A Hypothesis for the Cause of Complex Regional Pain Syndrome - Type I (Reflex Sympathetic Dystrophy): Pain Due to Deep-Tissue Microvascular Pathology

Abstract

Complex regional pain syndrome - type I (CRPS-I; Reflex Sympathetic Dystrophy) is a chronic pain condition that usually follows a deep-tissue injury such as fracture or sprain. The cause of the pain is unknown. We have developed an animal model (chronic post-ischemia pain; CPIP) that creates CRPS-I –like symptomology. The model is produced by occluding the blood flow to one hind paw for 3 hr under general anesthesia. Following reperfusion, the treated hind paw exhibits an initial phase of hyperemia and edema. This is followed by mechano-hyperalgesia, mechano-allodynia, and cold-allodynia that last for at least one month. Light- and electron microscopic analyses of the nerves at the site of the tourniquet show that the majority of these animals have no sign of injury to myelinated or unmyelinated axons. However, electron microscopy shows that the ischemia-reperfusion (I-R) injury produces a microvascular injury, slow-flow/no-reflow, in the capillaries of the hind paw muscle and digital nerves. We propose that the slow-flow/no-reflow phenomenon initiates and maintains deep tissue ischemia and inflammation, leading to the activation of muscle nociceptors, and the ectopic activation of sensory afferent axons due to endoneurial ischemia and inflammation. These data, and a large body of clinical evidence, suggest that in at least a subset of CRPS-I patients, the fundamental cause of the abnormal pain sensations is ischemia and inflammation due to microvascular pathology in deep tissues, leading to a combination of inflammatory and neuropathic pain processes. Moreover, we suggest a unifying idea that relates the pathogenesis of CRPS-I to that of CRPS-II. Lastly, our hypothesis suggests that the role of the sympathetic nervous system in CRPS-I is a factor that is not fundamentally causative, but may have an important contributory role in early stage disease.

Symptomology of CRPS-I

CRPS-I is a chronic pain syndrome that occurs following injuries such as sprains, fractures and crush injuries that are not accompanied by a clinically-verified nerve injury [1]. CRPS-I is a relatively rare disorder, with an estimated incidence of 26.2 per 100,000 [2]. Nevertheless, this condition has fascinated, perplexed, and frustrated clinicians for well over a century. Symptoms of CRPS-I include spontaneous pain (“burning” pain referred to the skin, and “aching” pain referred to deep tissues), and a variety of stimulus-evoked abnormal pain sensations, including mechano-hyperalgesia, mechano-allodynia, cold-allodynia and sometimes heat-hyperalgesia. Other symptoms include disorders of vasomotor and sudomotor regulation; trophic changes in skin, hair, nails, and bone; and dystonia and other motor abnormalities [3–9]. CRPS-II (causalgia) is similar in all respects except that a clinically-verified nerve injury is present [3,4].

Until recently, it has been generally accepted that most, if not all, the ancillary symptoms of CRPS-I are due to a hyperactive sympathetic outflow to the painful region, and that the sympathetic outflow is somehow causally related to the pain. This belief gained credence from the therapeutic benefit that is achieved in at least some patients following sympathectomy or local anaesthetic block of the sympathetic ganglia [10]. However, it is well known that not all CRPS-I patients benefit from sympathetic interventions, and those that do sometimes receive only partial or temporary pain relief. Not surprisingly, the role of the sympathetic innervation in the CRPS-I syndrome is not well understood [11].

It is generally believed that there is a temporal progression of symptoms in CRPS-I, resulting in three broadly defined stages of disease, with the first typified by hyperemia and edema, the second by cold, hyperhydrosis and cyanosis, and the third by dystonia and dystrophic changes [5,9]. There is evidence that this temporal progression is not present in all patients [8]. Clinical experience and animal experiments suggest that the sympathetic innervation may contribute in the early stage of CRPS-I, but that the underlying pathology evolves from sympathetically-maintained pain (SMP) to sympathetically-independent pain (SIP) [5,10,12,13].

