Abstract
Chronic daily headache (CDH), when defined as ≥ 15 headache days per month, affects 3-5% of the adult population. Major life changes are putative precipitating events for onset of chronic pain, including chronic headache. This study compared the occurrence of specific life events between CDH cases and episodic headache controls in a community sample. CDH cases (180+ headache days per year: n = 206) and episodic headache controls (2-104 headache days per year: n = 507) were identified from a randomly selected adult US population. Subjects were interviewed about the occurrence of certain major life changes or events (change of residence, employment status, marital status, related to their children, deaths of relatives or close friends, and ‘extremely stressful’ ongoing situations) occurring in a defined time period. Events that occurred during the same year or year before frequent headache onset in cases or in an equivalent time period in controls were considered to be antecedent events. Those that occurred after this time were considered subsequent events. Compared with episodic headache controls, CDH cases had more major life changes in the year before or same year as CDH onset. After adjusting for age, gender, headache type and year of event, the odds of CDH increased additionally with each antecedent event [odds ratio (OR) 1.20 (1.1, 1.3), P < 0.001], but not with subsequent events [OR 0.94 (0.8, 1.1), P < 0.4]. In secondary analyses, the association between antecedent events and CDH was significant only for the approximately half of CDH cases who were aged ≥ 40 years [OR 1.33 (1.2, 1.50) vs. OR 1.04 (0.9, 1.2), P < 0.05 for interaction by age]. These results suggest that major life changes are associated with the onset of chronic daily headache, particularly in middle age.
Introduction
Chronic daily headache (CDH) has been defined by convention as headache occurring on ≥ 15 days per month (1, 2). Whereas the prevalence of migraine and, to a lesser extent, tension-type headache (TTH) decreases with age, CDH appears to affect roughly 3–5% of the general population throughout early to late adulthood (1). However, the CDH population thus defined is a highly fluid one, with high rates of incidence and remission (3–7).
Primary forms of CDH can be subdivided into headaches of short duration (< 4 h per attack) and long duration (≥ 4 h per attack) (2). Primary CDH of long duration includes four disorders: transformed migraine, chronic TTH, hemicrania continua, and new daily persistent headache (2). The second edition of the International Classification of Headache Disorders (ICHD-2) includes variants of all four of these disorders, but uses the term chronic migraine to describe a disorder with ≥ 15 days of migraine per month (8). The classification of the very frequent headache disorders, and the CDH phenotype itself, is a subject of ongoing discussion and research, and these criteria may undergo further modification (9, 10).
Prior reports suggest that stressful or traumatic life events or major life changes are associated with chronic headache (11–21) and other types of chronic pain (22–25). For chronic headache, most studies have been based on clinical or other selected samples. Two population-based studies that considered this question did not distinguish between life events that occurred prior to CDH onset and life events that occurred after CDH onset (26, 27).
In this study, we have used data from the Frequent Headache Epidemiology Study to assess the potential role of major life events that immediately preceded or followed the onset of CDH. We compared the prevalence of specific major life changes in the period before CDH onset in CDH cases with an equivalent time period in an episodic headache group that is ‘at-risk’ for CDH.
Methods
The results presented here are from a larger nested case–control study (the Frequent Headache Epidemiology Study). The aims of this study were to (i) describe the short-term natural history of CDH in a non-clinical population, and (ii) determine whether certain exposures hypothesized to be associated with CDH in clinical samples were associated with CDH in the general population. Major life changes were a primary exposure of interest. Prior risk factors considered in this study population included habitual snoring, caffeine consumption and head or neck injury (3, 28–30).
The study population was nested within a cohort of adults who had participated in a general health telephone survey. Households in this survey were selected from a randomly ordered list of telephone numbers from the Atlanta, GA, Baltimore, MD, or Philadelphia, PA metropolitan areas. Participants were 18–65 years old. Potential cases with 180+ headaches per year (n = 2757) and potential controls with two to 104 headaches per year (n = 53 222) were identified from survey participants. Approximately 1 year after the health survey, efforts were made to contact all of the potential cases and an approximate 4% random sample of the potential controls (n = 2033) for a 30–40-min telephone interview on risk factors for CDH, including major life events (described below). Potential cases were eligible if they reported 180+ headache days per year at follow-up and CDH duration of ≤ 5 years. Potential controls were eligible if they reported two to 104 headache days per year at follow-up.
