Cerebrovascular Hemodynamic Responses to Breath-Holding in Migraine: A Longitudinal Functional Near-Infrared Spectroscopy Study Comparing a Calcitonin Gene-Related Peptide Monoclonal Antibody and Oral Preventive Treatment

Article information

Headache Pain Res. 2026;27(1):52-63

Publication date (electronic) : 2026 February 24

doi : https://doi.org/10.62087/hpr.2025.0029

Dong A. Yea¹, Jong Kwan Choi², Yoo Hwan Kim^,1

¹Department of Neurology, Hallym University Sacred Heart Hospital, Hallym University College of Medicine, Anyang, Republic of Korea

²OBELAB Inc., Seoul, Republic of Korea

Correspondence: Yoo Hwan Kim, M.D., Ph.D. Department of Neurology, Hallym University Sacred Heart Hospital, Hallym University College of Medicine, 22 Gwanpyeong-ro 170beon-gil, Dongan-gu, Anyang 14068, Republic of Korea Tel: +82-31-380-3740, Fax: +82-31-380-4118, E-mail: drneuroneo@gmail.com

Received 2025 December 30; Revised 2026 February 2; Accepted 2026 February 5.

Abstract

INTRODUCTION

Migraine is a common neurological disorder characterized by recurrent headache attacks. Increasing evidence suggests that migraine is better understood as a disorder involving abnormal activation of the trigeminovascular system rather than a purely neuronal condition.1-3 Cerebrovascular reactivity (CVR) and vasomotor reserve have been investigated in patients with migraine; however, previous studies have reported heterogeneous findings, reflecting variability in study populations, imaging modalities, and assessment paradigms rather than a consistent cerebrovascular abnormality.4-6

Calcitonin gene-related peptide (CGRP) plays a central role in migraine pathophysiology and is a potent vasodilatory neuropeptide released from trigeminal afferents during migraine attacks.7 Through its effects on trigeminovascular system, including neurogenic inflammation and modulation of cerebral arterial tone, CGRP contributes to migraine pain generation and associated vascular responses.8-10 On this basis, monoclonal antibodies targeting the CGRP pathway including fremanezumab have been developed and have demonstrated robust efficacy in reducing migraine frequency and headache-related disability.11,12

Despite these established clinical benefits, the physiological effects of CGRP monoclonal antibodies on cerebral vascular function remain incompletely understood. In particular, it is unclear whether clinical improvement following CGRP pathway inhibition is accompanied by measurable changes in CVR.13

Cerebral vasomotor reactivity refers to the capacity of cerebral arterioles to dilate or constrict in response to changes in arterial carbon dioxide tension, thereby contributing to the regulation of cerebral blood flow. This concept is distinct from cerebrovascular autoregulation, which refers to the maintenance of stable cerebral blood flow across changes in perfusion pressure.14,15 Altered CVR in migraine therefore reflects differences in CO₂-mediated vasodilatory capacity rather than heightened nociceptive sensitivity or exaggerated vascular responsiveness.16,17 Neuroimaging studies using positron emission tomography, single-photon emission computed tomography, and blood oxygenation level dependent functional magnetic resonance imaging have demonstrated alterations in cerebral perfusion and cortical excitability in migraine; however, their longitudinal applicability is limited by cost, accessibility constraints, and susceptibility to motion artifacts.18,19

Functional near-infrared spectroscopy (fNIRS) is a non-invasive optical neuroimaging modality that enables continuous monitoring of cortical hemodynamic responses through quantification of oxygenated and deoxygenated hemoglobin (HbR) concentrations. Owing to its portability, safety, and relative tolerance to movement, fNIRS is well suited for longitudinal evaluation of cerebrovascular hemodynamic responses in clinical environments.20-24

The breath-holding test (BHT) is an established physiological challenge that induces transient hypercapnia for the assessment of CVR. Rising arterial carbon dioxide tension during breath-holding (BH) results in cerebral arteriolar dilation and increased cerebral blood flow, which can be reliably captured using fNIRS derived hemoglobin signals. Compared with controlled carbon dioxide inhalation paradigms, the BHT is simpler to administer, well tolerated, and widely used as a screening tool for impaired vasomotor reserve.25,26

To date, longitudinal fNIRS-based investigations directly comparing CGRP monoclonal antibody therapy with conventional oral preventive treatments remain limited, and the relationship between clinical improvement and change in cerebrovascular hemodynamic responses has not been fully clarified. Therefore, this study aimed to longitudinally and descriptively compare fNIRS-derived cerebrovascular hemodynamic responses during the BHT between migraine patients treated with fremanezumab and those receiving conventional oral preventive medications at baseline and after 3 months of treatment.

