Research Submissions

309

Medical Oophorectomy With and Without EstrogenAdd-Back Therapy in the Prevention of

Migraine Headache

Vincent Martin, MD; Suzanne Wernke, MD, PhD; Karen Mandell, PharmD; Willie Zoma, MD;Judy Bean, PhD; Susan Pinney, PhD; James Liu, MD; Nabih Ramadan, MD; Robert Rebar, MD

Objectives.—To determine the preventive benefit of “medical oophorectomy” and transdermal estradiol inwomen with migraine.

Background.—Epidemiological studies have demonstrated that declines in serum estrogen levels occurringduring normal menstrual cycles can trigger headache in women with migraine. Prior to this study, no randomizedcontrolled trials have evaluated whether minimizing these hormonal changes pharmacologically can preventheadache.

Methods.—Twenty-one women with regular menstrual cycles and a diagnosis of migraine headache were en-rolled. After a 2.5-month placebo run-in phase, all patients received a subcutaneous goserelin implant (a gonado-

tropin-releasing hormone agonist) to induce a medical oophorectomy. One month later, while continuing goserelin,

participants were randomized to receive a transdermal patch containing 100

�

g of estradiol-17

�

(gonadotropin-releasing hormone agonist/estradiol group, n

�

9) or a placebo patch (gonadotropin-releasing hormone agonist/placebo group, n

�

12) during a 2-month treatment phase. The primary outcome measure was the headache index,which was defined as the mean of pain severity ratings (0 to 10 scale) recorded three times per day by daily diary.Secondary outcome measures included headache disability, headache severity, headache frequency, and the per-centage of headaches with a pain severity rating of 7 or greater.

Results.—The headache index was significantly lower during the treatment period in the gonadotropin-releasing hormone agonist/estradiol group than in the gonadotropin-releasing hormone agonist/placebo group

(

P

�

.025). Similar improvements were observed in the gonadotropin-releasing hormone agonist/estradiol groupfor all secondary outcome measures with the exception of headache frequency, which was unchanged between thegroups. Within the gonadotropin-releasing hormone agonist/estradiol group, there was a 33.7% reduction (95%confidence interval,

�

64.4 to

�

3.0) in the headache index during the treatment phase when compared with theplacebo run-in phase; no difference was seen between those phases within the gonadotropin-releasing hormone ago-nist/placebo group.

Conclusions.—Minimization of hormonal fluctuations with gonadotropin-releasing hormone agonist therapyalone is inadequate to prevent headache in women who are premenopausal with migraine. The addition of transder-mal estradiol to existing gonadotropin-releasing hormone agonist therapy provides a modest preventive benefit.

From the Departments of Internal Medicine (Drs. Martin, Wernke, and Mandell) and Environmental Health (Dr. Pinney), Uni-versity of Cincinnati (Ohio) College of Medicine; the Department of Obstetrics and Gynecology, Baystate Medical Center, Spring-field, Mass (Dr. Zoma); the Center for Epidemiology and Biostatistics, Children’s Hospital Medical Center, Cincinnati, Ohio (Dr.Bean); the Departments of Obstetrics and Gynecology and Reproductive Biology, University Hospitals of Cleveland and CaseWestern University School of Medicine, Cleveland, Ohio (Dr. Liu); Eli Lilly and Company and the Department of Neurology, In-diana School of Medicine, Indianapolis (Dr. Ramadan); and the American Society for Reproductive Medicine, Birmingham, Ala(Dr. Rebar).

Address all correspondence to Dr. Vincent T. Martin, Division of General Internal Medicine, University of Cincinnati College ofMedicine, 231 Albert Sabin Way, ML 6603, Cincinnati, OH 45267-4217.

Accepted for publication October 6, 2002.

310

April 2003

Epidemiological evidence suggests that migraineheadache is influenced by female sex hormones. Mi-graine headaches are 3 times more common inwomen than in men, and their prevalence is the high-est in women between the ages of 30 and 50 during theirpeak reproductive years.

1,2

Reproductive milestonessuch as menarche, pregnancy, and menopause influ-ence the frequency and severity of migraine.

3-6

Mi-graine headaches are reported to occur during theperimenstrual period in 60% of female migraineurs,

7

which is associated with abrupt decreases in serum es-trogen and progesterone levels. Therefore, it has beenpostulated that changes in female sex hormones pro-voke migraine headaches in susceptible women.

Somerville studied 8 patients with “true men-strual migraine” (patients who experienced migraineheadaches during the perimenstrual period and notduring other times of the menstrual cycle).

8

Migraineheadaches could be prevented by administration ofan intramuscular injection of estradiol valerate beforemenstruation, while injections of progesterone had noeffect. The estradiol injections prevented the apparentdecrease of estrogen that occurred shortly before theonset of menstruation, thus supporting the hypothesisthat “true menstrual migraine” was caused by estro-gen withdrawal.

Estrogen-containing medications have been re-ported both to prevent and precipitate headache inwomen with migraine. Oral contraceptives with ahigh estrogen and progesterone content have gener-ally shown a worsening of migraine headaches.

9-11

Mostheadaches occurred during the placebo week, whichrepresented a period of a rapid fall of serum estrogenlevels. In women who are postmenopausal, hormonereplacement therapy has been reported to improveand worsen headache.

12–15

These conflicting past studies led us to study therole of estrogen and medical oophorectomy in the pre-vention of headache. We postulated that changes inserum estrogen or other sex hormones, which occur

as part of a normal menstrual cycle, could triggerheadache in female migraineurs. We designed a clini-cal trial to determine if elimination of these hormonalfluctuations by administration of a gonadotropin-releasing hormone agonist (GnRHa), which inducesa medical menopause, would result in an improve-ment in headache when compared with a normal un-treated menstrual cycle. We also wanted to deter-mine if the addition of a transdermal estradiol patchwould provide further preventive benefit as com-pared to a placebo patch after induction of a medicalmenopause.

