Introduction

Chronic pain is a highly prevalent condition that may impact negatively on the individual’s quality of life; moreover, it is an expensive condition for society. Presently, chronic pain is considered as a bio-psychosocial phenomenon–a combination of physical dysfunction, beliefs, coping strategies, distress, illness behavior and social interactions. Pharmacotherapy is a cornerstone in the multimodal interdisciplinary treatment. However, chronic pain treatment still largely depends on local experience and traditions rather than individual patient characteristics [1, 2]. Chronic pain treatment, usually involving more than one drug, is seldom adjusted according to individual treatment response, but rather escalated or several therapies are tried in turn. Unfortunately, this strategy often leads to insufficient pain control, intolerable side effects, and psychosocial distress [3]. A deep concern has recently been raised about the uncritical prescription of opioids leading to growing abuse, addiction, and overdoses in the US [4]. Up to 50% of patients treated with morphine experience inadequate analgesia, despite escalating dose and/or experience intolerable or dose-limiting side effects [1]. This has driven a discovery-oriented research for biomarkers that can predict treatment response. However, due to the complexity of chronic pain, this approach has failed in clinical practice. Quantitative sensory testing, electroencephalography (EEG) recordings, and coping strategies have previously been used to identify patient subgroups in experimental studies and in highly selected chronic pain populations [5–8]. This includes morphine response in healthy volunteers, duloxetine efficacy in painful diabetic neuropathy, and pregabalin efficacy in painful chronic pancreatitis [9–11]. However, no studies have yet examined whether individual pain processing can predict efficacy of unselected opioids in a mixed patient population with chronic pain.

We hypothesized that the most promising techniques in previous experimental, postoperative, and clinical pain studies could serve as biomarkers of chronic pain and opioid analgesia. Hence, the objective of this study was to determine whether characteristics of altered pain processing in opioid-naïve patients could be used to predict clinically relevant treatment responses to opioids. To ensure general applicability, we collected longitudinal data from heterogeneous patient populations with moderate to severe chronic pain initiating opioid use at five centers in Europe.

Methods

Results

The five participating centers screened 73 patients for eligibility (Fig 1). A total of 63 patients were enrolled and examined at baseline; 59 patients (94%) completed the post-treatment follow-up assessment. Demographic and baseline characteristics of patients are depicted in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flowchart.

Numbers of patients screened for eligibility, examined, and analyzed for primary outcome.

https://doi.org/10.1371/journal.pone.0171723.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Demographic and baseline characteristics of patients.

https://doi.org/10.1371/journal.pone.0171723.t001

Outcomes

Table 2 shows changes in primary and secondary outcomes after 14-days of opioid treatment. Treatment was clinically successful in nine patients (16%) where average pain intensity was reduced by at least 30%; 12 patients (21%) demonstrated a clinically relevant change (score ≥10 from baseline) in quality of life assessed by QLQ-C30. Eleven patients (19%) stopped opioid treatment prematurely. The most frequently reported side effect was sedation (74%) followed by constipation (49%), nausea (46%), and dry mouth (44%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Changes in primary and secondary outcomes after 14-days of opioid treatment.

https://doi.org/10.1371/journal.pone.0171723.t002

Predictors of clinically successful treatment

Mixed-effects logistic regression output are shown in Table 3. These analyses indicated that clinically successful treatment in terms of average pain reduction to opioid treatment was associated with cold pain intensity (OR: 0.69; 95% CI: 0.53 to 0.90; P = 0.01), pain catastrophizing (OR: 0.82; 95% CI: 0.68 to 0.98; P = 0.03), relative delta (OR: 0.76; 95% CI: 0.59 to 0.97; P = 0.03) and beta EEG activity (OR: 1.18; 95% CI: 1.01 to 1.37; P = 0.04) induced by experimental cold pain. In contrast, none of the study variables were predictive of improvement in quality of life (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Associations of pre-treatment predictive factors with measures of clinically successful treatment.

https://doi.org/10.1371/journal.pone.0171723.t003

Discussion

In this cohort study on a heterogeneous study population of patients with various chronic pain conditions and different underlying pain mechanisms, initiating use of both weak and strong opioids, we found evidence to support the hypothesis that opioid efficacy can be predicted before treatment based on patients’ responses to experimental pain. Our results suggest that patients with a relative ability to inhibit pain-related thoughts and feelings, reporting lower pain intensity ratings, or expressing specific electrical brain activity patterns during painful stimulation benefit most from opioid treatment.