Recruitment results
Approximately half of the potential cases (49%) and controls (51%) were not available for the follow-up interview (i.e. invalid phone number, moved, or unreachable after 10 attempts). No difference in average headache frequency was observed between those who could not be contacted compared with participants who were contacted, and contact status was similar between cases and controls by age, race and marital status. However, participation was better with increased educational level for the controls but not the cases. An additional 10% of potential controls and 7% of potential cases either refused participation or were not eligible due to age, incompatible schedule, or inability to complete the interview for any other reason such as lack of knowledge of English, leaving a sample of 806 potential controls and 1220 potential cases. After screening, a total of 668 controls and 268 cases met eligibility criteria, of whom 507 controls and 206 cases participated in the follow-up interview. (Of the 1220 potential cases contacted at follow-up, 698 were ineligible because they no longer had CDH and, of the remainder, 254 were excluded because their duration of CDH was > 5 years. Of the 806 potential controls, 136 were excluded because their headache frequency was out of the predetermined range.)
Headache classification
Participants were grouped into four mutually exclusive headache groups in accordance with the 1988 International Classification of Headache Disorders (ICDH-1) (31) using methodology similar to previous studies (28, 32). At the first digit level, migraine and TTH have similar definitions in the revised system, ICDH-2, published after these data were collected (8). Individuals with multiple headache types were classified in the following order of precedence: (i) migraine with aura, (ii) migraine without aura, (iii) TTH, and (iv) all other headaches.
Life events
The life events history was a shortened version adapted for telephone of the Life Events Inventory used in the Baltimore Epidemiologic Catchment 1993 follow-up study (33). Subjects were asked about the occurrence of six major life events in reference to a defined time period. This time period started at the ‘recall year’ (defined below) up to the time of the interview. For the cases, the recall year was defined as the year before CDH onset. Controls were assigned a random recall year in the same range as the cases (e.g. ranging from 1 to 5 years before the interview). In order to improve recall, subjects were asked (not immediately before the life events interview) to recall one or two ‘landmarks’ (e.g. moving, vacations, starting or finishing school, changes in relationships, local or national events) that occurred during this year.
The life events interview focused on six types of life changes: work changes (e.g. changing jobs, retiring, finishing or starting school, getting fired or laid off, leaving a job due to health problems, or other important change), relationship changes (e.g. getting married, divorced, separated, widowed, or a similar change in an important relationship), major changes with children (e.g. having or adopting a child, having a child move in or move out of the home, having a serious problem or other major event with one's children), changes in residence, deaths of family members or close friends, and ongoing self-defined ‘extremely stressful situations’ (e.g. financial problems, ongoing illness in a close family member or friend, ongoing illness in self, or an ongoing abusive relationship). Respondents were asked, for each of the six event types, if this event had occurred to them since the recall year and, if so, the year(s) of occurrence and type of event(s). Because subjects were interviewed about up to two events per each of the six event types, details were collected on as many as 12 events.
Depression
Current depression was assessed with the Primary Care Evaluation of Mental Disorders (PRIME-MD) mood module (34). The PRIME-MD mood module is a nine-item questionnaire regarding symptoms occurring ‘nearly every day’ during the last 2 weeks and is based on the Diagnostic and Statistical Manual of Mental Disorders, 4th edn diagnostic criteria for major depressive disorder (35). Screen-positive major depressive disorder in the PRIME-MD is defined as endorsement of five or more symptoms, at least one of which must be anhedonia (‘little interest or pleasure in doing things’) or depressed mood (‘feeling down, depressed, or hopeless’).
Analyses
In the analysis, events were categorized as ‘antecedent’ or ‘subsequent’. Antecedent events were defined as those that occurred the same year or year before CDH onset for the CDH cases or the equivalent 2-year time period for the episodic headache controls. All other events were categorized as ‘subsequent’.
Univariate comparisons between CDH cases and episodic headache controls were made using non-parametric methods (χ2 or the rank sum tests). For our primary aim, logistic regression was used to estimate the risk of CDH associated with each type of life event and total life events, performing separate analyses for antecedent and subsequent events. To account for demographic differences between cases and controls that might be associated with life events, we adjusted analyses for age, gender and primary headache type. We also adjusted for the recall year, on the assumption that the quality of recall varied by the length of the recall period. Other confounders considered included body mass index (BMI), pre-CDH caffeine consumption, habitual snoring, lifetime or recent occurrence of head or neck injury, race, and educational level. In secondary analyses, we considered whether the risk of CDH associated with life events was modified by age (< 40, ≥ 40), gender, and primary headache type (categorized as migraine or non-migraine) by including appropriate interaction terms in the logistic models. The age cut-point of 40 years was chosen, as this was the approximate median age in the study sample.
