Serum calcitonin gene-related peptide in patients with persistent post-concussion symptoms, including headache: a cohort study

This cohort study was based on serum samples from participants included in a recently published open-label parallel-group randomized controlled trial (RCT) conducted at Hammel Neurorehabilitation Centre and University Research Clinic and The Research Clinic for Functional Disorders and Psychosomatics, Aarhus University Hospital in Denmark. The RCT study showed that a new interdisciplinary intervention, Get going After concussIoN (GAIN), significantly lowered PPCS symptoms compared with enhanced usual care (EUC) [15]. The 8-week interdisciplinary intervention was non-pharmacological, based on cognitive behavioral therapy and gradual return to activities, and included both group sessions and individual sessions with a physio/occupational therapist and a neuropsychologist [15]. As outlined in the preregistered analysis plan (referenced in the ‘Ethics Approval and Registrations’ section), we did not anticipate any influence of the non-pharmacological intervention on CGRP levels within the cohort. Consequently, both intervention arms were combined at follow-up akin to a cohort study for the primary outcome.

Participants were recruited between March 2015 to September 2017 from general practitioners and hospitals in the Central Denmark Region. Eligible patients were 18–30 years old and had experienced PPCS for 2–6 months. Concussion was defined according to the recommendations from the World Health Organization (WHO) Task Force [16]. Additionally, a direct head trauma was required to rule out pure acceleration-deceleration traumas. PPCS were defined as having a Rivermead Post-Concussion Questionnaire (RPQ) [17] score ≥ 20. Patients with severe brain injury, previous concussion with symptoms > 3 months, drug/alcohol abuse, severe somatic or psychiatric conditions that impeded participation in the intervention, and patients who could not speak Danish were excluded. The inclusion/exclusion criteria were thus identical to those of the previously published RCT [15], except that patients below 18 years of age were not included in this study.

Potential PPCS patients referred from general practitioners or hospitals were invited to Hammel Neurorehabilitation Centre and University Research Clinic by mail. A thorough assessment for eligibility was done by a neurologist and a psychiatrist and included:

Further details regarding the assessment procedure have been reported previously [18].

In 2022, we recruited a random cross-sectional sample of healthy, anonymous individuals aged 18–30 from the Blood Bank at Aarhus University Hospital, Denmark, matching the age group of the PPCS participants. The inclusion of healthy individuals allowed us to assess whether CGRP levels were altered in the PPCS participants at baseline, a crucial part of the primary outcome (outlined below). The healthy individuals did not participate in the RCT intervention, and the only demographic data available were age and sex. A sample size of 120 yielded a statistical power of 82% based on the effect size and standard deviations in a study in PTH [14].

The blood samples were drawn from the antecubital vein using serum tubes (Vacutainer Cat 368815 or Cat 367896) and were allowed to clot for 30 min at room temperature. Subsequently, the blood samples were centrifuged for 10 min at 2880 g at 20 °C. The serum was then collected and initially frozen to  – 20 °C, and within 24 h moved to  – 80 °C until analysis. The samples were stored in 1.5 mL polypropylene tubes (Sarstedt, Nümbrecht, Germany). The preanalytical methodology was the same in samples from healthy individuals and PPCS participants, except the latter samples underwent one freeze–thaw cycle due to the measurement of other biomarkers for another study (not yet published, ClinicalTrials.gov NCT05812742). No protease inhibitors were added, and all participants were non-fasting.

The primary outcome was serum CGRP concentrations at baseline and follow-up. The baseline blood samples were obtained in connection with the assessment of eligibility 2–6 months after the concussion. The follow-up samples were obtained approximately 7 months later after the completion of the intervention/reference treatment. In contrast, the samples from the anonymous healthy individuals (blood donors) were obtained during their routine voluntary blood donations and no follow-up samples were collected.

