BACKGROUND AND PURPOSE: The clinical characteristics of cluster headache (CH) are based mainly on retrospective attack descriptions of 'usual' attacks, but whether these reports are reliable is uncertain. The aim was to compare retrospective and prospective attack descriptions and describe the within- and between-patient variability of attacks.

METHOD: Fifty-seven CH patients underwent a semi-structured interview obtaining a retrospective account of usual CH attacks. Patients thereafter prospectively recorded the clinical characteristics of up to 10 attacks per patient in a headache diary. Four different attack characteristics were investigated: (i) severity, (ii) duration, (iii) number of autonomic symptoms and (iv) number of migrainous symptoms. Retrospective and prospective data were compared. Within- and between-patient variability of attacks was assessed.

RESULTS: Retrospective attack descriptions (n = 57) were significantly longer (P = 0.046) and more severe (P < 0.0001) for untreated attacks compared with prospective reports (n = 500). The number of autonomic symptoms was significantly higher in the retrospective reports compared to the prospective reports (P < 0.0001). Within-patient variability for attack duration, pain severity and number of autonomic and migrainous symptoms was low. Compared to men, more women reported longer (P = 0.026) and more severe (P = 0.028) attacks with more migrainous symptoms (P = 0.033).

CONCLUSIONS: Important differences were found between prospectively and retrospectively reported attacks with duration and severity of untreated attacks overestimated in retrospective attack descriptions. CH attacks display low within-patient variability, but the presentation of CH attacks varies between patients. The high prevalence of symptoms typically associated with migraine should raise more diagnostic awareness for CH, especially in women who are more often misdiagnosed as having migraine.