Comparing genuine and sham surgery for sacroiliac joint pain using self-assessments, pain testing, and neuroimaging

Supplemental Digital Content is Available in the Text.

There were differences in pain-relevant brain networks after genuine vs placebo surgery in patients with sacroiliac joint pain.

Abstract

Introduction:

Contrasting active treatment against a placebo has long been the gold standard in clinical medicine. The possible impact of placebo responses in surgery has recently been investigated using sham surgery. Despite indications that both genuine and placebo surgeries may lead to positive outcomes, no investigation into the differential routes to improvement has been performed.

Objectives:

To assess the mechanisms involved in improvements seen in patients with sacroiliac joint pain who undergo genuine or placebo surgery.

Methods:

This randomized controlled trial incorporated both subjective and objective assessments, including functional magnetic resonance imaging and experimental pain testing, at baseline and 6-month follow-up in a surgical trial including patients with chronic pain. Twenty-three patients were randomized to receive genuine surgery (sacroiliac joint fusion) or placebo (sham). An additional 7 patients were included as observational controls.

Results:

There was a significant reduction in weekly pain intensity for both the genuine and placebo groups at follow-up, with greater reductions in the genuine group compared with placebo (P = 0.04). The difference was driven by a few “super-responders” in the genuine group. Clinical improvements correlated with experimental pain outcomes at the operated sacroiliac joint. Functional brain connectivity between the somatosensory cortex and the default mode network decreased more in the genuine group compared with the placebo group.

Conclusion:

Preliminary findings indicate decreased connectivity between somatosensory and default mode networks for patients in the genuine vs sham group, demonstrating the first findings of differential neural processing in pain-relevant brain networks after genuine vs placebo surgery using objective measures. Understanding the active mechanisms of surgery may lead to personalized treatments, more effective pain reduction, and less side effects for patients with pain.

1. Introduction

Placebo responses significantly influences surgical outcomes,^14,38 as shown in clinical trials where patients are randomized to receive either real or sham (placebo) surgery. Sham surgery serves as a placebo control when there is uncertainty about the active element of a specific surgical procedure. It mimics all aspects of the real surgery except the intended therapeutic maneuver¹⁰ and has been endorsed in recent guidelines for control interventions in nonpharmacological trials.¹¹ Owing to ethical concerns, sham surgeries are typically limited to conditions that are not life-threatening, such as musculoskeletal pain or obesity.³⁸ A systematic review of sham-controlled surgical interventions for a range of medical conditions found no significant difference between genuine and sham surgery in approximately half of the studies.³⁸ However, a major limitation of this research is its reliance on subjective patient reports, such as pain severity and overall functioning. Such reports offer limited insights into the underlying mechanisms of improvement, as they may seem identical for both the genuine and sham groups.

Quantitative sensory testing (QST) and functional magnetic resonance imaging (fMRI) have increasingly been used in clinical trials, either to predict pain and disability^2,9 or evaluate the treatment outcomes.^13,22 QST is common for testing pain tolerance, thresholds, and pain inhibitory function.^5,8 Neuroimaging studies using fMRI can be used to assess which brain regions are used in pain processing, and resting-state techniques can display the communication between the brain's networks during rest. To the best of our knowledge, no studies have used QST and fMRI measures to evaluate outcomes in sham-controlled trials.

In recent years, minimally invasive surgical implants have become increasingly common in treating low back pain such as sacroiliac (SI) joint pain.³⁹ Low-back pain is a widespread health issue and the leading cause of disability worldwide, measured by years lived with disability.⁶ In 15% to 30% of patients with low-back pain, the source may dysfunction of the SI joint,³¹ although SI joint pain is challenging to diagnose as its symptoms often mimic those of other conditions.^3,7 Empirically supported treatments for SI joint pain are limited, with minimal evidence for the effectiveness of both nonsurgical interventions and open surgery.^24,32 Two randomized controlled trials (RCTs) comparing minimally invasive fusion surgery with nonsurgical treatments have showed promising results.^24,33 However, the contribution of treatment expectations and placebo effects is unknown due to lack of sham-controlled studies.

The goal of this randomized controlled study was to explore the differential routes to improvement in patients with chronic sacroiliac joint pain undergoing minimally invasive SI joint fusion vs sham surgery. To this end, a combination of subjective and objective measurements was used to link pain ratings, experimental pain testing, and fMRI measures.