Headache data were retrieved from a previously published headache phenotype study conducted in the same PPCS participants [19]. In brief, the PPCS participants filled out a comprehensive headache questionnaire in connection with the assessment of eligibility. It contained data on monthly headache days, duration, current headache pain measured on the visual analogue scale (VAS), questionnaire data on the Headache Impact Test (HIT-6) [20], and allowed classification according to The International Classification of Headache Disorders 3rd edition [21]. Furthermore, questionnaire data on the RPQ, the Short Form (36) Health Survey (SF-36) [22], the Bodily Distress Syndrome Checklist (BDS) [23], Whiteley-8 (WI-8) [24], the “limiting behavior” and “all-or-nothing behavior” subscales from The Behavioural Responses to Illness Questionnaire (BRIQ) [25], The Brief Illness Perception Questionnaire (B-IPQ) [26], Symptom Checklist 8 AD (SCL-8AD) [27], and the Perceived Stress Scale (PSS) [28] were obtained from the RCT-study [15]. Supplementary Table 1 provides an overview of each questionnaire.

We analyzed the serum samples using a commercial enzyme-linked immunosorbent assay (ELISA) from Bertin Bioreagent which targeted α-CGRP and β-CGRP [29] (Cat No: A05481, Montigny le Bretonneux, France, batch no. 121 & 123). To reduce matrix effects, all calibration curves and quality controls (QCs) were prepared in CGRP-free serum. We prepared CGRP-free serum by incubating a pool of samples from healthy individuals and PPCS participants with CGRP antibodies (Cat No: A19482, Bertin Bioreagent, batch no. 0222) overnight at 4 °C on a tilt shaker. This was then filtered in accordance with Bertin’s instructions. Analysis of three replicates of the resulting CGRP-free serum, using the same methodology as for the study samples, revealed that the depletion was successful (< 2 pg/mL). To assess freeze–thaw stability, we spiked CGRP-depleted serum with 125 pg/mL using the provided CGRP standard in the kit, followed by freezing at  – 80 °C. The CGRP concentration remained stable averaging 141.3 pg/mL (SD: 25.4) for three freshly spiked replicates at 125 pg/mL, compared with 122.9 pg/mL (SD: 10.4) after one freeze–thaw-cycle. The observed difference was thus well below the < 20% deviation criterion specified in method validation guidelines [30].

Subsequently, the study samples were thawed at room temperature and analyzed using a calibration curve (7.8–1000 pg/mL) and quality control (125 pg/mL), both freshly prepared in CGRP-free serum. The assay procedure was performed following the manufacturer’s instructions. In short, after loading the samples, the tracer was added, and the plate was incubated overnight (16–20 h at 4 °C). Subsequently, Ellman’s reagent was added, and the resulting color intensity was analyzed with a microplate spectrophotometer (Synergy H1, BioTek, Winooski, The United States (USA)) at 410 nm after 75 min of incubation. Samples from patients and healthy individuals were included on each plate in an equal ratio to ensure representation of both groups. The position of the samples was varied between plates to reduce any bias from plate position. Analyses were conducted by the same experienced bioanalytical technician who was unblinded to the sample group (healthy or PPCS), but was blinded to sex and the hypothesis of the study. A calibration curve was included on each plate and plotted using Graphpad Prism (Graphpad Software, San Diego, USA) from which data were interpolated. The best fit was obtained using a second order polynomial regression (R² ≥ 0.99, Suppl. Figure 1). CGRP values below limit of detection were replaced with the limit of detection reported by the manufacturer (2 pg/mL), and similarly, values exceeding the detection range were replaced with the highest value of the calibration curve (1000 pg/mL). Dilution of samples exceeding 1000 pg/mL was not performed due to insufficient sample material.

The mean QC concentration on all eight plates was 177.2 pg/mL (SD: 27.3) which was higher than the nominal concentration (125 pg/mL). However, the QCs were highly reproducible with an inter-assay coefficient of variation (CV) of 15.4% and an intra-assay CV ranging from 0% to 13% (median: 6.0%).

Since the data were not normally distributed (Suppl. Figure 1), non-parametric tests were employed. CGRP concentration differences between PPCS participants and healthy individuals were assessed with the Wilcoxon rank-sum test, and adjustment for gender was done by stratified analyses. Changes at follow-up in PPCS participants were analyzed with the Wilcoxon signed-rank test. Non-parametric 95% confidence intervals (CI) were generated by bootstrapping the median CGRP difference, computing the resulting 2.5th and 97.5th percentiles across 10,000 replicates. For the primary outcomes, the significance level was set at p ≤ 0.050.

The effect of the non-pharmacological intervention on delta CGRP values (follow-up minus baseline) was evaluated by comparing the delta CGRP values in the treatment arms using the Wilcoxon rank-sum test. This was not included as a primary outcome, since we did not expect the intervention to alter CGRP levels, which was stated in the preregistered analysis plan.

