	0.48	0.48	0.39
Verum	40	0.52	0.52	0.61
Stimulation	73				0.95	0.81	0.95	0.77
No	33	0.44	0.48	0.44
Yes	40	0.56	0.52	0.57
Sensation_dose1	73	0.901/0.667	0.717/0.5	0.855/0.667	0.67	0.38	0.83	0.53
		(0.8/1.25)	(0.682/0.667)	(0.72/1.5)
Sensation _dose2	73	0.832/0.5	0.576/0.25	0.758/0.4	0.54	0.28	0.75	0.46
		(0.881/1.38)	(0.766/0.75)	(0.835/1.58)
Sensation _dose3	73	0.908/0.375	0.629/0.286	0.675/0.167	0.52	0.29	0.38	0.87
	73	(1.04/1.87)	(0.799/1)	(0.919/1.54)

Groups are defined based on pain threshold tertiles (in kg: 1 = 0–1.81, 2 = 1.81–3.12, 3 = 3.12–6.5).

IQR = interquartile range.

Figure 2 presents box plots (median and 25th–75th percentile box, 1.5 interquartile range [IQR] outliers) summarizing the observed relationship between the types of acupuncture (verum vs sham), cumulative dose, and pressure pain threshold (tertile). While the univariate analyses show no significant treatment effects, Figure 2 suggests distinctly different patterns of clinical pain responses to sham and verum acupuncture. In particular, the pressure pain threshold plays an important role in modifying the dose-response relationship. Patients exhibiting a higher pain threshold (low sensitivity to pain, low tenderness) have improvements in clinical pain during verum acupuncture (a unit increase in pain threshold leads to a reduction in pain of 0.841 [VAS] per unit cumulative acupuncture dose [SE = 0.304, P = 0.006]), but not much, if any, with sham acupuncture. Conversely, patients with lower pressure pain thresholds (greater tenderness) have a positive response with sham acupuncture (a unit lower pain threshold shows a trend for a reduction in pain of 0.428 [VAS] per unit cumulative dose [SE = 0.229, P = 0.062]) and an adverse dose response relationship with verum acupuncture. With these relationships in mind, all final conclusions are based off of the model described below and shown in Figure 3.

Figure 2

Summary of clinical pain response by treatment (verum and sham) and pressure pain threshold. The relationship between the type of acupuncture (verum vs vham), cumulative dose (number of treatments since washout), and the level (tertile) of pressure pain threshold (category scale on the right; low: 0–1.81 kg; medium: 1.81–3.12 kg; high: 3.12–6.5 kg). Box plots show median line, 25–75th percentile box, and 1.5 interquartile range outliers. Zero dose corresponds to the washout periods. Patients with low pressure pain thresholds respond better to sham but not verum acupuncture. Patients with high pressure pain threshold respond better to verum and not sham acupuncture.

Figure 3

Predicted pain responses. Predicted mean visual analogue scale pain responses in a patient treated by different types of acupuncture and characterized by different pain threshold and sensation levels.

To study the dose response patterns in greater detail, multivariate analysis using mixed models was performed. A (cumulative) dose variable was constructed that counted the number of acupuncture treatments received since washout or start of study, whichever was most recent. This variable expresses the treatment effect of main interest.

Initially, the trial was analyzed in the conventional way using two groups (acutype = verum or sham), not including potential tenderness and sensation modifiers (Table 3). We see that although both types of acupuncture are associated with a significant improvement (P = 0.03 for sham, and 0.04 for verum), there is no significant difference between verum and sham (P = 0.83). None of the demographic variables showed any significance, and they were removed by backward elimination (race is removed with P = 0.42, sex with P = 0.87, age with P = 0.24). There was a carryover effect showing decreasing pain with time (P = 0.017).

Table 3

Linear mixed model of acupuncture and demographics effects on clinical pain

Model term	Value	SE	t	P
Intercept	57.697	2.584	22.332	<0.001
Time	−0.397	0.166	−2.389	0.017
Dose (sham)	−0.779	0.359	−2.169	0.031
Dose (verum vs Sham)	0.1	0.465	0.215	0.83
Dose (verum)	−0.678	0.329	−2.061	0.04

The results of a linear mixed model exploring the effect of cumulative dose (dose), type of acupuncture (verum or sham), carryover effect (time), and demographic variables (age, sex, and race). The model is shown after variable selection with nonsignificant effects and variables removed by backward elimination. Race is removed with P = 0.42, sex with P = 0.87, age with P = 0.24. Sample size n = 73 patients.