Finally, we performed three sensitivity analyses. The primary analysis was repeated after excluding controls with 24+ headache days per year (the median headache frequency in controls) because of concern that some controls might be misclassified cases. We also considered whether results were substantially affected by the inclusion of the comparatively subjective category of ongoing ‘extremely stressful situations’ by, first, not including those life events in this category involving an illness in the subject and, second, not including all events in this category. Analyses were performed using Stata (StataCorp. 2003, Stata Statistical Software: Release 8.2, College Station, TX, USA; StataCorp LP). The study design was approved by the Institutional Review Board of the Johns Hopkins School of Public Health. Verbal consent was obtained over the phone according to the Institutional Review Board-approved study protocol.
Results
The study sample consisted of 507 episodic headache controls and 206 CDH cases. Characteristics of the study sample are shown in Table 1. Compared with the episodic headache controls, CDH cases were slightly younger, more likely to be female, have less than a high school education, were more likely to have migraine headaches, and were more likely to screen positive for major depression based on the PRIME-MD (34). The female cases had a higher BMI than the female controls. The average start of the antecedent time period (e.g. the ‘recall year’) was approximately 4 years (range 2–6) before the interview, but was about 4 months more recent on average for CDH cases than for episodic headache controls (3.7 vs. 4.0 years, P < 0.01).
Characteristics of CDH cases and episodic headache controls
Wilcoxon rank sum test or χ2 test.
CDH, chronic daily headache; BMI, body mass index.
Approximately one-third (37%) of the study sample reported changing their residence one or more times during the total recall period. One or more changes in marital status or in an important relationship were reported by 25% of subjects, with the most common relationship change being a break-up (42%), getting married (16%), other unspecified changes (15%), separation (15%) and divorce (9%). About half (52%) of those with children reported one or more major changes with their children, including a child moving out (22%), a child being born (20%), an illness in a child (10%) and other unspecified changes (39%). Changes in work or school status were reported by 59%, with the largest category being a job change (44%), unspecified other work changes (20%), starting school (8%), being fired or laid off (7%), or quitting (7%). Deaths in the family or of close friends were reported by 62%, with the most common loss being that of a close friend (29%), other relative (21%), aunt/uncle (20%) and grandparent (14%). Half (52%) reported an extremely stressful ongoing situation, with the most common situations being financial (31%), other unspecified changes (29%), illness in someone else (28%), illness in self (9%), or an ongoing abusive relationship (4%). The median (mean) total number of life events (both antecedent and subsequent), adjusted for recall year, was 3.4 (3.6) for controls and 3.9 (4.2) for cases; mean adjusted antecedent/subsequent events for controls were 2.1/1.5 and for cases 2.8/1.5.
After adjusting for the recall year only, the odds of CDH increased additionally with each life event in the antecedent time period [odds ratio (OR) 1.21 (1.1, 1.3), P < 0.001], but not for the subsequent recall period [OR 1.01 (0.9, 1.2), P < 0.9]. Results were similar after adjustment for age, gender and primary headache type (Table 2—overall) and remained similar when the controls with more frequent headaches were excluded [OR 1.26 (1.1, 1.4), P < 0.001], when the category of ‘extremely stressful situations’ involving an illness in the subject was excluded [OR 1.20 (1.1, 1.3), P < 0.001] and when all ‘extremely stressful situations’ were excluded [OR 1.18 (1.1, 1.3), P < 0.002]. Further adjustment for educational level, race, current BMI, pre-CDH caffeine consumption, head or neck injury, and habitual snoring slightly attenuated the results [antecedent events OR 1.17 (1.0, 1.3), P < 0.005; subsequent events OR 0.97 (0.8–1.1)]. The calculated attributable fraction for life events was 27% (12–40%). The attributable fraction is a statistical concept which, in this analysis, represents the theoretical proportion of CDH cases due to antecedent life events assuming causality.
Adjusted odds of chronic daily headache (CDH) associated with each additional major life event in the pre-CDH (antecedent) or post-CDH (subsequent) time period
P < 0.017 for interaction by age.
