Abstract
Background: Medication-overuse headache (MOH) management usually includes a medication withdrawal. The choice of withdrawal modalities remains a matter of debate.
Methods: We compared the efficacy of in-patient versus out-patient withdrawal programmes in 82 consecutive patients with MOH in an open-label prospective randomized trial. The main outcome measure was the reduction in number of headache days after 2 months and after 2 years. The responders were defined as patients who had reverted to episodic headaches and to an intake of acute treatments for headache less than 10 days per month.
Results: Seventy-one patients had a complete drug withdrawal (n = 36 in the out-patient group; n = 35 in the in-patient group). The reduction of headache frequency and subjective improvement did not differ between groups. The long-term responder rate was similar in the out- and in- patient groups (44% and 44%; p = 0.810). The only predictive factor of a bad outcome 2 years after withdrawal was an initial consumption of more than 150 units of acute treatments for headache per month (OR = 3.1; 95% confidence interval 1.1–9.3; p = 0.044).
Conclusion: Given that we did not observe any difference in efficacy between the in- and out-patient withdrawals, we would recommend out-patient withdrawal in the first instance for patients with uncomplicated MOH.
Keywords
Introduction
Medication-overuse headache (MOH) is defined by the International Classification of Headache Disorders 2nd edition (ICHD-II) as a headache present on at least 15 days per month which develops following regular overuse of acute migraine treatments for more than 3 months (1). The IHS criteria definition has fixed the critical intake of symptomatic drugs required for a primary headache to become chronic at 10–15 days per month. Epidemiological studies suggest that MOH may affect about 0.7–1.7% of adults in the general population (2–7) and 0.5% of adolescents (8,9).
MOH management strategies include the following general measures: (i) optimizing preventive and acute symptomatic treatments of headache; (ii) identifying and treating comorbid psychological and medical conditions; and; (iii) discontinuation of the offending symptomatic medication (7,10–12). In practice, medication withdrawal strategies vary with time, and from one country to another (12–15). In Europe, the recently voiced professional consensus was to recommended in-patient withdrawal to patients overusing opioids, barbiturates or analgesics in large daily doses, to those with anxiodepressive disorders, to those with substance dependency, and to patients needing to be separated from their familial background (11,16,17). Conversely, given the cost and constraints of hospitalization other authors recommend out-patient withdrawal as an initial step (18–20). These two strategies are not supported by any medically based evidence because few prospective comparative trials have compared in- and out-patient strategies (21–25).
We therefore performed a long-term, prospective, randomized study to compare the efficacy of in-patient versus out-patient withdrawal programmes. This efficacy was assessed by the reduction of headache frequency at 2 months and 2 years. We also planned to assess medical and psychological factors that may be risk factors for poor outcome after medication withdrawal.
Methods
Patients
From March 2003 to December 2005 all the consecutive patients referred by their primary care physician to our pain centre with suspected MOH were systematically mailed a questionnaire and a 1-month headache diary to be completed before attending their first visit. Inclusion occurred during a second visit, 1 month later. Because this study was conducted before the publication of the International Classification of Headache Disorders 2nd edition (ICHD-II) (1), we defined MOH as a headache ≥15 days/month for at least 3 months combined with an intake of acute symptomatic treatments for headache (ATH) ≥15 days/month over the same period. This definition of MOH was more restricted than those published in the ICHD-II because we did not included the patients with an intake of analgesics only between 10 and 14 days per month. The other inclusion criteria were: (i) age ≥ 18 years; (ii) patients who accepted the principle of randomization into either the out-patient group or the in-patient group and; (iii) patients who agreed to halt their professional activity for at least 8 days in order to complete an abrupt drug withdrawal. Patients were excluded if: (i) they had suffered from any significant illness or major depression in the past month (according to Diagnostic and Statistical Manual of Mental Disorders, 4th edition, DSM-IV); (ii) they were pregnant; (iii) they were unable to precisely describe their headaches or their medication consumption; (iv) their medication overuse included World Health Organization step III opioids; and (v) if there had been no improvement of the headache after a previous well-conducted withdrawal. We did not include restrictions concerning barbiturates because they are not available in acute treatments for migraine in France. The setting-up of this study was approved by the direction of clinical research in our hospital and was validated by the local ethics committee. This open-label randomized trial was conducted over a 2-year period after inclusion.