METHODS

Patients.—

A randomized, placebo-controlled,pilot trial with a parallel-group design was conductedat the University of Cincinnati. The participants wererecruited from newspaper advertisements and throughthe offices of the principal investigator (one generalinternal medicine practice site and one subspecialtyheadache clinic). Consecutive women with a historyof migraine headache were screened. Inclusion crite-ria included: (1) women aged 21 to 45 years whomenstruated and who fulfilled the InternationalHeadache Society criteria for migraine with or with-out aura,

16

(2) regular menstrual periods occurringevery 25 to 35 days, and (3) self-reported moderate tosevere headaches on 3 or more days during the non-perimenstrual period. (This included all days exceptthose occurring from 2 days before to 7 days aftermenses.) Exclusion criteria included: (1) pregnancyor lactation, (2) previous adverse reactions to GnRHasor estradiol patches, (3) use of estrogen- or progest-erone-containing medications in the preceding 2months, (4) breast or endometrial cancer, (5) throm-boembolic disease or hypercoagulable states, (6) fam-ily history of a hypercoagulable state, (7) diseasesoften associated with headache (eg, sarcoidosis, lupus,etc), (8) refusal to use approved barrier contraceptivemethods during the trial, (9) daytime sleep habits

Key words: migraine headache, headache, estrogen, medical oophorectomy, gonadotropin-releasing hormoneagonists

Abbreviations: GnRHa gonadotropin-releasing hormone agonist, BDI-II Beck Depression Inventory-II

(

Headache

2003;43:309-321)

Headache

311

(8 am to 8 pm), (10) hemiplegic migraine or stroke,and (11) self-report of the intake of greater than 100analgesic medications per 30 days to abort headache.The protocol was approved by the Institutional Re-view Board at the University of Cincinnati. Patientsparticipated in the trial after they had given writteninformed consent. Recruitment for the study began inJuly 1997 and the study ended in March 2001.

Design.—

The study lasted 6.5 months with alead-in month followed by 3 subsequent phases of 2.5months’, 1 month, and 2 months’ duration, respec-tively (Figure 1).

The lead-in month provided a baseline and wasused to determine eligibility for the study. This phaseincluded all days of the first menstrual cycle. A men-strual cycle was defined as the period beginning withthe first day of vaginal bleeding of a given menstrualcycle and ending the day before the start of the nextcycle. The participants collected daily urine samplesand recorded headache severity/disability scores andtheir use of acute migraine medications in a dailyheadache journal. Women were excluded from par-ticipation after the lead-in month if they consumed:(1) 100 or more tablets of analgesics, (2) 50 grams ormore of aspirin, or (3) more than 2 mg of ergotamine

orally or 1 mg rectally per day averaged across themonth. They were also excluded if they were notcompliant in data collection, defined as: (1) greaterthan 10% of the data in the headache diary not re-corded or (2) missing the collection of more than 2consecutive daily urine samples or more than 3 sam-ples in the month. Participants fulfilling the above re-quirements were enrolled in the study.

Phase 1 was the placebo run-in phase and beganon the first day of the second menstrual cycle andended the day before the administration of the firstGnRHa injection 2.5 months later. This phase in-cluded all the days from menstrual cycles 2 and 3and the first 22 to 24 days of the fourth menstrualcycle. All participants received placebo transdermalpatches that were changed every 6 days. Participantsalso received 3 separate placebo GnRHa injections,which were given every 25 to 28 days at office visits 1,2, and 3.

Phase 2 was the period in which GnRHa therapywas initiated. The phase began on the day of the firstGnRHa injection during office visit 4 and ended theday before office visit 5, which was 24 to 27 days later.The GnRHa was administered by subcutaneousplacement of a 3.6-mg goserelin implant (Zoladex)

Fig. 1.—Overview of study protocol. MC indicates menstrual cycle; GnRHa, gonadotropin-releasing hormone agonist. Assumesthat the first two placebo injections were given during the first day of menstrual cycles 1 and 2. The rest of the injections were givenat 25- to 28-day intervals during office visits.

312

April 2003

under the skin of the lower abdomen. This implantcould not be felt or visualized by the patient. Partici-pants continued to apply their placebo estradiolpatches during this phase of the study.

Phase 3, the randomization period, began on theday of office visit 5 and ended 50 to 60 days later onthe day of office visit 7. The participants receivedgoserelin implants at 25- to 28-day intervals duringoffice visits 5 and 6. Participants were randomized toreceive either estradiol or placebo patches. Partici-pants randomized to the GnRHa/estradiol group re-ceived a 100-

�

g patch containing estradiol-17

�

(Cli-mara) applied topically to the skin every 6 days, andthose randomized to the GnRHa/placebo group re-ceived a placebo patch applied topically every 6 days.Placebo patches were identical in form and color tothe estradiol patches.

A double-blind randomized-block design wasused to allocate participants to the GnRHa/estradioland GnRHa/placebo groups. The block size was 4and they were allocated in a 2:2 ratio. The partici-pants were allocated to the treatment assignmentthrough use of a random number sequence devel-oped by the statistician. Participant identificationcodes were stored with the principal investigator,while the treatment allocation codes were stored inthe Department of Pharmacy at the Children’s Hos-pital Medical Center.

The investigators were not blinded to the admin-istration of goserelin or placebo injections duringphases 1, 2, and 3 or to the administration of placebopatches during phases 1 and 2. Investigators were,however, blinded to the administration of placebo orestradiol patches during phase 3. The participants re-mained blinded to the administration of all therapiesthroughout the study.