Odds ratios (ORs) estimate the odds of CDH associated with each additional antecedent or subsequent life event, after adjusting for age, gender, recall year and headache type. ORs are estimated overall and with interaction for age (< 40, ≥ 40 years), gender and headache type (migraine, non-migraine).
In secondary stratified analyses, the association between antecedent life events and CDH was similar for men and women and for those with migraine and non-migraine headaches (Table 2). However, there was interaction by age, with a stronger association for those aged ≥ 40 years (Table 2, Fig. 1), in whom the attributable fraction approached 50% [47% (25–63%)]. The association between number of antecedent life events and CDH is illustrated in Fig. 1a (adjusted for recall year only) and in Fig. 1b (adjusted for recall year, age, gender, primary headache type), and is shown separately by age groups. The association between life events and CDH, when stratified by age, was not significant for the younger group (Fig. 1a,b, Table 2), and there was the appearance of a threshold effect at four or more antecedent events for the older group.
(a) Minimally adjusted relative odds of chronic daily headache (CDH) by number of antecedent life events (adjusted for recall year). (b) Adjusted relative odds of CDH by number of antecedent life events (adjusted for age, gender, primary headache type, recall year). (A) Overall. (B) Age < 40 years. (C) Age ≥ 40 years. ∗P < 0.05.
Among the types of events measured, the strongest predictor was an ongoing ‘extremely stressful situation’ with an overall OR of 1.8 (1.4, 1.2) (Supplementary Table S1a), although results overall remained similar after excluding this category (above). Changes in work status were associated with increased odds of CDH after the age of 40 and decreased odds before age 40 [Supplementary Table S1a: age ≥ 40 OR = 1.75 (1.2, 2.5) vs. age < 40 OR = 0.65 (0.5, 0.9), P < 0.001 for interaction by age]. There were few differences in the occurrence of life changes subsequent to CDH onset and no significant interaction by age (supplementary Table S1b). However, CDH cases were less likely than controls to report the death of a close friend or relative [0.66 (0.5, 0.9)].
Discussion
In this community sample, individuals with CDH reported more major life events in the time prior to the onset of CDH than those with episodic headaches in an equivalent 2-year time period. No difference was seen for total life events in the period subsequent to CDH onset. Prior CDH risk factors measured in this cohort include habitual snoring, obesity, head or neck injury and caffeine consumption (3, 28–30). We note that the relationship between antecedent major life events and CDH remained significant, although attenuated, after adjusting for these other factors.
The effect of major life events on CDH onset was similar for men and women and for those with migraine and non-migraine headaches. However, when stratified by age (< 40, ≥ 40 years), the effect of major life events on CDH was significant only for the older group. There was the appearance of a threshold effect in the older group—with the effect evident for four or more antecedent events. One example of the differential response by age to life events was seen in the effect of changes in work status; this particular event was associated with increased odds of CDH after the age of 40 and decreased odds before age 40.
Life changes have been considered as a precipitant of chronic headache in clinical or other selected populations (11–21), although population-based data are scarce (26, 27). In a large school-based study from Taiwan (26), adolescent students with and without CDH were interviewed about the presence of childhood stressors using the Global Family Environment Scale (GFES) (36). The GFES measures negative family environmental factors and adverse family events such as parental absence, parental separation or divorce, or abuse. The time period over which events were assessed is not specified, but presumably includes events occurring both before and after CDH onset. Results showed that both physical abuse and parental divorce were more common in the families of the adolescents with CDH than the control group, with the CDH subjects overall having a 10% higher GFES.
In an earlier study based on a large Dutch population (27), the occurrence of life events in the year before the study was compared between a group with current frequent headache (defined as headache at least weekly) and a control group. Life events were measured using Paykel's life events scale (37). Results showed that both major and minor life events were more common in the frequent headache cases than in controls. In contrast to the results presented herein, the effect was evident only for the younger (< 50 years) age group. Differences in the age effect with our study may be explained by use of different life event history tools, case definition (i.e. headaches at least every other day in our study vs. at least weekly in the Dutch study), or classification and temporality of life events (i.e. life events occurring no more than 5 years before the interview in our study, stratified by pre- and post-CDH vs. all life events occurring within 1 year before the interview in the Dutch study). It is likely that the life events in the Dutch study were primarily subsequent events in our nomenclature.