We hypothesized larger improvements in the genuine group on subjective pain ratings (0–100 mm visual analogue scale [VAS]) and experimental pain testing. We also hypothesized that high variability of clinical pain at baseline would predict the difference in weekly average pain at follow-up with a larger effect in the genuine group than in the sham group. Furthermore, we hypothesized that the resting-state fMRI (rsfMRI) functional connectivity between nodes in the primary somatosensory cortex (hip/trunk/leg area) and the rest of the brain will be decreased from baseline to follow-up, with a larger effect in the genuine group.

2. Material and methods

2.1. Study design and setting

The study was part of a multicenter double-blind RCT in which patients with SI joint pain were randomly assigned to minimal invasive SI joint fusion (hereafter called “fusion”) or sham surgery (hereafter called “sham”). Patients were recruited at 2 study sites: Oslo University Hospital (Norway) and Karolinska University Hospital (Sweden). The primary results from the multicenter have been published.²⁸ Here, only patients recruited at the Karolinska University Hospital were included, as this was the only site where experimental pain testing and fMRI were performed. Data were collected between December 18, 2019, and April 7, 2022. The Consolidated Standards of Reporting Trials (CONSORT) guidelines were followed.³⁰ In addition to the overall ClinicalTrials registration for the multicenter trial (NCT03507049), a separate pre-registration of the hypotheses and statistical tests for the experimental part of the trial was performed before data analysis (https://osf.io/nxfcd). No interim analysis was performed. The study was approved by the Regional Ethics committee in Stockholm, Sweden (2018/1463-31).

2.2. Patient population

Patients with suspected SI joint pain were evaluated for eligibility after being referred from general practitioners as well as orthopedic surgery or rehabilitation medicine specialists. A total of 23 patients were randomized to fusion or sham surgery at the Stockholm site. An additional 7 patients were recruited as observational nonoperated controls to account for the natural course of SI joint pain.

For a full description of inclusion and exclusion criteria, see the Supplementary information, http://links.lww.com/PR9/A347.

2.3. Randomization and blinding

Right before surgery, patients were randomized to fusion or sham in a 1:1 ratio in blocks of 4 or 6. The randomization sequence was generated by an independent statistician using a tool at Viedoc.com. Only the operating surgeon and the staff in the operating room were aware of the treatment allocation and did not meet the patient postoperatively. Patients, staff at the ward and outpatient clinic, and researchers who conducted the assessments and statistical analyses were blinded to the treatment group allocation. After the 6 months' follow-up assessments, the patients were unblinded and patients in the sham group were offered genuine surgery. All statistical analyses were performed in line with the pre-registration. For a full description of the surgical procedures, see Supplementary information, http://links.lww.com/PR9/A347.

2.4. Procedure

At baseline and follow-up, patients underwent experimental pain testing to characterize different aspects of pain regulation as well as MRI scanning to assess resting-state functional connectivity (rsfMRI). Baseline assessments were administered 1 day before surgery. At the 6-month follow-up, patients were assessed again before being unblinded. All tests were performed at the Karolinska Hospital MR center in Stockholm, Sweden (Fig. 1).

Figure 1.

Schematic representation of the trial design and assessment timepoints.

2.5. Outcomes

2.5.1. Primary outcome

The primary outcome was the difference in patients' subjective rating of average pain intensity during the last week, from baseline to follow-up, using a VAS (0–100 mm VAS), anchored with no pain (0 mm) and worst possible pain (100 mm).

2.5.2. Secondary outcomes

2.5.2.1. Pain survey data

Prediction analyses were conducted of (1) variability of general pain intensity (0–100 mm VAS), defined as last week's maximum pain minus last week's minimum pain (Δ_max-min); (2) psychological measures of the patients' “desire to obtain pain-relief,” anchored by “no desire” to “highest possible desire” to get pain relief, measured on a (0–100 mm VAS); and “expectation of pain level after surgery” (measured on a 0–100 mm VAS) where the value indicates the expected pain at follow-up, with a low value thus indicating high expectations of analgesia.

2.5.2.2. Experimental pain testing

Pressure pain thresholds and supra-threshold pain (pain corresponding to 4/10 on a Numeric Rating Scale [NRS]) at the SI joint pain site and on the thigh (control site) were assessed using a Somedic Algometer version II (Somedic, Hörby, Sweden).

