Chronic headache (⩾15 days/month for at least 3 months) affects 3–4% of the general population and is a worldwide problem.1^–4 Medication overuse contributes to the problem.5 6 Detoxification often leads to an improvement in the headache; nevertheless, it is usually difficult to motivate patients to do this. Another challenge is the high relapse rate in those who have managed to stop their medication overuse.7 Patients with medication overuse headache (MOH) often experience tolerance and withdrawal (ie, reduced efficacy of the medication(s), need for higher doses to achieve analgesic effect and rebound headache).8 9 These are signs of physical dependence according to the DSM-IV criteria, and suggests that MOH has similarities with dependence on traditional addictive drugs.10 11 The revised criteria for medication overuse of the International Classification of Headache Disorders (ICHD-II) defines MOH exclusively on the basis of consumption of medication and a worsening of the headache during the time of medication intake.3 12 13 However, information about dependency-like behaviour might also be helpful in the management of MOH. The Severity of Dependence Scale (SDS) focuses on psychological aspects of dependence and provides a simple graded measure for assessing dependency on different prescribed and illegitimate drugs.14 Validation of the SDS and assessment of diagnostic cut-off limits have been performed for many different drugs of abuse, including alcohol, amphetamine, benzodiazepines, cannabis, cocaine and heroine.14^–20 Little is known about dependency-like parameters in patients with chronic headache and the scale has not previously been used in this setting. An instrument that can detect dependence-like behaviour in relation to headache medication overuse would be a powerful tool for general practitioners and other health care professionals for characterising and planning optimal management of their patients with chronic headache. If associated with MOH, such a tool may also make identification of these patients, among others with chronic headache, easier.

We investigated the SDS in a representative sample of people with primary chronic headache from the general population and tested the hypothesis that SDS score is associated with a diagnosis of medication overuse.

METHODS

Design

This was a cross sectional epidemiological survey. A short postal questionnaire screened for chronic headache (⩾15 days/last month and/or ⩾180 days/last year). Those with self-reported chronic headache were invited to a clinical interview and examination conducted by neurological residents. The interview focused on consumption of medication. To assist with this, participants were asked to complete a medication list at home, prior to the interview.

Classification

The criteria of the ICHD-II and relevant revisions for MOH were applied.3 12 21 22 Inclusion required a diagnosis of primary chronic headache (ie, chronic tension-type headache (CTTH), probable CTTH, chronic migraine (CM), probable CM and other rare primary chronic headaches). The criteria for CM include only migraine without aura, but participants that otherwise fulfilled the CM criteria for migraine with aura were also classified as CM. Box 1 shows the definition of CTTH, CM and MOH. When medication overuse is present, primary headache diagnoses are designated with the prefix “probable”, according to the original classification, since the primary diagnosis is regarded as uncertain until 2 months after medication overuse has ceased. For readability and as this paper has a main focus on medication overuse, we used the terminology with medication overuse rather than the prefix probable.

The Severity of Dependence Scale (SDS)

The SDS, which was completed during the interview, consists of five questions designed to measure psychological dependence (box 2).14 23 The questions apply to the headache medication taken within the past month. Each item is scored on a 4 point scale (0–3), and the total maximum score is 15. The screening questionnaire response rate was 71% and the participation rate of the interview was 74%. The method has been described in more detail elsewhere and will be described only briefly here.2

Setting

The study was conducted at Akershus University Hospital, Oslo, Norway.

Participants

A random sample of 30 000 people, aged 30–44 years, and stratified for age and gender from the general population of eastern Akershus County, Norway, was drawn from the Norwegian Population Register.

Main outcome measure

The mean SDS score in persons with primary chronic headache with and without medication overuse was the primary outcome measure.

Statistical analyses

The statistical analyses focused on CTTH with and without medication overuse as these two groups constituted 95% of those with primary chronic headache. Statistical analyses were performed using SPSS 15 for Windows. The level of significance was set at 5%. The Student’s t test and χ² tests were used to test two group differences for continuous and categorical data, respectively. Fisher’s exact test was used when appropriate. Analysis of variance (ANOVA) with Bonferroni corrections was used for post hoc comparisons and multigroup comparisons of continuous data. The normality assumption was assessed using the Komolgorov–Smirnov test. SDS scores were missing for two cases, one with and one without medication overuse. These were excluded from analyses.

