Treatment of Medication Overuse Headache: Effect and Predictors After 1 Year

A Randomized Controlled Trial

Louise N. Carlsen MD, PhD; Carolien Rouw MD; Maria L. Westergaard MD, PhD; Mia Nielsen MD; Signe B. Munksgaard MD, PhD; Lars Bendtsen MD, PhD, Dr.Med.Sci.; Rigmor H. Jensen MD, Dr.Med.Sci.


Headache. 2021;61(7):1112-1122. 

In This Article


Study Population

Patients with MOH referred to the Danish Headache Center were considered for participation. Patients were invited to participate if they fulfilled the following criteria: MOH diagnosis according to the International Classification of Headache Disorders, 3rd edition (beta version),[1] where information about headache frequency and acute medication use was obtained from detailed history or at least 1 month filled-out headache calendar; MOH arising from preexisting migraine and/or tension-type headache; minimum 18 years old and capable of providing informed consent; considered eligible for outpatient treatment based on the type of medication overuse (without daily or almost daily use of opioids or barbiturates); personal resources and motivation; capable of completing a headache calendar; no severe physical illness (e.g., severe comorbid pain, uncontrolled diabetes, serious heart disease, cancer) or psychiatric disorders (requirement of antidepressant medication, ongoing treatment by a psychiatrist, or in a psychiatric clinic); no alcohol or drug addiction; no pregnancy, breastfeeding, or planned pregnancy within the next 12 months; ability to provide information about medical history (without linguistic barrier); and no preventive headache treatments at baseline.

The study was approved by the regional Ethics Committee (H-16029763). All participants signed informed consent, and the study was registered at, identifier: NCT02993289.

Study Design

This study was a continuation of a prospective, longitudinal, open-label, randomized, controlled study comparing three treatment strategies: (a) W+P, (b) P+pW, and (c) W+pP. Patients were followed for 1 year. During the study period, visits were planned at baseline, 2, 6, and 12 months. Additionally, patients were contacted by phone at 1, 4, and 9 months (Figure 1). There were no deviations from the original trial protocol, which is found in Supporting Information 1. Results from baseline to 6-month follow-up have already been published.[6,7] This paper focuses on the planned secondary analyses on outcomes after 1 year.

Withdrawal Approach

Patients received advice on withdrawal and MOH from trained headache nurses before the start of withdrawal. Patients were recommended to completely discontinue acute analgesics and migraine medications for the first 2 months. Patients were offered rescue medication (levomepromazine or promethazine maximally 75 mg per day) and antiemetics (metoclopramide or domperidone 10 mg) during the withdrawal period. After 2 months, patients could use acute migraine-specific medication up to 9 days per month, or 14 days per month for simple analgesics alone.

Patients in the P+pW group were offered withdrawal therapy after 6-month follow-up if they still fulfilled the MOH criteria. It was voluntary to follow the advice on withdrawal. Importantly, patients in the P+pW group did not receive any specific advice or information about withdrawal in the first 6 months of treatment. Note that the change in the P+pW group was predefined in the study design.

Preventive Medication

Before starting the treatment, patients received information about specific preventive medications chosen according to current guidelines.[9] For migraine, the following preventive medications were considered: beta-blockers, candesartan, lisinopril, topiramate, amitriptyline, and botulinum toxin A injections according to the PREEMPT protocol.[10] For tension-type headache, amitriptyline or mirtazapine was considered. If patients experienced unacceptable adverse effects or lack of effect, preventive medication was changed. Patients in the W+pP group were offered preventive treatment at the end of the withdrawal period (2 months) if the treatment team identified an indication, and the patients consented to receiving the additional medication. Effective preventive treatment was continued throughout the study period.

Multidisciplinary Treatment

During the 1-year study period, all patients were offered voluntary therapy sessions with a psychologist, and sessions with a physiotherapist, as part of standard treatment at the Danish Headache Center. Therapy sessions with a psychologist focused on coping strategies for living with headache, trigger factors for headache, for example, stress and anxiety, and acceptance of a life with chronic pain. Sessions with a physiotherapist included, for example, correction of posture, biofeedback, active exercises, and relaxation techniques.