2.5.2.3. Brain imaging data

Change in rsfMRI functional connectivity between a seed region in the primary somatosensory cortex (S1) corresponding to the SI joint (hip/trunk/leg area) and the rest of the brain from baseline to follow-up was assessed. Secondary analyses also included rsfMRI connectivity from baseline to follow-up between our S1 seed region and nociceptive brain regions according to the predefined Neurologic Pain Signature (NPS)³⁶ and Stimulus Intensity Independent Pain Signature-1 (SIIPS1).³⁷ While NPS and SIIPS1 are both designed for stimulus-induced activation, we used them as masks to define pain-related brain areas. Directly after the MRI scan, while still in the scanner, patients were asked to rate their average overall pain level (not restricted to the site of operation) on a VAS (VAS 0–100 mm).

2.6. Statistical analysis

All researchers involved in data processing and analysis were blinded to the treatment allocation.

2.6.1. Survey data and pain testing outcomes

Restricted maximum likelihood linear mixed models were used to analyze the difference between the sham and fusion groups from baseline to follow-up according to our predefined analysis strategy. Linear regression was used for the prediction analyses. The researchers conducted additional exploratory analyses after being unblinded to reveal possible differences at baseline between responders and nonresponders to fusion surgery, and to elucidate correlations (Pearson r) between baseline variables and treatment outcomes as well as group differences from baseline to follow-up using repeated-measures ANOVA. All analyses were two-tailed. Statistical significance was determined at alpha-level 0.05. Cohen d effect sizes were calculated using observed means.

2.6.2. Brain imaging data

The functional connectivity analyses of the brain were performed on Blood-oxygen-level-dependent (BOLD) fMRI data during resting-state. Preprocessing of fMRI images was performed using the fMRIPrep processing pipeline with 36 regressors for motion correction. The hip/trunk/leg area of the primary somatosensory cortex (S1) was defined a priori as a brain region of interest and used for seed-based connectivity analyses to the rest of the brain. The brain areas used for connectivity analyses were defined according to the 100 node Schaefer parcellation²⁹ and Yeo allocation³⁴ of nodes into 7 canonical resting-state brain networks. For the first analysis (seed-based functional connectivity), Pearson correlations (r) were used to create Fisher transformed (r to z) connectivity matrices. The correlations consisted of outcomes for the seed region (bilateral) and the rest of the brain (99 remaining nodes). Delta-connectivity values (Δ_post-pre) were retrieved by subtracting the baseline connectivity values from values at the 6-month follow-up.

For the second analysis, we restricted the seed analysis to brain regions associated with pain processing, using the NPS and SIIPS1 brain masks to create a weighted pain parcel. The connectivity between the seed region (ipsi- and contralateral) and the pain masks were then correlated with each other at baseline and follow-up (Δ_pre-post). This provided a 4 × 4 symmetrical connectivity matrix between the S1 seed (ipsi-/contralateral) and the 2 pain-specific weighted masks. For both analyses, False Discovery Rate (FDR) correction was used to control for the number of statistical tests within each analysis. The significance threshold was set to P < 0.05 FDR corrected. To validate the brain imaging analyses on the network level, we used a null model approach to test whether a brain network feature was more prominent than would be expected by chance.³⁵

2.7. Blinding

Blinding of treatment type (fusion/sham) was assessed by interviewing patients at follow-up before unblinding. Patients were asked to guess their treatment assignment and were allowed to express uncertainty. The Bang Blinding Index (BI) was used to calculate the blinding index in R²⁵ using the R package BI version 1.1.0.

3. Results

Baseline and follow-up data were available for 22 patients in the operated group (randomized to fusion or sham) and 6 patients in the observational group. One operated patient and 1 patient in the observational group were lost at follow-up. Baseline characteristics are summarized in Table 1. The 2 randomized groups, fusion and sham, were part of our statistical assessments. The observational group was recruited independently and will therefore not be formally compared with the 2 randomized groups in statistical analyses. A detailed description of harm and adverse events is found in Randers et al.²⁸

Table 1.

Baseline characteristics.