All participants were instructed to follow a dietlimiting the amounts of caffeine (permitted to haveless than 2 beverages per day), dietary nitrates, andtyramine-containing products. Patients were permit-ted to remain on existing acute and preventive mi-graine medications that had been taken prior to en-rollment, but dosages of the preventive medicationswere required to remain constant. The initiation ofnew acute or preventive therapies was not permittedduring the course of the study.

Study Protocol.—

A total of 7 office visits wereconducted at the Clinical Research Center at theChildren’s Hospital Medical Center in Cincinnati,Ohio (Figure 1). The first office visit was conductedbetween the first and fourth days of the second men-strual period. Subsequent visits were scheduled at 25-to 28-day intervals for the remainder of the study. Anattempt was made to conduct the fourth office visit 23to 25 days after the start of vaginal bleeding of thefourth menstrual cycle. Placement of the implant dur-ing the mid-luteal phase of the menstrual cycle wasthought to minimize any flare of estrogen that couldhave occurred with the first GnRHa administration.

During the first office visit, baseline demographicinformation was collected and a complete history andphysical examination was performed. An abbreviatedphysical examination was conducted during officevisits 3, 5, and 7. An adverse event questionnaire wascompleted during all of the office visits evaluating thecommon side effects related to the use of GnRHaand estrogen therapy. The Beck Depression Inven-tory-II (BDI-II) was administered during office visits1, 3, 5, and 7. All participants completed a daily head-ache diary which included the following: (1) severity,disability, and frequency of headaches; (2) type andquantities of all abortive medications taken for head-ache; (3) days on which patches were changed; and(4) dates of vaginal bleeding. Headache severity wasrated on an 11-point scale with 0 representing “nopain” and 10 representing “pain as severe as it canbe.” Headache disability was rated on the following6-point scale: 0

�

no impairment (headache did notinterfere with functioning), 1

�

slight impairment(headache began to interfere with functioning but thedisability was slight), 2

�

mild impairment (mild in-terference with functioning and disability mild),3

�

moderate impairment (moderate interferencewith functioning and moderate disability, but was stillable to perform activities), 4

�

severe impairment(must have refrained from a social- or work-relatedactivity, but not forced to sleep), and 5

�

total inca-pacitation (had to go to sleep or withdraw into a darkroom). Severity and disability scores were recordedduring the morning (6

am

to 12

pm

), afternoon (12

pm

to 6

pm

), and nighttime (6

pm

to 12

am

). Participantswere instructed not to record the severity or disability

Headache

313

of headaches that occurred between 12

am

and 6

am

because of the difficulty in ascertaining the severityand disability of headaches during sleeping hours.They were asked to record the peak severity and dis-ability of all headaches lasting longer than a half hourduring each period.

Participants collected daily “early morning”urine samples that were frozen and later assayed forestrone glucuronide (E1) and creatinine. Estrone glu-curonide is an inactive metabolite of estradiol and itsurinary concentrations have been found to closelyparallel serum concentrations.

17

Estrone glucuronidewas measured by enzyme immunoassay using anestrone-conjugated horseradish peroxidase system aspreviously described. Urinary creatinine was mea-sured by colorimetric methods. Each urine samplewas assayed in triplicate and the mean of the 3 valueswas used as the daily hormone measure. Mean uri-nary E1 values were divided by the creatinine to cor-rect for the daily concentration variation in urinaryexcretion.

Two serum samples were obtained from the par-ticipants during phase 3 of the study and later assayedfor estradiol-17

�

during peak (within 48 hours afterthe placement of a new patch) and trough (within 48hours prior to a patch change) periods. Serum estra-diol levels were measured by a radioimmunoassaywith an assay sensitivity of 2.2 pg/mL and an intra-assay coefficient of variation of 8% and an interassaycoefficient of variation of 10%.

Statistical Analysis.—

Power calculations demon-strated an 83% probability of detecting an effect size of1.25 (effect size

�

[mean outcome variable

treatment

group

�

mean outcome variable

placebo

group

]/standard deviationoutcome variable

both

groups

) with 12 women per group.With an anticipated dropout rate of 20%, we ex-pected to enroll a total of 30 patients.

The primary efficacy measure of the study wasthe headache index defined as the mean of the head-ache severity scores within a given phase of the study.Only the last 38 days of phases 1 and 3 were used fordata analysis of each respective phase. This windowwas used to ensure sufficient time for the GnRHa tocompletely abolish ovarian steroidogenesis duringphase 3 of the study. The use of the last 38 days alsoguaranteed that both treatment groups had the same

number of menstrual periods during phase 1 of thestudy.

The primary between-group analysis comparedthe headache index between the GnRHa/estradioland the GnRHa/placebo groups during phase 3. Theprimary within-group analysis compared the head-ache index between phases 1 and 3 within each treat-ment group.

Analyses were performed on the following sec-ondary efficacy measures: (1) disability index (meanof disability scores), (2) headache frequency (meannumber of headache days), (3) headache severityscores (the mean of headache severity scores onheadache days only), and (4) percentage of headacheseverity scores with a score of 7 or greater. Anotheranalysis was performed to determine if the headacheindex varied on different patch days during phase 3 inthe GnRHa/estradiol and the GnRHa/placebo groups.

Analyses were performed to ensure that con-founding variables such as depression, vaginal bleed-ing, and acute medication use did not differ duringphase 3 between the treatment groups. For eachtreatment period, acute medications were furtherstratified into the mean number of triptans and totalmedications (triptans and other acute medications)taken per patient for each group.