The pathogenesis of CDH is not well understood, but is likely to be multifactorial and complex (38). Other risk factors for CDH have been described in this cohort and elsewhere (1), and these factors may operate independently or synergistically. It is likely that the importance of these factors varies through the life span due to different opportunity for exposure and changes in the resilience to such exposure. In particular, the likelihood of experiencing a traumatic or major life event varies considerably throughout the life span (39–41). Post-traumatic stress disorder (PTSD), although not assessed in this study, may be broadly relevant to the discussion of the findings presented here. PTSD, characterized by ‘re-experiencing the traumatic event’, ‘avoidance and numbing’ and ‘increased arousal’ (35), occurs in 15–24% of individuals after exposure to a traumatic event (40). Headache and other chronic pain disorders, although not a symptom of PTSD, commonly co-occur with PTSD (42–44). Why some individuals are at increased or decreased risk of developing PTSD after traumatic events is uncertain and is an area of active research (40, 43, 45–47). Endogenous factors that are associated with—and may mediate susceptibility to—PTSD after exposure to traumatic events include circulating levels of stress hormones (catecholamines) (45–48), depression (40, 48), and genetic variants that regulate serotonin or the likelihood of developing depression post trauma (49, 50). Perhaps these mechanisms mediate the response to less extreme events. The epidemiological literature is uncertain as to whether increased age is associated with greater or lesser risk of developing PTSD post trauma (39, 51). We note that our finding of increased risk of CDH associated with life events after age 40 was a secondary finding and needs to be replicated in other samples.
Psychological factors that may moderate the response to major life changes or traumatic life events include appraisal or meaning of an event, individual hardiness and self-efficacy, emotional well being, co-morbid depression or anxiety, and social support, among other factors. Self-efficacy has been defined as ‘the belief in one's capabilities to organize and execute the courses of action required to manage prospective situations’ (52). According to the moderating effect of self-efficacy noted by Marlowe (53, 54), it is possible that, for example, people who lose their jobs at an older age feel less control over their ability to secure a new job or successfully navigate a major change than a younger person. Donahue has reported on the relationship between involuntary or forced retirement and subsequent physical and mental health (55). Using data from the first 3 Waves (1992, 1994 and 1996) of the Health and Retirement Survey, a national, longitudinal, US survey, the author analysed data on 1069 individuals who had been employed during Wave 1 (1992) and retired at Wave 3 (1996); 308 of the subjects had been forced to retire and 761 retired voluntarily. The effects of forced retirement on several dimensions of both physical functioning and mental health were negative and statistically significant, even after controlling for baseline health status, sociodemographic and socioeconomic factors.
This study has the limitations of all case–control studies, including potentially differential error in recall of major life events. We attempted to minimize recall error in several ways. We limited our life events history to major life changes that would be presumably well-recalled. Cases were limited to those with CDH duration of ≤ 5 years due to concerns about recall. However, this limitation means that the results reported do not necessarily apply to those with CDH of longer duration. If there was substantial differential error in the recall of dates between cases and controls, this could theoretically have had the effect of exaggerating or attenuating the association between life changes and CDH onset. It is not possible to rule out this type of measurement error. For example, the association could have been exaggerated if CDH cases erroneously dated a life event to coincide with CDH onset or—alternatively—erroneously recalled the date of CDH onset to coincide with a life event. The association could have been attenuated if, for example, CDH cases were more likely than controls to have memory impairment due to medication, depression or other factors. Another limitation is that the major life events we assessed did not themselves comprise a validated measurement instrument. However, the items were selected to represent the more severe life events included in validated life events inventories such as the Social Readjustment Rating Scale (Holmes) (56) and Interview for Recent Life Events (Paykel) (37).
Strengths of this study include our population-based design, which increases the generalizability of the results. Cases and controls were drawn from the same well-defined sampling frame, thus increasing the likelihood that all participants were at the same theoretical risk of exposure to life changes. The finding that life events were more common before, but not after, the onset of CDH supports their role as a potential aetiological factor.
In summary, we have found that CDH cases were more likely than episodic headache controls to have experienced certain major life events in the 2-year period prior to CDH onset. Results further suggest that life events were more important as a CDH risk factor for CDH sufferers age ≥ 40 years.
Footnotes
Acknowledgements
Financial support for this study was provided by GlaxoSmithKline, the Migraine Trust, and the American Headache Society.