Based on the initial model, we studied potential treatment effect modification by baseline pressure pain tenderness. Another round of variable selection resulted in the model presented in Table 4. There was still a carryover effect showing decreasing pain with time (P = 0.015). Only the medium staircase pressure pain threshold measure remained in the model. The pressure pain threshold variable proved to be a significant modifier of the effect of acupuncture for all acupuncture types. In patients with low pressure pain thresholds, sham is of benefit (dose main effect, P = 0.007), while verum is an adverse effect modifier (dose*acutype interaction effect, P = 0.021). With increasing pain threshold, we see a reduction in the positive effect of sham (dose*tenderness interaction, P = 0.062) and an increase in the positive value of verum acupuncture (dose*acutype*tenderness 3-way interaction, P = 0.006). The model was also evaluated for the presence of nonlinear (quadratic) pain threshold terms, including effect modification, that were not significant, with P values ranging from 0.1 to 0.57.

Table 4

Linear mixed model of acupuncture and tenderness (pressure pain threshold) on clinical pain

Model term	Value	SE	t	P
Intercept	57.388	2.585	22.196	<0.001
Time	−0.413	0.168	−2.451	0.015
Dose (sham)	−1.775	0.66	−2.691	0.007
Dose (verum vs sham)	2.098	0.908	2.312	0.021
Dose (sham)*tenderness	0.428	0.229	1.872	0.062
Dose (verum vs sham)*tenderness	−0.841	0.304	−2.769	0.006

The results of linear mixed model exploring the effect of cumulative dose (dose), type of acupuncture (verum or sham), carryover effect (time), and pressure pain threshold (tenderness), excluding needling sensations. Analysis is based on 70 subjects with nonmissing values of the included variables.

We next assessed the relationship of needling sensation to analgesic response. At this third step of multivariate modeling, sensation variables (three variables, one per dose level) were added to the dose-acutype-tenderness model, presented in Table 4. After backward elimination, needling sensation at dose level 2 was the only sensation variable remaining in the model as the main effect, as well as a two- and three-way effect modifier. The final model is shown in Table 5.

Table 5

Linear mixed model of acupuncture, tenderness, and needling sensation effects on clinical pain

Model term	Value	SE	t	P
Intercept	55.426	3.261	16.996	0
Time	−0.41	0.167	−2.454	0.015
Dose	−1.08	0.568	−1.901	0.058
Sensation_dose2	2.575	2.765	0.931	0.355
Dose*sensation_dose2	−2.228	1.181	−1.887	0.06
Dose*tenderness (sham)	0.314	0.195	1.615	0.107
Dose*tenderness (verum)	−0.354	0.168	−2.108	0.036

The results of linear mixed model exploring the effect of cumulative dose (dose), type of acupuncture (verum or sham), carryover effect (time), pressure pain threshold (tenderness), and sensations. This model is shown after variable selection, with nonsignificant effects and variables removed by backward elimination. Cdose*ave_dose3 was removed with P = 0.72, ave_dose2 with P = 0.93, cdose*acutype*ave_dose3 with P = 0.80, cdose*acutype with P = 0.77, ave_dose3 with P = 0.25, cdose*sensation_dose1 with P = 0.47, cdose*acutype*sensation_ dose1 with P = 0.27, ave_dose1 with P = 0.13, cdose*tendernessmedium with P = 0.11, and cdose*tendernessmedium*acutypesham with P = 0.11. Sample size = 70 patients.

In order to visualize the complex three-way effect modification relationships expressed in Table 5, a pattern of predicted responses was generated for hypothetical patients with different characteristics (Figure 3), based on the final model. Sham and verum acupuncture groups show distinctly different patterns of effects. The sham group shows better clinical pain response with increasing dose, and those displaying increased sensation amplify this effect. In contrast, verum acupuncture shows a worsening of response with lower pressure pain thresholds that is further worsened with increased levels of sensation. In the verum acupuncture group, higher-threshold patients still show a positive pain response to increasing dose, and the effect is much less pronounced than in the sham group. Higher sensation leads to better clinical pain response in the sham group and a worse response in the verum group. Both patterns are amplified with increasing treatment dose. We note that the sensation variable was only borderline significant as a modifier of the dose effect (P = 0.06), so the presented sensation patterns should be treated as exploratory analysis. Sensation aside, effect modification by tenderness in this model is in agreement with the dose-acutype-tenderness model given in Table 4. We also assessed for a three-way interaction between acupuncture tenderness and needling sensation; however, the result was not significant, which was expected as this test generally requires a larger study population to find significance. While forward selection modeling might have yielded slightly different results, the quality of the prediction remains stable.