Characteristic	Fusion	Sham	Observational control	Total
No. of participants	11	12	7	30
Age, mean (SD), y	44.1 (6.6)	45.8 (7.4)	47.2 (10.1)	45.5 (7.6)
Sex, n female (%)	10 (90%)	12 (100%)	6 (86%)	28 (93%)
Using strong opioids, n (%)	4 (36%)	6 (50%)	1 (14%)	11 (37%)
Using weak opioids, n (%)	3 (27%)	4 (33%)	3 (42%)	10 (33%)
Using NSAIDs or paracetamol	10 (91%)	11 (92%)	7 (100%)	28 (93%)
No. of children: (median, 25th, 75% percentile)	2 (2; 3)	2.5 (2; 4)	3 (2; 3)	2.5 (2; 3)
Underlying cause of pain, n (%)
In relation to pregnancy	5	9	3	17
Trauma	1	0	0	1
Other	5	3	0	8
Unknown	0	0	4	4
Pressure pain threshold SIJ, mean (SD)	157 (46.2)	181 (78.9)	262.7 (103.6)	191.3 (83.9)
Pressure pain supra (4/10) SIJ, mean (SD)	229.2 (50.5)	267.9 (72.9)	433.4 (226.4)	293.2 (143.4)
Expectation, mean (SD)	13.2 (14.6)	18.7 (9.5)	n.a.	16.1 (12.2)
Hope, mean (SD)	74.5 (30.6)	81 (18.2)	n.a.	77.9 (24.4)

Strong opioids (eg, oxycodone); weak opioids (eg, tramadol, codeine). No patient was using strong and weak opioids simultaneously.

SD, standard deviation; SIJ, sacroiliac joint.

3.1. Primary outcome

There was a significant decrease in last week's average pain from baseline to follow-up across the fusion and sham groups, with a model implied mean decrease of −26.9 points (95% CI −42.5, −11.2; P ≤ 0.001, d = 0.66) on the VAS scale. There was also a significant difference between groups, favoring the fusion group. The model implied mean showed that the fusion group improved more, as they rated their weekly pain 22.4 points lower than the sham group at follow-up (95% CI 1.0, 43.7; P = 0.04, d = 0.80). The difference in last week's average pain was unequally distributed, with a few patients in the fusion group showing large benefits (Fig. 2). As a comparison, the observational group (not randomized) had no improvement in weekly average pain from baseline to follow-up (M = −1.6), with wide individual differences ranging −35 to 36 mm VAS.

Figure 2.

Average weekly pain. Difference in self-rated weekly average pain between baseline and follow-up (Δ_pre-post) for the Fusion, Sham, and No treatment (nonrandomized observational) groups, respectively. A positive value indicates that self-reported pain improved from baseline to 6-month follow-up.

3.2. Secondary outcomes

3.2.1. Pain variability

High variability of clinical pain at baseline (Δ_max-min) did not predict the difference in weekly average pain at follow-up in the fusion group (P = 0.677) or the sham group (P = 0.152).

3.2.2. Psychological variables—desire for pain relief and expectations of relief

Neither self-reported desire to get well nor positive expectations on pain relief from the present surgical treatment predicted the difference in weekly average pain from baseline to follow-up for the fusion group (desire P = 0.074, expectation P = 0.257) or the sham group (desire P = 0.572, expectation P = 0.746). The 2 variables, desire and expectation, were uncorrelated (Fig. 3).

Figure 3.

Desire to get well and treatment expectations. The 2 psychological variables “desire to get well” and “expectation of pain relief” were uncorrelated to each other. The expectation variable represents the pain intensity expected at follow-up (0–100 mm VAS), where a low value indicates expectations of analgesia. Conversely, the desire variable indicates no (0 mm) to highest possible (100 mm) desire to get well. Both variables were unrelated to the primary treatment outcome (weekly pain on a 0–100 mm VAS). VAS, visual analogue scale.

3.3. Objective and experimental pain testing outcomes

3.3.1. Pressure pain thresholds

The pressure pain threshold (kPa) on the operated SI joint side did not change significantly between baseline and follow-up across the fusion and sham groups (P = 0.254), and there was no interaction effect between time × group (P = 0.689). There was no significant change in supra-threshold pain (kPa) at a calibrated pressure-pain level of 4/10 NRS between baseline and follow-up (P = 0.193) nor any interaction effect between time × group (P = 0.389).

