Cluster headache and risk of chronic transformation

Excerpt

Cluster headache (CH) is characterized by excruciatingly painful attacks [1] and when it has a chronic course (chronic CH—CCH) provokes deep negative changes in patients’ lives. CCH may arise de novo from the onset of CH or may take the place of a pre-existing episodic CH (ECH). Episodic CH patients are scared to become CCH. The chronic CH form accounts for about 10% of cases, ranging from 5 to 20% [2]. Clinic-based case series have tried to evaluate the risk of developing a chronic form after an episodic onset, with estimated risk ranging from 2 to 19% [2]. Differences in patients’ selection, length of follow-up, statistical methods, and sample size can explain the discrepancy. Survival analysis comprises a set of statistical methods used to investigate the time until occurrence of an event of interest. The distinguishing feature of survival data is that at the end of the observation period, the event will probably not have occurred for all the patients. For these patients, the data is incomplete, and the time is said to be censored. We do not know when or whether the patient will experience the event, only that she/he has not done so by the end of the observation. The Kaplan-Meier (KM) method is a non-parametric method used to estimate the survival probability from observed survival times in the presence of censored cases. Another instrument in survival analysis, the log-rank test, tests the null hypothesis that survival curves of two or more groups do not differ, considering the whole follow-up period. In the present study, we estimated the probability of ECH to transform in CCH within 5, 10, 20, and 30 years of disease onset; then, we questioned if sex and age at onset (clearly definable, invariable, and easily accessible factors) could influence the transformation. Other proposed unstable risk factors for transformation from ECH to CCH include smoking and drinking habits, head injuries, high frequency of clusters, and shorter duration of remission periods [3]; people can start or stop smoking or drinking, or have multiple head injuries over the disease course and therefore these factors are not suitable for the KM method. Also, they are sometimes difficult to define or measure, or they are not suitable for long-term prediction (e.g., changes in cluster frequency). We analyzed the clinical documentation of all patients with a diagnosis of CH visited between 1/1/2011 and 19/3/2020 at the Headache Center of the Foundation IRCCS Neurological Institute C. Besta in Milan. This is a tertiary national referral center for cluster headache and related disorders. Clinical records of 539 CH patients were analyzed. Episodic CH were 357 (66.2%) and CCH were 182 (33.8%); 449 (83.3%) were male. CH course were ECH 334 (62%), pCCH 82 (15.2%), sCCH 97 (18%), and unknown 26 (4.8%) (of which 13 with disease < 1 year) (Table 1). The following were extracted from the clinical records: diagnosis at last observation, sex, age at onset of disease, disease course, and year of onset of chronic form. The disease course was defined as follows: (i) ECH, episodic form throughout all the disease; (ii) pCCH (primary CCH), chronic since the onset of disease; (iii) sCCH (secondary CCH), CCH developing after an episodic onset; (iv) unknown, disease length too short (less than 1 year) to define the clinical course, or clinical course not stated in the available documentation. First, we computed the disease length (last observation—onset) and the chronification latency (year of onset of chronic form—disease onset). Subjects with CH course pCCH or unknown, subjects with missing disease length and sCCH without latency were excluded. For robustness, the longest period of follow-up was determined by the disease length achieved by at least 10% of the sample. The state of the CH course (ECH or sCCH) was determined at timepoints 5, 10, 20, and 30 years. Next, we calculated the survival time at the different timepoints: if the state was sCCH (the patient has become chronic within the time period), then the survival corresponded to the chronification latency; if the state was ECH (the patient is still episodic at the end of the time period or hasn’t reached the timepoint), then the survival corresponded to the disease length (maximum = time period). The KM method was used to estimate the survival probability (event = sCCH) at 5, 10, 20, and 30 years. We compared the survival between sexes at 5, 10, and 20 years with the log-rank test. After binning of age at onset into decades (< 20, 20–29, 30–39, 40–49, 50–59, 60+), we compared the survival between decades of age at onset at 5 years with the log-rank test, with overall comparison and test for linear trend. All analyses were conducted in SPSS v25. Cases viable for survival analysis were 349 (261 ECH and 88 sCCH): 73 ECH and 9 sCCH had missing data. The cumulative proportion developing sCCH was 9.3%, 18.1%, 24.5%, and 36% at 5, 10, 20, and 30 years. In our study, age at onset affected the evolution of CH: patients with a later onset of CH had a higher risk of becoming chronic within 5 years (p, 0.015 for overall comparison and p, 0.05 for linear trend at log-rank test). Sex did not affect the survival distribution at 5, 10, and 20 years (p > 0.09 at all timepoints). The provided estimated risk of developing CCH at different disease durations may be of help when delivering full information on the disease course to CH patients and their relatives as well as when designing clinical studies. …