Abstract
Though symptomatic medication overuse is believed to play a role in progression from episodic headaches (EH) to chronic daily headaches (CDH), population-based data on this topic are limited. Our objective was to describe patterns of medication use among CDH and EH sufferers in a general population sample. We compared medications used to treat headache in CDH cases and EH controls identified from a large population-based computer-assisted telephone interview survey. CDH began within 5 years of the computer-assisted telephone interview. Questions on medication use focused on treatment prior to the onset of CDH for cases and on an equivalent period in the past for controls. We asked about the likelihood of treating, time waiting to treat, number of different medications used, first, second and third most frequently used headache pain medication, and total treatment days. Questions were also asked about the use of medication for non-headache pain. Current treatment patterns and past treatment patterns were assessed. Likelihood of use of specific medications was compared between CDH cases and EH controls after adjusting for age, sex, primary headache type and number of medications taken to treat pain. Our sample consists of 206 CDH cases and 507 EH controls. CDH subjects were more likely than EH controls to use over-the-counter/caffeine combination products, triptans, opioid compounds and ‘other’ prescription pain medications. Use of aspirin was protective. After adjustment, aspirin and ibuprofen were (negatively) associated with CDH [OR = 0.5 (0.3–0.9), OR = 0.7 (0.5–1.0)] and opioids remained positively associated with CDH [OR = 2.3 (1.3–3.9)]. For past use, CDH was positively associated with over-the-counter/caffeine combination products and opioid compounds and was negatively associated with use of aspirin. Only ibuprofen remained (negatively) associated with CDH after adjustment [OR = 0.6 (0.4–0.9)]. After adjusting for demographic factors, primary headache type and number of medications taken, CDH sufferers are more likely to use opioid-combination analgesics, and less likely to use aspirin or ibuprofen, than EH sufferers.
Introduction
Chronic daily headache (CDH) is a syndrome characterized by primary headaches of long duration (≥4 hours per day) that occur at least every other day (e.g. more than 15 days per month) (1). CDH is usually sub-divided by primary headache type (e.g. chronic migraine or chronic tension-type headache) (1,2). CDH affects 3–4% of adults worldwide (2–5), although the CDH population thus defined is highly fluid with high rates of incidence and remission (6–11). For example, the 1-year incidence of CDH/chronic migraine is about 3% among persons with episodic headache (Frequent Headache Epidemiology Study) (7)) or with episodic migraine (AMPP) (11).
Among the risk factors for CDH, overuse of acute medications intended to relieve headache attacks has received considerable attention (12–16). This syndrome has been referred to, among other terms, as medication overuse headache and medication induced headache. ‘Medication overuse’ has been defined [revised International Classification of Headache Disorders (ICHD)-IIR criteria] (17) as use for at least 3 months of any painkiller above a certain number of days per month dependent on type of medication (e.g. simple analgesics 15+ days/month or ergotamine; combination analgesics, triptans, opioids, or combination of acute medications 10+ days/months). Medication overuse, using various definitions, is evident in over 80% of patients with CDH from specialty care samples (12,15,18) and in about 30% of CDH sufferers from the general population (19).
Although excessive symptomatic medication (ESM) use and CDH are firmly associated, the causal sequence is not clear. The primary support for the belief that ESM may aggravate headache frequency is because patients in subspecialty care with CDH and ESM often improve after abrupt discontinuation of symptomatic medication use (15,20). However, as reviewed elsewhere (21,22), treatment success as described in these primarily uncontrolled studies may be explained in whole or part by regression to the mean or natural history of CDH, other interventions included in the study protocols such as initiation of prophylactic agents, and methodological problems such as excluding from the analysis patients who are unable to comply with medication withdrawal. There is no evidence at this time that limiting use of medication for headache to a certain number of days per week will prevent the development of CDH. Nonetheless, a recent large population-based observational study conducted within a group of episodic migraine sufferers suggested that, at least for specific classes of medications (e.g. opioids and barbiturates), use of these medications to treat episodic headache is associated with increased risk of incident CDH 1 year later (11). Addressing this controversial issue has important clinical and public health implications. If frequency or class of pain medication used to treat headache is a risk factor for CDH incidence, caution and monitoring are required to maintain the use below critical limits; otherwise, guidelines may be liberalized and pain under-treatment may be reduced. Pain under-treatment itself may aggravate headache frequency (23,24).
Observational studies inform this debate by describing the extent to which ESM is found in a less severely affected headache population than is typically seen in specialty care and, further, can consider if specific classes of pain medications are more or less strongly associated with CDH incidence or prevalence. Herein we describe patterns of medical consultation and medication use among episodic and chronic headache sufferers in a general population sample. CDH sufferers had CDH onset within 5 years. CDH cases and episodic headache controls were interviewed about current medication use and medication use in the year prior to the onset of CDH (CDH cases) and an equivalent time period in the past (episodic headache controls).
