Department of Basic Medical Sciences, Neurosciences and Sense Organs, ‘Aldo Moro’ University of Bari, Bari, Italy
Correspondence: G. Defazio, Department of Basic Medical Sciences, Neurosciences and Sense Organs, ‘Aldo Moro’ University of Bari, Policlinico, Piazza Giulio Cesare 1, 70124 Bari, Italy (tel.: +0039 080 5478511; fax: +0039 080 5478532;
In recent years, increasing attention has centred on pain in Parkinson's disease (PD). Pain in PD is heterogeneous in quality and body distribution. To clarify how the various pain types relate to PD and to propose plausible treatment strategies, in this paper we reviewed psychophysical, neurophysiological and imaging data reported in parkinsonian patients with and without pain. Most available evidence supports abnormal central nociceptive input processing that probably reflects an impairment in the lateral and medial pain pathways. Changes in central pain processing probably underlie all the different pain types and also intervene in patients with PD without pain. Thus, altered pain processing might predispose patients with PD to spontaneous pain that is variable in quality. These background pain-processing abnormalities may interact with additional factors (such as contractures secondary to marked rigidity/bradykinesia, dystonia and medical conditions associated with painful symptoms), thus causing pain to manifest itself clinically in various ways and providing candidate targets for pain treatment in PD.
Patients with Parkinson's disease (PD) frequently suffer from pain that is variable in quality and localization [1-3]. The widely used Ford criteria  distinguish pain arising in a dystonic body part (dystonic pain) from pain in a body part unaffected by dystonia (non-dystonic pain) that may manifest as muscular pain, rheumatic or arthralgic pain, peripheral neuropathic pain and central neuropathic pain. Several reports also describe unusual pain involving the face, head, pharynx, epigastrium, abdomen, pelvis, rectum and genitalia, all areas in which painful dystonia or musculoskeletal conditions are unlikely [5-8]. Patients often report experiencing more than one type of pain.
Several epidemiological studies estimated a pain prevalence in PD ranging from 60% to 83% [9-12], a figure significantly greater than estimates in age- and sex-matched control groups [9-12]. Further supporting pain as a PD non-motor symptom, a large population study showed greater chronic analgesic drug use in patients with PD than in the general population .
Possible risk factors for pain in PD include female gender [11, 14] , motor complications [14, 15], depressive symptoms and medical conditions (such as diabetes mellitus, osteoporosis, rheumatic disease, degenerative joint disease, arthritis and disc herniation) potentially associated with painful symptoms [10-12, 14-19]. Recent evidence also suggests a genetic contribution as variants within the SCN9A and FAAH genes were associated with an increased risk of pain in patients with PD .
Pain frequency in PD tends to increase as motor signs become more severe [11, 16]. Consistently, pain may manifest in the body site initially, or more severely, affected by rigidity or bradykinesia, may worsen during off episodes or in patients with dystonia or dyskinesia, and may improve in the on state, particularly in patients with motor and non-motor fluctuations [9, 11, 15-18, 21, 22]. Pain may also manifest independently of motor problems in 25–64% of patients [11, 16, 23].
Based on the variable relationships between pain and PD motor signs, several authors proposed distinguishing PD-related pain and PD-unrelated pain [9, 16, 23, 24]. Although one study did not show differences in frequency in PD-unrelated pain between patients with PD and controls , this classification has two disadvantages because it neither addresses a particular pain subtype  nor predicts an individual response to dopaminergic treatment [16, 21, 22, 24]. The relationship between the various pain types and PD therefore remains puzzling. Nor is it clear whether dopaminergic transmission plays a prominent role in PD pain [25, 26].
Information from several psychophysical, neurophysiological and imaging studies provided useful information on pain-processing mechanisms in PD with and without pain. These studies nevertheless yielded heterogeneous results, probably because of methodological factors. For example, some failed to differentiate between patients with PD with and without pain, whereas others tested patients with different kinds of pain, and investigated patients with different procedures. These and other limitations leave the mechanisms responsible for pain in PD uncertain and make it difficult to know whether a common pathophysiological mechanism underlies the different types of pain.
In this paper, we reviewed articles dealing with the mechanisms responsible for pain in PD that clearly differentiated patients with and without pain, and classified the type of pain and investigational procedures used. Understanding more about the pathophysiological mechanisms underlying pain would be useful in planning treatment strategies.
No ethics board approval was required for this review of the literature.
Among studies assessing pain processing in PD, our review identified studies on pain thresholds and pain tolerance [14, 27-38], and neurophysiological studies focusing on the nociceptive withdrawal reflex (NWR) to electrical stimuli (exploring pain processing in the spinal cord) [27, 28, 39] and on scalp CO2 laser-evoked potentials (LEPs), a tool that can non-invasively assess the functional status in cerebral structures (the cingulate gyrus and insula) responding to nociceptive inputs [33-36]. Finally, studies investigating the descending inhibitory control (DNIC) system assessed activity in the endogenous pain inhibitory system , and positron emission tomography (PET) provides information on pain-induced activation in cortical areas [30, 31]. Because studies enrolled patients under chronic dopaminergic treatment rather than naive patients, we referred to those assessing patients in the off condition.