In an exploratory analysis, we correlated serum CGRP levels with headache days, headache duration and current headache pain (VAS score) at baseline using Spearman’s rank correlation. Kruskal-Wallis test was used to test serum CGRP differences between headache phenotypes. Additionally, we assessed the correlation of CGRP levels to patient-reported outcomes (RPQ, HIT-6, SF-36, BDS, WI-8, BRIQ, B-IPQ, SCL-8AD, and PSS) using Spearman’s rank correlation. Finally, we assessed the correlation between delta CGRP and the delta value of the questionnaire data (follow-up minus baseline) using Spearman’s rank correlation. In total, 40 analyses were planned in the exploratory analyses, and the significance level was thus adjusted to p = 0.05/40 = 0.0013 (Bonferroni correction).

All statistical analyses were two-tailed and conducted using Stata 17 for Windows (StataCorp, College Station, USA), and graphs were created using Graphpad Prism.

Serum CGRP concentrations in PPCS participants at baseline of the RCT study and the healthy individuals are presented in Fig. 2 and Table 3. Median serum CGRP concentrations were doubled in PPCS participants compared to healthy individuals, with a median difference of 82.2 pg/mL (95% CI 17.7–145.6, p = 0.050). When stratifying for sex, it became evident that this difference was due to females with PPCS having a five times higher median concentration than healthy females (166.3 pg/mL vs. 32.1 pg/mL, p = 0.0006). No statistical difference was observed in males with PPCS (p = 0.83). Follow-up blood samples were collected at a median of 11.4 months (IQR: 10.3–12.7) after the trauma and following the completion of the non-pharmacological intervention/reference treatment. Figure 3 and Table 4 shows that CGRP levels decreased at follow-up in PPCS participants, with a median change of  – 1.3 pg/mL (95% CI  – 17.6–0, p = 0.024). The observed change was primarily attributed to females, who showed a median difference of  – 2.7 pg/mL (95% CI  – 37.4–0, p = 0.028) whereas males did not show any statistical difference (p = 0.54).

There was no effect of the non-pharmacological intervention on CGRP levels: The median delta CGRP concentration was  – 3.5 pg/mL (IQR:  – 64.8–0) in the reference treatment (EUC), and 0 pg/mL (IQR:  – 46.4–14.9) in the intervention arm (GAIN), and there was no statistical difference (95% CI of median difference:  – 15.7–47.3, p = 0.33).

CGRP concentrations ranged from 2 pg/mL to 1000 pg/mL, indicating large interindividual variations in CGRP concentrations (Suppl. Figure 1). However, within each individual, baseline and follow-up concentrations showed a strong correlation (rho = 0.95, p < 0.0001).

At baseline, high CGRP concentrations were associated with a more favorable outcome in several patient-reported outcome measures. For example, higher serum CGRP correlated weakly with fewer headache days, shorter headache duration, and less headache pain with p-values ranging from 0.038 to 0.050 at baseline (rho ≈  – 0.23) (Fig. 4). However, the Kruskal–Wallis test showed no CGRP concentration difference between headache phenotypes (p = 0.67). Further exploratory analyses (Figs. 5 and 6) showed that higher serum CGRP was correlated with a better physical function, physical health, and less bodily pain on the SF-36 questionnaire (rho = 0.29–0.35, p ≤ 0.0079). Likewise, higher CGRP levels were correlated with a smaller symptom burden as measured by the BDS and WI-8 questionnaires (rho =  – 0.29, p = 0.006). However, the only correlation that remained statistically significant after Bonferroni correction was the bodily pain domain on the SF-36 questionnaire (rho = 0.35, p = 0.0008).

An additional observation unrelated to PPCS was a higher median CGRP concentration among healthy males compared with healthy females as shown in Table 3 (201.8 pg/mL vs. 32.1 pg/mL, p = 0.0025), with a median difference of 169.7 pg/mL (95% CI: 14.2–496).