Assessment of adverse effects was performedcomparing the adverse event questionnaire obtainedduring office visits 3 and 7 between the GnRHa/estradiol and the GnRHa/placebo groups. All othernonserious and serious adverse events experiencedby participants were recorded throughout the study.

Compliance with treatment regimens was as-sessed by an analysis of the serum estradiol and uri-nary E1 levels. Serum estradiol levels and urinary Ellevels obtained during phase 3 were compared be-tween treatment groups.

Analysis of covariance (ANCOVA) or repeated-measure analysis of variance (ANOVA) was used toascertain differences in outcome measures duringphase 3 between treatment groups, while a percentchange analysis was used to assess differences be-tween phases 1 and 3 within treatment groups. Themean of the outcome measures during phase 1 wasused as a covariate for the ANCOVA analyses. Per-cent change was defined as: (outcome variable

phase

3

�

314

April 2003

outcome variable

phase

1

)/(outcome variable

phase

1

)

�

100.A percent change score was obtained for each indi-vidual within each treatment group and a mean ofthese scores was used to determine the percentchange for a given treatment group. The “differenceof the percent change” was defined as the percentchange in the GnRHa/estradiol group subtractedfrom the percent change in the GnRHa/placebogroup. This provided an assessment of the magnitudeof difference in outcome measures between groupsduring the treatment phase. Other statistical testsused, when appropriate, were pooled and 2-sample

t

tests for continuous variables and Fisher exact testsand chi-square tests for categorical variables.

Only patients who completed the study were in-cluded in the data analysis. One participant fromeach treatment group withdrew during phase 3, butneither completed sufficient time to ensure the aboli-tion of ovarian steroidogenesis. A

P

value of .05 was

considered significant for the primary analyses.When multiple analyses were performed, the level ofsignificance for the

P

values was corrected by theBonferroni method. Statistical Analysis Systems pro-gram software was used for the data analysis (version8.2, SAS Institute, Cary, NC).

RESULTS

Enrollment.—

A total of 31 subjects entered theinitial lead-in phase of the study. Twenty-three ofthese subjects were subsequently randomized to 1 ofthe 2 treatment groups during phase 3. Two partici-pants withdrew during phase 3 leaving a total of 21who completed the study (Figure 2).

Demographic Data and Baseline Characteristics.—

As shown in Table 1, baseline characteristics of thestudy participants were statistically comparable be-tween the groups at the time of enrollment. Head-ache outcome measures and the characteristics of the

Fig. 2.—Trial profile. GnRHa indicates gonadotropin-releasing hormone agonist.

Headache

315

menstrual cycles during phase 1 were also compara-ble (Table 2). There was no difference in the head-ache index between the first 19 days of menstrual cy-cles 2, 3, and 4 during the placebo run-in phase(phase 1,

P

�

.45) confirming no regression to themean of headache outcome measures.

Headache Index.—

After controlling for baselinedifferences during phase 1, the headache index was sig-nificantly lower during the treatment phase in theGnRHa/estradiol group than in the GnRHa/placebogroup (

P

�

.025). The ANCOVA was repeated omit-ting various intervals of perimenstrual days from the

Table 1.—Patient Characteristics*

FeatureGnRHa/

Estradiol GroupGnRHa/

Placebo Group

P

Age, mean, y 38.7 39.7 .60Type of headache, No. (%)

Migraine without aura 5 (56) 10 (83) .33Migraine with aura† 4 (44) 2 (17)Episodic/chronic tension-type 4 (44) 7 (58) .67

Use of triptan upon enrollment, No. (%) 7 (77.8) 7 (58.3) .64Use of preventives during study, No. (%) 6 (66) 6 (50) .66No. of past preventives, mean 4.33 4.33 1.00

*A Fisher exact test was used for the analysis of discrete variables and a two-sample

t

test for continuous variables. GnRHa indi-cates gonadotropin-releasing hormone agonist.†Individual participants that experienced both kinds of migraine headache were placed in the migraine with aura group.

Table 2.—Comparison of Treatment Groups During Phase 1*

GnRHa/Estradiol Group

GnRHa/Placebo Group

P

Headache index 2.2 1.7 .55Disability index 1.1 0.8 .44Frequency of headache days, %† 57.6 57.9 .98Headache severity, mean‡ 3.5 2.2 .09Abortive medications, mean§ 42.3 40.9 .94Baseline BDI-II, mean

�

12.6 5.8 .17Vaginal bleeding, days 5.4 7.7 .08Length of cycle, mean, days 26.9 27 .89Urinary estrone glucuronide, mean, ng/mg¶ 33.7 31.6 .10

*A repeated-measures analysis of variance, paired

t

test, or two-sample

t

test were used for the analysis of the above variables.Data were calculated from the last 38 days of phase 1. GnRHa indicates gonadotropin-releasing hormone agonist; BDI-II, BeckDepression Inventory-II.†Percentage of days with headache during the last 38 days of phase 1.‡Excluding all days without headache during the last 38 days of phase 1.§Administered per patient during the last 38 days of phase 1.

�

Completed during office visit 3.¶For menstrual cycles 2 and 3 during phase 1. Urinary estrone glucuronide concentrations were divided by the creatinine concen-tration to correct for the daily variations in urinary excretion.

316

April 2003

covariate for phase 1 to determine if the primary effectof the GnRHa/estradiol group was secondary to an im-provement solely in perimenstrual headaches (Table 3).

The percent change analysis demonstrated an im-provement of the headache index during phase 3when compared with phase 1 in the GnRHa/estradiolgroup, while no significant change was seen in theGnRHa/placebo group (Table 4).