MATERIALS AND METHODS

3. Breath‑holding test protocol

All participants underwent a standardized BHT consisting of three consecutive phases (Figure 1): (1) baseline resting phase, 30 seconds of spontaneous breathing; (2) BH phase, 30 seconds of voluntary BH; and (3) recovery (Rec) phase, 30 seconds of spontaneous breathing.

Figure 1.

Experimental setup and breath-holding test protocol for fNIRS-based cerebrovascular reactivity assessment. fNIRS was applied over the bilateral prefrontal cortex to continuously measure oxygenated and total hemoglobin concentrations. The breath-holding test involved a 30-second baseline resting phase, followed by a 30-second breath-holding phase to induce hypercapnia, and a 30-second recovery phase. All participants underwent the same protocol at baseline and at 3 months.

fNIRS, functional near-infrared spectroscopy; CGRP, calcitonin gene-related peptide; Ab, antibody; Fpz, frontal pole, midline position; F7, left frontotemporal electrode; F8, right frontotemporal electrode.

Participants were instructed to perform BH for approximately 30 seconds following normal expiration while avoiding deep inspiration to minimize Valsalva‑related reductions in cerebral blood flow. During BH, increasing arterial carbon dioxide tension induces cerebral arteriolar dilation and increased cerebral blood volume, which are detectable through fNIRS derived hemoglobin signals. In accordance with previous literature, CVR was characterized using task related changes in oxygenated hemoglobin (HbO), HbR and total hemoglobin (HbT) during the BH and Rec phases.26

For BH–induced hemodynamic responses, the fNIRS signal was segmented into five consecutive 10‑second intervals (resting, BH_A, BH_B, Rec_A, Rec_B). All segments were referenced to the resting interval to enable comparison of relative temporal trajectories of HbO, HbR, and HbT. Primary indices included the BH index (BH_B–BH_A) and Rec index (Rec_A–Rec_B), computed at baseline and 3 months and compared longitudinally to characterize potential alterations in CVR (Figure 2).

Figure 2.

Conceptual schematic of fNIRS hemodynamic response patterns during resting, breath-holding, and recovery phases. This figure provides a conceptual illustration of typical hemodynamic response patterns measured by fNIRS during resting, breath-holding, and recovery phases. The schematic illustrates overall temporal trends and directional changes rather than quantitative study data used for statistical inference.

fNIRS, functional near-infrared spectroscopy; HbO, oxygenated hemoglobin; HbR, deoxygenated hemoglobin; HbT, total hemoglobin; Hb, hemoglobin.

RESULTS

Baseline demographic characteristics did not differ significantly between the CGRP monoclonal antibody group and the oral preventive medication group. Age, sex distribution, height, body weight, and body mass index were comparable between-groups (all p>0.05) (Table 1).

Table 1.

Baseline demographic characteristics of the CGRP mAb group and oral preventive medication group

At 3 months, monthly headache days numerically decreased in both groups; however, the within group changes did not reach statistical significance (CGRP: Δ=–2.00, p=0.26; oral: Δ=–1.50, p=0.48), and there was no significant between-group difference (p=0.85). Acute medication days were significantly reduced in the CGRP group (Δ=–7.00, p=0.03) but not in the oral group (Δ=–3.50, p=0.19), with no significant between-group difference (p=0.36). HIT-6 and MIDAS scores improved in both groups; however, only the reduction in MIDAS was significant in the CGRP group (Δ=–21.25, p=0.02), while changes in the oral group were not significant (HIT-6: p=0.05; MIDAS: p=0.11), and no between-group differences were observed (both p=0.55). GAD-7 and PHQ-9 changes were modest and non-significant in both groups (all p>0.05). PGIC scores at 3 months were numerically higher in the CGRP group than in the oral group (4.50 vs. 3.42), without a significant between-group difference (p=0.15) (Table 2).

Table 2.

Changes in clinical outcomes from baseline to 3M in both treatment groups

During the BH phase, both the CGRP monoclonal antibody group and the oral preventive medication group showed numerical increases in HbO across multiple prefrontal channels. In channel 06, HbO increased from 0.003 at baseline to 0.117 at the end of BH in the CGRP group and from –0.011 to 0.006 in the oral preventive medication group. During the Rec phase, HbO values in both groups numerically decreased toward pre-task levels, reaching 0.033 in the CGRP group and 0.022 in the oral preventive medication group by the end of Rec. Similar patterns were observed across the majority of channels, although the magnitude of change varied by channel and time point (Figure 3).

Figure 3.

