Gabapentin in the prophylaxis of chronic daily headache

Methods.

Study design.

This was a 21-week multicenter, randomized, double-blind, placebo-controlled, crossover phase IIIB study of GPT in the prophylaxis of CDH (figure 1). The study was approved by each center’s human research ethics committee, recognizing that this represented an off-license use of GPT, which hitherto had only been approved in Australia for treatment of epilepsy.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Study design. Screening assessments: entry criteria, informed consent, history, examination, quality of life, Visual Analogue Scale, laboratory tests. *Titration to 2,400 mg gabapentin or placebo equivalent.

The commonly accepted criterion of efficacy for trials of headache prophylaxis—namely, a 50% reduction in frequency of headaches6–8⇓⇓—was considered inappropriate for this highly refractory headache form. Opinion leaders were canvassed and determined that anything less than an overall mean response of 7.5% difference in headache frequency between the active and placebo arms of the study would have no clinical relevance. This figure for the group response was accepted in the light of clinical experience that many patients with CDH are refractory to all forms of intervention and their lack of response could conceal substantial therapeutic effects in others.

It was calculated that the mean difference of 7.5% would be detected at the 5% level with 80% power if a total of 70 subjects were enrolled. To allow for attrition and ensure adequate power, a target sample size of at least 120 patients was selected.

Inclusion criteria were men or women aged 18 to 65 years who were considered capable of effectively completing the headache diary, complying with instructions, and providing informed consent. Subjects had to experience at least 15 headache days per month for each of the previous 6 months with a headache duration of greater than or equal to 4 hours per day. Female patients had to practice a reliable method of contraception and have a negative screening pregnancy test. Exclusion criteria included confounding CNS disorders; confounding concomitant illnesses; a history of illicit drug or alcohol abuse during the previous year; severe hepatic or renal insufficiency; consumption of more than 4 mg ergotamine per week, 300 mg of sumatriptan, or 15 mg of naratriptan or zolmitriptan; and use of other experimental drugs in the month prior to baseline.

At the initial visit patients were screened for inclusion and exclusion criteria and provided written informed consent. A medical history was obtained and a physical examination performed. Laboratory investigations included full blood count, liver function tests, and serum creatinine, and a pregnancy test was performed where appropriate. Patients completed a Visual Analogue Scale (VAS), recording their assessment of maximal pain severity during the preceding week (0 = no pain to 10 = extreme pain). Subjects were instructed in the use of the headache diary in which they were to record, each day, the occurrence and duration of headache, attack severity (six-point scale, appendix 2), associated features such as nausea, vomiting, light, and sound sensitivity, aggravation by movement and degree of disability associated with the attack (five-point scale, see appendix 2), and consumption of all medications, including nonprescription items. At week 4 ± 4 days, patients were reviewed and headache diaries were checked to confirm compliance with entry and exclusion criteria. Patients completed the Short Form-36 (SF-36)20 quality of life (QOL) questionnaire and a further VAS of headache severity. Those subjects satisfying entry criteria were randomized to either GPT or placebo and the dosage titrated over the following 2 weeks to either 2,400 mg of GPT or the placebo equivalent.

At week 12 ± 4 days, patients were reassessed and a medical history, physical examination, and laboratory tests were repeated. The headache diary was reviewed, QOL questionnaire and VAS completed, adverse events (AE) noted, and any changes in concomitant medications recorded. There was then a 1-week washout period after which patients were re-examined, the headache diary reviewed, a VAS completed, and any AE noted. The crossover medication was dispensed and subjects commenced a further 8 weeks of active therapy consisting of 2 weeks titration and 6 weeks stable dosing of 2,400 mg GPT or placebo equivalent.

At week 21 ± 4 days, patients underwent physical examination and laboratory screening, and the headache diaries were reviewed. The QOL questionnaire and VAS were completed, AE assessed, and any changes in concomitant medications noted. Patients were offered open label ongoing use of GPT.

Results.

Over a 13-month period, 133 patients were enrolled from 12 centers in Australia. The patient population comprised 41 men and 92 women with a mean age of 43 ± 12.1 years, of whom 96% were white. All 133 enrolled patients were eligible for the safety analysis, 121 of whom received placebo and 120 GPT. Withdrawal rates were similar in the two treatment groups (table 1) with similar reasons for withdrawal (table 2). Reasons for exclusion from efficacy analysis included inadequate diary (1), failure to satisfy diagnostic criteria for CDH (4), and fewer than 3 weeks of stable dosage (33: 13 placebo only, 12 GPT only, and 8 in both arms). Thirty of the 33 had less than the titration exposure (14 days) in at least one arm of the study and 14 in both. The data from those excluded were reviewed to consider an intention to treat (ITT) analysis but this proved inappropriate, as there was insufficient information within these patients’ records to allow ITT. After exclusion of these 38 patients, 95 subjects remained evaluable for efficacy.