I. A new animal model of CRPS-I

We have developed an animal model (chronic post-ischemia pain; CPIP) that creates CRPS-I-like symptomology [14]. The CPIP model is produced in rats under general anesthesia with a tourniquet placed around the ankle for 3 hr. Following reperfusion, the hind paw exhibits an initial phase of hyperemia and edema lasting for 2–12 hrs, followed by neuropathic pain (mechano-hyperalgesia, mechano-allodynia, and cold-allodynia) that lasts for at least one month. Light- and electron microscopic analyses of the hind paw nerves from CPIP rats show that nearly all have no sign of nerve injury due to the tourniquet [15]. Some of the animals have a small number of degenerating myelinated axons, which would be difficult or impossible to verify if present in a human patient. In addition, there are no changes in conduction velocity of the sural nerve at 5 and 7 days post-reperfusion [16]. However, there is a reduced density of the sensory fibers’ terminal arbors in the epidermis in the injured hind paw, as determined by PGP9.5 immunohistochemistry [16], similar to that observed in CRPS-I patients [17].

Electron microscopic analysis of the hind paw digital muscle and digital nerves reveals microvascular pathology that is indicative of the capillary slow-flow/no-reflow phenomenon [15] – a well known consequence of I-R injury in cardiac and skeletal muscle. The hind paw muscle also exhibits poor perfusion and a reduced density of viable capillaries for a week following reperfusion [16]. Single fiber recordings from primary afferent axons in CPIP rats at intervals of 2–9 days after tourniquet release have found spontaneously discharging Aβ, Aδ, and C fibers in every case examined [15]. Normal sensory fibers have little or no spontaneous discharge.

Slow flow/no-reflow in deep tissue microvasculature would be expected to produce a persistent inflammatory state. Data from the CPIP animals is consistent with this idea. Malondialdehyde, a product of free radical-induced lipid peroxidation, was significantly elevated in the CPIP rat hind paw, and CPIP allodynia was dose-dependently attenuated by the anti-oxidant N-acetyl-L-cysteine, and the free radical scavenger 4-hydroxy-2,2,6,6-tetramethyl-piperydine-1-oxyl, suggesting a key role of free radicals [16]. Furthermore, the pro-inflammatory cytokines, tumor necrosis factor alpha (TNFα), interleukin-1 (IL-1) and interleukin-6 (IL-6), and the related transcription factor, nuclear factor kappa B (NFκB), are all elevated in the CPIP rat hind paws early after reperfusion, and both an IL-1 receptor antagonist and an NFκB inhibitor, pyrrolidine dithiocarbamate, dose-dependently reduce CPIP allodynia, suggesting that an NFκB-dependent generation of pro-inflammatory cytokines plays a role in CPIP. Lactate is also increased in the muscle tissue of CPIP rat hind paws, and the levels of lactate are further elevated when the rats exercise. Exercise also decreases mechanical paw withdrawal thresholds in CPIP rats, and there is a direct correlation between mechanical allodynia and lactate levels in unexercised and exercised CPIP rats [16].

CPIP rats also develop a hypersensitivity to norepinephrine (NE), reflected by enhanced arterial responsiveness to NE, as well as enhanced painful responses to hind paw injections of NE [18]. Mice also develop CPIP after I-R injury, and CPIP mice have an up-regulation of endothelin A (ET-A) receptors in their hind paw muscles. Accordingly, CPIP mice exhibit enhanced sustained nociceptive behaviours following hind paw injections of ET-1[19].

II. A hypothesis for the pathogenesis CRPS-I

Our findings in CPIP animals lead us to propose a hypothesis that we believe to be applicable to both the CPIP animals and to at least a subset of CRPS-I patients. We propose that the fundamental cause of the pain is a persistent deep-tissue (muscle, bone, and nerve) ischemia and consequent inflammatory reaction produced by microvascular pathology subsequent to an I-R injury. We propose that the following processes (summarized in Fig. 1) are critical to the initiation, development, and maintenance of the CPIP and CRPS-I syndromes.

Fig. 1.

Schematic of the hypothesized pathophysiological mechanisms in the CPIP model and at least a subset of CRPS-I patients. In CRPS-I patients, deep tissue injury leads to edema and a compartment-like syndrome. Reperfusion (release of the tourniquet in the animal model) leads to injury of the microvascular endothelial cells induced by free radicals. These events lead to arterial vasospasms and slow flow/no-reflow in deep tissue microvasculature, which produces persistent ischemia that both spreads the I-R injury, and causes chronic inflammation. In muscle and bone, the resulting ischemia and inflammation (including generation of lactate) activates and sensitizes nociceptors. In nerve, the ischemia and inflammation cause ectopic nociceptor discharge. Injury to a neurovascular bundle may also evokes arterial vasospasm, and this may initiate the same vicious cycle in patients with CRPS-II (causalgia).

