Do expectations influence pain? Recognizing Irving Kirsch’s contribution to our understanding of pain

ABSTRACT

Pain is a universal experience that can take different forms, and it can be acute or chronic. Experimental pain, such as heat pain, can help us better understand the pain experience, as it induces transient, but robust central sensitization in participants. Central sensitization is considered a key underlying concept in the development and maintenance of chronic pain and is defined as an overly effective transmission of nociception in the central nervous system. Expectations can influence perceived pain intensity and treatment success. Irving Kirsch’s work in the field of experimental pain has greatly contributed to our understanding of how expectations influence the pain experience. In this article, we present some of Kirsch’s landmark studies in this area and discuss their (clinical) implications.

KEYWORDS:

• Expectations
• experimental design
• Irving Kirsch
• pain

Experimental pain as an opportunity to learn about the pain experience

Pain is a universal human experience, everyone experiences pain from time to time. The International Association for the Study of Pain (IASP) defines pain as “an unpleasant sensory and emotional experience associated with, or resembling that associated with, actual or potential tissue damage” (Raja et al., 2020). Pain can take different forms: acute pain serves as a warning function and advises us to end an operation that is potentially or actually harmful for us, such as removing one’s hand from the hot stove. Acute pain lasts seconds, minutes, days, or weeks. In contrast, chronic pain is defined as pain that lasts or recurs for three or more months. Chronic pain is a prevalent and debilitating condition that affects around 20% of the general population (Dahlhamer et al., 2018). Headache, migraine, and musculoskeletal pain such as chronic low back pain are the most prevalent pain conditions (Dueñas, Ojeda, Salazar, Mico, & Failde, 2016). Previous research has associated chronic pain with decreased quality of life, missed days at work, marital problems, and enormous costs for the health-care system (Buchbinder & Underwood, 2012; Vos et al., 2012). Since 1990, lower back pain has been the leading cause of years lived with disability (Vos et al., 2017), and it is therefore a common global problem (WHO Scientific Group on the Burden of Musculoskeletal Conditions at the Start of the New Millennium, 2003). The IASP definition of pain also takes into account that pain is always a personal experience, influenced by biological, social, and psychological factors (Raja et al., 2020). This definition is largely influenced by the biopsychosocial model of pain (Gatchel, 2004; Gatchel, Peng, Peters, Fuchs, & Turk, 2007).

In the area of both acute and chronic pain, Irving Kirsch’s work has made a substantial and clinically relevant impact. In this text, a fractional amount of his rich research portfolio will be described by taking a closer look at several studies and by discussing their impact on future research and clinical practice. First and foremost, Kirsch is a pioneer when it comes to the experimental induction of pain, and early on, Irving Kirsch made use of the experimental heat paradigm to study placebo effects, which allows to determine an individual’s pain threshold and tolerance. In the heat pain experiment, a Peltier element, i.e., a thermal control module in the form of a small disk, is placed on participants’ forearm. Patients can then increase the magnitude of the heat stimulus with a mouse-click. For each click, the temperature increases by 0.5°C. Pain threshold is reached as soon as participants’ feeling changes from hot to painful (Locher et al., 2017; Rief & Glombiewski, 2012). Pain tolerance, in contrast, is defined as the time from the onset of the pain stimulus to a participant’s withdrawal from the stimulus, while participants are instructed to continue with the task for as long as they can (Granot, Sprecher, & Yarnitsky, 2003).

The studies that Irving Kirsch conducted in the area of experimental pain have substantially improved researchers’ understanding of the impact of several variables that contribute to the pain experience (Birnie, Caes, Wilson, Williams, & Chambers, 2014). In addition to being well validated and standardized, the heat pain paradigm can also induce transient, but robust central sensitization in healthy participants (Shenker, Haigh, Mapp, Harris, & Blake, 2008). Central sensitization has been defined as an “increase in synaptic efficacy in nociceptive pathways in the central nervous system and/or reduced descending inhibition of pain leading to enhanced pain” (p. 213) (Bromberg, Schechter, Nurko, Zempsky, & Schanberg, 2014). As a result of genetic susceptibility, repeated trauma, infections, and inflammation, the central nervous system might become overly effective in transmitting pain signals, and less effective in inhibiting them. Consequently, pain experiences that are greater in amplitude, duration, and spatial extent than would be expected under normal circumstances might qualify as a result of central sensitization (Manresa, Mørch, & Andersen, 2010; Woolf, 2011). Importantly, central sensitization is considered a key concept in the development and maintenance of chronic pain. Thus, gaining a better understanding of factors being associated with it is crucial. Kirsch’s work not only influenced placebo research but also increased our understanding of a key mechanism in the development and maintenance of chronic pain, i.e., central sensitization, by using the heat pain paradigm in many of his studies.