Patients with PD with pain
When we sought information on pain mechanisms from studies in patients with PD with pain, we observed that patients with muscular and peripheral neuropathic pain have a lower pain threshold and pain tolerance to electrical stimuli, and lower pain threshold to heat thermode than healthy controls (Table 1) [14, 27]. No correlation was found between the reduced pain threshold to electrical stimuli, and intensity, quality and body distribution of muscular and peripheral neuropathic pain .
Table 1. Results from studies applying psychophysical and neurophysiological procedures to parkinsonian patients with and without pain under off conditions
Test, first author, year
Parkinsonian patients with pain versus healthy subjects
Pain-free parkinsonian patients versus healthy subjects
Parkinsonian patients with pain versus those without
NWR, nociceptive withdrawal reflex.
Pain type: aprimary central pain; bmusculoskeletal pain; cperipheral neuropathic pain.
Although data are lacking on the electrical pain threshold and pain tolerance in primary central pain, patients with this type of pain have reduced pain thresholds to cold pressor  and heat thermode [32, 33] (Table 1). No correlation was observed between the reduced heat-pain threshold and severity of primary central pain .
The NWR threshold to electrical nociceptive stimuli was reduced in patients with PD with musculoskeletal pain , thus suggesting abnormal nociceptive input processing in the spinal cord. No other pain types were tested by this procedure.
A study investigating scalp CO2 LEPs showed a reduction in N2/P2 LEP amplitude recorded at the vertex in patients with PD with muscular pain . No significant correlation was observed between the N2/P2 amplitude reduction and muscular pain intensity . Another study using a different methodology found an increased N2/P2 amplitude in patients with PD with central neuropathic pain . The different types of N2/P2 amplitude changes (decreased or increased) in patients with muscular pain  or in patients with central neuropathic pain  might reflect pathophysiological differences between different pain types. Methodological differences in the technique used, including number of averaged stimuli and duration of interstimulus interval, may however also explain the different types of N2/P2 abnormalities reported .
Others noted similar DNIC system activation in patients with PD suffering from pain of variable quality and healthy controls , thus suggesting that this system makes no contribution to pain in PD.
Collectively, studies in patients with PD with pain conclude that the different pain types arise from abnormal nociceptive input processing at various levels of the CNS. The studies we reviewed reported only mild changes in pain-processing mechanisms likely contributing to the intermittent type of pain often seen in patients with PD. The lack of correlation between pain-processing abnormalities and intensity/quality of pain [14, 32, 34] does not support an exclusive relationship between altered pain-processing mechanisms and spontaneous pain.
Patients with PD without pain
When we considered studies on patients with PD without pain, various groups reported that pain thresholds to electrical stimuli [14, 27-29], cold water  and laser CO2 heat [34-36], and pain tolerance to electrical stimuli  are lower in patients with PD without pain than in healthy subjects (Table 1). Similarly, the heat-pain threshold was slightly lower in pain-free patients with PD than in control subjects [32, 33] (Table 1).
Patients with PD without pain also have a reduced NWR threshold to electrical stimuli [27, 28].
Similarly, in several studies investigating scalp CO2 LEPs, we found a reduced N2/P2 LEP amplitude at the vertex in hemiparkinsonian patients without pain regardless of the clinically affected side [34-36]. Conversely, other investigators using a lower number of averaged stimuli and longer interstimulus intervals recorded normal LEP amplitudes in patients with PD without pain , possibly because their technique masked the difference between subjects.
A PET study showed abnormal pain-induced activation in the cortical areas involved in sensory discriminative processing of pain (such as insula/SII) and cortical areas subserving affective motivational processing of pain (anterior cingulate cortex and prefrontal cortex) .
Descending inhibitory control system activation was similar in patients with PD without pain and healthy controls , again arguing against DNIC system involvement in PD pain.
Overall, these findings indicate that the changes in central pain processing demonstrated by psychophysical and neurophysiological investigations in patients with PD with pain also intervene in patients with PD without pain.
Patients with PD with pain versus patients with PD without pain
When we compared data from studies focusing on patients with PD with muscular pain and patients without pain (Table 1), we found a similarly reduced pain threshold and pain tolerance to electrical stimuli in the two groups [14, 29]. Similar findings have been observed in patients with PD with peripheral neuropathic pain (Table 1) . No studies have tested electrical pain threshold and tolerance in central neuropathic pain. As compared with pain-free patients, patients with central neuropathic pain had similarly reduced pain threshold to cold pressor tests , but lower heat-pain thresholds (Table 1) [32, 33].