(1) CRPS-I is initiated when an inflammatory response to a deep-tissue injury produces a compartment-like syndrome that impairs blood flow to muscle, nerve and bone

At its onset, CRPS-I is often characterized by significant regional edema that is sometimes described as “exaggerated” [20–22]. Oedema develops due to the extravasation and accumulation of plasma in the interstitial space [23]. In the early stage of CRPS-I, there is plasma extravasation [20], increased density of perfused vessels, and higher capillary filtration capacity (an index of microvascular permeability) [24,25].

The pressure exerted by the interstitial accumulation of extravasated plasma within a relatively confined anatomical space occludes the capillaries of adjacent tissues and causes a compartment syndrome [26,27]. Pressures within a myofascial compartment of as little as 30–40 mm Hg will occlude capillary flow [28,29]. Compartment syndrome is a well-known complication of fractures, joint trauma and joint surgery [26,30–32]. CRPS-I also often follows fractures, joint injuries, and wrist, elbow, and knee surgery [33,34]. We think that it is extremely important to note that CRPS-I nearly always follows injury to deep tissues (fractures, sprains, surgeries, crush injuries, etc.). CRPS-I initiated by a strictly cutaneous injury (e.g., a laceration or burn) is exceedingly rare [35]. Oedema in the subcutaneous space does not lead to a compartment syndrome.

In its most severe form, musculoskeletal compartment syndrome leads to ischemia that is severe enough to cause tissue necrosis, but subtotal or episodic (exertional) ischemia lead to compartment-like syndromes that do not progress to tissue necrosis [36–38]. We propose that in the CPIP animal, a compartment-like syndrome is created by the period of extensive edema that follows release of the tourniquet. In man, both compartment syndrome and CRPS-I are known risks of excessive tourniquet exposure [39,40].

(2) The compartment syndrome leads to an I-R injury and persistent deep-tissue microvascular pathology

While the proposed compartment-like syndrome does not produce tissue necrosis, it does produce microvascular injury. Ischemic tissues accumulate oxidative enzymes, primarily xanthine oxidase and NADPH oxidase. Upon reperfusion, the accumulated oxidases reduce the returning molecular oxygen and the cells composing the microvessels are exposed to high levels of oxygen free radicals that damage both vascular endothelial and smooth muscle cells [41–48]. The onset of reperfusion is obviously well defined when a tourniquet is removed, but less clearly demarcated when the ischemia is due to a compartment-like syndrome. In the latter case, it is probable that reperfusion is episodic or partial, and may migrate from one part of the tissue’s capillary bed to another.

I-R injury is a multifactorial phenomenon characterized by several pathological mechanisms affecting arterioles, capillaries, and venules [48–55]. These mechanisms interact and generally contribute to positive-feedback loops that perpetuate and worsen the I-R injury. Although most studied as a consequence of ischemic insult to heart muscle, I-R injury phenomena are also known to occur in the microvasculature of skeletal muscle [52,56]. We propose here that I-R injury also occurs in the microvasculature of bone and peripheral nerve.

I-R injury, aterioles, vasospasm and the sympathetic nervous system

Following I-R injury, the arteriole’s endothelial cells release less nitric oxide, and the reduced amount of nitric oxide that is released is converted to toxic nitrogen free radicals after interaction with oxygen free radicals. The result is a deficit in the nitric oxide-mediated vasodilatation that normally modulates the vasoconstriction that is evoked by the sympathetic nervous system [49,57,58]. Recent data show that CRPS-I patients have abnormal vasodilatation responses after sympathetically-evoked vasoconstriction [59], and decreased levels of nitric oxide have been found in blister fluid from the affected region [60]. Endothelin-1 (ET-1) is a potent vasoconstrictor derived from vascular endothelial cells that produces it’s pressor effects by acting on ET-A receptors on vascular smooth muscle cells [61]. Following I-R injury, ET-1 production and release are increased, and so is the vasoconstrictor response that it evokes [62]. Increased levels of ET-1 are found in blister fluid from the affected extremity in early stage CRPS-I patients [60]. As we described above, CPIP mice have up-regulated ET-A receptors in hind paw muscle, and show enhanced painful responses to intraplantar ET-1 injections [19].