The difference in last week's average pain (VAS) correlated significantly with the difference in supra-threshold pain (kPa) from baseline to follow-up at the affected SI joint pain site across groups (r = 0.44, P = 0.019), but not with the control site on the patient's leg (nonpainful site), and did not differ between groups (Fig. 4).

Figure 4.

Subjective pain ratings and pressure pain outcomes. Improvement in weekly average pain (0–100 mm VAS) correlated significantly with the difference in supra-threshold (4/10 NRS) pain at the operated SIJ during pressure pain testing. A high pressure pain value (post-pre) and a high weekly VAS value (pre-post) indicate less pain. NRS, numeric rating scale; SIJ, sacroiliac joint; VAS, visual analogue scale.

The baseline variability in weekly pain was correlated to suprathreshold pain (kPa) changes from baseline to follow-up in the sham group (r = 0.71, P = 0.010), but not the fusion group (r = 0.01, P = 0.986) (Fig. 5).

Figure 5.

Pain variability and pressure pain. Pain rating difference for average weekly pain (Δ_pre-post) and pressure (kPa) needed to evoke pain during suprathreshold stimuli (Δ_post-pre) at 4/10 NRS. A high pressure pain value (post-pre) indicate less pain. NRS, numeric rating scale.

3.4. Exploratory self-reported outcomes

Ratings of pain on the affected SI joint site (0–10 NRS) decreased significantly from baseline to follow-up across both groups (95% CI −5.4, −1.2; P = 0.002), but there was no time × group interaction (95% CI −2.1, −2.1; P = 0.989), indicating that both the fusion and sham groups had comparable reduction of SI joint pain. The overall change in last week's average pain (Δ_pre-post) correlated with the change in self-reported pain on the affected SI joint site (Δ_pre-post) (r = 0.64, P = 0.002), indicating an overlap between ratings of specific SI joint pain and overall pain severity.

3.5. Pain during functional magnetic resonance imaging scanning

Exploratory analyses of self-assessed pain during fMRI scanning revealed a significant decrease from baseline to follow-up across groups with a model implied mean decrease of 24.9 points (95% CI −38.0, −11.8; P ≤ 0.001) on the 0 to 100 mm VAS scale. In addition, there was a significant difference of pain in scanner between groups, favoring the fusion group who had significantly less pain during scanning at follow-up. The model implied mean showed that the sham group had 18.9 points (95% CI 0.25, 37.6; P = 0.047) more pain during scanning at follow-up. As a comparison, the observational group had no improvement of self-rated pain in the scanner from baseline to follow-up, as the average pain at baseline was 52 mm VAS vs 55 mm VAS at follow-up.

3.6. Responder characteristics for fusion surgery

An exploratory visual inspection of baseline characteristics for patients in the fusion group indicated potential differences between responders (defined as improvement ≥20 on VAS average pain) and nonresponders. Means and standard deviations are summarized in Table 2. Because the number of individuals was small, these potential differences were not tested statistically.

Table 2.

Means and standard deviations of baseline characteristics for responders and nonresponders in the fusion group.

Characteristic	Responders (n = 5)	Nonresponders (n = 5)	All (=10)
Age, y (SD)	43.4 (7.6)	45.4 (6.7)	44.4 (6.8)
Sex, n female (%)	5 (100)	5 (100)	10 (100)
VAS pain (SD) kPa	89.8 (9.7)	71.1 (13.7)	80.5 (14.9)
NRS SIJ (0–10)	8.6 (1.1)	7.2 (1.9)	7.9 (1.7)
PPT leg (SD) kPa	358.8 (116.6)	238.8 (115.4)	298.8 (126.4)
Supra PPT leg (SD) kPa	560.7 (104.3)	395.4 (178.2)	478.1 (162.9)
PPT SIJ (SD) kPa	172.5 (56.1)	140.1 (38.8)	156.3 (48.6)
Supra PPT SIJ (SD) kPa	239.7 (74.1)	222.5 (24.4)	231.1 (52.8)
Expectation pain (0–100)	6.8 (6.4)	19.2 (18.9)	13.7 (15.4)
Desire pain-free (0–100)	63.5 (44.6)	80.3 (19.7)	72.8 (31.9)
General health (0–100)	25 (20)	42 (29.5)	33.5 (25.4)
Family pain yes/no	3/2	3/2	6/4

NRS, numeric rating scale; PPT, pressure pain threshold; SD, standard deviation; SIJ, sacroiliac joint; VAS, visual analogue scale.