How does expectancy influence pain? Landmark studies by Kirsch and colleagues

In a paper from 1999, Kirsch and colleagues (Price et al., 1999) set out to examine two factors that might contribute to placebo analgesia, by means of experimentally induced heat pain: They were interested in the desire for treatment to significantly relieve pain, i.e., an individual’s desire for a therapeutic agent to provide pain relief. Further, they looked at expectancy, which in this context would mean that expectations for pain reduction after placebo intake might directly cause subsequent pain relief. Another purpose of the study was to compare the magnitude of placebo effects based on concurrent (i.e., at the moment of placebo treatment, immediately after pain induction) and retrospective (i.e., approximately 2 minutes later) ratings of pain. During the trial, participants were told that a new local anesthetic called Trivaricaine was used to evaluate its effect on pain reduction and that preliminary studies had confirmed its effect in reducing pain.

In a first step, participants underwent a calibration procedure to control for individual differences in pain perception. Then, participants received 10 stimulus trials with “strong” Trivaricaine, 10 trials with “weak” Trivaricaine, and 10 trials with a control solution. In order to establish expectations, participants were told that in the first two conditions, two different strengths of Trivaricaine were used and that neither the participants nor the investigator knew which bottle contained which strength of Trivaricaine. In a next step and in order to induce high desire for treatment to significantly relieve pain, participants were told that they would receive 30 trials of the highest pain level they had experienced in the study so far. They were further told that this could potentially be difficult to endure if the Trivaricaine was not working. To induce low desire for significant pain relief, another group of participants was told that even though the pain might be high, there were only six trials to go through and that most other participants had found this easy to endure.

The authors found expectancy, but not desire for pain relief, to contribute to the magnitude of placebo analgesia. Further, placebo effects assessed by remembered pain intensity (i.e., as rated two minutes post pain exposure) were over three times greater than placebo effects based on concurrent pain ratings. As placebo effects of remembered pain ratings were also strongly associated with expectancy ratings, the authors concluded that these “combined results further establish expectancy as a causal factor in placebo analgesia and extend previous work by showing that expectancy has a role in both placebo effects based on concurrent and remembered pain” (p. 154).

This paper has been cited more than 640 times to date (June 2022) and has inspired more research on the role of expectancy in (experimentally induced) pain: for example, a study from 2001 and with Irving Kirsch as senior author found that pain expectancy mediated the relation between depression and pain experience and partially mediated the relation between catastrophizing and pain experience in a study using the cold pressor procedure (i.e., an experimental procedure during which participants immerse one arm in ice water for one minute while providing ratings of their pain experience) (Sullivan, Rodgers, & Kirsch, 2001).

The role of expectancies in hypnotic analgesia

Expectancies not only influence the pain experience but also hypnotic analgesia (Baker & Kirsch, 1993). Typically, hypnotic analgesia includes direct suggestions for pain reduction, for example, the suggestion that the respective body part is numb and lacks feeling (Chaves, 1993). Three studies by Kirsch and colleagues contributed substantially to our understanding of the mediating effect of expectancies on hypnosis and placebos in experimental pain (Baker & Kirsch, 1993; Milling, Kirsch, Allen, & Reutenauer, 2005; Milling, Kirsch, Meunier, & Levine, 2002). In one of those studies, healthy participants were assigned to one of the three groups: two experimental and one control group (Milling et al., 2005). In the “placebo first” group, participants first received a placebo solution on the location of the painful stimulus – the Trivaricaine, known from an earlier study (Price et al., 1999) – followed by nonhypnotic imaginative analgesia suggestions and finally hypnotic imaginative analgesia suggestions. The hypnotic analgesia suggestion consisted of an audio tape that first provided positively framed educational content about hypnosis and an imaginative suggestion, i.e., participants were told that their hand was insensitive and numb, as if they were wearing a thick glove, adapted from Spanos and colleagues (Spanos, Perlini, & Robertson, 1989). The nonhypnotic imaginative analgesia suggestion included the same audio tape, except that the educational content that was designed to foster a positive attitude toward hypnosis was missing. In the second experimental condition, the “placebo last,” participants first received nonhypnotic imaginative suggestions, followed by the hypnotic imaginative suggestions, and finally the Trivaricaine placebo. The third group was a no treatment control condition. All three groups underwent the same experimental pain induction by a stimulus that consisted of a circular weight and was placed on participants’ finger. All groups underwent four different rounds of experimental pain, whereby the experimental groups had different orders of interventions (i.e., “placebo first” vs “placebo last” condition). One of the main aims of the study was to evaluate the contribution of expectancies to pain reduction across experimental groups.