Headache Severity and Disability.—

Headache se-

verity during phase 3 was lower in the GnRHa/estradiolgroup than in the GnRHa/placebo group (

P

�

.003).After controlling for baseline severity scores inphase 1, the percentage of pain severity scores in the7 to 10 range during phase 3 was less in the GnRHa/estradiol group than in the GnRHa/placebo group(

P

�

.02).The headache disability index was significantly

lower during phase 3 in the GnRHa/estradiol groupthan in the GnRHa/placebo group (

P

�

.035). Thepercent change analyses for these outcome measuresare shown in Table 4.

Headache Frequency.—

Headache occurred in 49%of days in the GnRHa/estradiol group and in 55% ofdays in the GnRHa/placebo group during phase 3.There were no significant differences in the overallfrequency of headache days during phase 3 betweenthe 2 treatment groups (

P

�

.70).

Headache Index During Different Patch Days.—

Within the GnRHa/estradiol group, the headacheindex from patch days 1 and 2 was found to be sig-nificantly higher after Bonferroni adjustment thanthose from patch day 6 (

P

�

.001 using a repeated-measure ANOVA). There were no significant dif-ferences between the headache indexes from patchdays within the GnRHa/placebo group (Figure 3).

Table 3.—Analysis of Covariance of the Headache Index After Omission of Perimenstrual Days From

Phase 1 of the Covariate

Perimenstrual Days Omitted*

P

None† .0250 to 2 .045

�

2 to

�

2 .051

�

2 to

�

4 .053

*Day 0 was defined as the first day of menstrual cycle bleeding.For example, 0 to 2 included the first day of vaginal bleedingplus the 2 following days.†This represented the primary analysis of the study.

Table 4.—Percent Change Analysis of Outcome Measures*

Outcome Variable Percent Change Analysis† Difference of Percent Change‡

Headache indexGnRHa/estradiol group

�

33.7 (

�

64.4 to

�

3.0)

�

39.1 (

�

78.3 to 0.0)GnRHa/placebo group 5.1 (

�

22.7 to 33.6)Headache severity

GnRHa/estradiol group

�

23.6 (

�

38 to

�

9)

�

45.9 (

�

82 to

�

10)GnRHa/placebo group 22.2 (

�

8 to 53)Headache disability

GnRHa/estradiol group

�

38.8 (

�

70.0 to

�

7.5)

�

59.9 (

�

117.3 to �2.5)GnRHa/placebo group 21.1 (�25.9 to 68.2)

*A two-sample t test was used to calculate the 95% confidence intervals (in parentheses). All values were rounded to one decimalpoint. GnRHa indicates gonadotropin-releasing hormone agonist.†Defined as (outcome variablephase 3 � outcome variablephase 1)/(outcome variablephase 1) � 100. A negative percent change repre-sented an improvement in the outcome variable during phase 3 when compared to phase 1.‡Defined as the percent changeGnRHa/estradiol � percent changeGnRHa/placebo. A negative difference of percent change represented animprovement in the outcome measures in the GnRHa/estradiol group as compared to the GnRHa/placebo group.

Headache 317

Vaginal Bleeding, Beck Depression Inventory-II, and Abortive Medication Use.—Vaginal bleeding oc-curred on 6.5 days during phase 3 in the GnRHa/estradiol group and on 1.3 days in the GnRHa/placebo group. The number of days of vaginal bleed-ing occurring during phase 3 was more common inthe GnRHa/estradiol group than in the GnRHa/placebo group (P � .04). The episodes of vaginalbleeding in the GnRHa/estradiol group were not typ-ical of normal menstrual periods and were likely theresult of breakthrough bleeding from unopposed es-trogen therapy.

The mean values for the BDI-II during phase 3were 16.9 in the GnRHa/estradiol group and 4.9 inthe GnRHa/placebo group. After controlling for theBDI-II during phase 1, there was no difference inthe BDI-II between the 2 groups during phase 3(P � .09).

The use of total abortive medications and trip-tans alone were not significantly different duringphase 3 between the treatment groups (P � .73 and.74, respectively). There was a nonsignificant de-crease in the use of triptans during phase 3 as com-pared to phase 1 within the GnRHa/estradiol group(P � .045). There was no difference in the use of totalabortive medications within the GnRHa/estradiolgroup (P � .84). There were no differences within theGnRHa/placebo group between phases 1 and 3 re-garding the use of total medications or triptans(P � .25 and .31, respectively).

Serum Estradiol and Urinary Estrone Glucuronide Levels.—The serum estradiol levels were signifi-cantly lower in the GnRHa/placebo group than in theGnRHa/estradiol group during phase 3 (P � .0001for peak levels and .002 for trough levels [Table 5]).The mean urinary E1 levels were also significantlylower in the GnRHa/placebo group (P � .006).

Adverse Events.—The following severe adverseevents were reported in 9 participants during thestudy: (1) an episode of urticaria during phase 1; (2)four episodes of worsening depression, two duringphase 1 and two during phase 2; and (3) four episodesof worsening headaches, one during phase 1, two dur-ing phase 2, and one during GnRHa/placebo treat-ment during phase 3.

Fig. 3.—Headache index during different patch days. GnRHa indicates gonadotropin-releasing hormone agonist. Patch day 1 wasthe day of the patch change and day 6 was the day before the next patch change. Bars represent standard errors.

Table 5.—Mean Serum Estradiol Levels During Phase 3*

IntervalGnRHa/Estradiol

Group†GnRHa/Placebo

Group‡

Peak 51.1 (37.7-64.5) 13.1 (8.1-18.1)Trough 40.4 (24.8-55.9) 8.6 (5.8-11.4)

*Values are pg/mL (95% confidence intervals). To convert topmol/L, multiply the above values by a conversion factor of 3.671.GnRHa indicates gonadotropin-releasing hormone agonist.†Peak levels were obtained in 8 of the participants and troughlevels in 7.‡Peak levels were obtained in 10 of the participants and troughlevels in 11.