Table 3 presents the headache diagnoses at screening and demonstrates that almost 2/3 of patients experienced both migraine and tension-type headaches. Length of headache history was 14.6 ± 11.7 years, range 9 months to 51 years (see table 3). The mean headache-free rate (percentage headache-free days over the month) prior to randomization was 10.9% ± 15.5 and the mean VAS of headache severity was 6.5 ± 1.9 (see table 4). The mean headache severity, prerandomization, of 2.5 represented a mild to moderate intensity but 35% reported their headaches to be moderate or more severe (severity scale ≥ 3).

The results of the statistical analysis of the primary efficacy data are presented in table 5. Both nonparametric and parametric results are provided.

Primary efficacy results revealed a mean difference in the headache-free rate between GPT and placebo of 9.1% ± 20.9, which exceeded the minimum mean clinically relevant value of 7.5% specified in the protocol (see table 5, figure 2). A nonparametric sign test of the difference between GPT and placebo in the headache-free rate was significant (see table 5), indicating that patients experienced a higher headache-free rate while on GPT than on placebo.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Headache-free rate (HFR) (% headache-free days per period). Prerandomization vs placebo, p < 0.001; prerandomization vs gabapentin, p < 0.0001; placebo vs gabapentin, p < 0.0005.

The efficacy data were also examined in terms of actual headache days per month (table 6; supplementary figure E-1 [available at www.neurology.org]) and revealed a definite placebo effect (p = 0.0009). There was a greater response to GPT when compared to placebo (p = 0.0004) and prerandomization (p < 0.0001). Similar results were obtained when analyzing those headache days in which headache lasted 4 hours or more (see table 6). Almost 60% of patients had headaches every day (30.4 days per month) during the prerandomization period, as compared with almost 50% who did so while on placebo (p = 0.134) and 40% on GPT (p = 0.0003) (supplementary figure E-2 [available at www.neurology.org]). Of the 95 evaluable patients, 34 subjects had headaches every day in all three phases of the trial. Of the remaining 61 patients, 22 (36%) had a greater than 50% reduction in headache frequency while on GPT compared with 7 (11%) on placebo, p = 0.0006 (McNemar test).

The headache-free rate in each period was also examined using an ordinary least squares generalized linear crossover analysis. No significance was found in this model for either the period or carryover effect, and the difference between GPT and placebo was found to be significant.

Headache duration, averaged over the whole treatment period, revealed significant reductions on both placebo and GPT and GPT over placebo (see table 6). This was found to reflect the effect of reduction of headache frequency because averaging headache duration solely on those days in which headaches occurred only showed significance of GPT over prerandomization duration of headaches (see table 6).

Analysis of those subjects in whom treatment response resulted in their no longer satisfying diagnostic criteria for CDH revealed significant increase among those on GPT over placebo (figure 3). These data further emphasized the effect of frequency on the averaged duration of headaches with no significant difference in duration when averaged for headache days (see figure 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Loss of diagnostic criteria for chronic daily headache. Headache days/month (Hpm) and duration p = 0.0218; <15 Hpm any duration p = 0.0073; <4 hours/day overall p = 0.0412; <4 hours/day on headache days = not significant. *McNemar test for paired categorical data.

Twenty-five subjects were identified as analgesic overusers. Subset analysis of their responses to placebo and GPT showed inferior responses as compared to non-overusers, but this did not achieve significance.

Severity of headaches was significantly diminished by GPT over placebo and both over prerandomization when averaged over all days in the treatment period (table 7) but this partially reflected the impact of the reduction of headache frequency. Table 7 provides the results when severity was assessed only on headache days and the placebo effect was no longer significant. There was a definite GPT effect on headache severity when compared to placebo (p = 0.028). VAS scores also demonstrated reduction in severity for treatment with GPT compared with placebo (p = 0.0006) (see table 7). The effect of treatment on the use of analgesics reflected the efficacy of GPT over placebo (see table 7).

In an attempt to determine if prerandomization headache frequency influenced the magnitude of the response, a regression analysis was preformed, plotting percentage reduction of headache days per month against prerandomization headache days per month in those with a positive outcome. The negative slope of the regression line revealed a tendency for those with fewer headache days in the prerandomization period to experience a slightly greater reduction in headache days both on placebo and GPT but favoring GPT. The slope of the line was not significantly different from zero, indicating that the treatment effect occurred across the entire spectrum of headache frequencies within this sample population (supplementary figure E-3 [available at www.neurology.org]).

GPT also demonstrated a beneficial effect, compared with placebo, on headache-associated symptoms including nausea (p = 0.0271) and photophobia or phonophobia (p = 0.0444), but there was no significant effect on vomiting or the aggravation of headache by movement. Analysis of the degree of disability revealed a similarly beneficial effect of GPT over placebo with respect to the percentage of days in which the headaches affected normal activities (p = 0.024) and headaches necessitating bed rest (p = 0.001). There was also a benefit of GPT over placebo for some QOL domains including bodily pain (p = 0.009), current general health and vitality (p = 0.0001), social function (p = 0.006), and reported health transition (p = 0.0002).