Data Acquisition

Patients filled out headache calendars continuously for 1 year. Information about headache days, migraine days, pain intensity, and use of acute medication was obtained from these headache calendars. Patients filled out the following validated questionnaires at baseline and 12 months: (a) the Severity of Dependence Scale (SDS) questionnaire was used as an indicator of dependence-like behavior, (b) the WHO Quality of Life BREF (WHO QoL-BREF) questionnaire measured quality of life (QoL), (c) the Headache Under-Response to Treatment (HURT) questionnaire addressed disability and patients' perceptions of control of headache, (d) the Hospital Anxiety and Depression Scale (HADS) was used to identify patients at risk of having depression or anxiety (cutoff value of 8), and (e) the Perceived Stress Scale (PSS) estimated stress level.[11–16] Table S1 summarizes information about the questionnaires.


The Sealed Envelope homepage was used to randomize the patients into three groups in blocks of nine patients.[17] This process was performed by a project nurse from another research group at the Danish Headache Center.


The primary endpoint was change in monthly headache days from baseline to 1-year follow-up. Secondary outcomes for 1-year follow-up were as follows: (a) change in monthly migraine days from baseline; (b) change in days per month with acute medication from baseline; (c) change in monthly pain intensity from baseline, where the total monthly score ranged from 0 to 90 (sum of daily scores from 0 = no pain, 1 = mild pain, 2 = moderate pain, to 3 = severe pain); (d) number of patients who reverted to episodic headache; (e) number of patients with resolved MOH; (f) change in SDS score from baseline; (g) change in QoL scores from baseline; (h) change in HURT score from baseline; (i) change in HADS depression and anxiety scores from baseline; (j) number of patients with anxiety or depression from baseline according to HADS score ≥8; and (k) change in PSS score from baseline. All endpoints were reported as change for the total study population and for comparisons between the three treatment strategies. Dropout rates were calculated for each group.

Prognostic Factors for Still Having Chronic Headache After 1 Year

The total study population was divided into patients who reverted to episodic headache (<15 headache days/month) and those who still had chronic headache (≥15 headache days/month) after 1-year follow-up.

We tested the following baseline characteristics as possible predictors of still having chronic headache after 1 year: age (years), sex (woman/man), type of preexisting headache (chronic migraine, episodic migraine and tension-type headache, chronic tension-type headache), headache frequency (days/month), migraine frequency (days/month), use of acute medication (days/month), pain intensity (0–90/month), duration of medication overuse (years), SDS score (0–15), overall QoL (1–5), self-rated overall health (1–5), HURT score (0–24), depression or anxiety (HADS ≥ 8), PSS score (0–40), and daily caffeine consumption (based on coffee, tea, cola, and energy drinks, under/over 200 mg/day).


R statistical software (R Foundation for Statistical Computing, Vienna, Austria, version 1.1.463) was used for statistical analyses. Sample size was calculated using one-way ANOVA and Cohen's f (0.35) for comparison of three groups, based on estimates from previous literature and clinical experience (highest mean value, 14.0; lowest mean value, 6.0; standard deviation (SD) between groups, 7.5; SD for the complete group, 10.5). We used an alpha of 5% and arrived at a sample size of 102 patients to achieve a power of 80%. The efficacy variable was reduction in headache days per month. Because we assumed that the dropout rate would be approximately 15%, we aimed to include 120 patients with MOH.[6]

Normally distributed continuous data were presented as means with SDs, or 95% confidence intervals (CIs); whereas skewed continuous data were shown as medians (ranges). Paired t-test was used for comparing outcomes at baseline to 1-year follow-up. One-way ANOVA was used for comparison of continuous outcomes between the three treatment strategies. Model control, F-test, and Levene's test were conducted to test the assumptions underlying the statistical method. In case of heterogeneity of variance, Welch's test was performed. Categorical variables were presented as percentages and numbers, and the chi-squared test was used for statistical comparisons. If expected counts were fewer than five, Fisher's exact test was used. Bonferroni's correction was used to adjust for multiple testing when comparing a variable between three groups.

Logistic regression models were used to investigate prognostic factors. All covariates were tested in univariate analyses and afterward corrected for age, sex, and treatment strategy in a multiple regression analysis (additive model).

All p-values were two-tailed, and p < 0.05 was considered statistically significant.