A neurobiologist’s attempt to understand persistent pain

1.1 Complex problems require an integrated approach

2.1 What pain is and what it is not

Unfortunately, the “pain” literature is often conceptually unclear due to a lack of an explicit distinction between pain as a sensation (experience) on the one hand and its causes and underlying mechanisms on the other. Pain is a sensation, and in common with other sensations (e.g. itching) it has a bodily location. Nevertheless, pain is obviously not a thing that can be physically localized, in contrast to neurons and their activities (pain is not in the brain), inflammation, a herniated disc, and so forth. Neither is pain a perception: an object or event exists regardless of whether it is perceived or not, whereas a pain (e.g. in the knee) exists only as it is felt [17]. A perception may be falsified (I thought my pain was caused by a torn meniscus, but it turned out to be something else), while a sensation cannot. Whether pain is felt and the intensity of suffering, however, depend critically on how the person perceives the situation. In other words, we must distinguish between the experience of pain and the meaning that the person gives to it. It obviously does not make sense to say that “my knee hurts but I do not feel it”, but the person’s beliefs about the cause of pain may be right or wrong. It is an example of conceptual confusion when the doctor questions the patient’s report of pain because he does not find a plausible cause by his examination. The pain is exactly as the person describes it (if we exclude persons that for some reason lie); the cause of pain, however, may be located somewhere else or not be what the person believes. As pointed out by Bennett and Hacker [17, p. 123]: “So-called referred pains (e.g. sciatica, referred toothache) are not pains which the subject mistakenly thinks are where he points or assuages, but rather pains that are felt in places other than the locus of the injury, infection, etc.... So, the location of the cause of a pain must be distinguished from the location of the pain itself.”

2.2 Homeoceptor rather than nociceptor?

The use of words such as “nociceptor”, “pain system”, “pain receptor” and “pain matrix”, may restrict our thinking unduly. For example, the basic properties of nociceptors enable them to constantly monitor the composition of the tissues and the mechanical and thermal forces they are subjected to, whereas they evoke pain only under certain conditions and in certain contexts. That is, the main task of nociceptors appears to be to detect and signal homeostatic threats [18,19], and only when the threat requires conscious action do we feel pain. Consequently, “homeoceptor” might be a more appropriate term than nociceptor. Daily activities that are not experienced as painful such as using scissors, opening a jar, playing the guitar, riding a bicycle etc. nevertheless must be expected to excite many mechanosensitive nociceptors [20]. Also, when we turn over in bed, change our position in a chair, or move a leg to alter the strain on ligaments it is most likely initiated by signals from polymodal nociceptors. Furthermore, microneurographic studies [20,21] show that considerable firing of human C-fibres can be evoked by non-noxious stimuli. Even in controlled, experimental situations the relationship between C-fibre firing-frequency and pain is far from linear [21]. Thus, firm mechanical stimulation of the skin (pinching, pressing with a pointed object) can make C-fibres fire with a high frequency (10/s) without evoking pain, whereas radiant heat causes pain with a C-fibre firing-frequency of 2/s. Even with radiant heat the nociceptors begin increasing their firing rate at temperatures well below the threshold for eliciting pain [21]. In conclusion, all sensory signals must be evaluated in a broad context to be useful, and signals from nociceptors are no exception. As expressed by Prescott and Ratté [22, p. 632]: “...noxious stimulation, although crucial for normal (i.e. nociceptive) pain, is neither necessary nor sufficient to evoke pain...There is nothing innate to primary afferent nociceptors that endow them, and only them, with the capacity to evoke pain.”