The authors found that hypnotic and nonhypnotic imaginative suggestions did not differ significantly from each other and were more effective than no treatment in reducing pain. When the placebo was administered before the suggestions, i.e., “placebo first” condition, suggestions were more effective than placebo. When the placebo was given after the analgesia suggestions, i.e., “placebo last,” there was no difference in effectiveness when compared to the suggestions. Interestingly, pain reduction was mediated by expectancies, measured as expected pain before each trial in all groups.

Knowing that patients’ experiences play a crucial part in the experience and management of pain is also of great importance for clinical practice. For example, before the administration of a pain medication, patients could be asked about their expectations of the treatment: do they think it will be beneficial? Do they expect to experience side effects? If a patient expresses doubts, a physician would have the possibility to take these considerations seriously and to be more specific about the factors why the drug might be beneficial, i.e., to set realistic expectations.

What about clinical samples?

Kirsch’s work in the field of experimental pain was also the basis for the translation of the findings into the field of clinical pain. In a series of studies, Vase and colleagues (Vase, Riley, & Price, 2002; Vase, Robinson, Verne, & Price, 2003) further explored the role of expectancy and desire for pain relief. In these studies, they recruited patients who suffered from irritable bowel syndrome, which is defined as abdominal pain or discomfort that cannot be explained by structural or biochemical abnormalities, for at least 12 weeks in the preceding 12 months, with at least two of the three following features: pain is relieved with defecation, pain onset is associated with a change in frequency of bowel movements, or pain onset is associated with a change in the form of the stool (Horwitz & Fisher, 2001; Thompson et al., 1999). In contrast to the studies using experimental pain in healthy participants, in these samples with actual patients, desire for pain relief was more important and – together with expectancy – accounted for a large amount of variance (up to 81%) in pain variance (Vase et al., 2002, 2003).

Response expectancies in pharmacological treatment

In another study inspired by the work of Price, Kirsch and colleagues looked at the role of response expectancies in the clinical pharmacological setting, aiming to compare placebo analgesia in double-blind vs. deceptive administration of placebo (Pollo et al., 2001). In this interesting study, three groups of patients undergoing thoracic surgery were studied during the first few days after the surgery, when they were still in the hospital. All patients received analgesic treatment (i.e., buprenorphine) once they recovered from anesthesia. On the first day, an infusion of saline solution was started and continued for three consecutive days, and analgesic treatment was based on verbal request. The three groups differed regarding the instruction they received about the saline solution, i.e., they differed with regard to their expectancy.

In detail, patients in the first group were not told that the infusion had any analgesic effect, as this group was used to study the natural history. The second group was treated as a small double-blind study, in which they were told that the medication could either be a painkiller or a placebo, and they actually received either buprenorphine or placebo. Only those patients who received placebo in this group were included in the analyses. Finally, the third group was told that the infusion contained a powerful painkiller, when in fact they received either placebo or buprenorphine (i.e., their chances of actually receiving a painkiller were 50%). In sum, while participants in the second group were unsure whether they would receive a painkiller or a placebo, the patients in the third group were convinced that they would receive a powerful painkiller, even though their actual chances of receiving an analgesic agent were exactly the same as the ones of patients in group two.

The results of this study showed that different instructions that induced varying levels of expectancies led to a significant reduction in opioid intake, despite the same time course of pain across all groups. The placebo hence created a real analgesic effect. Also, when participants were certain that they received a powerful painkiller (group three), the placebo effect they experienced was stronger than when they expected to receive either a placebo or a painkiller. This study also confirmed Kirsch and Weixel, who predicted stronger placebo effects in a deceptive administration (i.e., as in group three – patients are convinced to receive painkiller, but actually their chances are only 50%) compared to a classic double-blind administration, as the expectations differ between the two settings (Kirsch & Weixel, 1988).

