Serum CGRP, VIP, and PACAP usefulness in migraine: a case–control study in chronic migraine patients in real clinical practice

Calcitonin gene-related peptide (CGRP), vasoactive intestinal peptide (VIP) and pituitary adenylate cyclase activating polypetide-38 (PACAP-38) have relevant roles in migraine pathophysiology. Their serum levels have been proposed as biomarkers for migraine. Our aim was to assess their diagnostic value in real clinical practice in a cohort of chronic migraine (CM), episodic migraine (EM) and healthy controls (HC). We recruited subjects with CM, EM and HC at two medical centers. Blood samples were drawn under fasting conditions in the interictal period, immediately centrifuged and stored at − 80 ºC. Serum levels were determined by ELISA. Neuropeptide levels, the effect of preventatives, correlations with clinical and demographic variables, and their diagnostic value were studied among clinical categories. 296 age- and sex-matched subjects (101 CM, 98 EM and 97 HC) were included. All three neuropeptide serum levels were higher in CM [median and IQ for CGRP = 18.023 pg/ml (14.4–24.7); VIP = 121.732 pg/ml (48.72–186.72) and PACAP = 204.931 pg/ml (101.08–597.64)] vs EM [CGRP = 14.659 pg/ml (10.29–17.45); VIP = 75.603 pg/ml (28.722–107.10); and PACAP = 94.992 pg/ml (65.77–128.48)] and vs HC [CGRP = 13.988 pg/ml (10.095–17.87); VIP = 84.685 pg/ml (35.32–99.79), and PACAP = 103.142 pg/ml (59.42–123.97)]. Using multinomial modeling, only VIP (OR 1.011, 95% CI  1.003–1.018, p = 0.005) and PACAP (OR 1.003, 95% CI 1.001–1.005, p = 0.002) increased the risk for CM, but not for EM. CGRP did not predict CM or EM. This model could correctly classify only 62/101 (61.38%) of CM, 75/98 (76.53%) of EM, and 5/97 (4.12%) of HC [globally 147/296 (49.8%)]. Individually, PACAP performed the best for classifying clinical categories [global accuracy 150/296 (50.67%)]. In CM, neuropeptide levels were higher in those OnaBT-treated than in no-treated patients. Although interictal serum CGRP and VIP were higher in CM than both EM or HC, their utility to discriminate migraine categories was low. Contrary to other studies, PACAP serum levels were also higher in CM than in EM or HC and had more discriminative capability to distinguish CM from EM and HC. Further investigation is needed for determination technique standardization.