3.1 Network activity and the feeling of pain

The brain correlate of our mental life is most likely synchronized, oscillatory electrical activity in distributed neuronal networks [23,24]. This pertains also to the feeling of pain [25, 26,27], and numerous neuroimaging studies (PET, fMRI) have revealed a consistent pattern of activated cortical sites when people are subjected to experimental pain [28, 29,30]. The assumption that these sites constitute nodes in a “pain network” is strengthened by the demonstration that the nodes are functionally connected [26]. The anterior cingulate gyrus, the insula, the second somatosensory area (SII), and the thalamus (less consistently some other parts) show enhanced activity during acute, experimental pain (Fig. 1). As might be expected, these sites are also targets of nociceptive signals carried in the spinothalamic tract [31]. However, activation of the typical “pain network” correlates with pain experience even in situations with no nociceptive input, such as in persons under hypnotic suggestion [32,33]. Indeed, excruciating pain can be experienced in situations where a person mistakenly perceives a serious injury [34]. Furthermore, merely anticipation of pain activates the “pain network” [35,36], whereas perceived ability to control the intensity of an impending nociceptive stimulus reduces the network activity and subsequent pain experience [37]. The link between network activity and pain sensation is further strengthened by the finding that network activity increases in parallel with reported pain intensity rather than with stimulus intensity [36]. Fig. 2 gives a highly schematic presentation of various inputs that may drive the “pain network” and thereby evoke pain sensation.

Fig. 1

Regions of the cerebral cortex showing increased activity during experimental pain, as studied with functional MRI. Based on [25].

Fig. 2

The cortical pain networks and modes of activation. Schematic to showhow quite different inputs can activate the network – from peripheral nociceptors triggered by tissue injury to purely mental processes.

From Brodal P. The Central Nervous System 5th Ed. New York: Oxford University Press; 2016. With permission.

Due to the complexity of clinical pain, the responsible networks are harder to identify than those related to experimental pain. Accordingly, results vary among studies and are often difficult to interpret. Nevertheless, it seems safe to conclude that persistent clinical pain does not necessarily have the same cerebral “signature” as acute experimental pain [38]. The involvement of the medial prefrontal cortex [38,39] seems particularly interesting, since this region is thought to be implicated in making meaning of sensations from one’s own body [40]. Less consistently, several cortical and subcortical regions outside the “pain network” have been associated with persistent pain (e.g. the amygdala, the accumbens nucleus, and the cerebellum). This should perhaps not be surprising, since persistent pain influences most cognitive and emotional processes.

Numerous MRI studies have reported widespread cortical grey matter reductions (less often increases) in patients with persistent pain [41,42]. Regions with alterations differ considerably among studies, however, and the significance of such findings is unclear. Broadly speaking, the findings indicate that persistent pain leads to changes in large-scale networks, and that core pain-related regions may be affected in most persons with persistent pain whereas alterations in other regions may be pathology-specific [42]. The grey matter alterations may be a result of persistent pain rather than a cause [43,44], since they seem to disappear when the pain is relieved (e.g. after surgery for painful hip osteoarthrosis, [43]).

3.2 Is the “pain network” specific for pain?

The foregoing account strongly suggests that the feeling of pain is correlated with activation of certain cortical networks. These networks are most likely not specific for pain, however. Thus, the nodes of the pain network are activated in conjunction with other mental phenomena than pain. For example, the anterior cingulate gyrus has been implicated, among other tasks, in self-control and monitoring of goal-directed actions [45]. The insula shows increased activity in tasks evoking emotions, responds to inputs from various sensory modalities, and is related to bodily awareness [46,47]. Both the anterior cingulate gyrus and the insula seem to be involved in attentional processes [45, 46,47]. Overall, the “pain network” activity seems mainly to reflect the salience of a stimulus regardless of its modality [47,48,49]. Accordingly, both the anterior cingulate gyrus and the insula are included in the so-called salience network [50,51], characterized by its activation by any event that catches our attention and increases arousal. Certainly, pain is a highly salient feeling not meant to be ignored.