Methods
Details of the study methodology have been provided in earlier publications (7,25,26). 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 of age. 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 or less. Potential controls were eligible if they reported two to 104 headache days per year at follow-up. More details of the study cohort, including the change in headache status between the first and second interviews, are provided in an earlier publication (7).
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 as a result of 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 more than 5 years. Of the 806 potential controls, 138 were excluded because their headache frequency was out of the pre-determined range.
Headache classification
Participants were grouped into four mutually exclusive headache groups in accordance with the 1988 ICHD-1) (27) using methodology similar to previous studies (2,28). At the first digit level, migraine and tension-type headache have similar definitions in the revised system, ICHD-2, published after these data were collected (29). Individuals with multiple headache types were classified in the following order of precedence: (i) migraine with aura; (ii) migraine without aura; (iii) tension-type headache; and (iv) all other headaches.
Medication use and treatment patterns
Questions on medication use for headache (below) focused on likelihood to treat, time waiting to treat, number of different medications used and first, second and third most frequently used, total treatment days and use of preventives. Questions were also asked about the use of medication for non-headache pain.
Respondents were first asked about the likelihood to treat headaches by asking ‘When you have a headache, how likely are you to take medicine for the pain? Please use a scale from 0 to 10 where 0 means never and 10 means every time you have a headache.’ Time to treatment was assessed by asking ‘When you do take medicine for headache pain, how long do you usually wait until after the pain begins before taking the medicine?’ Respondents were then asked how many different medications they took to treat their headaches in the 3-month period immediately prior to the interview (current use of medication). They were then asked to name the first, second and third most frequently used treatment for headache attacks (acute medications). Finally, they were asked, ‘Overall, on about how many days in the last 3 months did you take any medication for headache pain?’ The interview also assessed the most frequently used (if any) medication taken to treat a non-headache pain condition.
Subjects were then asked the same questions, prompting them to recall their past consumption of these substances during a defined time period (starting at the ‘recall year’) (26). For CDH cases, this time period was the year before CDH onset (defined during the screening interview). Because cases were limited to those with CDH duration of no more than 5 years, this recall period ranged from 1 to 5 years before the interview. For controls, this time period was a randomly assigned year in the same range (e.g. 1 to 5 years before the interview). To facilitate recall, all participants were asked to define a temporal anchoring event by naming two memorable things that occurred during this year.
Statistical analyses
All analyses were performed using STATA (Intercooled Stata 10.0). Summary statistics for the independent variables were calculated for the cases and controls and measures of association were used to look for differences in the distribution of these variables. Pearson chi-square was used to test categorical variables and a non-parametric test (Rank sum test) was used to compare continuous variables. Multivariate regression was used to compare the prevalence (logistic regression) or means (robust linear regression) between cases and controls. Adjustment was made for differences between cases and controls that are plausibly associated with treatment patterns, including age, sex, headache type (migraine, non-migraine) and number of medications taken. Adjustment was also made for the recall year (the year of CDH onset for cases and the randomly assigned year for controls) on the assumption that the quality of recall was related to the recall year.
The study was approved by the Institutional Review Board of The Johns Hopkins University School of Public Health.
Results
Demographic characteristics of study participants
Wilcoxon rank sum test or chi-square test.
Medical consultation and functional impact for chronic daily headache cases and episodic headache (HA) controls
Chi-square or rank sum test.
Treatment patterns for chronic daily headache cases and episodic headache controls
Chi-square or rank sum test.
Predicted means and proportions calculated using robust linear regression (continuous variables) or logistic regression (percentages). Adjusted by age, sex, headache type (migraine, non-migraine), recall year.
Among those using pain medications for non-headache conditions.
Current and past pain medication used for headache and other pain conditions in chronic daily headache cases and episodic headache controls
Odds of CDH in those with any use of indicated medication relative to those without any use of indicated medication. Adjusted ORs are adjusted for age, sex, primary headache type (migraine, non-migraine), recall year and number of medications.
OTC/Caf/Combo, over the counter caffeine-containing combination analgesic; First, second, third choice, percentage of subjects who named the indicated medication as their first, second or third most frequently used medication for headache, limited to those who reported a first (or second or third) most frequently used medication; Other pain, percentage of subjects naming the indicated medication as their most frequently used pain medication for non-headache pain, limited to those who reported using medication for non-headache pain; Any use, percentage of subjects with any use of the indicated medication, limited to those reporting any use of pain medication for headache or other conditions.
The most commonly named first-choice medications for headache were acetaminophen, ibuprofen and over-the-counter (OTC)/caffeine combinations (Table 4). The most commonly named medications for non-headache pain were ibuprofen, acetaminophen and opioid combinations. For current use, in crude comparisons, CDH cases were more likely than EH controls to report any use of OTC/caffeine combination analgesics [OR = 1.9 (1.3–2.9)], triptans [OR = 2.3 (OR = 1.0–5.3)], other prescription medications [OR = 2.5 (1.1–5.7)], or opioids [OR = 3.0 (1.8–5.0)] and were less likely to report use of aspirin [OR = 0.6 (0.3–0.9)]. After adjusting for age, gender, headache type, recall year and number of medications, use of opioids [OR = 2.3 (1.3–3.9)], aspirin [OR = 0.5 (0.3–0.9)] and ibuprofen [OR = 0.7 (0.5–1.0)] was associated with CDH status.