Background
Migraine is a complex neurovascular disorder with lacking fully understood pathophysiology. There is enough evidence on the activation of the trigeminal-vascular system (TVS) that leads to the release of vasoactive peptides at their nerve terminals leading to the phenomenon known as neurogenic inflammation, that has been hypothesized to be responsible for migraine throbbing pain [1,2]. Despite the high frequency of this disorder, affecting 15% of the general population and around 2% in its chronic form [3,4], diagnosis is based on clinical criteria due to the absence of validated biological markers [5]. Accuracy of migraine diagnosis is troublesome since different types of headaches and comorbidities frequently overlap, especially in chronic headaches [6], something that would be improved with biochemical, radiological and neurophysiological biomarkers.
Searching for useful migraine biomarkers, several authors have investigated plasmatic or serum levels of migraine-involved neuropeptides, such as calcitonin generelated peptide (CRGP) [2,[7][8][9][10][11][12][13][14][15][16][17][18][19], vasoactive intestinal peptide (VIP) [11,[19][20][21], and pituitary adenylate cyclase activating polypeptide-38 (PACAP-38) [21][22][23][24][25][26] (Table 1). CGRP, so far the best known migraine neurotransmitter, is a powerful vasodilator of cranial vessels, released by sensory fibers of the trigeminal nerve, that also exerts neuromodulatory effects on central pain circuits [2]. It is known to be released during migraine attacks [2] and this has been reflected in several studies as an elevation of plasma levels in the external jugular vein and/or peripheral blood in migraine patients, both during and outside attacks [9][10][11][12][13]. It was also seen that these CGRP levels decreased as pain was relieved with triptans [13]. CGRP role in migraine pathophysiology is strongly supported by the fact that its infusion in migraineurs has shown to trigger migraine attacks [27], and that new drugs that antagonize CGRP or its receptors could abort crises [28][29][30]. It might also have some role in the persistence of pain as has been suggested by the finding of significantly elevated plasma levels in chronic migraine (CM) patients outside attacks [18]. Onabotulinum toxin (OnaBT), a most effective treatment for CM, has proved to reduce plasmatic CGRP levels [16], and higher levels predicted a better response to treatment [17,19]. All these data have led to proposing its usefulness as a biomarker for CM. However, this hypothesis could not be validated by others [7], addressing the need for further studies. VIP and PACAP-38 are two parasympathetic peptides released by the efferent arm of the trigeminal-facial arch that also have vasoactive functions. VIP serum levels have been found higher during migraine attacks in patients who had parasympathetic activation [11], and also higher interictally in episodic and CM [19][20][21]. Peripheral blood VIP levels also help, though to a lesser degree than CGRP, in predicting response to OnaBT [19]. Nevertheless, VIP intravenous infusion to migraineurs, despite causing marked dilation of cranial arteries, does not induce migraine [31]. In view of these results, it seems that VIP could be a marker of parasympathetic nervous system activation, but VIP induced vasodilation is by itself not enough to cause a migraine attack. PACAP-38 has been found higher in plasma from the cubital vein [23], as well as from the external jugular vein [24], during attacks, with plasmatic decrease in responders after sumatriptan administration [24]. Interictal measurements were lower in migraineurs as compared with healthy controls (HC) [23,24], and tension type headache (TTH) [25], suggesting a difficult to explain PACAP-38 role in migraine pathogenesis. Administered intravenously, it induces migraine-like attacks in 58% of migraineurs [26]. However, unlike CGRP and VIP, interictal PACAP-38 has shown no changes in peripheral blood in CM [21], suggesting a low utility as a CM biomarker. Nevertheless, PACAP-38 plays an important role in migraine, not only in premonitory phase, but also in activation and propagation of trigemino-vascular reflex and neurogenic inflammation, which could exacerbate migraine phenotype [32]. The finding that over 80% of CM patients recognized at least one cranial autonomic parasympathetic symptom [33] could be the manifestation of a more complex phenotype of CM.
Therefore, there are some data pointing to a possible usefulness of these neuropeptides as biological markers of CM, but more validation studies are needed in order to solve some pending unknowns such as optimal detection method, sensitivity, specificity, reproducibility [34], and behavior of these neuropeptides under real conditions. We hypothesized that CGRP, VIP, or PACAP-38 (from now only PACAP) serum levels, or a combination thereof, could serve as biomarkers for migraine, especially for CM. To do this, we determined and compared CGRP, VIP and PACAP interictal serum levels in a case-control study of CM, EM, and HC, and assessed their possible diagnostic value in daily clinical practice, including patients treated or not with oral preventatives and OnaBT.

Methods
We conducted a case-control study approved by the Ethics Committee of Cantabria, Spain. We prospectively recruited migraine patients between 18 and 65 years old in two different medical centers in the Northern Spain (one tertiary center and one primary care center). In consideration of real clinical practice, we gathered patients treated with OnaBT and other preventatives, and also naïve patients to seek the

Study participants
All participants signed written informed consent before study entry. Subject selection followed a pre-specified scheme 1:1:1, age-and sex-matched for CM, EM, and HC. At the end of the inclusion period, some subjects were excluded for analysis because they did not fit matching criteria. We classified subjects according to the International Classification of Headache Disorders 3β [35] in CM for at least the last year, EM and HC. HC were recruited among healthy volunteers in the primary care clinic and among family members (8 male patients' spouses), patients´ friends, and healthy personnel. HC were asked for any type of acute headache or chronic pain. All subjects were instructed for blood drawing under fasting conditions and having fulfilled headache diary. Among patients, blood samples were drawn from those who had not had a migraine attack in the previous 72 h nor they had consumed acute anti-migraine medication (triptans or NSAID). Exclusion criteria included headache not fitting migraine criteria and insufficient headache information from the previous month. Patients having CM or EM could also have previous medication-overuse headache (MOH) or TTH. Psychiatric comorbidities were not an exclusion criterion. We collected patients' information on demographic variables (age, age at migraine onset, sex and body mass index), clinical variables (migraine with or without aura, total days of headache/term, medication abuse, allodynia), comorbidities (hypertension, hyperlipidemia, smoking, alcoholism, ischemic heart disease, fibromyalgia), contraceptive intake and MIDAS and HIT-6 scales.

