Two small open studies reported an improvement of the headache in migraine patients using an angiotensin-converting enzyme (ACE) inhibitor 12. Indirectly, a beneficial effect of angiotensin II receptor blockers (ARB's) on headache is shown in a meta-analysis on side effects reported in placebo controlled trials including over 12 000 patients 3. Two randomized, placebo controlled studies conducted by our research group have evidence for efficacy of an ACE inhibitor (lisinopril) and an ARB (candesartan) in migraine prophylaxis 45. This and other evidence points in the direction of involvement of the renin-angiotensin system (RAS) in migraine pathophysiology. (For further discussion on possible mechanisms see reference 6).

The human angiotensin converting enzyme (ACE) gene consists of either an insertion (I) allele or a deletion (D) allele forming three possible genotypes: II, ID or DD. Many studies have suggested an association between the ACE-D allele and cardiovascular diseases 7. For migraine an Italian (Paterna) 8, an Australian (Lea) 9, and a Japanese (Kowa) 10 study has demonstrated different results regarding whether an association between the ACE polymorphisms and this condition exists (Table 1).

The objectives of the present study were two-fold. Firstly we wanted to examine whether a beneficial effect in the above mentioned migraine prophylactic studies 45 could be predicted by ACE genotype, a question that has also been raised in a recent publication 11. Secondly we wanted to investigate the ACE genotype as a possible risk factor for migraine with (MwA) and without (MoA) aura in a Norwegian population.

Included in the study were 347 migraine patients aged 18–68 (155 MwA, 187 MoA and 5 missing aura subgroup data, based on ICHD-2 criteria 12) and 403 healthy non-migrainous controls > 40 years of age. The migraineurs were recruited partly from the lisinopril 4 (n = 49) and candesartan 5 (n = 59) studies, and the remaining group (n = 239) from the outpatient clinic of the Department of Neurology, Trondheim University Hospital. The patients and the controls were recruited from the same area and only subjects with Nordic ethnic background were included. The diagnosis was confirmed by an experienced clinical neurologist. Responder status in the candesartan and lisinopril studies was defined as a reduction in days with headache of at least 50% in the treatment period compared to the placebo period. Non-responders were the subjects not defined as responders and with both genotype and response data available. No patients were included in both the lisinopril and candesartan studies. The control group was recruited in collaboration with the Department of Immunology and Transfusion Medicine and criteria for inclusion were no present or former history of migraine or other types of chronic headaches, no history of epilepsy or of hypertension in need of medical treatment, and age > 40 years (since status as "non-migraineur" cannot be determined with relative certainty before this age). No direct interview was made in the control group, but the participants filled out a questionnaire to determine eligibility for participation. In addition to not having migraine the control group was required to have no other headache condition and less than one headache day per month.

The migraine group had a mean age of 41 years (standard deviation (SD): ± 12 years) and consisted of 268 women and 79 men. Median age of migraine onset was 16 years and median attack frequency was 4.0 attacks per month. In the control group with 233 women and 170 men, mean age was 50 years (SD: ± 7 years).

311 of the samples were genotyped by AstraZeneca, R&D Genetics, UK, and 439 samples were genotyped at the Department of Neurology, Trondheim University Hospital, Norway.

The observed genotypes in the control population did not deviate significantly from the Hardy-Weinberg equilibrium (p = 0.98). With regard to the genotype and allele distributions, no significant differences between migraine patients and controls were detected, even though the ACE-D allele tended to be more frequent (p = 0.058) among responders than non-responders in the candesartan group (Table 2). Furthermore, there was no significant difference between the controls and the MwA or MoA subgroups, nor between responders and non-responders to lisinopril and candesartan, and no difference was detected when stratifying by sex. Within the migraine group differences in genotype could not explain the presence of aura (n = 342, missing data = 5, p = 0.64), of coexisting tension-type headache among migraineurs (n = 343, missing data = 4, p = 1.0), differences in age of debut (n = 342, missing = 5, p = 0.69) or frequency of migraine (n = 342, missing = 5, p = 0.52) or in headache frequency as recorded in the placebo period in the candesartan study (n = 56, missing = 3, p = 0.77).

Frequencies of the genotypes and alleles for the different studies are presented in Table 1. There are large differences in genotypes and alleles among the controls. E.g. the II genotype varies between 12.4 and 41.5 and the D-allele between 35.5 and 62.4%.

In the present Norwegian sample, there is no difference in ACE genotype or allele frequency in a migraine group compared to a control group. Associations between ACE polymorphism and migraine reported in other studies are not consistent and have been detected in different diagnostic or sex categories. The results of these studies are shown in Table 1. In addition a recently published study from Taiwan found no differences in ACE allelic frequencies between migraine patients and controls, but stratified by gender the DD frequency was significantly lower in male migraineurs than controls (not included in Table 1 because only the abstract was published in the English language) 16. Findings that the DD genotype is more frequent in MoA 8 and MwA 10 or less frequent in male migraineurs 16 are not supported by our data. Our population which is the largest to date used to study the relationship between ACE polymorphism and migraine (MoA and MwA) should have >80% power to detect an association of the same magnitude as in the study by Paterna et al 8. Our study also did not find a relationship between ACE genotype and response to prophylactic drugs influencing the RAS. The allele frequency in the responders versus non-responders in the candesartan group had a p-value of 0.058 and with low numbered groups the risk of a false negative result is present.

The purpose of looking at the association between responders in the two clinical trials and ACE genotype was not to detect a small theoretical association, in which case this subgroup analyses would be underpowered, but to see whether there was an association so strong that it would be valuable in clinical use predicting response in migraine – prophylactic treatment. That is whether it would be clinically beneficial to use ACE genotype to predict whether the patient would respond to the drug or not. Our opinion is that in order for an association to be clinical valuable at least 75–100% of responders to a migraine-prophylactic drug should have a specific ACE-genotype. The power of our study to measure a percentage of 75% DD among the candesartan responders with the control population as reference is > 80%.

A limitation of the study is that the control group was not directly interviewed increasing the risk of migraineurs self-reporting themselves as non-migraineurs and thereby increasing the risk of type II errors. In order to minimize this problem participants in the control group were required to have no other headache condition and less than one headache day per month.

Population stratification refers to differences in allele frequencies between cases and controls due to systematical differences in ancestry rather than in the association of genes with disease 1718. There are large differences in the frequencies of the ACE-alleles in different populations (Table 1). Hence, due to the problem with population stratification we did not find it meaningful to perform a statistical analysis of the merged data of all these studies. This might have been misleading also because the way diagnosis were made, both of migraineurs and controls, may differ somewhat between the studies.

There was no difference in ACE genotype distribution between a migraine and a control population in our material. Our study also indicates that ACE genotyping will not be a valuable tool for predicting clinical response of drugs influencing the angiotensin system in headache treatment. It is, however, important that these findings should be confirmed in other studies with more patients and among different ethnic groups.

RAS, renin-angiotensin system; ACE, angiotensin converting enzyme; MoA, migraine without aura; MwA, migraine with aura; ARB's, angiotensin II receptor blockers.

Co-authors Amanda Gladwin and Katryn Owen are AstraZeneca staff.

No disclosures from the rest of the authors.

ET, LJS, GB, LRW and HS were involved in designing the study. AG, KO and LRW were responsible for the genotyping. ET and LJS conducted the statistical analyses. All authors were involved in either drafting the manuscript or revising it.