For the clinical practice, these findings are of great importance for both patients and clinicians. One important reason for this is the steady rise in prescription of opioids for chronic non-cancer pain in the US (Schuchat, Houry, & Guy, 2017) and in Europe (Berterame et al., 2016) over the last 20 years and associated deaths of drug overdose involving opioids (Seth, Scholl, Rudd, & Bacon, 2018). Therefore, new (adjunct) treatments are needed to taper down opioid doses. For chronic non-cancer pain, clinical guidelines recommend non-opioid analgesics as first-line treatment, and warn against potential risks of opioids (Busse et al., 2018). For context, chronic non-cancer pain is an umbrella term that includes painful conditions that persist or recur for more than 3 months and are not associated with a malignant disease (Busse et al., 2017; Merskey & Bogduk, 1994). In fact, for low back pain, for around 85–95% of people presenting to primary care providers, no pathoanatomical origin can be detected for their pain. The new International Classification of Diseases, 11th Revision (ICD-11), took this into account and created a new diagnostic entity for chronic pain, namely chronic primary pain. Chronic primary pain is defined as pain in one or more anatomical regions that lasts or recurs for three or more months and is associated with significant emotional distress or functional disability (Nicholas et al., 2019; Treede et al., 2015). Unless another diagnosis would better account for the symptoms, the diagnosis is appropriate independently of identified biological or psychological contributors.

Can we make use of expectancies in treatment?

Studies have shown large placebo analgesia effects in a range of chronic pain conditions, especially in irritable bowel syndrome (Kaptchuk et al., 2010, 2008; Klinger, Soost, Flor, & Worm, 2007; Petersen et al., 2012; Price, Craggs, Verne, Perlstein, & Robinson, 2007; Vase et al., 2002, 2003; Verne, Robinson, Vase, & Price, 2003). Additionally, in chronic low back pain trials, placebo effects have been shown to explain a significant percentage of large and clinically meaningful improvements (Chaparro et al., 2014; Henschke et al., 2010). Therefore, a group around Irving Kirsch set out to test the differences between treatment as usual (TAU) and treatment as usual + open-label placebo (Carvalho et al., 2016). Open-label placebo (OLP) is placebo without deception, the placebo is given with the patient’s full knowledge of the inert nature of the pill, but accompanied (in many cases) with the provision of an explanation of the effects and mechanisms of placebos (Kaptchuk et al., 2010). In Kirsch’s study, 83 adults with chronic low back pain completed a randomized controlled trial in which they received TAU with or without OLP for 3 weeks. Compared to TAU alone, the TAU + OLP condition led to greater pain reductions. Effect sizes were large for OLP vs. TAU for pain (g = 0.76) and pain-related disability (g = 0.74) (Carvalho et al., 2016). Interestingly, when the TAU group received OLP for 3 weeks after the initial treatment period, they showed comparable reductions in pain and pain-related disability compared to patients in the experimental group. Pain medication intake also decreased significantly in two-thirds of the sample while taking OLP (Carvalho, Kirsch, & Kaptchuk, 2017). This study was met with a lot of interest and critical discussion (Braillon, 2017; Mestre & Ferreira, 2017; Traeger & Kamper, 2017), also in the media (e.g., Schwartz, 2021).

In a follow-up study 5 years after the initial randomized controlled trial, the original participants were re-contacted and their pain, pain-related disability, and pain medication intake were assessed (Carvalho et al., 2021). Of the original sample, 82% took part in the follow-up assessments, and data on pain, disability, and medication use were not significantly different between those in the original sample and the follow-up sample. The results showed that the reductions in pain and pain-related disability that were obtained at the end of the trial were maintained at the five-year follow-up. In addition, the use of pain medication had decreased substantially, and participants reported an increase on the use of alternative approaches to manage pain (non-pharmacological approaches and self-care pain management). When asked if they would take OLP again if their doctor prescribed it, 65% of participants said yes. This finding reveals that patients considered OLPs to be acceptable, one of the most important clinical outcomes.