Study design
During the first visit (V1), a neurologist, expert in the headache field, explained to the patient the benefits of undergoing a detoxification programme. The participants in both groups received written general information about MOH, the study design, and withdrawal procedures and gave their informed consent before taking part in the study. Then, they were randomly assigned to the out-patient group (n = 41) or the in-patient group (n = 41). In order to reinforce the patients’ motivation, we presented the theoretical advantages of hospitalization to those patients who were randomized in the in-patient group as avoiding being alone at home, having the possibility of discussing the pain and their feelings regarding medication withdrawal, having constant medical attention, and better management of their withdrawal symptoms, and that they would not have to work and could rest. We similarly presented the advantages of ambulatory withdrawal to those included in the out-patient group as being quiet at home without the noise and constraints of the hospital, the possibility of adapting their activity to their clinical state, and the lack of disorganization to their family life.
Medication withdrawal was programmed to begin 1 month later. This waiting time between the decision and the start of the programme gave patients time to psychologically prepare themselves for the drug withdrawal. During this period, patients were told to complete a daily headache diary, without changing their preventive or acute drug treatments for their headaches. This diary gave us baseline data on headache frequency and medication consumption.
A second visit (V2) took place at the end of this first month, just before drug withdrawal during which the subject was interviewed to (i) check the eligibility criteria and; (ii) recall data about the patient. Conducted by a neurologist and a psychologist, the interview guide included: (i) history of headaches; (ii) qualitative and quantitative evaluation of acute treatments of headaches (ATH) during the last month (number of days with intake, and number of units per month); (iii) frequency and types of headaches during the last month with the patients having to choose between the following items to best describe the temporal course of their usual headaches: headache attacks (with clear headache-free periods between the attacks), persisting headaches without any definite headache attacks (persisting headache was defined by the absence of any headache-free period), or persisting headaches overlaid with headache attacks (these attacks differed from simple exacerbation of their headaches by their longer duration and greater intensity) (26); and (iv) anxiety and depression. The latter dimension was assessed using a validated French version of the widely used 14-item hospital anxiety and depression scale (HADS) (27,28). Consumption of anxiolytics, sedatives or antidepressants was also recorded.
Medication withdrawal
Patients were asked to halt their professional activity for 8 days. They also had to stop their consumption of ATH completely and abruptly. In-patient withdrawal was performed under the control of the neurological service. The out-patient withdrawal patients were told to consult their general practitioner if needed. In both groups the patients were given the telephone number of the pain centre. To minimize the withdrawal syndrome, an oral treatment by amitriptyline was systematically proposed (20 to 40 mg/day for 8 days, followed by a progressive decrease over the next month) and the patients also received metoclopramide against nausea. Should they experience severe rebound headaches, they had the possibility of using cryotherapy, transcutaneous electrical neurostimulation and, as an emergency treatment, non-steroidal anti-inflammatory (NSAI) tablets or nefopam solutions (without exceeding 2 days/week). A preventive treatment was introduced on the first day of the withdrawal. The neurologist chose this treatment during visit V2, with reference to the previous preventive treatments already used by the patient and in accordance with French national recommendations (16,29). At the end of their withdrawal, patients received a prescription of ATH, usually triptans or NSAI, with the instructions: ‘Do not take this medication for more than eight days per month’.
A report book was delivered to each patient. This report book was in two parts. The first was completed by the neurologist and featured all the relevant information concerning the chosen method of withdrawal and the preventive treatments and advice concerning the intake of ATH. The second part was a diary allowing the patients to report daily on the intensity of their headaches using an 11-point numerical scale (0, no headache to 10, maximal headache), and their consumption of ATH.
Follow-up
An initial follow-up consultation (visit V3) was programmed 2 months after the withdrawal. Patients had to score (i) the severity of their withdrawal symptoms using an 11-point numerical scale graded from 0 (no withdrawal symptoms) to 10 (the most severe withdrawal symptoms imaginable); (ii) their craving for ATH during withdrawal on a numerical scale graded from 0 (no craving for ATH) to 10 (maximal imaginable craving for ATH); and (iii) the percentage of subjective improvement (the baseline period was the period before withdrawal). Data were also collected from the headache diary. We asked patients if they had felt any psychological distress related to the drug withdrawal.
During the 2 years after drug withdrawal, all the patients were told to consult their general practitioner or their usual neurologist at least every 3 months, and to attend a consultation in the Pain Centre at least once a year. We told them to continue completing their headache diaries. A second follow-up visit was programmed 2 years after the drug withdrawal so that we could collect data regarding the subjective clinical improvement compared with the baseline state, the number of days with headache, days with an intake of ATH and the presence of other chronic pain that may have justified the intake of analgesics. A telephone interview was conducted by a psychologist unconnected with this study and trained to collect medical data about headache. We chose this method in order to evaluate the maximal number of patients rather than just the most compliant among them, as this would not necessarily have been representative of the whole population.