Studies by Tinazzi et al. [34-36] showed a decreased N2/P2 LEP amplitude at the vertex in patients experiencing muscular pain as compared with pain-free patients. Although the greater amplitude reduction in LEPs recorded in painful patients might suggest that pain leads to additional changes in nociceptive processing, the finding can also be the consequence of the so-called segmentary inhibitory effect , a phenomenon that well explains the LEP amplitude reduction by experimental muscle pain . Using a different methodology, the N2/P2 LEP amplitude observed by Schestatsky et al.  was increased in patients with central neuropathic pain and normal in pain-free patients.
A recent PET study  showed similar brain activation patterns in patients with PD with central neuropathic pain and patients without pain, even though patients with pain had lower pain activation in the right prefrontal cortex and posterior insula and a higher pain activation in the right anterior cingulate cortex than pain-free patients.
In conclusion, although some observations suggest that spontaneous pain is associated with additional changes in nociceptive processing mechanisms [33, 34], the majority of the studies indicate that similar changes in central pain processing intervene in patients with PD with and without pain.
Pain-processing mechanisms and dopaminergic medication
An acute levodopa challenge had either no effect or normalized the decreased pain threshold observed in patients with PD under off condition [27, 28, 30-33, 36]. These variable findings might reflect methodological issues, placebo effect or confounding by other factors, including the involvement of monoamine systems other than the dopaminergic system and the presence of levodopa-induced motor complications.
Support suggesting that the decreased pain threshold observed in patients with PD involves other monamine systems comes from a recent study showing that an acute apomorphine challenge had no effect on electrical and heat-pain thresholds as compared with placebo in patients with PD without pain .
Comparing the levodopa-induced change in cold-pain threshold and tolerance among stable responders, fluctuators without dyskinesia and dyskinetic patients (all groups included patients with muscular pain and patients without pain), Lim et al.  noted a larger threshold increase in dyskinetic patients than in stable responders. After patients received levodopa, the cold-pain threshold increased also in fluctuators without dyskinesia, though to a lesser extent . Previous studies providing variable findings on levodopa-induced changes in pain thresholds failed to specify whether patients were stable levodopa responders, fluctuators or had dyskinesia. It is worth noting that the studies yielding a significant levodopa effect on pain thresholds [27, 28] enrolled patients with a long disease duration, and presumably therefore included patients with levodopa-related motor complications.
Acute levodopa medication increased the NWR threshold to electrical stimuli in patients with PD without pain , but a later study failed to confirm the finding . Levodopa left the N2/P2 amplitude unchanged in patients with muscular pain and in patients without pain [34, 36]. Finally, levodopa reduced nociceptive activation in the right posterior insula and right anterior cingulate cortex in pain-free patients with PD, whereas apomorphine did not [30, 31].
Overall, the foregoing findings imply that dopaminergic mechanisms may contribute to pain-processing abnormalities. The greater levodopa effectiveness in patients with motor complications is consistent with the reported pain relief by dopaminergic medication in patients with PD with motor complications rather than in the stable responders [21, 22]. The finding by Nebe and Ebersbach  that both muscular pain and other pain types (such as pain involving the epigastrium, abdomen and genitalia) probably unrelated to muscular conditions undergo partial or complete remission during the on medication state might suggest the hypothesis that levodopa relieves fluctuating pain by interfering with pain-processing mechanisms rather than merely improving rigidity or bradykinesia. Additional studies, however, are needed to confirm the observation. Finally, the greater levodopa effectiveness in relieving pain threshold abnormalities in patients with motor complications , its absent effect on N2/P2 amplitude  and the lack of effect of apomorphine on pain-induced cerebral activation patterns  may suggest that levodopa induces its anti-nociceptive effect and interferes with abnormal pain processing partly through neurotransmitter systems other than dopamine.
Pain-processing mechanisms and risk factors for pain
Six studies assessed whether the demographic or clinical features associated with spontaneous pain also correlated with the pain-processing abnormalities found in patients with PD [14, 30, 32, 35, 37, 39]. In these studies, pain-processing abnormalities measured by pain threshold or LEP assessment correlated neither with sex, motor complications nor medical conditions associated with painful symptoms.
Three relatively small studies also found no correlation between the pain threshold to cold/warm stimuli and severity of motor signs [30, 32, 39]. Despite these findings, a recent study enrolling 101 patients with PD, therefore providing greater statistical power, observed a significant inverse correlation between the pain threshold to electrical stimuli and severity of motor signs or depressive symptoms . Consistently, the NWR threshold abnormalities were mild during early PD stages and tended to increase with motor severity . However, no significant correlation was found between the N2/P2 amplitude lowering and motor symptom severity in a study assessing a small group (n = 12) of patients with PD .
