The distinct changes we found in cortical responses to low and high numbers of sensory stimuli in patients with MOH suggest that the underlying brain mechanisms are altered and differ from those acting in patients with episodic migraine without aura. Low numbers of median nerve electrical stimuli (block 1) disclosed sensory cortex sensitization in patients with MOH and in episodic migraineurs recorded ictally, whereas amplitude changes over sequential block averagings were consistent with habituation in healthy volunteers and episodic migraineurs recorded ictally, but not in MOH patients and episodic migraineurs recorded interictally. In MOH patients, SEP amplitude was lowest in those with the longest history of migraine, whereas it was highest in those with the longest period of headache chronification, suggesting that the electrophysiologic changes reflect chronification. Patients who overused triptans had lower SEP amplitudes than those who overused NSAIDs or both anti-migraine medications combined, indicating that sensitization varies according to the drug overused.
The combination of an initial SEP amplitude increase (sensitization) along with the subsequent lack of habituation suggests that the electrophysiological pattern underlying MOH differs from that underlying episodic migraine. In episodic migraine, SEP recordings show two characteristic changes: a lack of habituation on interictal recordings, and sensitization during the attack. The habituation deficit normalizes during attacks, whereas sensitization disappears between attacks, but in the immediate pre-ictal phase both sensitization and absent habituation may co-exist [9–11]. The electrophysiological pattern we found in MOH may therefore suggest that the sensory cortex is locked in a pre-ictal state associating both hyper-sensitivity (due to sensitization) and hyper-responsiveness (due to deficient habituation), which contrasts with episodic migraine where these cortical states alternate. It is likely that the disclosure of this peculiar electrophysiological pattern was made possible by the fact that we avoided to record MOH patients during a full-blown migraine attack. The SEP pattern associating sensitization and lack of habituation that we compared with a "persistent pre-ictal state", closely resembles the response patterns generated by central sensitized neuronal circuits. Sensitization refers to a facilitatory process that competes with its opposite, habituation to determine the final behavioural outcome after stimulus repetition. This has been called the "dual process" theory [15, 16]. Illustrative of central sensitization are the plastic changes in neural structures belonging to the "pain matrix"  that result in decreased nociceptive thresholds and increased responsiveness to noxious and innocuous peripheral stimuli . Studies in animals  and humans  show that SEP amplitudes increase when transient intense activation of nociceptive afferents induces central sensitization, as happens in clinical pain conditions including chronic headache. Our study shows that sensitization, as reflected by increased initial SEP amplitudes, is common to MOH and migraine attacks, although we did not record MOH patients during an attack. A clinical consequence of central sensitization is cutaneous allodynia. It was shown to be prevalent during episodic migraine attacks at cephalic and extracephalic sites [23, 24], but even more so in chronic migraine . It is associated with increased nociceptive reflexes [26, 27], but, interestingly, in MOH trigeminal evoked potentials were increased, whereas nociceptive blink reflexes remained unchanged, suggesting as in our study that sensitization takes place at supraspinal levels .
Our finding that the SEP amplitude increase in MOH is proportional to the duration of headache chronification suggests that medication overuse and increased headache frequency promote or reinforce central sensitization, but leaves open the question of the culprit. Conversely, since total duration of the migraine disorder correlates inversely with SEP amplitudes, the SEP amplitude increase is likely related to factors other than migraine duration and simply repetition of attacks. In keeping with this interpretation, patients who overused triptans alone had no initial SEP amplitude increase indicating that the major culprit for central sensitization in MOH could be NSAIDs. The neurobiological underpinning for this difference remains to be determined. An observation that might favour of NSAIDs consumption as a factor promoting sensitization is that NSAIDs increase spinal expression of inducible cyclo-oxygenase-2 , an enzyme that contributes to sensitization in a rat model of inflammatory pain .