I-R injury also evokes an up-regulation of the expression of α-adrenoceptors on arterial smooth muscle cells, resulting in a three-fold increase in the contractile response to NE [63]. As a result, arteries will spasm in response to normal levels of sympathetic discharge (vasoconstrictor “tone”), to the myriad of sympathetic reflexes that are evoked by daily activity, and perhaps to catecholamines (NE and epinephrine) that arrive via the circulation. As discussed above, CPIP rats exhibit enhanced arterial vasoconstrictor responses to NE. Both CRPS-I patients and CPIP rats display abnormal pain responses to intraplantar injections of NE [18,64]. There is clear evidence that there is dramatically reduced sympathetic reflex activity in early CRPS [65]. There is also evidence that this reduced sympathetic outflow itself results in a hypersensitivity of the smooth muscle adrenoceptors, i.e., functional denervation super-sensitivity [66]. There is evidence for increased vascular α-adrenoceptor responsiveness in CRPS-I patients [67–69]. The situation is likely to be very complex and variable from patient-to-patient, or from time-to-time in a given patient: increased vascular α-adrenoceptor responsiveness that might be driven by circulating catecholamines and/or the NE released in the context of decreased activity at the postganglionic sympathetic fiber’s synapse. In any case, the result is that sympathetic activity would be contributory, rather than fundamentally causative, and exacerbate ischemia, inflammation and consequent pain.

Nerve injury evokes sympathetic fiber sprouting in the dorsal root ganglion. However, there is evidence that this sprouting may not contribute to CRPS-II pain (and, by implication, CRPS-I pain) [70]. There is also evidence that nerve injury evokes the de novo expression of α-adrenoceptors on primary afferent nociceptors, which suggests a direct link between NE and epinephrine (via the circulation or via the sympathetic postganglionic synapse) and nociceptor activation [71]. The presence of such a mechanism is not incompatible with the hypothesis that we propose.

Vasopasm in pre-capillary arterioles exacerbates the ischemia and leads to further I-R injury, and may contribute to the spread of microvascular dysfunction. The resulting vicious cycle contributes to the maintenance and worsening of the ischemic state. Such spreading dysfunction may account for the “contiguous” spread of CRPS-I symptoms, i.e., the tendency for edema and pain to gradually spread outwards from the initially symptomatic region [72]. There is extensive evidence for such a spreading phenomenon in the case of ischemic heart disease [73]; we propose here that the same thing occurs in skeletal muscle. A long forgotten paper by Foisie [74] suggested that arterial vasospasm might contribute to causalgic-like pain after crush injuries and soft tissue injuries that did not injury a nerve (i.e., CRPS-I). Arterial vasospasm after I-R injury is sometimes relieved by a brief tourniquet application [75]. CRPS-I patients sometimes report temporary pain relief after brief tourniquet application [76,77].

I-R injury to capillaries

Free radicals also damage capillary endothelial cells and stimulate them to release various pro-inflammatory mediators. Free radicals increase the expression of selectins, intracellular and extracellular adhesion molecules [78], complement factors [79], leukotriene B₄ [80,81], and platelet activating factor [82,83]. Many of these molecules are chemotaxic and recruit monocytes, leukocytes, and platelets which accumulate and occlude the capillary lumen, and release TNFα, IL-1 and IL-6, which produce toxic effects that might spread tissue injury to adjacent regions [84,85].

Occlusion of capillaries generates the phenomenon known as slow-flow/no-reflow, which we believe to be a key feature of both the CRPS-I and CPIP syndromes. Slow-flow/no-reflow is characterized by the swelling of microvascular endothelial cells, platelet aggregation, and plugging of the capillary lumen by leukocytes and erythrocytes [45,48,49,51,52,55,86–95]. “Slow-flow” refers to the condition where the lumen is partly occluded; complete occlusion is “no-reflow”. In cardiac muscle, the onset of slow-flow/no-reflow is detectable immediately after reperfusion, but it worsens significantly during the following hours and days [55,96]. It is important to recall how easy it is to block capillary flow. The smallest capillaries are formed from a single endothelial cell, and their lumen diameters are nearly the same as the shortest diameter of a red blood cell. The lumens’ diameter is restricted further at the level of the endothelial cell’s nucleus. Red blood cells must deform and squeeze through the capillary lumen. The pain of a sickle cell crisis is due to ischemia that occurs because of an abnormality of the erythrocye’s membrane that makes it too stiff to deform and squeeze through the capillary lumen.