Operationalizing the results of this paper, future treatments could be assigned with QST in mind to maximize the benefit for the patient. Table 6 shows model-predicted benefit after eight acupuncture sessions and with no treatment, by type of acupuncture and QST pain threshold (dichotomized using the median into high and low as in our previous 2013 paper). Using a common standard deviation of pain of 25 (VAS) as predicted by the analysis, we assessed the sample size of QST-based personalized treatment trial required to achieve at least 85% power. Using the simple randomization design with two equally sized groups, each assigned eight acupuncture treatments of a single type (no crossover), or no treatment, we found that 160 patients per group are needed for the largest contrast of verum vs no treatment in high-threshold patients. Such patients benefit from verum more than they do from sham, and 280 patients per group would be needed to prove verum vs sham superiority. The effect of sham vs no treatment is quite small in high-threshold patients, with 2,500 patients per group required to detect it. Verum acupuncture is predicted to not bring any benefit in low-threshold patients. A comparison of sham vs no treatment (or verum) in low-threshold patients will require 350 patients per group to detect the benefit of sham. Given the large number of participants needed to distinguish verum from sham acupuncture in a meta-analysis of previous clinical trials (around 1,000 to 2,000; [10]), these data suggest that addition of QST outcomes at baseline would reduce the number of patients needed in an acupuncture clinical trial of chronic pain to find verum vs sham differences.

Table 6

Expected pain (VAS) after 8 treatments by treatment and sensitivity to pressure pain

Treatment	Pain threshold	Pain VAS
Verum	High	44.986
Sham	Low	49.558
Sham	High	53.889
Verum	Low	57.141
No treatment	Low	57.394

VAS = visual analogue score.

Discussion

Here, we show that experimental pressure pain testing is associated with differential clinical pain outcomes for verum and sham acupuncture. This analysis largely replicates our previous work [17] in that patients with greater tenderness were more likely to respond to sham acupuncture. However, in this analysis, we also found that verum acupuncture had the opposite effect, or a worsening of pain for patients that had greater tenderness. This was not observed in our previous Harte et al. study noted above [17]; however, no needling sensations were obtained in that study, so the degree of needling sensation may not have been identical between the two trials. That said, the current work supports the previous hypothesis offered by Lundeberg and Lund in a paper anecdotally suggesting that some patients sensitive to needling may have exacerbated pain in response to verum acupuncture [11]. This hypothesis is supported by the fact that endogenous pain modulation in FM is impaired [24–26], which could result from an imbalance between the inhibitory and stimulatory tracts [27]. Therefore, strong acupuncture stimulation paired with a decreased inhibitory pathway may result in exacerbated pain.

We hypothesize that pain amplification mechanisms triggered by increased peripheral pain during verum acupuncture may be the root of the adverse response to this type of acupuncture in pain-sensitive patients. Evidence suggests that in patients suffering with central sensitivity syndromes, such as FM, the endogenous processing of nociceptive signals is dysfunctional [28–30]. Patients with the lowest conditioned pain modulation tend to have the most pain-related morbidity [31]. Patients in this study with the lowest pain pressure threshold could have actually felt more localized pain upon insertion and stimulation of needles due to a decreased descending analgesic pathway. Previous neuroimaging studies have localized acupuncture-induced sensation (controlling for sham acupuncture sensation) to the dorsomedial prefrontal cortex [32], and connections between this region and descending analgesic brain systems (i.e., periaqueductal gray, rostroventral medulla) may support the association between needle sensations and analgesia in more sensitive FM patients. In light of the findings here, previous clinical trials could be reevaluated using pressure pain threshold as an additional variable in the analysis if it was obtained.

Our results, if replicated in future well-powered prospective studies, could help explain an issue in acupuncture clinical pain trials. Multiple studies have shown little benefit of verum over sham acupuncture, and this is also true for FM [6]. Some have attributed these results to the use of inappropriate sham protocols. Indeed, it has not been elucidated which factor(s) specifically lead to efficacious results for acupuncture [33]. Interestingly, most randomized controlled clinical trials of acupuncture for chronic pain have failed to assess sensitivity to experimental pain at baseline. Our results are consistent with the hypothesis that patients are predisposed to the effects of acupuncture based on their endogenous sensitivity to experimental pressure pain. With a mixed population of FM patients in a study, the specific clinical effect of verum acupuncture might be missed if some patients are more sensitive to needling than others and this variable is not taken into account.