(A) HbO time series during the breath-holding test across prefrontal channels in the CGRP and oral preventive groups at baseline and at 3 months. Solid lines indicate baseline recordings, and dashed lines indicate 3-month follow-up recordings. (B, C) Three-dimensional maps showing the mean HbO concentration at 3 months in (B) the CGRP monoclonal antibody group and (C) the oral preventive medication group.

CGRP, calcitonin gene-related peptide; mAb, monoclonal antibody; HbO, oxygenated hemoglobin.

Two-way ANOVA revealed no significant group×time interaction effects across channels after correction for multiple comparisons for the HbO-based BH index (Table 3). Although channel 06 showed a nominally significant interaction at the uncorrected level (F(1,16)=8.448, p=0.010), this effect did not remain significant after family-wise error or false discovery rate correction (both corrected p=0.155), and no other channels demonstrated significant interaction effects.

Table 3.

Two-way ANOVA results for the HbO-based breath-holding index

Post hoc analyses focusing on channel 06 (Figure 4) demonstrated a significant difference between the CGRP group at baseline and the oral treatment group at 3 months (mean difference=0.169, p<0.001), which remained significant after Bonferroni correction. In contrast, the within group change from baseline to 3 months in the oral treatment group was significant only at the uncorrected level (mean difference=–0.069, p=0.029) and did not remain significant after correction for multiple comparisons.

Figure 4.

Treatment×time effects on the HbO-based breath-holding index on Channel 6. Tukey-adjusted p-values control the FWE rate for a family of four comparisons. Bonferroni-adjusted p-values control the FWE rate for a family of six comparisons. Green dotted line: decreasing within subject change from Base to 3M; orange dotted line: increasing within subject change from Base to 3M. Asterisk indicates a statistically significant (*p<0.05; ***p<0.001).

CGRP, calcitonin gene-related peptide; Base, baseline; 3M, 3 months; IQR, interquartile range; HbO, oxygenated hemoglobin; SE, standard error; df, degrees of freedom; FWE, family-wise error.

Analyses of HbR- and HbT-based BH indices showed no significant group×time interaction effects after correction for multiple comparisons, with only nominal uncorrected effects in a few channels. Detailed results are presented in the Supplementary Tables 2, 3 (available online).

For the HbO-based Rec index, two-way ANOVA identified no significant treatment×time interaction effects across channels after correction for multiple comparisons (Table 4). Channel 10 showed a significant interaction at the uncorrected level (F(1,20)=8.242, p=0.009); however, this effect did not persist after family-wise error or false discovery rate correction (both corrected p=0.142), and no other channels showed significant interaction effects.

Table 4.

Two-way ANOVA results for the HbO-based recovery index

Post hoc analyses in channel 10 (Figure 5) revealed a significant difference between the CGRP group at baseline and the oral treatment group at 3 months (mean difference=0.150, p=0.001), which remained significant after Bonferroni correction (p=0.007). Additionally, a significant within group change was observed in the oral treatment group from baseline to 3 months (mean difference=–0.097, p=0.002), which also remained significant after Bonferroni correction (p=0.010).

Figure 5.

Treatment×time effects on the HbO-based recovery index on Channel 10. Tukey-adjusted p-values control the FWE rate for a family of four comparisons. Bonferroni-adjusted p-values control the FWE rate for a family of six comparisons. Green dotted line: decreasing within subject change from Base to 3M; orange dotted line: increasing within subject change from Base to 3M. Asterisk indicates a statistically significant (*p<0.05; **p<0.01).

CGRP, calcitonin gene-related peptide; Base, baseline; 3M, 3 months; IQR, interquartile range; HbO, oxygenated hemoglobin; SE, standard error; df, degrees of freedom; FWE, family-wise error.

Analyses of HbR and HbT based Rec indices revealed no significant treatment×time interaction effects after multiple comparison correction; detailed results are reported in the Supplementary Tables 4, 5 (available online).

DISCUSSION

This longitudinal fNIRS study examined whether clinical improvement following migraine preventive therapy is accompanied by changes in CVR during a hypercapnic challenge. Although both the CGRP monoclonal antibody group and the oral preventive medication group demonstrated clinically meaningful improvements over 3 months, these benefits were not uniformly associated with normalization of cerebrovascular hemodynamic responses. This dissociation suggests that symptomatic improvement in migraine may occur, at least in part, independently of restoration of altered CVR.