Outcome measures
The main outcome was the reduction in number of headache days after 2 months and after 2 years, compared with the baseline period. Other outcomes were the psychological distress related to drug withdrawal, the severity of withdrawal symptoms and the degree of craving for ATH, the percentage of clinical subjective improvement compared to baseline (2 months and 2 years after withdrawal) and the number of patients with episodic headaches (<15 days of headache/month). Some authors have proposed to define responders as patients who had reverted to episodic headaches and to an intake of ATH less than 10 days per month (23). We used this last criterion to define responders 2 months and 2 years after withdrawal.
Power calculation and statistics
Given that we expected a mean reduction in headache frequency of 50 ± 38% in accordance with results obtained in a previous study (30) and to highlight a minimal difference of 25% between the two groups with an α risk of 0.05 and a power of 0.8, we calculated that we needed to include 36 subjects per group. Taking into account the risk of drop-outs, we chose to include 41 patients in each group. Quantitative data were analysed by calculating their mean and standard deviation (SD). Non-parametric tests were used for comparisons between groups (Mann-Whitney U-test for quantitative data and χ2 Fisher exact test for categorical data). Statistical significance was taken to be p < 0.05. In a second step, we identified, the factors differentiating responders from non-responders (χ2 Fisher exact test; odd-ratio with 95% confidence interval (CI)) from among all the patients who had undergone complete drug withdrawal; patients who dropped out were classified as non-responders. The predictors tested in the present study included the sex, the duration of MOH greater than 3 years, a history of migraine, a history of tension-type headache (TTH), the presence of persisting daily headaches, HADS score ≥ 18, the intake of triptans, the intake of opioids, an initial consumption greater than 150 units of ATH per month, the In or out-patient withdrawal, the severity of their withdrawal symptoms (numerical scale ≥ 5), the psychological distress related to the drug withdrawal. Statistical analyses were performed using SPSS 11.0.4 software for Mac OS X (Chicago).
Results
Among the 82 patients who were referred to the pain centre for probable MOH, seven were excluded before the withdrawal began (Figure 1). Among the out-patient group, one patient had spontaneously decreased his medication consumption between the first two visits and reported a resolution of his MOH; another one refused the withdrawal modalities because she finally felt unable to limit her drug consumption without being hospitalized. Among the in-patient group, three patients had spontaneously decreased their medication consumption before visit 2 and reported a resolution of their MOH; a fourth refused hospitalization and a fifth had to be operated on for a different pathology before withdrawal began.
Flow Diagram.
Four more patients were subsequently excluded because they did not fully withdraw their medication: three women (8%) from the out-patient group and one woman (3%) from the in-patient group. They were able only to decrease their daily ATH intake but did not completely withdraw from the drug owing to their overwhelming craving for ATH. The final statistical analysis was thus conducted on the 71 patients who had undergone complete withdrawal and who were seen in consultation in the pain centre 2 months after their drug withdrawal: 36 patients in the out-patient group and 35 patients in the in-patient group.
Demographic and headache characteristics of the study population
Comparison was performed using a Mann-Whitney U-test for quantitative data and a Fisher exact test for the qualitative data. Statistical significance was taken to be p < 0.05 (bold). Persisting headache was defined by the absence of any headache-free period.
ATH: acute treatment of headaches, HADS: hospital anxiety and depression scale, MOH, medication overuse headache, TTH: tension type headache.
Characteristics of overused medication in each group
WHO step II opioids marketed in fixed combinations are classified as ‘Analgesic in combination’. Given the small patient sample for some subtypes of MOH, we did not perform any statistical comparisons between groups.
Characteristics of prophylactic treatment in each group 2 months after withdrawal
Primary and secondary outcome measures
Data obtained from 36 and 35 patients who carried out complete drug withdrawal in the out-patient and in-patient groups, respectively. †Data obtained from 34 and 32 patients at long-term follow-up in the out-patient and in-patient groups, respectively. The percentage of subjects with episodic headaches and the rate of responders was calculated among the whole population because we performed an analysis with the intention of treating patients who had initially performed a complete drug withdrawal.
ATH: acute treatments of headaches.