For past use, in crude comparisons, CDH cases were more likely to report use of OTC/caffeine combination analgesics [OR = 1.6 (1.0–2.6)] and opioids [OR = 2.2 (1.1–4.1)], and were less likely to report use of aspirin [OR = 0.6 (0.3–0.9)]. After adjustment, only ibuprofen was (negatively) associated with CDH status.
Discussion
We measured and compared treatment patterns in a population sample of adults with EH and CDH. Medications that were most consistently associated with risk of CDH were opioid-combination analgesics (risk factor) and aspirin (protective). The data further suggest a possible protective effect of ibuprofen. We note that these results refer to any use of the specified medication, not necessarily overuse.
Previous clinical series suggest that frequent use of specific classes of medications is associated with increased risk of subsequent development of CDH, particularly in individuals with a past history of migraine (12–16,30,31). For example, in a cohort study based on a clinical sample, Katsavara et al. showed that frequent acute medication use was a risk factor for incident CDH, even after adjusting for baseline headache frequency (18). Population-based data regarding specific medication use or overuse and CDH are scarce. In a large population-based European cohort study, incident CDH was more than seven times more likely among those with high analgesic exposure in comparison with those without it (OR = 7.5, 95% CI, 6.6–8.5) (32). This result is limited by the fact that CDH status was unknown at baseline; thus, some of the CDH sufferers at follow-up may have already had CDH at the first assessment. This study did not examine the influence of specific classes of medication.
Some results of the present study are consistent with results from the American Migraine Prevalence and Prevention Study (AMPP), a longitudinal population-based study that assessed the influence of medication on the development of CDH in persons with episodic migraine (11). Results of the AMPP suggested that use of opioids or butalbital-containing medications was associated with increased risk of CDH 1 year later. Use of aspirin was not assessed in this study. The magnitude of the odds ratios for opioids and barbiturates was similar to the present study, although herein the findings were not significant for barbiturates. Thus, there is now evidence from a previous population-based cohort study and the present population-based case-control study that opioids and barbiturates are associated with CDH. The similarity of findings in the case-control and cohort design makes the findings more credible.
There are several ways in which to interpret these findings. One possibility is that the apparent association is attributable to a methodological artifact such as recall bias, reverse causality or confounding by indication. Because this was a case control study, reports of medication consumption may be affected by errors in recall. Second, we chose episodic headache sufferers as our comparison group (rather than the general population) because we were interested in measuring treatment patterns for headache. Third, we chose a wide range of headache frequency for the control group owing to uncertainty about the definition of a ‘normal’ headache frequency. It may be that our upper cutoff of 104 headache days/year for the control group resulted in some misclassification.
Confounding by indication is probably the strongest challenge to study validity in observational studies of this type. We cannot determine if these medications are themselves risk factors or if they are preferentially prescribed or taken by those at increased risk for CDH. For example, headache sufferers with comorbid pain disorders may preferentially take opioids. Headache sufferers may also prefer medications that affect mood if they suffer from depression. As comorbid pain disorders and depression are likely risk factors for CDH (21,33), medications preferred by these individuals may spuriously appear associated with CDH risk. In the AMPP, opioids and barbiturates were associated with the new onset of CM but that does not eliminate the possibility of confounding by indication. These associations remained even after adjusting for attack frequency and disability and a dose-response curve was demonstrated, but residual confounding cannot be excluded. Similarly, while aspirin may reduce the risk of CDH through an anti-inflammatory effect, it is also possible that those with frequent headache may avoid aspirin because of side-effects from frequent use.
There are also biologically plausible mechanisms that could account for the apparent protective effect of aspirin. Anti-inflammatory drugs have been reported to protect against neurologic disorders other than CDH, including Parkinson's disease and Alzheimer's dementia (34). It is possible that in headache and in these other conditions aspirin blocks the release of inflammatory mediators, preventing neurologic damage related to disease. The release of inflammatory mediators and activation of glia have been suggested in migraine pain (35,36). Furthermore, there are similarities in ibuprofen and aspirin, regarding the Cox1/Cox 2 profile, and important differences with naproxen (37). While three large clinical trials (38–40) support a role for aspirin in the prophylaxis of migraine, to our knowledge, there are no such trials of aspirin to prevent the development of CDH.
Our study extends prior clinic-based studies (30,31,41), and a longitudinal population-based study (11) suggesting that opioid use is associated with increased risk, and aspirin is protective, for CDH. Randomized trials in patients at high risk for CDH may be considered to investigate the protective effect of aspirin or of dose limits for opioids. Also, while difficult to design and implement, a controlled clinical trial of medication withdrawal as a treatment for CDH will establish the efficacy of this intervention.