Logistic regression analysis was performed to determine if the SDS score is a good predictor of medication overuse. The SDS score was treated as a continuous predictor variable and medication overuse as a dichotomous outcome variable. The method of data splitting was used to avoid hypothesis fishing.24 The data were first stratified with respect to the confounding variables age, gender and co-occurrence of migraine, and were then randomly split into two parts. The first data set was used to identify a model. Confounders were left in the model regardless of their significance. The second data set was used to test the model. When both data sets showed significant relationships between the predictor and the outcome variables, we used the entire data set for the final analysis.

Linear regression analysis was used to test the hypothesis that SDS score was associated with number of days of medication intake per month. Again, data splitting was used.

A receiver operating characteristics (ROC) analysis was performed to evaluate the relationship between sensitivity (proportion of true positives) and specificity (proportion of true negatives) of the SDS score for the identification of the presence of medication overuse. The area under the ROC curve (AUC) gives the diagnostic utility of the test (ie, the proportion of cases that are correctly discriminated by the SDS score). The optimal cut-off score was defined as the one that yields the highest χ² value.

Ethics

All participants gave informed consent. The Regional Committee for Medical Research Ethics and the Norwegian Social Science Data Services approved the study.

RESULTS

Sample characteristics

Of the 405 people (88 men and 317 women) with primary chronic headache, 95% had CTTH, 4% had CM and <1% had other primary chronic headaches.

Of the 386 people (83 men and 303 women) with CTTH, 44% (40 men and 130 women) had medication overuse, and 47% (26 men and 157 women) had co-occurrence of migraine (42% in those without medication overuse and 55% in those with). Significantly more women than men had CTTH (p<0.001) but the gender distribution was similar in those with and without medication overuse. Co-occurrence of migraine was significantly more common in those with than in those without medication overuse (p = 0.011). Of the 17 people (two men and 15 women) with CM, 82% (14 of 17) had medication overuse (p<0.002 compared with other primary chronic headaches). Two of four people (two men and two women) with other primary chronic headaches had medication overuse and none of these had co-occurrence of migraine.

Pattern of medication overuse

Chronic tension-type headache

Table 1 shows the pattern of medication overuse. Simple analgesics and combination analgesics accounted for more than 90% of the medication overuse. Overuse of ergotamines, triptans, opioids and a combination of acute medications was rare. Paracetamol, ibuprofen and phenazone–caffeine were the most frequently overused medications and only those with co-occurrence of migraine had overuse of ergotamines and triptans. There was a significant difference in the pattern of overuse between men and women (p = 0.036), with the largest differences for triptans (more men) and combination analgesics (more women) (fig 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Gender specific distribution of medication overuse in 40 men and 130 women with medication overuse headache.

The pattern of medication overuse was similar in those with and without co-occurrence of migraine. The mean duration of medication overuse was 7.8 years (6.9–8.6). It tended to be shorter for ergotamines, triptans and opioids and longer for simple analgesics and combination analgesics (table 1)

Chronic migraine and other primary chronic headaches

Among those with CM and medication overuse, triptans, combination analgesics, simple analgesics, combination of acute medications and ergotamines were overused by 36%, 29%, 21%, 7% and 7%, respectively.

The two participants with other primary chronic headache and medication overuse used simple analgesics and opiates, respectively.

In general, data are presented with 95% confidence intervals (CI) although CIs are not given for groups where n<5.

Severity of dependence scale

Chronic tension-type headache

Using the data splitting approach, it was first tested and then verified that mean SDS score is a significant predictor for medication overuse when corrected for the confounders age, gender and co-occurrence of migraine (p<0.001 in both cases). As these two analyses gave similar results, we subsequently analysed the entire data set as one. Figure 2 shows a significant difference in mean SDS score in those with and without medication overuse (5.6 (5.3–5.9) vs 2.7 (2.4–3.1)).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Severity of Dependence Scale (SDS) scores in people with chronic tension-type headache with and without medication overuse headache (MOH). Values are mean (95% confidence interval).