Laboratory procedures
All blood samples were drawn at the Headache Clinic of the University Hospital Marqués de Valdecilla, in order to minimize timing for sample handling. A total of 5 ml blood samples was obtained in tubes without anticoagulant and with separating gel, under fasting conditions. Samples were immediately centrifuged, put on ice, aliquoted, and stored at -80º C until analyses, taking a total time of ten minutes. Peripheral blood samples were taken from the antecubital vein from 296 age-and sex-matched subjects ( ml for CGRP, 6.17-500 pg/ml for VIP and 0-1000 pg/ml for PACAP. All ELISA were performed by the same experienced technician who was blinded for the clinical diagnosis. All samples were analyzed continuously in the same laboratory, under the same environmental conditions, and using the same batch for samples from different clinical groups (CM, EM and HC), in order to avoid a possible batch effect that could condition differences in results in the different groups.

Statistical analysis
Data are reported as the median and interquartile range (IQ), or mean and standard deviation (SD). The t-test for independent samples or Kruskal-Wallis (or Mann-Whitney U-test for comparison of two categories) were used depending on the normal distribution for continuous variables by Kolmogorov-Smirnov test. Pearson's "r" statistic was used for bivariate correlation analysis for normal variables, Spearman' "rho" test for non-normal variables, Chi2 for categorical variables. All tests were considered significant for twotailed p < 0.05. To evaluate the relationship among clinical groups and CGRP, VIP and PACAP, individually, we used a general linear model analysis. A multinomial logistic regression analysis, with clinical groups as polychotomous dependent variable, was performed to evaluate the risk of having CM or EM using serum levels of CGRP, VIP and PACAP as independent variables, adjusted for sex and age. Other confounding variables such as migraine aura, hypertension, diabetes mellitus, alcohol, smoke, hyperlipidemia, or previous ischemic heart disease had been discarded for logistic regression as they had no effect on neuropeptide levels (data not shown). A p value < 0.05 was considered to indicate statistical significance. Odds ratios (ORs) are expressed with their 95% confidence interval (CI). Additionally, we evaluated the power of CGRP, PACAP, and VIP to correctly classify clinical groups, obtaining proportions for the global sample and for specific clinical group. We have taken into consideration the potential collinearity of independent variables analyzing the variation inflation factor and tolerance. In a second model, we used only migraine categories, serving EM as reference.
All tests were performed with the SPSS package (IBM SPSS Statistics for Windows, v22.0. Armonk, NY: IBM Corp.).

Sample size calculations
For this purpose, we used published data for CGRP, PACAP, and VIP means and SD [18,20,21]. We calculated sample size necessary to discriminate CM and non-CM groups in the worst scenario. Thus, we obtained a power of 90% for CGRP (n = 90; alpha = 0.01) and PACAP (n = 81; alpha = 0.01) for mean differences of 20%; and a power of 80% (n = 87; alfa = 0.05) for VIP mean differences of 30%.

Clinical results
This is the primary analysis of present data. A total of 296 subjects, 30  . There were no differences between EM and HC for all these neuropeptides. We observed that ranges were wider in CM than EM and HC, especially for PACAP (Fig. 1).
Migraine aura, hypertension, diabetes mellitus, alcohol, smoke, hyperlipidemia, or previous ischemic heart disease had no effect on neuropeptide levels (data not shown).

Effect of preventatives on CGRP, PACAP, and VIP serum levels.
We have also investigated the relationship of CGRP, PACAP, and VIP with preventatives in CM and EM patients. In CM patients, median and IQ differed from those under OnaBT treatment (n = 42) and other preventatives (n = 27) or no treatment (n = 32.

Multinomial logistic regression modeling
We also evaluated the association of CGRP, PACAP and VIP and clinical categories, using a model adjusted for age and sex. This model could explain a third of the total variance [Pearson's goodness-of-fit χ 2 = 554.617 (p = 0.344), and Nagelkerke's pseudo R 2   Although all three neuropeptides showed moderate correlation in the covariance matrix among them (CGRP-PACAP, r = 0.504, p < 0.001; CGRP-VIP, r = 0.558, p < 0.001; and PACAP-VIP, r = 0.435, p < 0.001), the degree of collinearity was low (variance inflation factor = 1.145 for CGRP-VIP and CGRP-PACAP, and 1.226 for PACAP-VIP) and tolerance was higher than 0.8 for all correlations.