Nevertheless, the question of whether cortical networks show some degree of pain specificity is not finally settled. For example, some experimental and clinical findings may indicate that the posterior parts of the human insula at least show some degree of pain specificity [52, 53,54]. Furthermore, Wager and coworkers [29] identified what they interpreted as a “neurologic signature of pain”, based on fMRI and machine-learning analysis. They found that experimental pain is associated with a specific pattern of cerebral activation that differs from activation patterns caused by other salient sensory signals (see, however [55] for a critical discussion of Wager and coworkers’ conclusions).

Finally, it should be emphasized that the fact that certain patterns of network activity are correlated with pain experience does not prove that its activation is necessary for the feeling of pain. Indeed, a patient suffering a cerebral lesion (due to herpes simplex encephalitis) encompassing the amygdala, anterior cingulate, and the insula reported normal pain sensations to nociceptive stimulation [45,56]. This may probably be an example of network degeneracy [11,12], as discussed above.

4.1 Many false alarms preferable to a few missed ones

The importance of the ability to feel pain is demonstrated by the sad fate of individuals born without pain sensations [57]. Pain may be viewed as one among several alarms that serve to protect our physical and mental integrity by forcing us to change behaviour. In contrast to other sensory events that tend to adapt when a stimulus continues unaltered, alarms would not serve their purpose if they adapt, and the neural elements responsible for producing pain sensitise rather than adapt. Also, the relationship between pain and a specific event must be quickly learned and remembered. This requires that responsible neuronal groups are highly plastic. It follows also that the detector must be very sensitive; no threats should be missed, and it is better for survival with many false alarms than a few missed ones. This further implies that the system cannot be very specific; genetic pre-specification of all possible noxious life events would be impracticable. Instead, the system is highly adaptable due to its capacity for fast learning. These properties make the system prone to learning from alarms that were meant to be transitory, and may help, at least in part, to explain why the system so often goes awry. A misunderstanding concerning the cause and consequences of an event or situation is remembered (subconsciously), and continues to influence behaviour.

4.2 Pain, learning, and memory

When pain becomes persistent without a continuous nociceptive input, learning (in the form of widespread synaptic changes) must have taken place. In natural situations, much of the learning is conditioned rather than a direct response to a nociceptive input [58]. For example, in a situation requiring movement of a recently injured body part, the alarm may sound even before movement has started. The person learns to avoid everything that is expected to cause the pain [59,60]. Memory traces are presumably consolidated every time pain is temporally linked with the intention to move – it need not be an overt movement. To the person, the pain (or merely the expectation of pain) means that the body part is injured, regardless of the doctor’s assertions to the contrary. However, while learning avoidance behaviour is an important factor in maintaining pain [59], complex natural situations certainly include more varied learning opportunities [61]. It is how the person perceives the total situation – including nociceptive input, other sensory inputs, social context, previous experiences, expectations, mindset, and so forth – that will determine what is learned and the subsequent pain-related behaviour (Fig. 2). We will return to this under “The meaning of pain”.

A role of memory in persistent pain is exemplified by case reports of a few persons who, after becoming amnesic due to (most likely) transitory brain ischaemia, experienced relief of persistent pain and opiate dependence [62,63]. It should be noted that the patients had somatic diseases (bone metastases and complications from abdominal and back surgery) that seemingly explained their severe pains as caused by nociceptive signals. The kind of amnesia in these patients strongly suggests injury to the hippocampal formation. That this part of the brain should be involved in persistent pain may perhaps seem surprising since it is related mainly to explicit (declarative) and episodic memory [64]. However, further support for hippocampal involvement in persistent pain comes from a study showing that transition to persistent back pain was associated with altered hippocampal functional connectivity, notably with the medial prefrontal cortex [65]. Interestingly, connections between the medial prefrontal cortex and the hippocampal formation are involved in extinction of contextual fear [66]. Finally, animal experiments indicate that reduced adult hippocampal neurogenesis is related to development of persistent neuropathic pain [67]. Collectively, it seems plausible that deficits in contextual processing [66] – depending on disturbed interactions between the medial prefrontal cortex and the hippocampus – contributes to making pain persistent. Unfortunately, we cannot yet answer the clinically most relevant question: what might cause a disturbed contextual processing in patients with persistent pain?

