Graham and Wolff noted that compression of the ipsilateral common carotid artery or the superficial temporal artery provides relief during many migraine attacks, and that ergotamine tartrate also causes arterial vasoconstriction, which led them to the vascular theory of migraine pain (46). Vascular models include studies of the anatomy and neural innervation of the cranial vasculature, studies of cranial vessels (both in vitro and in vivo), and studies of arteriovenous anastomoses (AVAs) (22). The underlying purpose has been one of examining the receptor pharmacology and physiology of the cranial circulation to characterize novel vasoconstrictor mechanisms that would reverse the assumed vasodilatation associated with acute migraine attacks. The study of cranial circulation has been rewarding; indeed, vascular studies, such as those on the AVAs, have been absolutely pivotal to therapeutic advances in the field (90).

Although the theory that the pain of migraine is caused by arterial distension has driven much of the clinical work in the field (99), the relationship of vascular change to headache is now hotly debated and much less certain than Wolff had considered. For at least one third of patients, there is no vascular quality to the pain (20), which may reflect the fact that both the vessels and dura mater are pain producing (87), and that pain originating within dura mater origin may not have a throbbing or pounding quality (26). In some patients, the blood flow velocity of the large cerebral vessels is reduced, believed to be caused by dilatation of large vessels, but regional cerebral blood
flow is unchanged (27). Indeed, administration of sumatriptan is associated with an increase in blood flow velocity (from vasoconstriction) in the middle cerebral vessels (10), although these changes in blood flow velocity may not be temporally related to resolution of the headache (68).

A major shortcoming of the vascular model is that drugs designed to have vascular effects necessarily have safety concerns (19). The question of cardiovascular safety continues to plague the triptans as a class (70), and subclassspecific triptans have not shown efficacy against migraine at doses that do not imply a vascular mode of action (43,45).

Pain signals pass from the periphery to the CNS via pain-responsive afferent neurons. The so-called nociceptor has been the subject of intense study in recent years. Models of peripheral nerve activity have included surrogates, such as plasma protein extravasation (PPE) or neuropeptide release, and direct measurements, such as those of nerve fiber activity or trigeminal ganglion neuronal activity. Our view is that migraine should be considered a neurovascular headache to place the role of both nerves and blood vessels in proper perspective.

Markowitz et al. noted that neurogenic inflammation, the activation of nociceptors in the periphery that triggers the vascular changes leading to PPE, could be observed in the dura and conjunctiva but not in the brain after electrical or chemical stimulation of trigeminal afferents (71). The model of sterile neurogenic inflammation in the pathophysiology of migraine (77) has dominated the development of migraine drugs since then, as PPE can be blocked by ergot alkaloids, indomethacin, acetylsalicylic acid, the serotonin (5-HT)-1 agonists, sumatriptan, almotriptan, alniditan, eletriptan, naratriptan, rizatriptan, zolmitriptan, γ-aminobutyric acid agonists, benzodiazepines, neurosteroids, substance P antagonists, and highly potent analogs of sumatriptan, CP122,288 (16), and zolmitriptan 4991w93 (55).

Although the clinical activity of many antimigraine drugs correlates with the ability to block PPE in model systems, the converse has not proven true; potent inhibitors of PPE developed under this model do not all treat the pain of migraine. Such is the case with bosentan, an endothelin antagonist (72); four trials of nonpeptide neurokinin-1 (substance P) antagonists (12,18,44,80), and the conformationally restricted analogs of sumatriptan, CP122,288 (88), and zolmitriptan, 4991w93 (21), all of which are able to block PPE but do not show efficacy in treating acute migraine attacks. Indeed, the pathophysiologic correlates of PPE in clinical migraine are elusive. Gadolinium-enhanced magnetic resonance imaging of the brain and dura during migraine do not suggest the changes in vascular permeability predicted by PPE (79), nor can a correlate of PPE be seen in the retina of patients during acute migraine or cluster headache, despite the fact that PPE can be demonstrated in the rat retina (73).