Discussion
To our knowledge, this is the first time that a joint assessment of CGRP, VIP and PACAP serum levels and correlations among them have been studied in patients with CM and EM. There is a long debate about whether serum levels in peripheral blood of these neuropeptides reflect the activation of the TVS and whether they can help us to differentiate migraine clinical categories (CM vs EM) [7,18,20,21,34,36]. This latter question has been the rationale for our study.

Neuropeptide levels in chronic migraine
Firstly, we found that interictal serum CGRP, VIP and PACAP were significantly higher in CM compared to EM and HC (Fig. 1). Regarding CGRP, these findings concur with some previously reported [18], but disagree with others [7]. Some authors have found higher CGRP levels in MA than in migraine without aura [12,18], so a lower prevalence of MA within the CM group has been suggested to condition no differences between CM and EM [7]. However, we found no effect of MA on CGRP serum levels. Actually, mean CGRP values for the different clinical categories varied throughout published studies. Our observed median values of CGRP ranged from 14 pg/mL in HC to 20.4 pg/mL in CM; those observed in Lee et al. [7] from 75.7 pg/ml in HC to 64.9 pg/ml in CM; and those in Cernuda-Morollón et al. [18] from 33.74 pg/ml in HC to 74.9 pg/ml in CM. Given that the procedure and ELISA-based assay kits used were the same, it is possible that differences in ELISA performance, instability of the detecting antibody, or other, unreliable conditions could render disparate results across laboratories.
Regarding VIP, according to our results, other researchers also found higher interictal serum levels in CM [20]. Nevertheless, the present study is the first to show elevated interictal serum PACAP levels in CM [21].
In the univariate analyses, we also observed that ranges of the concentrations of neuropeptides in the CM group were strikingly wider than in EM and HC groups, especially for PACAP (Fig. 1). These data suggest that there is a large  interindividual variation to the detriment of accurate clinical diagnosis. Such a variation of these neuropeptide serum levels does not conform to the hypothesis of a persistent TVS activation in CM, as previously suggested [18], against the explanation that frequent migraine attacks increase the probability for detecting higher neuropeptide levels. But again, there is variability among studies. The ranges of CGRP, VIP and PACAP concentrations are also wider in CM in Cernuda-Morollón et al. studies [18,21], while Lee et al. [7] found greater variation in CGRP levels in the EM group.

Neuropeptide levels in episodic migraine
In the present study, serum CGRP, VIP and PACAP were not significantly higher in EM than in HC. These results support an episodic and less frequent activation of the TVS in these patients. Regarding CGRP, this concurs with Lee et al. findings [7], but disagree with two other studies that determined CGRP also in peripheral blood outside episodes in EM patients [10,12]. These studies recruited subjects solely from specialized clinics, which could condition a selection bias of patients with more frequent migraines than those recruited in our study, both from a specialized clinic and primary care. However, stratified analysis according to headache frequency, performed by Lee et al. [7], as well as by Ashina et al. [12], did not find any such correlation. In the present study, only PACAP showed some correlation to total headache days. Our results reflecting no higher PACAP levels in EM than HC, concur with the preceding ones [21,23,24], unlike those of VIP, whose levels have been reported higher in EM compared to HC [20]. It should be noted that VIP may preferentially rise in those patients who experience marked autonomic symptoms [37], and the proportion of patients with autonomic symptoms was not assessed in this study.

Correlations
We found moderate correlations between VIP and PACAP, VIP and CGRP, and PACAP and CGRP. The activation of the TVS is believed to result in the release of CGRP by the trigeminal nerve terminals, and VIP and PACAP by the efferent arm of the trigeminal-facial arch. Although in some individuals the release of one peptide may predominate over another, there seems to be a certain correlation among their levels, indicating that all of them may participate somewhat in most migraine pain. The inverse correlation obtained between age and CGRP, though marginal, could correspond to changes in headache severity in migraine attacks that occur with aging, with decreasing throbbing, pressure and stabbing [38], which might mean less TVS activation during attacks and lower neuropeptide release. This differs from the studies by Cernuda-Morollón et al. and Lee et al. in which CGRP concentrations were not influenced by age [7,18]. From our explanation, we would have expected to find an inverse correlation with PACAP and VIP as well, which, like others [21], we did not. In the case of VIP, its levels may depend on the presence and intensity of autonomic symptoms [37]. A further study is warranted to analyze the correlation between VIP and the presence of autonomic symptoms.