These results on the effect of OLP on pain, pain-related disability, and medication intake in patients with chronic low back pain are interesting, no doubt. But what about spontaneous improvement or the natural course of the pain? A comparison of OLP, natural course, and double-blind placebo was lacking for a long time – until Kirsch and colleagues set out to compare the three conditions in a sample of patients with irritable bowel syndrome (IBS) (Lembo et al., 2021). This was the first study to do so, and hence, its results were eagerly awaited, and even the protocol (Ballou et al., 2017) had been cited 46 times (as of June 2022). In the study, participants were randomly assigned to receive OLP, double-blind placebo, peppermint oil (to create ethical conditions for the double-blind placebo, half as many participants were randomized to receive peppermint oil), or “no pill control” (i.e., a control group who did not receive placebo and only received usual care). A total of 262 participants were randomized into three groups of interest, namely OLP, double-blind placebo, and no pill control (the group that received peppermint oil was not considered for the purpose of this analysis). Compared to no pill control, OLP significantly improved symptom severity and global improvement outcomes, with similar improvements in patients receiving double-blind placebo. The study design further allowed to observe that many participants with irritable bowel syndrome report extraintestinal symptoms at baseline in clinical trials, such as excessive sleepiness, fatigue, irritability, weakness, nasal congestion, or headache. However, and this is a new finding thanks to the OLP study that Kirsch and colleagues conducted (Lembo et al., 2021), these symptoms generally improve in severity over the course of a trial, and, interestingly, this is irrespective of the treatment arm (Ballou et al., 2022). The authors conclude that this might be due to various reasons, namely natural history, regression to the mean, Hawthorne effects (improvement due to the attention associated with enrollment in a trial), and participants’ expectations of improvement.

Clinical implications of Kirsch’s work on expectancies in experimental pain

As mentioned above, research with experimentally induced pain holds the potential to inform aspects of the (chronic) pain experience, as experimental pain is thought to induce transient central sensitization. Therefore, the last part of this paper focuses on discussing potential clinical implication of Kirsch’s work on expectancy in experimental pain.

The idea that expectancy regarding treatment success could be an important factor in patients with chronic pain started to gain interest in the late 1990s, when Kole-Snijders and colleagues found that treatment credibility ratings strongly predicted outcomes in a trial of cognitive-behavioral treatment for patients with chronic low back pain (Kole-Snijders et al., 1999). Kole-Snijders and colleagues’ assessment of credibility was done by asking participants about the extent to which they believed that the program would help them cope with their pain on a visual analog scale – which is similar to Price, Milling, Kirsch et al.’s question (“What do you expect the pain intensity to be with the Trivaricaine?,” on a visual analog scale) (Price et al., 1999). A year later, a validation study of the credibility/expectancy questionnaire revealed a distinction between the two constructs, in that expectancy refers to patients’ belief regarding the improvements that will be achieved by the treatment, while credibility refers to how logical and believable the treatment seems to the patient (Devilly & Borkovec, 2000).

Nevertheless, a few years later, Kalauokalani and colleagues found pre-treatment expectancy to predict improved function after massage and acupuncture, irrespective of which treatment patients received (Kalauokalani, Sherman, & Cherkin, 2001). In an analysis of two pooled randomized controlled trials on cognitive-behavioral treatment for patients with fibromyalgia (a condition characterized by widespread pain, sleep problems, fatigue, emotional distress, and sensitivity), pre-treatment expectancy of how well the treatment will work was an important predictor: Participants with higher pre-treatment expectancies reported significantly higher tolerance of activity despite the pain, higher health-related quality of pain, less catastrophizing thoughts and negative emotions, and better abilities to cope with their pain (Goossens, Vlaeyen, Hidding, Kole-Snijders, & Evers, 2005). Even though the explained variance of expectancy was considerably lower compared with other studies in the field of cognitive-behavioral therapy, the results suggest that one of the tasks of a therapist might be to assess and potentially modify the expectancies of patients.

Now that the research of Kirsch and colleagues intriguingly has demonstrated the role of expectations for the pain experience, one might ask: What are the mechanisms that steer expectancies of recovery in patients with chronic pain? A large proportion (37%) of variance of expectancy is explained by pain frequency and a smaller proportion by pain-related fears, and pain, functional disability, expectancy, and negative affect are closely interrelated (Boersma & Linton, 2006). Negative expectations might therefore be the logical consequence of the experience of frequent and intense pain, combined with the belief that pain signifies harm and danger (Boersma & Linton, 2006).

Do expectancies influence the pain experience?

Yes, they do!

In sum, expectancies seem to play a significant role in the experience of pain and pain-related disability, but also when predicting the success of a treatment. Hence, Kirsch’s contribution in the area of expectancies is crucial to better understand which expectations patients might have with regard to their future pain or the success of a new treatment. Given that chronic pain is among the health conditions that are most expensive and account for most years lived with disability (Vos et al., 2017), it is imperative that we begin to understand which treatment works for whom – and the assessment (and potentially, modification) of patients’ expectations might be a key mechanism to consider.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung [P5R5PS_203064; PZ00P1_201972].