As well as occurring in skeletal muscle, we have evidence from the CPIP animals that slow-flow/no-reflow also occurs in the nerve’s capillaries, and we think it likely that it also occurs in periosteal and intramedullary bone capillaries. There is no logical reason to suppose that slow-flow/no-reflow does not also occur in cutaneous capillaries. However, because of their important role in thermoregulation, the cutaneous capillary beds in the distal extremities of man are extremely dense and highly anastomotic; this renders the skin’s capillary bed relatively resistant to slow-flow/no-reflow [97].

I-R injury to venules

Free radical damage to the endothelial cells of venules resembles that seen in capillaries. However, venules have lumens that have a greater diameter than that of capillaries and venules are thus relatively resistant to slow-flow/no-reflow. However, I-R injury causes damage at post-capillary venules, which causes leakage of plasma through resultant gaps between adjacent endothelial cells [98]. It is significant that plasma extravasation occurs mostly at the level of the venules, not the capillaries. Free radical damage to venules is thus the immediate cause of edema formation. As we discuss below, it is significant to note that a leaky venule has nothing to leak if its upstream capillary supply is blocked.

Compensatory reactions to microvascular pathology

Although arterial vasospasms and slow flow/no-reflow would produce areas of tissue ischemia, the system responds with attempts to compensate via reactive hyperemia and arteriovenous shunting [52,56]. Limbs in early stages of CRPS-I often have high arterial flow, yet at the same time have elevated venous pressure and arteriovenus shunting [24,25]. These phenomena are likely to be due to slow-flow/no-reflow in the affected capillary bed. The alternation between ischemia and reperfusion provides a way to spare tissue from lethal injury, but it also creates a vicious cycle of I-R injury which maintains microvascular dysfunction.

(3) Microvascular pathology leads to persistent ischemia which in turn leads to persistent inflammation

We propose that microvascular pathology leads to an ischemic state that evokes persistent inflammatory responses.

Microvascular pathology and ischemia

Several observations are consistent with the presence of impaired blood flow in CRPS-I patients. Skin capillary hemoglobin oxygenation (HbO₂) is lowered and skin lactate is increased in CRPS-I limbs, suggesting that there is both impaired nutritive blood flow and enhanced anaerobic glycolysis [99,100]. Nail bed capillary flow is decreased [101], and reactive hyperemia in the cutaneous vasculature is impaired in CRPS-I patients [66]. Additionally, there is an impairment of high-energy phosphate metabolism in CRPS-I muscles [21,102], consistent with an impaired blood supply. Experimentally-evoked plasma extravasation is exaggerated in CRPS-I patients [103,104].

Several lines of evidence suggest that at least some CRPS-I patients have a persistent inflammatory condition. Necropsy studies of amputated limbs from patients with severe CRPS-I find lipofuscin deposits, atrophic fibers, and severely thickened capillary basal membranes in muscle and subcutaneous tissue [105,106], consistent with the presence of ischemia-evoked inflammation. Serum levels of calcitonin-gene related peptide and bradykinin are elevated in the venous drainage of CRPS-I limbs, consistent with the presence of an ongoing ischemic and inflammatory state in deep-tissues [107,108].

Eisenberg et al. [109] have detected greatly elevated levels of inflammatory by-products (e.g., malondialdeyhde) and cellular anti-oxidants in the serum and saliva of CRPS-I patients with typical disease. We suggest that these markers originate in tissues made ischemic and inflamed by slow-flow/no-reflow. There are several reports of increased levels of pro-inflammatory cytokines in CRPS-I patients [60,110–112], as would be expected if there was an ongoing inflammatory reaction. These findings are consistent with our results that malondialdeyhde and pro-inflammatory cytokines are increased in the hind paw muscle of CPIP rats [16], as well as numerous reports that these mediators are elevated after skeletal I-R injury (see above).

(4) The ischemic and inflammatory state in muscle results in nociceptor activation

In muscle, an ischemic state would be expected to be a direct cause of nociceptor discharge and nociceptor sensitization due to acidosis [113]. It has long been known that tissue acidosis associated with increased lactate causes muscle pain in ischemic tissue [114]. It is significant that mechanical allodynia in CPIP rats is directly correlated with lactate levels in CPIP hind paw muscle [16]. These findings are consistent with that the reports of CRPS-I patients that the pain produced by infusing a low pH solution into muscle in the affected limb is far more painful than the same infusion into a contralateral muscle [99]. Moreover, they report that the pain produced by an infusion into the normal, contralateral limb is qualitatively similar to their ipsilateral CRPS-I pain, i.e., their CRPS-I pain feels like inflammatory muscle pain [99]. Differential activation of sympathetic vasoconstrictor reflex responses to muscle and skin shows that the response in muscle, rather than skin, causes the greatest pain [115], consistent with ischemic muscle being a primary cause of CRPS-I pain.