4.3 Pain as result of an interpretation of the state of the body

In their book “Phantoms in the brain” [68, p. 54] Ramachandran and Blakeslee aptly states: “..pain is an opinion of the organism’s state of health rather than a mere reflexive response to an injury”. Accordingly, pain is felt (the alarm bell rings) whenever a homeostatic threat is considered sufficiently serious by our bodily surveillance systems. To reach an optimal conclusion, many kinds of information must be evaluated and integrated (presumably by the salience networks): is bodily homeostasis threatened? How serious is the threat? What does the threat mean? For a correct interpretation of the multitude of sensory signals arising in the body, a proper balance among the various somatosensory modalities seems necessary. Loss of sensory information – regardless of whether it concerns low-threshold mechanoreceptors, thermoreceptors, nociceptors, or all – may cause persistent pain [69, 70,71]. Furthermore, to feel the body as normal, motor commands need to be matched by expected sensory feedback from the moving parts [72,73]. Phantom pain, for example, may probably be understood in the context of distorted sensory and motor information processing. Thus, there is no flow of afferent signals from the missing body part, and motor commands are not matched by the expected sensory feedback. Often the person experiences that the missing body part is in a painful position, which cannot be altered voluntarily [74]. The above explanation is supported by the relief of pain with mirror visual feedback [75]. Presumably, this induces an illusion of normality and movement control of the missing body part, and perhaps “resets” networks that monitor bodily homeostasis so that the alarm is shut off. The illusion of normality created by mirror therapy seems to depend both on visual and proprioceptive feedback [76]. Mirror visual feedback has been found to reduce pain also in paraplegic patients with spinal cord injuries [77,78].

4.4 Disturbed body image and persistent pain

The foregoing account suggests that persistent pain can be reduced by therapies that alter (normalize) the person’s perception of her own body. Indeed, patients with persistent pain often show evidence of disturbed ownership and body image (the latter term used here in a broad sense, including both implicit and explicit body knowledge) [79, 80, 81, 82,83]. It is plausible that neural networks processing bodily information are changed after deafferentation, but as mentioned above, grey matter alterations occur in virtually all kinds of persistent pain conditions [41, 42, 43,44], presumably as an expression of network plasticity. For example, most patients with complex regional pain syndrome shows neglect-like symptoms and a distorted perception of the affected limbs [81], and patients with chronic back pain showed a disrupted body image of the back [82].

4.5 Network locking and persistent pain

So far, we have discussed evidence that persistent pain is correlated with plastic changes in neural networks, including all levels of the CNS. At the highest level the changes seem to concern networks related to sensory processing, to stimulus salience, memory, body image, and body ownership. The widespread alterations are most likely functionally correlated, and may help to explain core features of persistent pain at a mechanistic level. However, we still have only partial answers to what drives the plastic changes in the first place. After all, most people do not develop complex regional pain syndrome after an injury, not all persons with deafferentation develop persistent neuropathic pain, not all diabetics with peripheral neuropathy experience pain, only some suffer from persistent postoperative pain, and so forth. Many gene polymorphisms [84,85], personality characteristics [86,87], and neuroimaging data [55,88] are associated with individual variations in the tendency to develop persistent pain. Nevertheless, we are far from being able to predict with any clinically useful degree of certainty who will and who will not develop persistent pain. As concluded by Denk and McMahon [85]: “Just like life itself, some aspects of chronic pain may end up remaining irreducibly complex and ultimately unpredictable.” It may be that just one small added negative influence – that would be impossible to identify–is the final straw initiating an avalanche that alters networks responsible for homeostatic surveillance. A hypothesis based on the Ising model from physics [89] explains how acute pain becomes chronified when small changes (of one or more among numerous factors) suddenly leads to large changes. The networks may then be brought to a state that, for example, locks the “homeostat” at a high and inflexible level of sensitivity (Fig. 3). It should be realized that neither neurons nor complex interactive networks behave linearly [90], and that brain functional networks are continuously shaped and reshaped by synchronization (by phase and frequency locking) and desynchronization. The normal functioning of the brain is characterized by constantly shifting activity within and between large-scale networks [90,91].