The correlations between the change in CGRP concentrations (delta CGRP) and the change in questionnaire scores (delta score) from baseline to follow-up were examined (follow-up minus baseline). Unlike the baseline data, the follow-up questionnaire data and blood samples were obtained at different days, and the median time difference was 27 days (IQR: 9–50). Changes in headache days, duration, current headache pain, and changes in the SF-36 questionnaire scores were not correlated to changes in the CGRP concentration (Suppl. Figures 2 & 3). However, Fig. 7 shows that a reduction in CGRP at follow-up was associated with an improved RPQ and BDS-score, and the latter finding remained significant even after Bonferroni correction (rho = 0.54, p < 0.0001). A post hoc analysis revealed that the significant difference was mainly driven by an improvement in general symptoms such as headache and dizziness as well as cardiopulmonary/autonomic symptoms (Suppl. Table 2).

Unexpectedly, we observed that higher CGRP levels correlated with fewer headache days, shorter headache duration and reduced current headache pain at baseline (rho ≈  – 0.23). Similarly, elevated CGRP levels correlated with a lower symptom burden, although bodily pain measured by the SF-36 was the only finding that remained statistically significant after Bonferroni correction. The fact that high CGRP concentrations are correlated with a more positive outcome, contradicts findings in migraine [31] and the majority of pain-related studies [37]. However, it could align with the findings in the previously mentioned PTH study [14]. This earlier study included patients refereed to a specialized headache clinic, indicating more severe headaches. Interestingly, their study demonstrated lower CGRP levels in PTH than controls and the same tendency toward an inverse relationship with monthly headache days although it was not statistically significant (rho =  – 0.11, p = 0.27). This tendency was shown in a recent study conducted in post-deployment soldiers as well (rho =  – 0.12, p = 0.063) [38]. It is thus possible that more severe PTH are associated with lower CGRP levels. An explanation for lower CGRP given in the previous PTH study was that constant headache may result in the depletion of CGRP in trigeminal afferents [14]. This speculation was based on the finding that CGRP tissue levels were depleted in rats following capsaicin injection in the paw skin and sciatic nerve [39]. Whether capsaicin-mediated pain is comparable to the pain in headache is unknown.

However, other explanations can account for our findings as well. CGRP is a neuropeptide with several physiological functions unrelated to headache [13]. In the aforementioned animal study, CGRP inhibition with antibodies in concussed female rats did not alleviate cephalic pain hypersensitivity, raising questions about the role of peripheral CGRP in headache in females [34]. In contrast, there are several studies showing that CGRP inhibition could alleviate symptoms in rodents with migraine-like behavior in both males and females [40]. Furthermore, in a recent study on severe traumatic brain injury in humans, high CGRP concentrations were correlated with a lower risk of mortality, indicating a potential beneficial role for CGRP [41]. Additionally, an animal study showed that CGRP may have a favorable impact on peripheral nerve regeneration [42]. Further studies are needed to establish the role of CGRP in concussion and PTH.

A noteworthy observation, unrelated to PPCS, in our study was that healthy males had higher serum levels of CGRP than healthy females. This align with a recent study in post-deployment soldiers [38], but contrasts with a previous study that showed the opposite in females, particularly among those using oral contraceptives [43]. Studies investigating sex differences in CGRP among healthy individuals are limited, and it was not the primary aim of this study; further research is needed to draw definite conclusions.

At follow-up we showed that a reduction in CGRP levels was correlated with improved symptom levels measured by the BDS questionnaire. The post hoc analysis revealed that the reduction in CGRP correlated especially with an improvement in headache, dizziness and autonomic/cardiopulmonary symptoms (Suppl. Table 2). This suggests that a CGRP reduction could be linked to a positive physiological response. A major limitation in the follow-up data was that the questionnaire data and blood samples were collected at different time points, with a median time difference of 27 days. This could have contributed to the lack of an association between delta CGRP and change in headache days, duration, and pain, particularly when considering the limited response rate in the follow-up headache questionnaire (Suppl. Figure 2). Finally, the RCT intervention did not show any effect on CGRP concentrations at follow-up. This was expected and stated in the preregistered analysis plan since the RPQ-difference (symptom levels) was only 7 points between the intervention arms [15].

In conclusion of our exploratory analyses at baseline and follow-up, we found an association between CGRP levels and patient-reported outcomes measured by questionnaires, which is a rare finding. Future studies should replicate these findings, and additional research is needed to establish the role of CGRP in PPCS/PTH and concussion.