Studies of cranial venous neuropeptide levels have proved useful markers of trigeminovascular activation in primary neurovascular headaches (24). In both humans and animal models, stimulation of the trigeminal ganglion leads to an increase in cranial levels of the vasoactive neuropeptides substance P and calcitonin gene-related peptide (CGRP) (35). With a particular interest in pain-producing intracranial structures, stimulation of the superior sagittal sinus triggers the release of CGRP but not substance P (101), and trigeminal ganglion stimulation in the rat triggered increased CGRP release into the blood within the superior sagittal sinus (9). Providing a coherent link from animal models back to humans, CGRP, but not substance P, is elevated in the external jugular vein blood during a migraine attack in adults (36) and adolescents (29). Treatment with sumatriptan reduces CGRP levels in humans as migraine subsides and in experimental animals during trigeminal ganglion stimulation (32).

Elevated venous CGRP release has also proven an interesting marker for the pain of other primary headache disorders, such as in spontaneous (33) and provoked (25) acute attacks of cluster headache, and similarly during the pain of chronic paroxysmal hemicrania (34). In terms of understanding the pathophysiology of these primary headache disorders, it is also noteworthy that vasoactive intestinal polypeptide, a marker for cranial parasympathetic nerve
activation, is elevated in both cluster headache and paroxysmal hemicrania (33,34).

Venous neuropeptide levels are easily measured markers for migraine, and may be reliable predictors of antimigraine activity. For example, avitriptan, has much less of an effect on PPE than does sumatriptan, yet this potent 5HT_1B/1D agonist (89) blocks CGRP release in experimental animals and is clinically effective in treating migraine (13). In converse, CP122,288 is a potent blocker of neurogenic PPE (see above), but it is not effective in blocking CRGP release at doses sufficient to effectively inhibit PPE (62). Similarly, the zolmitriptan analog 4991w93 is ineffective at low doses in blocking CGRP release in animals but is effective at doses in which it has 5-HT_1B/1D receptor activity (63).

The success of the nonpeptide CGRP receptor antagonist BIBN4096BS in treating acute migraine attacks (81), although itself not having significant vasoconstrictive properties in healthy volunteers (84) or in animal models of cerebrovascular activity (85), has revitalized the hope of finding antimigraine medications without direct cardiovascular or cerebrovascular side effects. Although the discovery of an antimigraine drug that prevents the potent vasodilator activity of CGRP (23,76) does not immediately suggest a clue to the physiologic origin of migraine, it has greatly enhanced the interest in the complex biology of the CGRP receptor (47) and its possible role in the CNS pathophysiology of migraine pain (93).

Direct recordings from trigeminal afferents innervating the dura have addressed some of the most basic issues in persistent and frequent headache syndromes. In humans, the intracranial large vessels, and great sinuses, as well as the dura mater are pain producing (26,74,82,83). Stimulation of the superior sagittal sinus in humans is painful (26,75). Recordings from fibers of the nasociliary branch of the ophthalmic nerve in the guinea pig have been used to determine that primary afferents innervating the dura are a relatively homogeneous group of unmyelinated nociceptors that respond to lowered pH, application of an inflammatory soup, capsaicin, and a range of higher and lower temperatures (4). Furthermore, a mixture of inflammatory mediators applied to the rat dura lowers the threshold for mechanical stimulation of the dura mater, in a process generally referred to as peripheral sensitization (97).

In migraine, a process like peripheral or central sensitization (see next two sections) permits the transduction of the normally innocuous physiologic pulsatile distention of vascular structures to be experienced as the throbbing pain of migraine, even in the absence of abnormally distended blood vessels. In an in vitro preparation of trigeminal ganglia and trigeminal nucleus, sumatriptan reduces the spontaneous firing rate of trigeminal neurons without changing the amplitude of the excitatory postsynaptic potential, documenting a presynaptic mechanism of action (56). Although the presence of both 5HT_1B and 5HT_1D receptors on peripheral afferents is well recognized (3,100), the finding that 5HT_1D receptors localize specifically within peptidergic nociceptors (69,86) supports a role for 5HT_1D receptor activity by triptans in the inhibition of peripheral nociceptors (56,66).