An increase in SDS score by 1 increased the odds ratio (OR) for medication overuse by 1.72 (p<0.001) when adjusted for the confounding variables (unadjusted OR was 1.69 (p<0.001)). Men were less likely to have medication overuse compared with women, with an OR of 0.44 (p = 0.011).

Linear regression analysis with split file methodology confirmed that SDS score is also a significant predictor for number of days of medication overuse when corrected for the confounders age, gender and co-occurrence of migraine (p<0.001, R² = 0.39). A 1 point increase in SDS score resulted in a mean increase of 2.1 days per month with intake of headache medication.

The area under the ROC curve was 0.81 (0.77–0.86), indicating that the SDS correctly discriminates medication overuse in 81% of cases (fig 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Receiver operating characteristic (ROC) curve depicting relationship between sensitivity and specificity at different Severity of Dependence Scale cut-off scores for predicting medication overuse.

An SDS score of 5 or above had the optimal sensitivity and specificity for predicting medication overuse (table 2). An analysis by gender showed that the optimal cut-off for men was 4 or above while that for women was 5 or above; 78% of men and 79% of women with medication overuse and 16% of men and 24% of women without medication overuse had values at or above these SDS cut-off scores. The positive and negative predictive values were 0.82 and 0.80, respectively, for men, and 0.71 and 0.83 for women. Medication intake was also analysed in subjects without medication overuse. Those without medication overuse and with values at or above the SDS cut-off scores had a significantly higher mean number of days with medication intake per month than those with SDS score below the cut-off (men 8.4 days (3.4–13.3) vs 2.5 days (1.4–3.7) (p = 0.001); women 8.6 days (7.0–10.2) vs 4.8 days (4.1–5.5) (p<0.001)). The mean SDS score was 3.4 (2.8–4.0) in men and 4.2 (3.9–4.5) in women (p = 0.021) and 4.5 (4.1–4.9) in those with and 3.5 (3.2–3.9) in those without co-occurrence of migraine (p<0.001). We did not find significant differences in mean SDS scores between the three 5 year age groups, 30–34, 35–39 and 40–44 (p = 0.070).

The mean SDS scores were 5.2 (4.9–5.6), 6.2 (5.5–6.9) and 6.7 (5.1–8.3) in those with analgesic overuse, combination analgesic overuse and triptan overuse, respectively (p = 0.011). Post hoc analysis with Bonferroni corrections showed that the mean SDS score was significantly higher in overusers of combination analgesics than simple analgesics (p = 0.026).

Chronic migraine and other primary chronic headaches

The mean SDS scores of persons with CM were 6.2 (4.9–7.5) and 4.5 in those with and without medication overuse, respectively. The mean SDS scores for other primary chronic headaches were 8.5 and 5.5 in those with and without medication overuse.

DISCUSSION

Our principal finding was that the SDS score was a significant predictor for medication overuse in people with chronic headache in the studied age group from the general population. The vast majority overused simple analgesics or combination analgesics. Regardless of the type of medication overuse, the SDS scores were significantly higher in those with than in those without medication overuse (MOH vs not MOH according to ICHD-II). The SDS score also predicted the numbers of days with intake of headache medication in those using medication less than 15 days per month (ie, those without medication overuse). Thus the SDS is a simple and valid tool that identifies people who are likely to overuse acute headache medication. To our knowledge, this study is the first to analyse the applicability of an established dependency rating scale for describing dependency-like behaviour in relation to medication overuse among people with primary chronic headache.