No published information exists on the possible relationship between genetic factors associated with an increased risk of pain in patients with PD  and pain-processing abnormalities.
Our review identified several studies describing abnormal nociceptive input processing in PD. The main findings reported were a decreased pain threshold/tolerance to various stimuli, a reduced NWR threshold to electrical stimuli, changes in N2/P2 LEP amplitude and abnormal pain-induced activation in cortical areas involved in pain processing [14, 28, 30-32, 34, 35]. Overall, these abnormalities point to increased activity in both the ascending lateral and medial pain pathways .
Abnormalities in pain processing may arise from decreased basal ganglia dopamine levels. Studies in animals and humans document a role for basal ganglia dopaminergic neurotransmission in modulating pain perception and natural analgesia within supraspinal regions involved in the pain pathways, including insula, anterior cingulate cortex, thalamus and periaqueductal grey . Some evidence nevertheless suggests that neurotransmitters other than dopamine may contribute to the abnormal pain processing . Hence, changes in pain-processing mechanisms probably also reflect PD-induced neurodegeneration in non-dopaminergic structures mediating pain processing in the spinal cord, brainstem, diencephalon and limbic system .
The changes in central pain processing demonstrated by abnormalities in pain threshold or in N2/P2 amplitude do not correlate with the pain intensity in the various types of pain, including muscular, arthralgic, peripheral neuropathic and central neuropathic pain [14, 32, 34], and intervene also in patients with PD without pain [14, 27-29, 32, 35, 36]. In addition, evidence is scarce about consistent differences in pain processing between patients with and without pain. These findings imply that abnormal pain processing in PD predisposes patients to spontaneous pain of variable quality, but does not necessarily mean that they will experience it.
If abnormal nociceptive input processing acts as a predisposing background, then additional factors are probably necessary for pain to manifest. Female gender, motor complications and medical conditions associated with painful symptoms displayed no significant influence on sensory thresholds but nevertheless correlated with spontaneous pain [10-12, 14-16, 18, 19]. These factors, along with severity of motor symptoms that correlated with abnormal pain-processing mechanisms [14, 19] as well as with spontaneous pain [11, 14, 16], may therefore play a role in triggering pain in predisposed patients with PD.
Pain in PD may have variable quality, and several evidences raise the possibility that the various pain types share abnormal nociceptive processing mechanisms. Support for common background mechanisms comes from studies showing abnormally low pain thresholds and pain tolerance in patients with different pain types [14, 29, 32, 33]. Further support suggesting that pain, though differing in quality, shares the same predisposing mechanisms comes from the observation that one of the genetic variants increasing the risk of pain in PD is associated with different pain subtypes .
We believe that pain in PD cannot only be attributed to an abnormal nociceptive mechanism processing, as demonstrated by neurophysiological and other types of experimental studies. Our hypothesis is that the shared abnormalities in nociceptive input processing present in patients with various pain types and also in patients without pain would be responsible for a common abnormal physiological background indicating the substrate on which other mechanisms may operate to elicit spontaneous pain. Additional loco-regional factors, such as dystonia, marked rigidity or bradykinesia, and medical conditions, such as osteoporosis, rheumatic disease, degenerative joint disease, arthritis and disc herniation, might interact with the underlying abnormal pain processing thus producing dystonic or muscular or arthralgic pain in specific body parts (Fig. 1). The variability in factors interacting with abnormal pain processing to produce pain might also explain why pain in PD manifests clinically in several ways. Given that the factors possibly triggering central neuropathic pain remain unknown, however, this model may only partly apply to this type of pain.
Following this hypothesis, we propose that pain treatment should target the mechanisms underlying the abnormal nociceptive input processing in the CNS; and the loco-regional factors (rigidity or bradykinesia and medical conditions, such as osteoporosis, rheumatic disease, degenerative joint disease, arthritis and disc herniation) potentially associated with certain painful symptoms. Drugs acting on non-dopaminergic neurotransmitter systems known to contribute to pain-processing mechanisms (e.g. serotonin. norepinephrine and GABA) might be useful in managing pain in PD and need to be tested in controlled studies. An exploratory uncontrolled study showed that duloxetine, a dual reuptake serotonin inhibitor, provides some central neuropathic pain relief in patients with PD . In addition to treatments acting on pain-processing mechanisms, clinicians should also be ready to adopt further measures targeting locoregional factors contributing to pain appearance in predisposed patients with PD. Exercise and physical therapy may be beneficial for maintaining range of motion, mobility and preventing contractures; botulinum toxin injections can effectively treat dystonic pain; and conventional non-steroidal anti-inflammatory drugs may provide additional symptom relief in co-morbid rheumatological and orthopaedic conditions.
Disclosure of conflicts of interest
The authors declare no financial or other conflicts of interest.