Muscle lactate has also long been known to increase following exercise [116], so it is not surprising that lactate increases in exercised CPIP rats. Importantly, mechanical allodynia is also increased after exercise in CPIP rats, and mechanical allodynia is directly correlated with muscle lactate levels in CPIP rats [16]. This could explain the often reported phenomenon that the pain of CRPS-I is increased during exercise [20,117]. In sum, these mechanisms would be expected to cause a persistent deep, aching pain sensation. CRPS-I patients frequently (perhaps always) report the presence of a deep, aching pain, and they also commonly report that movement worsens their pain, as would be expected if their muscles were ischemic and inflamed.

(5) The effects of an ischemic and inflammatory state in bone

Prolonged ischemia of bone has been shown to produce bone inflammation and injury [118,119]. Three-phase bone scans reveal that CRPS-I patients often have a late-phase periarticular accumulation of tracer [33], consistent with an I-R injury-induced injury to venules leading to intramedullary plasma extravasation. The abnormality may not be confined to periarticular bone, as is generally believed, but may affect the bone shafts as well [120]. Using scintigraphical studies of CRPS-I patients over time reveals an early hyper-perfusion of bone that is followed by later hypo-perfusion [121].

Some CRPS-I patients also have radiographic evidence of osteoporosis, although this may be transient and migratory [33,120,122–126]. Osteoporosis is an expected consequence of impaired medullary perfusion. Clinical reports of pain relief in at least some CRPS-I patients treated with bisphosphonates or calcitonin are consistent with the idea of pain arising from ischemic and inflamed bone [127–130]. The bone pathology of CRPS-I was originally attributed to an inflammatory process [122]. All of the observations noted above are likely results of edema, ischemia, and inflammation secondary to slow-flow/no-reflow in periosteal and intramedullary microvessels. Ischemia in bone would cause deep, aching pain. It is our experience that CRPS-I patients often insist that their bones hurt.

(6) The effects of an ischemic and inflammatory state in peripheral nerves

There is no reason to doubt that endoneurial capillaries are also prone to the slow-flow/no-reflow phenomenon. Small, distal nerves would be especially vulnerable, simply because they have few capillaries. We have shown evidence of slow-flow/no-reflow in endoneurial microvessels in hind paw digital nerves in CPIP rats [15]. An ischemic and inflammatory state within a peripheral nerve should produce spontaneous ectopic discharge in sensory fibers. We have shown that ectopic discharge is present in A-fibers and C-fibers in CPIP rats [15].

Oaklander et al. [17] have shown degeneration of intraepidermal nerve fibers (IENF) in CRPS-I patients, and we have shown similar reduction of IENF density in the skin of CPIP rats [16]. No-flow in the capillaries of small, distal nerves might account for such degeneration.

Endoneurial ischemia and inflammation would be associated with increased levels of pro-inflammatory cytokines that are known to produce neuropathic (“neuritic”) pain [131,132]. Ectopic discharge in nociceptors due to endoneurial microvascular pathology would cause spontaneous pain referred to the tissue innervated by the nociceptors within the nerve. Such pain is properly categorized as neuropathic, while the pain from ischemic and inflamed muscle and bone is not. We thus propose that both inflammatory and neuropathic pain mechanisms are present in CRPS-I.

(7) Nociceptor activation leads to central sensitization

Deep-tissue C-fiber nociceptor discharge and the ectopic discharge of C-fibers traveling within ischemic and inflamed nerves will initiate and maintain N-methyl-D-aspartate receptor-mediated central sensitization [133]. Importantly, input from muscle C-fiber nociceptors is a much more potent initiator of central sensitization than input from cutaneous C-fiber nociceptors [134]. Also, while injecting inflammatory agents into skin produces allodynia that lasts at most hours, the same injections into muscle induce central sensitization and cutaneous allodynia lasting up to several weeks [135,136].

We hypothesize that the central sensitization evoked by deep-tissue nociceptor discharge is the primary cause of cutaneous allodynia and hyperalgesia. The presence or absence, severity, and spatial extent of cutaneous hypersensitivity might thus be expected to vary from patient-to-patient and from time-to-time, depending on fluctuations in the intensity of deep-tissue pain. Fluctuating cutaneous hypersensitivity is a common observation in CRPS-I patients [137,138], and cutaneous allodynia can be suppressed with the persistence of ongoing deep-tissue pain [12,139]. Cutaneous allodynia and hyperalgesia might thus be categorized as epiphenomena.