Cerebral blood flow is tightly regulated by arterial carbon dioxide tension, and within a physiologically relevant range, variation in pCO₂ produces dose‑dependent vasodilatory or vasoconstrictive responses. The resulting sigmoidal relationship reflects the degree of preserved vasomotor reactivity within the remaining autoregulatory reserve of the cerebrovascular system.14 In this physiological context, the BHT provides a transient hypercapnic stimulus suitable for evaluating vasodilatory capacity across resting, BH, and Rec phases. In individuals with preserved vasomotor reserve, rising pCO₂ during BH is expected to induce arteriolar dilation and increases in HbO and HbT, whereas in those with reduced reserve, hypercapnia‑induced increases may be attenuated or plateaued, reflecting limited residual dilatory capacity.26,32

In our cohort, the CGRP monoclonal antibody group showed a longitudinal decrease in both the BH and Rec indices, whereas the oral preventive medication group demonstrated an opposite pattern, with an increase in these indices from baseline to 3 months. These divergent temporal trajectories suggest that cerebrovascular responses may evolve differently over time between the two treatment groups. However, we caution against inferring a direct mechanistic effect of CGRP-targeted therapy on CVR from these findings alone.33 The attenuation of the CO₂‑driven hemodynamic response in the CGRP group may reflect modulation of late‑phase vasodilatory dynamics, variability in vasomotor reserve, or other unmeasured physiological or behavioral factors.34,35 Whether this pattern represents attenuation of excessive vasoreactivity or a relative reduction in cerebrovascular reserve remains uncertain and warrants confirmation in larger, mechanistically oriented studies. Accordingly, the present findings should be interpreted primarily in terms of longitudinal change patterns rather than absolute CVR values at individual time points.

From a clinical perspective, these findings imply that improvement in migraine burden may not necessarily require concurrent normalization of CVR. Nociceptive network modulation and cerebrovascular autoregulatory adaptation may therefore represent partially independent physiological domains. Accordingly, fNIRS-based cerebrovascular indices should be interpreted as complementary rather than surrogate markers of treatment response, particularly within short term therapeutic windows.36-38

One plausible interpretation is that CGRP monoclonal antibody therapy primarily exerts therapeutic benefit through modulation of trigeminovascular nociceptive signaling rather than direct correction of cerebrovascular autoregulatory dysfunction.1 CGRP blockade may effectively suppress migraine attacks without immediately normalizing interictal neurovascular abnormalities. Alternatively, cerebrovascular adaptations may require longer treatment duration or cumulative exposure before measurable changes in vasomotor responsiveness emerge. This interpretation aligns with prior reports indicating that vascular and perfusion abnormalities may persist interictally despite clinical improvement.13,39

It is also possible that distinct migraine phenotypes exist some characterized predominantly by nociceptive network dysregulation and others by greater cerebrovascular dysautoregulation. Future studies incorporating subgroup- or responder-stratified analyses, extended longitudinal observation, and multimodal vascular assessment will be required to clarify these relationships and determine whether CVR trajectories carry prognostic or mechanistic relevance in migraine.16

From a methodological standpoint, the BHT was used as a practical hypercapnic challenge for fNIRS-based assessment of cerebrovascular reactivity. Although controlled CO₂ inhalation offers a more standardized stimulus, it requires specialized equipment and monitoring, which limits feasibility for routine or longitudinal studies. By contrast, the BHT induces hypercapnia with minimal procedural burden and allows repeated assessments. Prior reports showing good correspondence between BH derived and CO₂-based CVR measures support its suitability as a pragmatic approach for longitudinal evaluation, despite inter-individual variability in BH capacity.6,25,40

In the present study, fNIRS measurements were confined to the prefrontal cortex. Although this region is not traditionally regarded as a primary nociceptive hub, it plays an important role in autonomic regulation and cognitive-affective integration of pain, and it has demonstrated robust CVR to hypercapnic challenge in prior fNIRS studies.41 Therefore, the present measurements likely capture a meaningful component of cortical vascular regulation relevant to migraine pathophysiology, although additional cortical and subcortical regions may contribute to cerebrovascular adaptation in ways not captured in our dataset.4

Several limitations should be acknowledged. Episodic and chronic migraine patients were analyzed together without stratification by migraine subtype. Given the known differences in disease severity between these subtypes, this heterogeneity may have influenced the observed treatment-related changes. Because treatment allocation was not randomized, potential confounding by indication cannot be excluded. The relatively small sample size may have limited statistical power to detect subtle between-group differences in cerebrovascular responses. The recording montage was restricted to the prefrontal cortex and may not reflect hemodynamic changes in deeper or posterior brain regions. Finally, the 3-month follow-up period may be insufficient to capture delayed or progressive cerebrovascular adaptations; given that vascular remodeling and tone regulation may occur on longer temporal scales than symptomatic improvement, extended longitudinal observation will be essential in future work. Studies incorporating larger cohorts, longer treatment exposure, expanded cortical coverage, and multimodal vascular assessment will be necessary to further elucidate the temporal relationship between clinical improvement and CVR in migraine.

SUPPLEMENTARY MATERIAL

Article information Continued