318 April 2003

During phase 3, the frequency of minor adverseevents reported on the adverse event questionnaireshowed no significant difference between groups withthe exception of hot flashes, which were more com-mon in the GnRHa/placebo group.

COMMENTSTo our knowledge, this represents the first ran-

domized placebo-controlled trial evaluating the effi-cacy of GnRHas with and without estrogen add-back therapy in the preventive treatment of head-ache in women with migraine. After creation of amedical menopause, the headache index was foundto be 39% lower in the GnRHa/estradiol group ascompared to the GnRHa/placebo group. This sug-gests a preventive role for transdermal estradiol inthis population. When compared to a normal men-strual cycle, there was a 34% improvement in theheadache index during the treatment phase of theGnRHa/estradiol group but no change in the GnRHa/placebo group.

The headache index was primarily a measure ofthe “overall effect” of all types of headache on eachtreatment group. It is unknown if the improvementwithin the GnRHa/estradiol group was secondary toits effect on migraine or tension-type headache as ourpatients also suffered from tension-type headache.This distinction may be less important since a recentstudy suggested that both of these headache typesmight be a spectrum of one headache disorder whenthey both exist in the same patient.18

The decrease in the headache index in the GnRHa/estradiol group was secondary to an improvement inheadaches throughout the menstrual cycle and notjust secondary to an effect on perimenstrual head-aches. This was supported by the ANCOVA in whichremoval of combinations of perimenstrual days fromcovariate had a minimal effect on the level of signifi-cance of the analysis.

Two case series and one case report evaluatedthe role of GnRHa with and without estrogen add-back therapy in the prevention of menstrual mi-graine.19-21 All three studies suggested that both theGnRHa/estrogen and GnRHa/placebo treatmentsprevented true menstrual or menstrually associated

migraine headaches. These prior studies were notdouble-blinded randomized trials and included pa-tients only diagnosed with menstrual migraine head-ache. Both the GnRHa/estradiol and the GnRHa/placebo therapies could be effective in the treatment ofmenstrual migraine, while only the GnRHa/estradioltherapy is effective in the treatment of headachesthroughout the menstrual cycle as demonstrated inour study.

Our secondary analyses demonstrated significantimprovements in headache severity and disability inthe GnRHa/estradiol group. The headache severityscores were reduced by 46% and the disability scoresby 60% in the GnRHa/estradiol group when com-pared with the GnRHa/placebo group. The reductionin severity occurred primarily through a reduction inthe more severe headaches with a pain rating of 7 orgreater. Compared with a normal menstrual cycle,there were reductions of 24% in the headache sever-ity scores and 46% in the disability scores in theGnRHa/estradiol group, while there were no changesin the GnRHa/placebo group.

Our study may have important implications forthe role of estrogen in the pathogenesis of headache.First, minimization of fluctuations in serum estradiolalone is insufficient to prevent headache in female mi-graineurs. Both of our treatment regimens minimizedfluctuations in serum estrogen levels when comparedto a normal menstrual cycle, but only the GnRHa/estradiol group was effective in headache prevention.Second, estradiol may have the ability to prevent andprovoke headache. After induction of a medical meno-pause, transdermal estradiol decreased headache com-pared both to a normal menstrual cycle and the treat-ment phase of the GnRHa/placebo group. It alsoprovoked headache during the first 2 days after thechange of an estradiol patch in the GnRHa/estradiolgroup. Third, small changes in serum estradiol levelsthat occur as part of the transdermal delivery systemcould be provocative of headache. The headache indexwas significantly higher on the day of and the day aftera patch change when compared to the day before apatch change in the GnRHa/estradiol group, yet theserum estradiol levels ranged only from 40 to 51 pg/mL during peak and trough periods. This range, how-ever, may have underestimated the true change in es-

Headache 319

tradiol levels observed with the transdermal patchsince peak and trough levels were not measured at the“exact” time of a patch change.

The effect of estrogen replacement therapy onheadache may be dependent upon the type of estro-gen used and its delivery system.22 Studies in womenwho are postmenopausal have demonstrated thatregimens containing a transdermal estradiol patchhave resulted in fewer migraine headaches than thosecontaining oral conjugated estrogens.23 The benefit ofthe transdermal estradiol patch may be secondary toits ability to maintain constant serum estradiol levelsthroughout the course of the patch, which couldprove advantageous if fluctuating estradiol levelstrigger headaches.24 Another possibility would bethat conjugated estrogens themselves may be morelikely to provoke headache than estradiol. The stabil-ity of serum estrogen levels may also be important, asregimens that interrupt dosing may be associatedwith an increased frequency of headache.15

Ours is the first study to examine the effects oftransdermal estradiol alone in the prevention of head-ache. Studies in postmenopausal women have used atransdermal patch along with an oral progestin toprevent endometrial hyperplasia. These studies dem-onstrated no change in the frequency of migraineheadache after the administration of the transdermalestradiol/progestin therapy when compared with abaseline period.23,25 The addition of progestin therapycould negate a preventive effect of the transdermalestradiol since the combination may precipitate head-ache in susceptible patients.14,26

There are some potential limitations to ourstudy. The patient population represented a selectgroup of premenopausal migraineurs with frequentheadaches who were primarily recruited from a head-ache clinic. Therefore, these results may not be appli-cable to women who are postmenopausal or to thosewith less frequent or severe headaches. Our resultsmay not generalize to other types or doses of estro-gen replacement therapies. Also, the patients mayhave been unblinded to the treatment period sincetheir patterns of vaginal bleeding were altered duringphase 3. This should not have influenced the between-group comparisons because both groups had alter-ations of vaginal bleeding.