3.7. Blinding index

The BI gave an estimate of −0.16 (95% CI: −0.62, 0.28) for the fusion group and 0.10 (95% CI: −0.41, 0.61) for the sham group, indicating successful blinding according to the BI index on which a score between −0.2 and 0.2 indicates sufficient blinding.

3.8. Neuroimaging

3.8.1. Seed-based functional connectivity

There was no significant difference in seed-based functional connectivity between the fusion and sham groups at baseline. However, when comparing the connectivity between baseline and follow-up and between the fusion and sham groups (timepoint × group), there was a difference between the S1 seed connectivity to the Default Mode Network (DMN), (t = 2.5, df = 15, P = 0.03, Cohen d = 1.15), indicating decreased seed to DMN connectivity in the fusion group. None of the other 6 networks differed between groups; while the ventral attention network showed a trend toward significance (P = 0.08), see Figure 6, this did not survive FDR correction for multiple comparisons.

Figure 6.

Brain connectivity outcomes. (A) Anatomical representation of the primary somatosensory cortex (S1) seed, corresponding to the hip, trunk, and leg, used in connectivity analyses and (B) brain regions pertaining to the Default Mode Network (DMN). (C) Connectivity matrix for the seed-based connectivity (Δ_pre-post) between [surgery − sham]; *P < 0.05. (D) Within-group representation of correlation matrices (Δ_pre-post) between all 7 canonical resting-state networks for the fusion group and sham, respectively (Yeo7 parcellation).

3.8.2. Seed connectivity to predefined pain processing brain regions

Extraction of the NPS and SIIPS1 pain signatures revealed no significant difference from baseline to follow-up (NPS [F(1, 17) = 4.33, P = 0.519]; SIIPS1 [F(1, 17) = 4.3, P = 0.123]) or interaction between time × group. Furthermore, there was no difference in S1 seed connectivity to the NPS brain regions [F(1, 17) = 1.2, P = 0.296] or to the SIIPS1 brain regions from baseline to follow-up [F(1, 17) = 0.87, P = 0.364]. There was no significant interaction between time × group for the S1 to NPS or SIIPS1.

4. Discussion

This randomized sham-controlled trial compared fusion surgery with sham surgery in chronic SI joint pain using subjective and objective measures. Both groups showed improvements, but differences in pain ratings and brain connectivity suggest mechanistic distinctions. An observational group provided a preliminary estimation of the natural course of SI joint pain from baseline to follow-up and indicated that spontaneous remission was unlikely to explain these results.

There was a significant group difference in weekly average pain from baseline to follow-up, with the fusion group reporting lower pain at follow-up. The difference was largely driven by 3 super-responders in the fusion group and should thus be interpreted with caution. In line with the primary outcome of the overall multicenter trial,²⁸ the rating of SI joint-specific pain on the operated side did not differ between groups from baseline to follow-up. The overall measure of weekly average pain and the SI joint specific pain were correlated, but are likely to represent slightly different aspects of the treatment response.

Exploratively, we sampled patients' ratings of pain while in the MRI scanner, which they perceived as uncomfortable and painful. There was a significant difference in pain ratings from baseline to follow-up, with the fusion group having comparably less pain at the 6-month follow-up. We believe these assessments complement weekly average pain scores, as they simulate challenges patients with SI joint pain may face in daily life—specifically, asymmetrical movements such as transitioning between lying and standing positions, as well as remaining still for approximately 30 minutes.¹⁸

Patients in both the sham and fusion groups improved from baseline to follow-up. Part of the response in this trial may be attributed to the placebo effect. Most placebo-controlled trials cannot distinguish placebo effects from the natural disease course because of the lack of a third observational group.¹² A placebo response refers to all improvements seen in patients who received sham treatment whereas the placebo effect refers to improvements directly attributable to placebo mechanisms, excluding factors such as spontaneous remission and regression to the mean. This was highlighted by Karjalainen et al.¹⁵ in a review of 62 sham-controlled trials, where only 3 included nonoperative controls. The present trial included a small observational group to contrast the sham and fusion groups, revealing no improvement in the observational group. This suggests that the sham group's improvement stemmed from placebo effects rather than spontaneous remission or other contextual influences. As there were only minor differences between treatment responses in the sham and fusion groups, a major part of the treatment outcomes seen in this trial may thus be attributed to placebo-related mechanisms.^14,38 However, the observational no-treatment group was not randomized, and the sample size was insufficient to allow for a formal comparison between the sham and observational groups and determine a placebo effect.

