Our literature analysis (Table 1) shows that studies based on CGRP determinations are highly variable in terms of measuring method and study design, including sample source, sample processing, inclusion/exclusion criteria for patients and controls and aim of the study [14, 15, 19, 31, 39, 42, 60, 66, 68]. Data analysis and presentation of laboratory determinations is also changeful, which hinder the comparison of the data. Despite all the difficulties, it results obvious that the overall outcomes and the conclusions drawn from them are inconsistent across works. Some authors have hypothesized that methodological differences might be the reason for such discrepancies [55, 56], and, although this is likely to be the case, there is not to date a consensus of how CGRP determinations should be carried out.

If we analyse the methods used to measure CGRP in migraine patients we can see there have been mainly based on two different techniques, RIA and ELISA. RIA was the first, and until the late 2000s, the only one employed. RIA is based on the competitive incubation for specific antibody sites to form antigen–antibody complexes of radio-labelled and native unlabelled antigen. At equilibrium, the complexes formed are separated from the unbound antigen with a resulting ratio between these two. The bound/free antigen ratio is dependent on the amount of native antigen present in the sample as the radio-labelled is always added at a stable known concentration [72]. Therefore, this technique relies on the antiserum used, which has to provide an appropriate specificity in order to detect the antigen but no other analogues, and a proper affinity to do so in the range of interest.

The use of different antisera across all the CGRP-measuring studies based on RIA is a main source of variability among articles (Table 1). Works employing the same protocol, antiserum, and sample source usually have similar peptide concentrations [14, 39, 47], with some exceptions [48], while the use of different brands containing different antiseras and protocols show differing concentration ranges even when performed with same sample source [15, 39, 63, 64], and even if they were done by the same specialist technician with the same samples [48]. Another problem is that even though studies with the exact same quantification method obtain similar concentration ranges they arrive to clashing conclusions, such as the presence of differences in CGRP concentrations between interictal migraine patients and healthy controls [17, 65].

ELISA technique first appears to be used to determine CGRP concentration in migraine patients in 2007 [21]. ELISA is an immunological assay based on the interaction between the antigen and a primary antibody against the antigen of interest. These will interact, forming a complex that is later confirmed through the enzyme-linked antibody catalysis of an added substrate, which can be quantitatively measured using readouts from either a luminometer or a spectrophotometer. ELISA techniques are broadly classified into direct, indirect, sandwich, and competitive ELISA. For CGRP determinations only competitive and sandwich ELISA have been employed. Competitive ELISA involves a competition between the sample antigen and the plate-coated antigen for the primary antibody, followed by the binding of enzyme-linked secondary antibodies (Fig. 5). Sandwich ELISA technique includes a sample antigen introduced to the antibody-precoated plate, followed by sequential binding of detection and enzyme-linked secondary antibodies to the recognition sites on the antigen (Fig. 6) [73]. In both cases, and similarly to what has been pointed out for RIA, the techniques rely on the specificity and sensitivity of the antibodies included in the kit. This is the reason why ELISA-based studies are also subjected to the exact same issues associated with RIA-based works. As it has been described, investigations using the same brand also reports similar peptide concentration ranges [25, 26, 30, 44, 49, 67], even though this is not always the case [32], but, most importantly, those using different kits clash in the range of concentrations [23, 61, 62] on top of the conclusions drawn [33, 61]. For this point we need to explain that kits from USCN Life Sciences and Cloud Clone Corp., and from Peninsula Laboratories and BMA Biomedicals have been considered as only two brands since these companies have merged or have been acquired by the other at some point in their history. Moreover, and this last point serves as an example, there is a lack of information by part of the researchers regarding the kits used, because sometimes the brand cited offers more than one kit or two different brands over the history have been in charge of its production, and with the given information it cannot be inferred which one it was [27, 34]. This could be the reason why across studies using kits from the same brand they obtained different concentrations. Also, this lack often comes from the manufacturers, which most of the times do not report essential information to the user such as the specific epitope recognised by the antibodies or their cross-reactivity for analogues of CGRP. This has caused some controversies such as works employing kits specifically designed, according to the manufacturer, for the detection of beta-CGRP reporting results as total-CGRP [35, 36, 59] without proving in their papers whether the technique recognises alpha, beta, or total-CGRP.

Schematic representation of a competitive ELISA protocol

Schematic representation of a sandwich ELISA protocol

CGRP has been analysed in a broad number of samples sources including plasma and serum from the peripheral circulation and jugular vein, CSF, saliva, tear fluid and GCF. Due to the enormous variability of concentrations found within the sources (Table 1) and the fact that results are not homogenous even when the same technique and sample source were used, we thought the comparison between sample sources did not make sense.