The precise magnitude of the treatment benefitof transdermal estradiol is difficult to ascertain. Thesmall sample size of the study led to broad confidenceintervals of our outcome measures for the GnRHa/estradiol group. Therefore, the magnitude of thetreatment effect of transdermal estradiol can likelyonly be categorized in the mild to moderate range.One should consider, however, that the primary out-come measure included both migraine and tension-type headaches when interpreting the results of thisstudy. A large treatment effect (greater than 50%—such as is required for an end point of “frequency” inpreventive trials of migraine headache) would be dif-ficult to obtain in a population of patients withchronic daily headache with this type of “combined”end point. It is likely that the GnRHa/estradiol ther-apy did indeed preferentially affect more severeheadaches (possibly migraine) as it significantly re-duced the frequency of headaches with a severity of 7or greater.

We do not believe that confounding variablessuch as acute medication use, depression, or days ofvaginal bleeding influenced our results. The use ofacute medications and presence of depression (as as-sessed by the BDI-II) were not significantly differentbetween the groups during phase 3 of the study. Dayswith vaginal bleeding during phase 3 were more fre-quent in the GnRHa/estradiol group than in theGnRHa/placebo group. Vaginal bleeding has beenthought to increase the frequency of headache aroundmenstruation possibly secondary to the release ofprostaglandins from a shedding endometrium.27,28

Therefore, it would be expected that episodes of vag-inal bleeding would increase the frequency of head-ache in the GnRHa/estradiol group, which would notexplain our study results.

Despite the fact that menstrual migraine was notan inclusion criterion, women who enrolled in thisstudy tended to have their more frequent, severe, anddisabling headaches during the first 6 days after theonset of menstrual bleeding.29 Women who believedthat their headaches were “hormonally related” mayhave been more likely to enroll in this study. There-fore, our study population was likely composed of pa-tients with menstrually associated migraine headache(patients with migraine occurring in the perimenstrual

320 April 2003

period as well as during nonperimenstrual periods),which may influence the interpretation of the studyresults.

Which patients, if any, would be candidates formedical oophorectomy and estrogen add-back ther-apy? One must evaluate the risks and benefits of suchtherapy. Medical oophorectomy is not a benign ther-apy; the surgery comes with the risk of osteoporosis,depression, irritability, hot flashes, decreased libido,and vaginal dryness. Many of these side effects couldbe negated by estrogen add-back therapy, but a re-cent randomized controlled trial evaluating an estro-gen/progestin combination in women who weremenopausal demonstrated a slight increased risk ofbreast cancer, thromboembolism, cardiovascular events,and cerebrovascular accidents.30 Our study investi-gated short-term therapy with a GnRHa and trans-dermal estradiol, but long-term therapy would requirethe addition of a progestin (either daily or cyclically)to prevent endometrial hyperplasia or cancer or both.The addition of a progestin to existing GnRHa/estra-diol therapy could have an effect on migraine head-aches. A recent case series found that combined ther-apy with a GnRHa, transdermal estradiol, and dailymedroxyprogesterone for 6 months significantly re-duced the headache index in 5 patients with “truemenstrual migraine headache,” when compared to anatural menstrual cycle. Until further larger-scale studieshave been conducted with a GnRHa/estrogen/progestincombination documenting the long-term benefit andsafety of such therapy, we would not recommend med-ical oophorectomy with estrogen add-back as a pre-ventive therapy for the typical female migraineurwith menstrual migraine headache.

The question remains as to whether this therapycould be used as a preventive regimen in women withmore refractory chronic daily headache. Our studytended to recruit women with the following charac-teristics: (1) mean age 39 years, (2) suffering fromchronic daily headache, (3) worsening headachesaround menstruation, and (4) toward the end oftheir reproductive years. Future studies of medicaloophorectomy and estrogen add-back therapy mightfocus on such a population. One might envision thistherapy being used during the turbulent years of peri-menopause (ages 40 to 50) as a bridge to menopause

in women with chronic daily headache and perimen-strual exacerbations of their headaches.

Can these results be extrapolated to women withestrogen add-back therapy after surgical oophorec-tomy? Our study did not address this question. How-ever, a past case series reported freedom from mi-graine headache at 2 months in 17 of 29 patients withmenstrually associated migraine after treatment witha GnRHa and transdermal estradiol.19 Thirteen ofthese 17 women later underwent surgical oophorec-tomy and continued transdermal estradiol therapyfor an additional year. The authors reported recru-descence of migraine headache in only 1 of the 17 pa-tients after surgical oophorectomy. Clearly, furtherstudies need to be done before it can be concludedthat surgical oophorectomy with estrogen add-backtherapy is a viable preventive therapy for migraineheadache.

Our results would suggest that medical oopho-rectomy alone had no impact, either positive or neg-ative, on migraine headache. The reintroduction ofa transdermal estradiol patch after medical oopho-rectomy provided a mild to moderate preventivebenefit, although the precise magnitude of such aneffect was difficult to quantify because of the smallsample size of the trial. Despite an overall preven-tive benefit, even small changes in estradiol levelsthat occur as part of the transdermal delivery systemcan precipitate headaches to a lesser degree. Theseare intriguing preliminary findings that require fur-ther study and confirmation before being used as abasis for treatment recommendations in femalemigraineurs.