Our previous studies have shown that fibromyalgia patients have a decreased μ-opioid receptor binding potential [34], which is associated with a reduced cingulate response (presumably antinociceptive) to experimental pressure pain, resulting in hyperalgesia (lower pressure pain thresholds) [35]. We also showed that verum and sham acupuncture have divergent opioid receptor mechanisms, with sham acupuncture likely increasing endogenous opioid release and verum acupuncture potentially increasing μ-opioid receptor number [36]. We propose here that fibromyalgia patients who exhibit a higher pain pressure threshold and respond better to verum acupuncture may benefit from increasing receptor numbers. Conversely, we would expect that patients who display lower pressure pain thresholds would experience the most analgesia from sham acupuncture, presumably through increased endogenous opioid release. However, no study has explicitly tested these relationships.

This data also foreshadows that acupuncture could become even more personalized moving forward as pressure pain sensitivity could help the practitioner provide the appropriate level of stimulation to their patient: strong stimulation for less sensitive patients and light stimulation for sensitive patients. It could also help discriminate patients who might not yield any analgesic affect from verum acupuncture. Quantitative sensory testing could be a future mechanism for predicting the analgesic effects of verum and sham acupuncture in FM patients. Traditional Chinese Medicine practitioners already often use a similar technique in which manual pressure is applied to the hand using the patient’s response to help to determine treatment type [1]. Additionally, patient feedback regarding sensitivity to needle-evoked sensations during acupuncture also helps guide clinicians’ decision of how vigorously needles should be manipulated during a clinical treatment, thus enabling personalized medicine. Our data could validate these different screening tools, though, to date, no formal trial has evaluated such methods.

Analysis of this data has inherent limitations that must be appreciated. All subjects in this trial had FM, so these results may not be generalizable to other chronic pain conditions. Future studies in other conditions would be needed to replicate this relationship. Of note, other chronic pain conditions also present with hyperalgesia and allodynia to painful stimuli. The majority of participants were women, necessitating further studies with a larger cohort of males. Our study also utilized a small number of participants, so a larger sample size might allow for more complex statistical interactions to be described. We only used pressure pain thresholds, so is it unknown if our results are generalizable to other QST modalities. Statistical analysis was completed using cumulative dosing over time; this effect can be seen over the time course of this study but might not hold true over different time periods. We also acknowledge the debate surrounding the sham protocol that was employed. While some studies have looked at needle placement and stimulation in FM, controversy still exists about the most appropriate sham acupuncture procedure to employ in clinical trials as the effects of various aspects of needling (penetration, manipulation, depth, stimulation) are still largely unknown. Lastly, we recognize that as this was a new analysis of a previously run trial, the design of this study was not ideal to answer all of the proposed questions and our analyses were post hoc.

Our results suggest that pressure pain sensitivity can be used to help differentially predict treatment response to both sham and verum acupuncture. These results suggest sensitivity to experimental pain and needling sensations may be a confounding factor in previous clinical trials that has not been taken into consideration. We encourage future acupuncture clinical pain trials to include some form of QST as well as needling sensations. We propose that these methods could help future studies find specific needling location effects that differentiate verum from sham acupuncture and can help to provide personalized medical treatment to patients with FM seeking pain relief through acupuncture. Future studies are needed to determine if these effects can be generalized to other QST modalities and other chronic pain conditions. The effects found in this study indicate that fewer patients (280 per group selected to have high pain threshold) would be needed to achieve 85% power to detect the difference between sham and verum acupuncture in future trials.

Author notes

*

Funding sources: This project was supported by the National Institutes of Health (National Center for Complementary and Integrative Health [NCCIH]) grant R01 AT00004. REH is supported by National Institutes of Health (NCCIH) grants K01 AT01111-01 and 5R01AT007550-04. VN is supported by National Institutes of Health (NCCIH) grants P01-AT006663 and R01-AT007550; (National Institute of Arthritis and Musculoskeletal and Skin Diseases) R01-AR064367; and (National Institute of Mental Health) R21-MH103468. SDM is supported by National Institutes of Health (NCCIH) K23 AT006392. AT is supported by National Institutes of Health (NCCIH) grant 1-R01-AT-007550-01.

*

Disclosure and conflicts of interest: NAZ no conflict, AT no conflict, SDM no conflict, SC no conflict, VN no conflict, REH no conflict. None of the authors have competing financial relationships that would constitute a conflict of interest with this work.