III. The role of the sympathetic nervous system and the stages of CRPS-I may depend on a progression from slow-flow to no-reflow

The duration of the initial, edematous stage of CRPS-I is highly variable. Some patients remain in this state, have episodes of exacerbated edema, or progress to a non-edematous stage characterized by cyanosis and/or atrophy of subcutaneous tissue. Patients with slow-flow would be expected to exhibit edema, but there can be no plasma extravasation from leaky venules if there is no-reflow in their upstream capillaries. Thus, at later times, when no-reflow predominates, the patient would be in the cyanotic/atrophic phase. We propose that the progression from early-stage to late-stage CRPS-I reflects the evolving dominance of no-reflow over slow-flow in deep tissue capillaries. The variability in the progression from predominately slow-flow to predominately no-reflow is likely to be the basis of the argument over whether CRPS-I is a staged condition [140].

Ischemia in the edematous stage of CRPS-I is likely to be due primarily to slow-flow and arteriole vasospasm. There is evidence that both phenomena may persist long after an I-R injury [55]. The exacerbation of slow-flow by arteriole vasospasm is likely to significantly worsen ischemia in the edematous stage of CRPS-I. Sympathetic activity (whether normal, increased, or decreased) would significantly reinforce arterial vasospasm and worsen slow-flow, thereby increasing ischemia and pain. Interruption of the sympathetic outflow to the affected region would relieve the arterial vasospasm, increase capillary flow and decrease pain; such patients would be classified as having sympathetically-maintained pain (SMP) [141]. In patients with relatively mild I-R injury, the absence of sympathetic exacerbation might allow microvascular recovery. In those patients with more severe I-R injury, the pain relief would be expected to be partial and/or temporary. Prolonged ischemia would likely lead to the gradual accumulation of capillaries with actual no-reflow. Sympathetically-mediated arteriole vasospasm would make little or no contribution to downstream blood flow if the capillaries are already occluded. The patient would thus be unlikely to benefit from interruption of sympathetic outflow to the affected region; i.e, the patient would have sympathetically-independent pain (SIP). Thus, we propose that the transition from microvascular slow-flow to no-reflow is also the basis for the transition from SMP to SIP.

IV. In causalgia (CRPS-II), injury to the neurovascular bundle induces an arterial vasospasm that produces I-R injury similar to that of CRPS-I

Vasospasm is a likely consequence of a traumatic injury to a major nerve because major nerves travel alongside a large artery and vein (i.e., the neurovascular bundle). Interestingly, Mitchell’s publications describing causalgia [142,143] also include the first detailed descriptions of severe long-lasting nerve injury-evoked arterial vasospasm. However, Mitchell made no mention of a possible link between vasospasm and causalgia. In CRPS-II, we suggest that injury to the neurovascular bundle evokes arterial vasospasm and initiates tissue ischemia and subsequent slow-flow/no-reflow in deep-tissue microvessels. The end result is the same as the I-R injury evoked by a compartment-like syndrome in CRPS-I.

Injuries to the brachial plexus at the level of the thoracic outlet sometimes lead to CRPS-like symptoms in the hand and arm. In such cases it is possible that irritation to the subclavian artery or to the sympathetic postganglionic fibers in adjacent nerves evoke vasospasm and the I-R phenomena discussed above. Relief of arterial and neural irritation via decompression of the scalene triangle is beneficial in these patients [144,145].

V. Clinical implications and novel therapeutic avenues

Our hypothesis suggests new treatment options for the pain of CRPS-I, and suggests possible mechanisms of action for therapies that are in use.

α-Adrenoceptor antagonists like phenoxybenzamine [69,146] may be more effective than sympathectomy or ganglionic blocks, especially in early stage CRPS-I, since they would prevent the enhanced vasoconstriction evoked by NE from the sympathetic postganglionic neuron, or NE and epinephrine from the circulation, from acting on up-regulated smooth muscle cell α-adenoreceptors. Arterial vasospasms, reduced vasodilatation, and ischemia associated with slow-flow/no-reflow may also be alleviated by treatment with nitric oxide donors or other vasodilators. The nitric oxide donors, nitroglycerine [147] and isosorbide dinitrate [148], have been used to alleviate CRPS-I.