Pressure algometry is a common method for evaluation of pain and determination of therapeutic effects in musculoskeletal conditions,²⁶ and ratings were hypothesized to indicate sensitivity to pain in the fusion group after surgery. The pain tests were not able to separate the groups per se. However, pressure pain was correlated with subjective ratings of weekly average pain, indicating that subjective ratings of improvement were linked to our experimental pain outcome. In addition, the baseline variability of weekly pain (Δ_max-min) was correlated to pressure pain outcomes. High variability of clinical pain at baseline (in contrast to static pain) has previously been associated with positive treatment outcomes and to placebo responses in patients with long-term pain.¹⁷ The assumption is that high pain variability, ie, having both good and bad days, may indicate a less persistent pain pattern with higher flexibility and greater chances for remission through endogenous pain regulation. Here, we found a relationship between pain variability and pressure pain outcomes in the placebo group but not in the genuine group. This may suggest that endogenous pain regulation may have accounted for more of the improvement in the sham group whereas reduction of nociceptive signaling may have been an additional component in the fusion group, in line with the surgery rationale.

Using objective measures of fMRI BOLD signaling during rest, we were able to demonstrate changes in intrinsic brain connectivity from baseline to follow-up between the fusion and sham groups, but our findings did not survive correction for multiple comparisons. Intrinsic brain connectivity reflects spontaneous neural and metabolic activities that occur in a resting state when no specific task is performed. Brain connectivity during rest occurs within distinct networks of the brain such as the DMN: an aggregate of brain structures associated with self-referential cognition. It has previously been established that intrinsic brain networks are disrupted in relation to the spontaneous pain experienced by patients with long-term pain conditions.²¹ In line with our hypothesis, we found that the S1 seed (representing the SI joint area on S1) displayed decreased connectivity to the DMN in patients in the fusion group compared with sham from baseline to follow-up. The DMN has previously been suggested as a biomarker for pain perception in long-term pain,¹⁹ and decreased connectivity between the DMN and the other brain regions has been demonstrated after treatment in fibromyalgia²³ and several other conditions. When comparing individuals with long-term pain and healthy controls, differences in brain connectivity are typically observed between DMN and areas implicated in attentional and somatosensory processing and regulation, primarily the insula and somatosensory cortices.⁴ It is not clear whether the differential resting-state functional connectivity seen in this study is a sign of functional reorganization or a marker of ongoing spontaneous pain in patients suffering from long-term pain, yet it has been suggested as a proxy for disease burden.⁴

There was a large response diversity in the present trial, and only some individuals had significant improvements in pain levels. Visual inspection of baseline differences between responders and nonresponders may generate future hypotheses about predictors of response, which can be tested clinically. For example, responders in the fusion group had higher expectancies as to the efficacy of the surgery (as they expected lower levels of pain at follow-up), which has previously predicted treatment outcome in surgery.^1,16 Our findings indicate that expectancies may be important to investigate before treatment decisions. Another observation is that the responders had more severe symptoms at the beginning of treatment compared with patients with no treatment response. This is in contrast with studies showing high preoperative pain to be linked to worse treatment outcomes after surgery.^20,27,40

A potential limitation of this study is the variability in opioid use among participants. Although opioid use was documented and did not differ systematically between groups, it remains a possible confounding factor in postoperative pain assessments. Future studies should consider stratifying or adjusting for opioid use to better isolate treatment effects on pain outcomes. The small sample size limits the statistical power of this study. The results should therefore be replicated in future, larger studies. Furthermore, the observational group was recruited outside of the RCT and could not be formally compared with the fusion and sham groups. Finally, the generalizability is limited by the predominately female sample.

Disclosures

The authors have no conflict of interest to declare.

Supplemental digital content

Supplemental digital content associated with this article can be found online at http://links.lww.com/PR9/A347.