Acknowledgments: This study was supported in

part from the National Institutes of Health grant

M01RR08084 and by grants obtained from the Ameri-

can Headache Society, the National Headache Founda-

tion, the American College of Obstetrics and Gynecol-

ogy, Astra-Zeneca, and Merck. Berlex provided the

estradiol and placebo patches for the study. The phar-

maceutical companies did not have any role in the de-

velopment, analysis, or writing of this research project.

Appreciation is expressed to Dr. Stephen Silberstein for

his review of the manuscript, Stacy Poe for statistical

advice, and Victoria Martin for her work as a research

coordinator.

Headache 321

REFERENCES

1. Stewart WF, Lipton RB, Celentano DD, Reed ML.Prevalence of migraine headache in the UnitedStates. Relation to age, income, race, and other so-ciodemographic factors. JAMA. 1992;267:64-69.

2. Lipton RB, Diamond S, Reed M, Diamond ML,Stewart WF. Migraine diagnosis and treatment: re-sults from the American Migraine Study II. Head-ache. 2001;41:638-645.

3. Stewart WF, Linet MS, Celentano DD, Van NattaM, Ziegler D. Age- and sex-specific incidence ratesof migraine with and without visual aura. Am J Epi-demiol. 1991;134:1111-1120.

4. Somerville BW. A study of migraine in pregnancy.Neurology. 1972;22:824-828.

5. Welch KM. Migraine and pregnancy. Adv Neurol.1994;64:77-81.

6. Waters WE, O’Connor PJ, Celentano DD. Epidemi-ology of headache and migraine in women. Neuro-surg Psychiatry. 1971;34:148-153.

7. Nattero G. Menstrual headache. Adv Neurol. 1982;33:215-226.

8. Somerville BW. Estrogen-withdrawal migraine. I.Duration of exposure required and attempted pro-phylaxis by premenstrual estrogen administration.Neurology. 1975;25:239-244.

9. Ryan RE. A controlled study of the effect of oral con-traceptives on migraine. Headache. 1978;17:250-251.

10. Whitty CW, Hockaday JM, Whitty MM. The effectof oral contraceptives on migraine. Lancet. 1966;1:856-859.

11. Dalton K. Migraine and oral contraceptives. Head-ache. 1976;15:247-251.

12. Campbell S, Whitehead M. Oestrogen therapy andthe menopausal syndrome. Clin Obstet Gynaecol.1977;4:31-47.

13. Martin PL, Burnier AM, Segre EJ, Huix FJ. Gradedsequential therapy in the menopause: a double-blindstudy. Am J Obstet Gynecol. 1971;111:178-186.

14. Greenblatt RB, Bruneteau DW. Menopausal head-aches—psychogenic or metabolic? J Am Geriatr Soc.1974;22:186-190.

15. Kudrow L. The relationship of headache frequency tohormone use in migraine. Headache. 1975;15:36-40.

16. Headache Classification Committee of the Interna-tional Headache Society. Classification and diagnos-tic criteria for headache disorders, cranial neuralgiasand facial pain. Cephalalgia. 1988;8(suppl 7):1-96.

17. Munro CJ, Stabenfeldt GH, Cragun JR, Addiego

LA, Overstreet JW, Lasley BL. Relationship ofserum estradiol and progesterone concentrations tothe excretion profiles of their major urinary metabo-lites as measured by enzyme immunoassay and ra-dioimmunoassay. Clin Chem. 1991;37:838-844.

18. Lipton RB, Stewart WF, Cady R, et al. Sumatriptan forthe range of headaches in migraine sufferers: results ofthe Spectrum Study. Headache. 2000;40:783-791.

19. Lichten E. The use of leuprolide acetate in the diag-nosis and treatment of menstrual migraine: the roleof artificially induced menopause. Headache Q.1995;6:313-317.

20. Murray SC, Muse KN. Effective treatment of severemenstrual migraine headaches with gonadotropin-releasing hormone agonist and “add-back” therapy.Fertil Steril. 1997;67:390-393.

21. Holdaway IM, Parr CE, France J. Treatment of a pa-tient with severe menstrual migraine using the depotLHRH analogue Zoladex. Aust N Z J Obstet Gynae-col. 1991;31:164-165.

22. Silberstein SD. Hormone-related headache. MedClin North Am. 2001;85:1017-1035.

23. Nappi RE, Cagnacci A, Granella F, Piccinini F, Po-latti F, Facchinetti F. Course of primary headachesduring hormone replacement therapy. Maturitas.2001;38:157-163.

24. Stumpf PG. Pharmacokinetics of estrogen. ObstetGynecol. 1990;75:9S-14S; discussion 15S-17S.

25. Facchinetti F, Nappi RE, Granella F, Cagnacci A,Tirelli A, Viazzo F. Effects of hormone replacementtreatment (HRT) in postmenopausal women withmigraine. Cephalalgia. 2001;21:452.

26. Magos AL, Brewster E, Singh R, O’Dowd T, BrincatM, Studd JW. The effects of norethisterone in post-menopausal women on oestrogen replacement ther-apy: a model for the premenstrual syndrome. Br JObstet Gynaecol. 1986;93:1290-1296.

27. Benedetto C, Allais G, Ciochetto D, De Lorenzo C.Pathophysiological aspects of menstrual migraine.Cephalalgia. 1997;17(suppl 20):32-34.

28. Silberstein SD, Merriam GR. Estrogens, progestins,and headache. Neurology. 1991;41:786-793.

29. Martin VT, Wernke S, Zoma W, et al. Does thephase of the menstrual cycle have an effect on head-ache? Headache. 2002;42:38-39.

30. Writing Group for the Women’s Health InitiativeInvestigators. Risks and benefits of estrogen plusprogestin in healthy postmenopausal women: princi-pal results from the Women’s Health Initiative.JAMA. 2002;288:321-333.