Another possible link between central sensitization, migraine and anti-migraine drugs is monoaminergic transmission in the central nervous system (CNS). Although both triptan and NSAID overuse lead to headache chronification, only the latter is accompanied by SEP sensitization. We hypothesize that this difference is due to a more profound decrease of 5-HT transmission after NSAID overuse. Between attacks, migraine patients have low blood 5-HT levels whereas the reverse is true ictally . Serotonin synthesis in the brain increases during attacks, and this increase is partly counteracted by acute triptan treatment . Chronic administration of triptans in rats, however, increases 5-HT synthesis in several cortical projection areas of the dorsal raphe nucleus  possibly reflecting down-regulation or desensitization of 5-HT1 receptors. By contrast, in rats chronically treated with analgesics, 5-HT2A receptors are down-regulated  and the 5-HT transporter is up-regulated in the cortex  and in platelets . Upregulated platelet 5-HT transporters  and decreased whole blood 5-HT levels  tend to normalize after drug withdrawal. Collectively, these experimental data suggest that anti-migraine drug overuse can disrupt central 5-HT transmission. In chronic triptan overuse both pre- and postsynaptic 5-HT1 receptors may become desensitised with the ensuing net effect that serotonergic transmission may be only mildly impaired. During analgesic and NSAID overuse, however, the combination of receptor desensitisation and transporter upregulation may lead to serotonergic hypoactivity. Together with noradrenaline and dopamine, serotonin is crucial for tuning cortical excitability including sensitization and habituation processes and its effect in animals varies with concentration and duration of application . A more severe hypofunction of 5-HT transmission after NSAID overuse may thus explain the SEP sensitisation observed in this subgroup of MOH patients. Whether the difference between the drug classes with regard to central sensitisation is related to the clinical observation that withdrawal headache is much shorter after triptan than after analgesic overuse  remains to be determined in a properly designed prospective study comparing clinical outcome and electrophysiological patterns.
The association of electrophysiological sensitisation, i.e. increased 1st block SEP amplitude, and lack of habituation.in MOH patients overusing NSAIDs is intriguing. It is at odds with the electrophysiological pattern associating high amplitude in 1st block and normal habituation found during migraine attacks [10–14], but, as mentioned before, it has been described in the pre-ictal phase [9–11]. One possible explanation for the lack of habituation in episodic migraineurs between attacks is the "ceiling theory"  postulating that there is a low preactivation level of sensory cortices, also responsible for the low 1st block amplitudes, would allow a larger range of activation before habituation occurs [6, 8]. The habituation deficit in NSAIDs overusers cannot be explained by the "ceiling theory" since their high 1st block amplitude indicates rather that the somatosensory cortex is sensitised. There is at present no straight forward explanation for this pattern. It is likely, however, that other neurobiological mechanisms that participate in the production of habituation are impaired. For instance, inhibitory interneurons could be hypofunctioning because of the reduction in serotonergic transmission induced by the prolonged NSAID overconsumption. This hypothesis can be tested experimentally by searching if habituation normalizes during full-blown attacks in MOH patients like in episodic migraine and by exploring inhibitory cortical interneurons with dedicated neurophysiological studies such as that of cortical silent periods using transcranial magnetic stimulation. Given the similar neural mechanisms underlying sensory and behavioural sensitization , the interesting question arises whether the sensory sensitization in patients with MOH parallels behavioural sensitization. Behavioural sensitization is paradigmatic of how the serotonergic, dopaminergic, and noradrenergic systems interact and contribute to central sensitization . Brain circuits involved in addictive behaviour include ventral and dorsal striatum, amygdala and orbitofrontal cortex and are heavily modulated by dopaminergic projections from the ventral tegmental area of the midbrain, serotonergic projections from the median and dorsal raphe nuclei, and noradrenergic projections from the locus coeruleus [4, 42]. According to DSM-IV criteria, many MOH patients manifest a dependence behaviour . The latter has been associated with orbito-frontal cortex hypoactivity , an abnormality also found in subgroups of MOH patients . The orbito-frontal cortex is thought to modulate habituation mechanisms  and orbito-frontal lesions induce SEP sensitization and lack of habituation , precisely the two sensory abnormalities we found in patients with MOH. Our findings along with current knowledge on the neurobiology of drug overuse therefore suggest that future studies seeking correlations between electrophysiological and metabolic measures should focus on the orbito-frontal cortex. In our study we did not control for associated depression and anxiety. Despite the evidence that cortical pain-related evoked potentials in MOH do not differ between subgroups of patients with or without depressive symptoms , it may still be appropriate to control for psychiatric comorbidity in future studies.