Relief of deep-tissue ischemia may underlie the beneficial effects of hyperbaric oxygen therapy [149], spinal cord stimulation [150], and physical therapy [151,152] in CRPS-I patients. The key role of free radicals and cytokines after I-R injury suggests that antioxidants, free radical scavengers and anti-cytokine drugs maybe useful as therapeutic agents in CRPS-I. There are reports of successful treatment of CRPS-I patients with free radical scavengers and anti-oxidants [95,153–156] and anti-TNFα therapies [110,157,158]. Proposed damage to endothelial cells and associated vascular leakage through post-capillary venules, and the role of leukocytes and neutrophils in plugging capillaries leading to no-reflow, suggests that pain associated with CRPS-I may be alleviated both by anti-inflammatory drugs, and by novel agents that inhibit endothelin, leukocyte, and neutrophil attraction and migration, and platelet aggregation.

VI. Limitations of the hypothesis

We propose a hypothesis that we believe is relevant to at least a subset CRPS-I patients, and perhaps also to CRPS-II patients. We do not claim that the pain mechanisms that we describe here are present in all CRPS-I patients. CRPS-I may be a term that encompasses pathophysiologically distinct conditions [1,140]. We are not the first to emphasize that CRPS-I might be an abnormal inflammatory condition [21,22,122,159], but we present a mechanism that would initiate and maintain such a state – deep-tissue ischemia and inflammation due to persistent microvascular pathology subsequent to an IR injury. Moriwaki et al. [160] presented the idea that early-stage CRPS-II pain might have both neuropathic and inflammatory pain mechanisms. We agree and think that this is also true for CRPS-I, and we suggest that the neuropathic component is due to ectopic nociceptor discharge subsequent to endoneurial microvascular pathology.

We have not extended our hypothesis to account for the motor abnormalities that occur in CRPS-I [7]. We have also not accounted for the deficits in somatosympathetic reflexes that characterize CRPS-I [65,161]. Although, it may be that decreased sympathetic activity is a homeostatic reflex that increases perfusion to ischemic tissue [162]. We have also not accounted for the occurrence of CRPS-I following stroke [163]. Our hypothesis does not address the question of why CRPS-I symptoms sometimes spread to the contralateral extremity [71,72,164,165]. However, we do see contralateral allodynia in CPIP rats [14], and contralateral symptoms are common in response to muscle afferent activation [134] or muscle inflammation [135,136].

The incidence of CRPS-I is 3–5 times higher in women than in men. Our hypothesis suggests that this may be related to the increased prevalence of other vascular conditions in women (e.g., varicosis and Raynaud’s disease). Muscle mass in women is typically less than in men, and lower muscle mass is likely to be associated with a less extensive and less anastomotic capillary bed that is more susceptible to the ischemia produced by slow-flow/no-reflow. There is evidence that smoking is a significant risk factor in CRPS-I [166], and this may be related to nicotine-evoked damage to microvessels. There is accumulating evidence that suggests that CRPS-I is accompanied by changes in CNS regions that regulate somatosensation, motor control, and the autonomic nervous system [71,167–169]. A role for CNS mechanisms as consequences of the peripheral pathology described here is not a contradiction.

VII. Conclusions

For many years now, the cutaneous manifestations of CRPS-I have been the primary focus of investigation. We propose that as an underlying mechanism cutaneous pathology may be of decidedly secondary importance. We propose instead that deep tissue pathology (endothelial dysfunction and persistent ischemia and inflammation) in muscle, bone, and peripheral nerve is the key to the syndrome’s pathophysiology, and we suggest that the deep-tissue pathology is driving the cutaneous abnormalities. We are not the first to suggest that deep tissue pathology may be of fundamental importance, but we have proposed a distinct and testable hypothesis for its cause. Our hypothesis grew from observations of an animal model of CRPS-I. However, our ideas can be tested in CRPS-I patients, and are consistent with many reported clinical observations and investigations. The recent discovery of markers of inflammation in the serum and saliva of CRPS-I patients with typical disease [109], suggests a likely selection criterion for studies seeking to detect abnormal deep tissue perfusion, i.e., the patients with the highest suspicion of inflammation. It is relatively easy to measure the perfusion of capillary beds in skin (and the nail bed), but difficult to measure perfusion in deep tissues. However, new methods like near-infrared spectroscopy [170] may allow direct measurements muscle capillary function.