Neuropeptide levels and preventive treatment relationship
We have observed that CGRP, VIP, PACAP and age were all higher in CM patients under OnaBT treatment than those with no preventatives in univariate analysis, but only age and VIP increased the risk of being in the OnaBT treatment group after adjustments. CGRP levels decreased from 76.85 pg/mL to 52.48 pg/mL in CM patients considered responders in one study [16]. It is noteworthy that even after OnaBT treatment, CGRP still remained higher than that observed in controls in a previous study of the same group [18]. There are no previous reports analyzing PACAP or VIP as treatment response. Our study was not specifically designed for studying treatment effect in CGRP, PACAP, or

Diagnostic value
No other study has performed a multinomial regression modeling including these 3 neuropeptides to assess their overall effect on migraine, which was globally as low as 49.8%. Individually, CGRP was the worst at classifying clinical groups, being 44.93% of subjects correctly classified, and only 59.4% of CM patients. Using ROC curves, Cernuda-Morollón et al. found that a CGRP concentration of 58.22 pg/ml correctly classified 85.7% of CM [18]. Although in the latter study, diagnostic value of CGRP would be higher, both show a non-negligible percentage of CM patients (40% and 15%) that would not be correctly classified by CGRP. It has been hypothesized that CM patients whose CGRP levels are in the range of HC may suffer from other headaches mimicking CM, or that the pain in these cases is secondary to the release of other substances [36]. Our opinion is that these values depend on the moment in which they are determined and on the procedure of determination. Surprisingly, in the present study PACAP levels correctly classified more than 92% of EM, a diagnostic value that had not been previously reported. None of the neuropeptides was useful to discriminate HC from migraine patients. This study has several limitations. Although we fitted with preliminary sample size calculations, present sample size maybe needs to be increased to get enough power for multiple comparisons and stratified analyses. Regarding sampling, there were 81 patients under preventative treatments who were not withdrawn as they formed part of our secondary objectives. This could have influenced neuropeptide levels in the EM and CM groups. Though this effect has not been demonstrated for oral preventatives [18], CGRP has shown to decrease in good responders to OnaBT treatment [16]. However, in this study, as we already mentioned, serum levels of OnaBT-treated patients continued to be higher than those untreated patients in the CM group. Regarding laboratory procedures, sample handling and peptide determination are unsolved issues that need further studies to ensure uniformity analyzing these neuropeptides. In the present study, samples were taken and stored at -80ºC in less than 10 min as recommended. Although degradation time after centrifugation is not known [7], we have made, as previous researchers, great efforts to minimize the time until analysis. Finally, some of the kits used are not specific for the peptide under analysis. For instance, PACAP kit, measures both PACAP-27 and PACAP-38, but PACAP-38 is the most prevalent isoform in mammals [22], and this kit is the same that previous researchers used [21].
The present study also has some strengths, such as a bigger sample size than most similar previous studies, age-and sex-matched samples in order to minimize possible age and sex variations [39] and the joint analyses of three neuropeptides, something which allowed us to study correlations to help better understand the process of TVS activation. A subset of subjects was recruited in a primary care center, thereby avoiding any possible selection bias of only the most severely impaired patients recruited in specialized clinics. Additionally, unlike some previous studies [18], we did not exclude patients with psychiatric comorbidities, something which makes our sample more representative of the reality in a Headache Clinic.

Conclusions
We found that interictal serum CGRP, VIP, and, for the first time PACAP levels, were significantly higher in CM than both EM and HC, regardless of preventative treatments. These findings support the role of the three neuropeptides in migraine, but more specifically in CM. However, with currently available techniques, CGRP, VIP and PACAP serum levels are hardly useful individually or jointly to discriminate migraine clinical categories. Only PACAP showed a significant capacity to correctly classify the EM. Although there is a prevailing need for a migraine biomarker feasible worldwide using standardized techniques, nowadays, being CGRP the best candidate as a migraine biomarker [36], there is no consistent method for its analysis. Sample handling and determination process of these neuropeptides needs to be standardized. Our results regarding the PACAP role in CM need further replication.