All the patients who achieved complete withdrawal were seen in consultation 2 months later. Overall, 2 months after withdrawal the frequency of ATH intake was less than 10 days per month for 82% of the whole population (n = 58), between 10 and 14 for 17% (n = 13), and equal to 30 days per month for 1% (n = 1). Fifty-six percent of them had reverted to an episodic pattern of headache (n = 40); 48% were considered as responders considering Rossi’s criteria – that is, patients who had reverted to episodic patterns of headache and intake of ATH for less than 10 days per month (n = 34). We observed an overall reduction of headache frequency of 46 ± 33% and a rate of subjective improvement reaching 60 ± 28%. We did not observe any differences between the out- and in-patient groups (Table 4).
2 years after drug withdrawal, two patients in the out-patient group refused the final interview and three patients dropped out in the in-patient group. The long-term follow-up concerned 34 patients in the out-patient group and 32 in the in-patient group (Figure 1). Overall, 49% of our patients had reverted to episodic headaches (n = 35); 42% were considered as responders (n = 30). We observed an overall reduction of headache frequency of 54 ± 34% with 44 ± 35% of subjective improvement. Neither the subjective nor the objective endpoints differed significantly between the two groups (Table 4).
By considering the whole patient population who achieved complete drug withdrawal (n = 71), the only factor differentiating responders and non-responders was an initial consumption greater than 150 units of ATH per month (OR = 3.1; CI = 1.1–9.3; p = 0.04). The long-term success did not depend on the type of primary headaches or the type of overused medication.
Discussion
Our study showed a similar efficacy between the out-patient and in-patient detoxification programmes proposed to MOH patients, whatever the duration of the follow-up (2 months and 2 years) and whatever the outcome criteria (objective or subjective). The modality of withdrawal had no predictive value on the rate of responders. Literature on comparisons between in- and out-patient withdrawal is scant. In the past, clinicians suggested that out-patient detoxification programmes were not sufficient in cases of severe MOH and recommended hospitalization of 10–14 days (31). Since then, a progressive decrease in ergotamine overuse concurrent with an increase in triptan overuse has been observed in tertiary centres (32,33). Because withdrawal symptoms are less severe and less frequent with triptans (34,35), medication withdrawals seem easier to perform nowadays. In this context, different practitioners have proposed an abrupt out-patient drug withdrawal as a first-line pragmatic treatment of MOH (36–38). The first comparative study suggested a higher rate of relapse in a group of 29 out-patients than in a group of 73 in-patients (38% relapse vs 25%, p < 0.005) but there was no difference in efficacy on the headache index; moreover, this study was not randomized and the long-term follow-up was short (4 months) (21). Suhr et al. compared withdrawal in a group of 41 in-patients with a group of 60 out-patients for 5.9 ± 4 years. They observed that the long-term relapse rate was higher in the out-patients than the in-patients (25% of relapse rate vs 15%) but this difference did not reach significance and their randomization was only partial (the authors wrote that it depended ‘on the personal situation of the patient’) (22). More recently, Rossi et al. showed in a prospective strictly randomized study that the rate of non-responders 3 months after withdrawal was similar between a group of 39 patients with in-patient withdrawal and another of 39 patients with out-patient detoxification programmes (23% vs 28%, not significant) (23). Among the responders, the relapse rate 1 year later was also similar in both groups (respectively 25% vs 23%, not significant) (24). This absence of superiority of in-patient vs out-patient withdrawal was demonstrated only in a population of patients with low medical needs. Our data confirmed these results in a mildly affected population of MOH patients.
Half of our MOH patients had reverted to an episodic pattern of headache 2 years after drug withdrawal. These results were similar to those recently obtained by specialized Headache Centres (6,11,17,18,39). However, by restricting the definition of responders to patients with both episodic headache and an ATH intake on less than 10 days per month, only 42% of our patients were considered to be long-term responders. A responder rate of 76% has recently been published using the same criteria, but this study included only migrainous patients with low medical needs (without any TTH history, anxiety disorders, substance addiction disorders or anxiolytic overuse) (24). In this study, the one-year follow-up was carried out only with those patients considered to be responders 2 months after withdrawal.
Various different factors may explain the poor outcome of MOH patients. In cases of severe chronic migraines or chronic TTH, the medication overuse may be a consequence rather than a causal factor of the chronic nature of these headaches (1). Severe substance dependency may lead to persistent medication overuse (40–43). Interestingly, in our study, 5% of the patients who began withdrawal reported an overwhelming craving for ATH and were unable to carry out complete withdrawal. In practice, it is crucial to ask MOH patients if they are really ready to undergo complete withdrawal. If so, it is often then useful to delay the onset of withdrawal to enhance their motivation. Patients also need to be advised that drug withdrawal represents only one step in MOH management (43). Regular follow-up is systematically required in order to optimize the preventive and acute symptomatic treatment of headache and to help those patients with high risks of substance dependency.