The CGRP concentrations in this study varied substantially with concentration levels ranging below detectable levels (< 2 pg/mL) to above 1000 pg/mL (Suppl. Figure 1). This wide concentration range is larger than those observed in migraine studies [31]. Apart from differences in inclusion criteria and demographic characteristics (such as sex), variations in assays and methodology can affect CGRP concentrations [44]. It is worth noting that a previous study using a similar methodology and same ELISA kit as our study showed a comparable concentration range, mean, and data distribution in healthy individuals (Suppl. Figure 4) [45].

This study had several strengths. The cohort was well-characterized, the concussion diagnosis was validated using the WHO criteria, and we had a relatively large sample size.

We used an ELISA-kit with no cross-reactivity with calcitonin, a peptide fragment of CGRP (CGRP position 8-37), amylin, and substance P according to the manufacturer. Furthermore, the assay was based on two antibodies for CGRP (sandwich ELISA) indicating its specificity for CGRP. In contrast, most other commercial ELISA kits have not investigated cross-reactivity and are only based on one antibody. Moreover, the QC’s and the calibration curves were constructed in study representative serum matrix, which should minimize the risk of matrix effects. Finally, we checked freeze–thaw stability of CGRP in serum to ensure that the reported concentrations in PPCS participants were accurate.

This study also had limitations. The QC’s systematically produced higher than nominal concentrations (170 pg/mL vs. 125 pg/mL), which indicates a slight overestimation of the concentrations in general. However, more importantly, the QC’s were reproducible with an inter-assay and intra-assay CV ≤ 15%. Furthermore, although the manufacturer reported no cross-reactivity between similar molecules, it cannot be ruled out that cross reactivity exists with a related peptide biomarker. However, this seems unlikely since CGRP was completely depleted in CGRP-free serum which was a pool of samples from both patients and healthy individuals. Since samples from patients and healthy individuals were evenly distributed on each ELISA plate, we do not expect these assay related factors to affect the conclusion of this study.

Preanalytical stability of CGRP is another point of concern. Prolonged storage and lack of protease inhibitors may decrease CGRP concentrations [45], although there is conflicting evidence on this matter [46]. In our study, patient samples underwent one freeze–thaw cycle and had a longer storage duration (up to 8 years), in contrast to the samples from healthy individuals (1 year of storage). Despite these possible limitations, our study still revealed significantly higher CGRP levels in PPCS patients than in healthy individuals. A further limitation of the study was the presence of multiple symptoms in addition to headache among PPCS participants, a consequence of the inclusion criteria requiring an RPQ score of 20 or higher. This diversity of symptoms prevented a clear identification of the exact causes of increased CGRP levels in PPCS participants.

Finally, the healthy individuals consisted of anonymous blood donors, for whom no demographic data were available. Danish blood donors tend to have a higher self-reported health and healthier lifestyle than non-donors, which may introduce a selection bias known as “the healthy donor effect” [47]. Although the mechanisms are poorly understood, lifestyle-related factors, such as higher weight, blood pressure [48], and exercise [49], might increase CGRP-concentrations. Since we lacked data on lifestyle factors in both groups, it cannot be ruled out that the observed differences might be partly explained by variations in lifestyle factors rather than PPCS/PTH. The potential confounding effects of preexisting migraines must also be considered since the headache questionnaire utilized did not allow accurate classification of headaches prior to the concussion [19]. However, 75% of the PPCS participants reported less than 15 headache days a year pre-trauma, and none reported using migraine medications before the concussion (Table 1). Furthermore, since migraine is generally not a contraindication for becoming a blood donor, the group of healthy individuals may also include individuals with migraine, thereby reducing the likelihood that this condition skews our findings. Regardless of potential confounding variables, the observed fivefold increase in median CGRP concentration in females with PPCS, compared with healthy individuals, is substantial. For this difference to be solely attributed to confounding factors, these factors would have to exert a strong influence on CGRP levels.

In conclusion, our data are strongly suggesting a role for CGRP in PTH. Future studies should aim to independently verify whether this is the case, preferably in a population with blood samples available before the head trauma (which can be done in athletes), to clearly establish a causal link between CGRP and concussion/PTH in humans. Furthermore, future studies should investigate whether CGRP targeted therapies are effective in PTH in a placebo-controlled RCT design.