Nonetheless, and because our group has focused on the determinations in serum with ELISA, we have done a specific analysis of the studies matching these two criteria. There seems to be a consensus range achieved by most of the studies, independently of the brand employed, and which approximately goes from 15 to 150 pg/mL for total and alpha-CGRP, because the data from the literature exhibits that most of the measured CGRP is the alpha isoform, and from 2 to 10 pg/mL for beta-CGRP.

Because there are examples of different works employing the same method, specific technique, sample source and similar inclusion/exclusion criteria whose results are contradictory [14, 33, 47, 59], we cannot conclude that all the problematic with CGRP measurements is related to the quantifying method and/or the sample chose by the authors. There has to be other factors playing a role in the discrepancies, such as fluctuations with the circadian [74] or with the menstrual cycles [58, 59, 75], effect of resting/exercise [76, 77], fast degradation of the peptide due to its short half-life [78], long-term storage stability [55], migraine and other comorbidities [69, 71, 79,80,81], and the effects of pharmacological treatments [26, 28, 38,39,40,41,42,43,44]. From our review we could not analyse these parameters, because they were not displayed with enough accuracy in most articles.

Here, in an effort to provide more detailed information about the suitability of serum from peripheral blood for CGRP determinations, we carried out a series of experiments in order to shed light on some of the main questions regarding the lack of consistency with CGRP quantifications beyond the data already discussed from our review.

We have found that the specific ELISA kit employed has a crucial effect on the CGRP measurements, showing completely different concentration ranges depending on the reference.

Besides the differences in the range we have obtained some alarming results. One of the kits assayed, from Biorbyt, did not meet the reproducibility criteria, which automatically should make this kit unsuitable for any kind of research. On top of that it did not conserve the linearity when diluting the samples which adds more doubts to its reliability. The one from BMA Biomedicals, even though a kit from this brand has been used for a published work when the company had the name Peninsula Laboratories [24], showed for 4 different times results below the detection limit (20pg/mL), contradicting the data of the cited article. Once again, these data call for a more exhaustive description of the methodology, not only by the researchers but also the companies.

The other two kits assayed fulfilled all the quality requirements and presented a range of measurements which fit the range observed in studies using the same sample source. Because the kit from CUSABIO is specific for beta-CGRP we have considered that the objective range for this kind of determination is different to the range for the Abbexa kit, which detects alpha-CGRP. This comes with no surprise because notwithstanding we have not displayed it, in our previous works the exact internal validations were performed and our researches already shown that these kits were reliable and were in accordance with the results published in the past by other groups [25,26,27, 31, 35, 36, 49].

Overall, the analysis of the kits performed here acts as a probe that the determination methodology needs to be carefully assayed and critically analysed as this is the ultimate guarantee of the validity of the data. Because we have already done so with the 4 kit references listed in this study, we would like to encourage researchers to share their internal validation data with other kits they might have been using, as well as to invite the companies to share more details about their products, as we believe it has been a huge limitation in the field and this would produce a significant advance, saving a lot of time and money to the research.

Throughout the literature many different studies have acknowledged the reported short half-life of CGRP [38, 52, 55, 56, 82] as a main limitation for their works. Still, many fail to describe precisely enough their methodology for sample processing so readers can infer how this limitation took place. This problem has been pointed out before and the latest works have included a more accurate description of the sample processing [35, 36, 38, 55]. To avoid this rapid degradation of the peptide Messlinger et al. [55] proposed buffering the sample with peptidase inhibitor, but they concluded that immediate freezing was the most effective way to preserve CGRP content.

We did not add peptidase inhibitors as we were using serum as sample and the addition of a peptidase inhibitor needed to be done right after centrifugation but we opt to freeze the samples immediately. Our results show that the degradation of the peptide did not happen, at least in the first 24 h, when samples were stored at 4°C. This complies with the instructions of most of the ELISA kits our group has assayed and which provide a window of time for sample storage depending on the temperature, specifying that samples can be stored at 4°C for up to 24 h before being analysed. These data appear to be contradicting the results of Kraenzlin et al. [78] regarding the half-life of CGRP. One could argue that the content of serum and plasma is different and the differences found in these studies could be accounted for the binding of CGRP to cellular compartments or to fibrinogen, effectively modifying its degradation. However, the cited article, performed in 1985, is not exempt from limitations and should be reconsidered when analysing the stability of the peptide, at least in isolated biological fluids. First, this pharmacokinetic (PK) study fails to achieve some critical points that are currently required for this kind of works [83]. CGRP concentration should achieve a steady state in order to be able to extrapolate the half-life as it at this point when the phenomena of absorption, distribution, metabolism and excretion have reached and equilibrium and therefore stopping the infusion will give the information about the actual elimination half-life. Moreover, results from human in-vivo PK studies are not necessarily equivalent to those obtained from in-vitro or animal models in-vivo [