Relief of persistent pain by seemingly nonphysiologic measures, such as caloric or galvanic vestibular stimulation [92] transcranial motor cortex stimulation [78,92], and electroconvulsive therapy [93] may perhaps be due to a sudden input that releases a locked network and restores plasticity [94].

5.1 The need for a higher level of explanation

Much of the discussion in the foregoing section dealt with mechanistic explanations of persistent pain. Knowledge about neurons, transmitters, and neural networks is, however, insufficient when trying to understand and alleviate the suffering of real people. More and more detailed knowledge down to the molecular level does not answer the clinically most important questions, namely what “drives” the pain network in one person, and what determines that nociceptor signals are prevented from evoking pain in another person with very similar injury or disease? To answer such questions, we need to go beyond the biologic and mechanistic level and search for the meaning the person ascribes to the pain (Fig. 3).

Fig. 3

Persistent pain and meaning. Possible steps and factors that may differentiate the normal situation, in which pain abates in pace with the healing, from situations with development of persistent pain.

From Brodal P. The Central Nervous System 5th Ed. New York: Oxford University Press; 2016. With permission.

5.2 Evaluation of the total situation determines pain experience

Eric J. Cassell in his book “The nature of healing” [95, p. 2] puts pain in a broad context: “Sickness and its manifestations are inextricably bound up with the phenomenon of meaning. Everything that happens to people – events, relationships, every sight and sound, everything that befalls the body – is given meaning.” A pertinent example is the apparently paradoxical finding that a large proportion (37%) of persons with acute injuries arriving at an emergency clinic do not feel pain [96], and a much higher number among wartime casualties have a pain-free period [97]. Henry K. Beecher, studying pain in wounded soldiers evacuated from the frontline in the Second World War, could find no relationship between the extent of the injury and experienced pain, and concluded [97, p. 1609]: “The intensity of suffering is largely determined by what the pain means to the patient.” With our present day knowledge, we can explain the freedom from pain at a mechanistic level as due to descending inhibitory connections that effectively shut off transmission of nociceptive signals [98], but this begs the question of what controls the descending inhibitory system. P.D. Wall discussed what the meaning might be in the examples with injury and no pain, and proposed [99, p. 260] “..that injury may be followed by circumstances in which treatment of the injury does not have the highest biological priority. Pain does not occur in this phase. The three obvious high priority behaviours are fighting, escaping and obtaining aid.” Perhaps the most clear-cut example of this kind of meaning is life-threatening situations in which continued physical activity is more important for survival than passivity. Likewise, an athlete may feel less pain because the goal of winning is more salient than the fear of later physical incapacity [100].

The importance of meaning for pain experience is not limited to the examples of acute injuries discussed above [101,102]. For example, whatever the cause of pain, fear of future consequences will influence its intensity and unpleasantness: Does the pain mean prolonged suffering or death? Will the condition prevent me from pursuing activities and goals that are dear to me? Will I lose my job and face economic problems and social isolation? Fear and anxiety are intimately coupled with the meaning the person attributes to the pain – they are aspects of the same phenomenon [95,99].

Experimental studies also support the importance of meaning for pain experience, and also the impact of verbal suggestions. For example, ischaemic arm pain in healthy volunteers was tolerated considerably longer when the person was told that ischaemia is good for muscle cells [103]. It was also shown that this effect could be abolished completely with administration of naloxone and rimonabant (blocking opioid and cannabinoid receptors). One study using hot and cold stimuli applied to the neck showed that expectation of tissue damage increases the experience of pain [104]. This is of course just two among many studies demonstrating effect of expectation – placebo and nocebo – on pain experience [for a recent review see 105].