The identification of factors predicting relapse may help clinicians to optimize the withdrawal programmes and follow-up. Here, we found that the risk of withdrawal failure after 2 years was three times higher if the initial consumption of acute treatments for migraine was above 150 units per month (150 units represents, for example, 5 units per day, 30 days per month, a situation that is frequently encountered when patients use analgesics or combined analgesics). This risk factor has already been observed by other authors but with a lower cut-off: 30 units per month (44). In practice, this may allow us to better identify patients at risk of developing refractory headache. Some authors implicated other factors for poor outcome such as: smoking and alcohol consumption (44), codeine or barbiturate consumption (21,22,45–47), ergotamine consumption (38), male gender, long duration of medication overuse (48), underlying TTH (49), high disability score for chronic headache (estimated by the Migraine Disability Assessment Score, MIDAS) (38), high self-reported bodily pain (as measured by the quality of life tool SF-36) (44), low self reported sleep quality (47) and high scores on the HAD scale (50). Conversely, patients overusing Triptan have a better prognosis than those abusing analgesics (49). The fact we did not find evidence for any such predictors of relapse may be related to an insufficient sample size, to the frequent association between several kinds of medications or the different kinds of primary headaches.
Although we expected that the absence of daily medical supervision would generate greater anxiety in the out-patient group, we observed that withdrawal was more frequently judged as psychologically difficult among the hospitalized patients. Interestingly, none of the patients in the out-patient group called their physician during the withdrawal period, although they had been told to call them if necessary. Our hypothesis is that out-patient therapy may lead patients to develop more active coping strategies whereas in-patient therapy may enhance dependent and/or passive behaviours. The importance of coping strategies has been shown previously (51,52). The question is whether these coping strategies may influence long-term withdrawal success.
Unexpectedly, we observed that 5% of the 82 enrolled patients spontaneously stopped their medication consumption as soon as they were warned of the detrimental effects of medication overuse. They did this by changing their way of using their medication without any medical assistance, and without any preventive treatments. This highlights the importance of a doctor’s advice as a first approach to patients with medication overuse (23). It is probable that a combination of patient education, information about MOH and optimization of migraine may be sufficient to treat MOH in a greater number of patients than would have been expected (53).
Several biases prevent us from generalizing our results about the comparison between out- and in-patient withdrawal. Firstly, our results concerned a population of MOH patients referred to our specialized centre, so the reproducibility of our findings needs to be verified in primary care. Even though a recent study failed to find a difference in results when patients were followed by a neurologist or by a primary care physician (25), this factor may influence the long-term follow-up after the drug withdrawal. As some patients would only accept telephone interviews 2 years after drug withdrawal the 2-year results are less precise than those obtained 2 months after drug withdrawal. However, we are convinced that this pragmatic attitude allowed us to increase the compliance of patients during the long-term follow-up, and so to avoid biases related to drops-outs. Moreover, because the number of telephone interviews was similar in both groups, this should not affect the comparisons between groups. One of the strengths of our study is that we followed all the MOH patients who had completed drug withdrawal and not only those who had been improved by this withdrawal. Hence our results reflect what we observe in clinical practice. Moreover, a two-year follow-up is sufficient to provide a real estimation of the long-term relapse rate (49).
In conclusion, given that hospitalization did not improve either the short or the long-term prognosis of these patients, and given that it did not increase the psychological comfort described by the patients during the withdrawal therapy, we would recommend out-patient withdrawal in the first instance for patients with MOH. This would minimize the financial costs related to medication withdrawal and may encourage primary care physicians to manage their own MOH patients and propose medication withdrawal programmes earlier than they have done in the past (54). However, it is important to note that the choice of a drug withdrawal programme must be done in accordance with the patient's preference to improve compliance with the programme (55). Moreover, we must not ignore the fact that constant medical care may be required especially when the medication overuse concerns WHO step II opioids or when the drug withdrawal poses a risk of decompensation for patients with important psychiatric comorbidity. Interestingly, instead of answering the question of ‘how should we withdraw this drug?’ we should ask ‘is drug withdrawal useful for this patient, and when should we do it, so that the patient will successfully withdraw from the drug?’
Footnotes
Acknowledgements
We thank L Bert, M Godot, Dr V Rousselon, W Lipski and the University Hospital of Saint Etienne for their technical support.
Funding
This work was supported by grants from Fondation de France and Fondation CNP.