AE were experienced by 47 (39%) patients on GPT compared with 17 (14%) on placebo. AE experienced by more than 5% included dizziness (21% GPT vs 3% placebo), somnolence (18% GPT vs 5% placebo), ataxia (8% GPT vs 1% placebo), and nausea (7% GPT vs 2% placebo). Nine patients on GPT and five on placebo discontinued treatment as a consequence of AE. Of the nine on GPT, three withdrew due to vomiting and the others due to one or more of dizziness, dysarthria, somnolence, or constipation. There were two serious AE—a cardiac death and a brachial nerve palsy. Both occurred while the patient was on GPT, but neither was considered due to the medication.

Discussion.

CDH does not represent a diagnostic category of head pain but is a descriptive term denoting a temporal profile of headache.1–3⇓⇓ Although CDH is not listed in the International Headache Society (IHS)’s headache classification,21 the term is gaining wide acceptance through recognition of the fact that CDH has a particularly high prevalence in headache clinics5 and its resistance to medications makes CDH a daunting therapeutic challenge.22 Despite this, there are very few double-blind controlled trials in its management and none that specifically examines GPT notwithstanding its increasing impact in the treatment of pain.16,17⇓

The demographics of the 133 patients in this study were comparable to those reported in the literature, including the ratio of women to men of 2.2 in the current study (compared with 1.7 to 4.422–24⇓⇓), age at onset of 29.4 years (compared with 30 to 36.923–26⇓⇓⇓), mean duration of CDH of 14.6 years (compared with 6.4 to 1023,24,26,27⇓⇓⇓), and range of durations of 0.75 to 51 years (compared with 0.5 to 51 years22). Although the mean headache severity was in the mild to moderate range, 35% of the study population reported a greater mean headache severity (compared with 32%28). Subset analysis of treatment response with regard to headache type could not be performed as 64% of subjects experienced mixed headache forms, many of which could not be characterized in terms of IHS criteria. Similar difficulties have been encountered by others in attempting to identify underlying headache types in CDH.22,25–29⇓⇓⇓⇓⇓

Although a minimum of 15 headache days per month with headaches of ≥4 hours duration satisfies the definition for CDH,2 the current study population included almost 60% who had a headache every day. The overall mean frequency was 27.4 ± 4.5 headache days per month in the prerandomization period, demonstrating the intrusive nature of CDH. A similar mean frequency has been reported by others (26.9 ± 4.0).30

Of the 133 patients recruited, only 95 were evaluable for efficacy, representing a redundancy of 29%, which highlights the difficulty of undertaking a double-blind, placebo-controlled, crossover study in CDH. This problem was anticipated and the recruitment target was set accordingly to ensure adequate power and this is the largest reported controlled trial of prophylaxis in CDH.15 A crossover design was adopted for ease of recruitment and internal control but was acknowledged to carry the risk that those who perceived active treatment in the first arm may choose not to proceed with the second phase. Examination of attrition (see table 1) did not support this suggestion as most of the 33 withdrawals occurred very early in the study, prior to completion of titration in the first arm. This made ITT analysis impossible, as there were insufficient data in the excluded records.

The mean of 9.1% reduction in headache days/month on GPT, over placebo, exceeded the selected criterion to demonstrate clinical relevance, thus confirming its benefit in the prophylaxis of CDH. Although the mean reduction in headache days per month of 2.3 on placebo, over prerandomization, and a further 2.7 of GPT, over placebo, may seem modest, this result must be seen in the context of a population in which almost 60% (56/95) had headaches every day in the prerandomization period and more than half (34/56) continued to do so in all phases of the trial. This emphasizes the recalcitrant nature of CDH. The benefit of GPT is demonstrated by the fact that, of the patients who did not experience headache on every day in all three phases of the trial, 36% achieved a greater than 50% reduction in headache frequency, compared with 11% on placebo. As CDH represents a heterogeneous group of headache disorders, of which the individual forms may have variable responsiveness to medications, the criterion of 50% reduction in headache frequency, advocated in migraine trials,31 may be inappropriate for CDH.

Some further benefits of GPT, over placebo, were seen with respect to headache severity, disability, and QOL, but headache duration did not appear to be modified by GPT when compared to placebo, although it did show benefit when compared to prerandomization. Although there are other studies examining prophylaxis in CDH, they are almost invariably open label designs.5,9-11,13–15,32⇓⇓⇓⇓⇓⇓⇓ Although there were more AE on GPT than on placebo, no serious AE were attributed to active therapy in the double-blind study.

The current study indicates that the treatment of CDH can and hence should be evaluated using stringent controlled trial methodology. The results demonstrated a significant placebo response in CDH. GPT proved superior to placebo in terms of reduction in headache frequency, severity, and disability, and an enhanced QOL. Consequent to these benefits there was a reduction in analgesic usage. Whereas GPT appeared to have a greater efficacy in those with lower prerandomization headache frequency, the benefit was seen across the frequency spectrum including those with headaches occurring every day.

Appendix 1

Members of The Australian Gabapentin Chronic Daily Headache Group: R.G. Beran, A.B. Black, C. Bladin, P.J. Brimage, D. Crimmins, G. Danta, G. Donnan, J. Heywood, J. Joubert, G.J. Schapel, P.J. Spira, F.J.E. Vajda, C. Yiannikas, and A. Zagami.

Appendix 2