Our study sample was recruited independently of previous contacts with the health care system (ie, contacts with general practitioners, neurologists, other specialists or hospitals) for a headache complaint. This supports the fact that our sample of subjects is a good representation of primary chronic headache in the general population. Additional strengths of our study were the large sample size, high participation rate and the fact that physicians conducted the interviews and examinations, including detailed information about medication use.1 2 The female preponderance of primary chronic headache was similar in those with and without medication overuse, making bias based on gender composition less likely. The mean SDS score was influenced by age, gender and co-occurrence of migraine, and for that reason our analyses were corrected for these confounders. The split file methodology is applicable in large data sets. The first step is to generate a hypothesis, and the second step confirms or rejects this. When a hypothesis generated using the first half of the data is confirmed in the second, the model may be applied on the entire data set. This approach has been advocated as a powerful tool for assessing relationships in epidemiological studies without the risk of finding significance by chance through “hypothesis fishing”.24

Previous studies of MOH have generally been performed on selected clinical populations, small samples of the general population or been questionnaire based. Findings from tertiary headache clinics cannot be extrapolated directly to the general population. For example, the pattern of medication overuse differs between selected and general populations25^–27; 20% of MOH patients from a Danish tertiary referral headache clinic overused triptans and 29% overused simple analgesics. In contrast, we found 4% with triptan overuse and 59% with simple analgesic overuse. In addition, opioids, ergotamines and combination analgesics are more frequently overused in the clinic population than in the general population.25 Two Taiwanese population based studies of adolescents and the elderly found that the overuse pattern was dominated by over the counter painkillers such as non-steroidal anti-inflammatory drugs in combination with caffeine, aspirin and paracetamol.27 28 In agreement with our results, there was practically no overuse of triptans or opioids.

The SDS has not been validated against other measurements of dependency in MOH sufferers and, indeed, there is still much discussion on whether MOH represents dependence. Being fully aware of this, we have used the SDS score not to attempt to define dependency but to address the issue of whether SDS score distinguishes between subjects with chronic headache with and without medication overuse. By analogy with the use of the scale in other settings and based on the questions included in the scale, we suggest that a high SDS score in MOH sufferers identifies “dependency-like” behaviour which is the word we have used throughout. Whether the high SDS scores found here reflect behaviour similar to outright addiction or a worry about losing control over medication intake due to frequent and strong headache cannot be determined by our cross sectional study. Neither can we determine how the traditionally “non-addictive” drugs lead to dependency, although pathways involving central sensitisation have been suggested.10 Nevertheless, the SDS has widespread use and validation for defining the degree of dependence in other types of medication and drug overuse.14^–19 The use of a well established dependency scale has the advantage of enabling a comparison with other types of dependency. The mean SDS score in those with medication overuse is about 2 points higher than the optimal SDS cut-off scores found in those with dependency on alcohol, amphetamines, cannabis and cocaine.14 16^–19 Only benzodiazepine overusers have a higher mean SDS score than our persons with medication overuse.15 Thus it is clear that SDS scores among participants with medication overuse are high compared with even those with addiction to more well established addictive drugs.

Two recent studies have evaluated dependency in MOH patients from headache clinics using DSM-IV dependency criteria.8 9 Both studies found levels of dependency among MOH patients remarkably similar to ours. These findings support the involvement of dependency in MOH and that the SDS indeed addresses a relevant issue. Furthermore, one previous study compared the Leeds Dependency Questionnaire scale in patients with MOH and drug addicts, and found that the dependency score was similarly increased in both groups.29 It is also notable that number of years of medication overuse was lower and SDS scores higher for more centrally acting medications (opioids, triptans, ergotamines) compared with simple analgesics and combination analgesics. This may also be taken as an indication that SDS is related to the addictive potential of specific drugs.

Our results are of general importance for physicians, and in particular for general practitioners who see the majority of patients with headache, including those with chronic headache.30 The management of patients with chronic headache commonly represents a “headache” for the physician. The presence of medication overuse and dependency-like behaviour will necessarily affect both acute and prophylactic medication and follow-up strategies. Not even simple analgesics are “safe drugs” in the management of headache. In fact, all headache medication requires awareness of the possibility of developing medication overuse. Thus the question of medication overuse must be taken into account in the management of primary frequent and chronic headaches. The question of dependency-like behaviour in MOH should be addressed in future studies of chronic headache. The SDS is a simple and valid tool for identification of people who are likely to have an overuse of acute headache medication.