There is emerging evidence from experimental pain studies that nociceptive characteristics can serve as predictors of treatment response [5]. Further, a few clinical studies have evaluated the role of pain processing in the prediction of clinically successful chronic pain treatment, but only in highly selected patient populations [9, 11]. Yarnitsky et al. showed that duloxetine, a serotonin-norepinephrine reuptake inhibitor (SNRI) believed to improve a malfunctioning pain modulation mechanism, was most beneficial in alleviating the pain in patients with painful diabetic neuropathy demonstrating inefficient conditioned pain modulation [9]. Olesen et al. found that patients with pain due to chronic pancreatitis with segmental hyperalgesia of the pancreatic viscerotome had a superior pregabalin response [11]. Our finding that the cold pressor pain response predicted opioid treatment response further stresses the potential of using quantitative sensory testing in routine clinical practice. Brain function is a promising biomarker for chronic pain and a neural signature of experimental pain has recently been defined in humans [22]. EEG recorded as evoked brain potentials has proven a viable tool for analyzing changes in cortical activity following administration of different analgesics [2, 23]. Specifically the activity in the delta band was changed in experimental and clinical studies following opioid administration [24]. In experimental studies the EEG response to experimental pain was also shown to predict the effect of morphine and the delta band was the main frequency band involved in differentiating between responders and non-responders to morphine analgesia [10]. This was also the case in the present study and thus it seems the delta band is closely connected to opioid analgesia, and thus a possible biomarker to be used in personalized clinical pain management interventions. The fact that the beta band also predicted successful opioid treatment has not been reported in previous studies. However, as relative EEG indices are used there is a chance that differences in the delta band could carry over to the beta band. Nonetheless, the beta band should be further investigated for its ability to predict pain and analgesia, as it has previously been shown to be involved in pain processing [25]. EEG might be a particularly interesting technology because it makes little demands on patients’ cognition and cooperation. Moreover, reliable EEG is now easily recorded using cheap commercial headsets, thus facilitating more widespread clinical use [26]. Typically, patients with high levels of pain catastrophizing rate their pain higher than non-catastrophizers. As such, catastrophizers are more likely than non-catastrophizers to continue to suffer from high levels of persistent pain in spite of treatment [27]. We aimed at investigating pain catastrophizing with reference to cold pressor pain, because more robust correlations with pain-related outcomes have been reported for situational measures of pain catastrophizing compared with dispositional measures [21, 28, 29]. The mechanism by which pain catastrophizing interferes with analgesic treatment response is unclear, but there is evidence to suggest that it may be related to a disruption in the endogenous modulation of pain [30]. Pain catastrophizing shares variance with negative affect constructs, such as anxiety and depression; other psychological tests might also have been suitable to elucidate a patient’s pain and inform pain-treatment strategies. However, we anticipated that additional questionnaires would only complicate and increase the length and difficulty of testing without necessarily producing better treatment prediction.

There are limitations to our study. The study was intentionally observational and representative of the real-world sample in order to mirror the daily clinical setting. Hence, we acknowledge, that our study group was very varied, representing many chronic pain related diseases and conditions that could all have influenced the treatment responses to opioids. Along this line, it is worth mentioning that, multiple pain mechanisms and outcome-relevant patient characteristics may be active to varying degrees in different patients within a chronic pain related disease or condition–leading to marked inter-subject variation in treatment effects [31].This variability in phenotypic presentation of different chronic pain related diseases and conditions has been found to be greater between patients than between different chronic pain related diseases and conditions [32–34], indicating that successful treatment is likely to be based at the level of the individual rather than at the level of the disease [31]. Furthermore, focusing on a single pain related disease or condition would have limited the generalizability of the study and reduced the sample size that could be achieved in the available study period. It is also worth mentioning that some patients will benefit from any type of intervention you offer. Thus we cannot exclude that multimodal treatment approaches usually involving more than one drug or even a placebo treatment would have had the same effect. Future studies might further standardize concurrent interdisciplinary treatment to clarify if our findings are specific enough for opioids. Another limitation is the lack of long-term follow-up. How this might affect prediction of the efficacy of a longer-term opioid prescription with a fixed mode of administration/dose is debatable. It is also uncertain whether our results can be generalized to all opioids. Our opioid response may seem low; however complete pain relief is unlikely and we may have set the threshold too high by defining a response as ≥30% pain relief. Based on the observed change from baseline in ratings of BPI pain interference items and QLQ-C30 functional scales, we found some evidence that opioids improved function; however, the number of reported side effects mostly supports the extensive evidence showing the possible harms of opioids. There are diseases and clinical circumstances under which improvement in physical function is a more realistic goal than reductions in pain and vice versa. Nonetheless, a single outcome measure including improvement in both pain relief and function is highly warranted.

In summary, the analgesic action of opioids remains unpredictable whereas most patients suffer from side effects. Hence, there is an unmet need for more personalized clinical interventions in pain management. Our results show that alterations in pain regulatory systems may be linked to successful chronic pain treatment. Our data also suggest that a test battery to investigate pain processing can easily be implemented in daily practice and may pave the road for improved pain treatment strategies.

Supporting information