Abstract
Although migraine is more common among women than men, the only two large, randomized trials of low-dose aspirin for migraine prophylaxis have been conducted in men. As part of the Women's Health Study, an ongoing randomized trial of low-dose aspirin and vitamin E among 39 876 female health professionals aged 45 and older, 1001 women with frequent migraine attacks were assigned to 100 mg of aspirin every other day (n = 525) or aspirin placebo (n = 476). Migraine frequency, as well as severity, duration, and degree of incapacitation, were assessed by self-report on questionnaires 12 months and 36 months after randomization, and also by monthly diaries kept before and after randomization. Women assigned to aspirin reported small and consistent decreases in migraine frequency (59.6% vs. 56.4% assigned to placebo reporting improvement at 36 months; odds ratio 1.13, 95% confidence interval, 0.86–1.48), as well as decreases in severity, duration, and migraine-related incapacitation. These reductions were not, however, statistically significant. These data are compatible with a small treatment effect of low-dose aspirin in the prophylaxis of migraine among middle-aged women.
Introduction
Migraine is a common medical problem. In a mail survey conducted in the United States, 17.6% of females and 5.7% of males reported having one or more migraine attacks per year (1). The prevalence varied with age as well as gender, peaking between the ages of 35 and 45 (1, 2).
One medication that has been assessed as a possible preventive agent for migraine is aspirin. Results from small trials have been inconsistent (3–13). In two large-scale trials of aspirin in the primary prevention of cardiovascular disease, the British Doctors' Trial and the Physicians' Health Study, the effect of low-dose aspirin on migraine prophylaxis in men was evaluated in secondary analyses comparing the numbers self-reporting subsequent migraine in the two treatment groups (14, 15). Both found statistically significant reductions in migraine associated with aspirin use. It must be noted, however, that randomization was not done according to headache diagnosis. No similar large-scale trial has evaluated low-dose aspirin for prophylaxis against migraine among women, even though migraine is much more frequent in women than men. Further, the two large trials in men and many of the smaller trials that have been conducted did not provide details regarding migraine severity, duration, and degree of incapacitation, nor did they attempt to verify self-reported information on migraine using established standardized criteria (16).
Many of the medications currently indicated for migraine prophylaxis, which include beta-blockers, calcium-channel blockers, anticonvulsants, serotonin inhibitors, and tricyclic antidepressants, have significant side-effects while aspirin has a good and well-defined safety profile. Furthermore, low-dose aspirin is being increasingly used in the secondary prevention of cardiovascular disease (17), as well as in the primary prevention of myocardial infarction in men (18). Because it is widely used, safe, and inexpensive, low-dose aspirin would be an attractive agent to consider for prophylaxis against migraine.
The pathophysiology underlying migraine is not yet clearly understood, nor is it clear by what mechanism aspirin can act on migraine prophylaxis. Changes in serotonin concentrations have been implicated as a possible trigger for migraine (19), and aspirin may diminish serotonin-mediated firing of neurones in the hypothalamus and brain stem (20). The finding of activated platelets in migraine sufferers, while not consistently demonstrated (21–23), raises the possibility that regular use of a platelet antagonist such as aspirin may help limit or prevent migraine (20). Aspirin's effects on the central nervous system (24, 25) may also be involved in migraine prophylaxis.
To examine the association between use of low-dose aspirin and migraine in women, we initiated the migraine substudy of the Women's Health Study (WHS), an ongoing randomized, double-blind, placebo- controlled trial testing the balance of benefits and risks of low-dose aspirin and vitamin E in the prevention of cardiovascular disease and cancer among healthy women. The substudy was designed primarily to evaluate the effect of low-dose aspirin on the frequency of migraine among women reporting frequent headaches at baseline, as well as to explore the effects of aspirin on their severity, duration and degree of incapacitation.
Subjects and methods
The subjects and methods of the Women's Health Study have been described in detail elsewhere (26). Briefly, a total of 453 787 female health professionals responded to a mailed invitation to participate in the WHS. Of these, 39 876 women who met the main eligibility criteria for the study, i.e. no prior history of cardiovascular disease or cancer other than nonmelanoma skin cancer and who demonstrated good compliance during the trial's run-in phase, were randomized using a two-by-two factorial design to low-dose aspirin (100 mg every other day) or aspirin placebo as well as to vitamin E (600 IU every other day) or vitamin E placebo. Of the randomized women, 1279 who had reported having at least one migraine per month on the WHS baseline questionnaire were sent an invitation to participate in the migraine substudy along with a migraine substudy questionnaire. A total of 1101 women responded and provided detailed information on the written questionnaire about their migraines. Of these, 1001 reported they had migraine attacks at least monthly and were thus eligible for the study – 525 were allocated to active aspirin and 476 to aspirin placebo. These 1001 women were asked to complete monthly migraine diaries to more accurately classify migraine occurrence and symptoms.
To examine the association between use of low-dose aspirin and migraine, we collected three sets of data: reports of migraine collected from all 39 876 women in the WHS on follow-up questionnaires; responses to detailed questionnaires from the 1001 participants in the migraine substudy collected at baseline and 12 and 36 months after randomization; and monthly diaries kept by the 1001 participants in the migraine substudy.
Analysis of entire randomized group
On the 12-month follow-up questionnaire sent to all 39 876 women randomized in the WHS, we collected information about the occurrence of migraine in the prior year. We calculated using logistic regression the odds ratio (OR) and 95% confidence interval (CI) of reporting migraine on the 12-month questionnaire for women allocated to aspirin compared with women allocated to aspirin placebo. In addition, we categorized all randomized women at baseline according to occurrence of menopause (yes or no) and current use of hormone replacement therapy (yes or no). We repeated the analysis for each of these four groups (premenopausal without hormone replacement therapy, premenopausal with hormone replacement therapy, postmenopausal without hormone replacement therapy, and postmenopausal with hormone replacement therapy), adjusted for age and allocation to vitamin E.
Analysis of the migraine substudy group using questionnaires
On the baseline questionnaire for the migraine substudy, women were asked to answer detailed questions about migraine frequency, severity, duration, functional limitation, and location and characteristics of pain. They were also asked about the frequency and severity of symptoms including aura, visual disturbances or other visual phenomena, nausea and/or vomiting, sensitivity to light, sensitivity to sound, behavioural or personality changes, food cravings, bloating or fluid retention, dizziness or vertigo, sensory symptoms such as numbness or tingling, and unilateral weakness in the face, arms, or legs. This questionnaire information was compared with the International Headache Society (IHS) criteria for categorizing migraine (16). We did not ascertain the number of migraine attacks each woman experienced prior to joining the Women's Health Study, and thus could not establish the additional IHS requirement of at least 5 attacks of migraine without aura and 2 attacks of migraine with aura during one's lifetime. However, since the women in our study reported having migraine at least monthly, they were likely to fulfil the IHS criteria for lifetime frequency.
At 12 months after randomization, the 1001 women in the migraine substudy were sent follow-up migraine questionnaires that asked about migraine frequency (During the past year, do you believe that you had more, fewer, or the same number of migraine headaches than previously?), migraine severity (During the past year, do you believe that your migraine headaches were more severe, less severe, or of similar severity than previously?), migraine duration (During the past year, do you believe that your migraine headaches were longer, shorter, or the same duration than previously?), and degree of incapacitation from migraines (During the past year, do you believe that you were incapacitated due to headaches for more, fewer, or the same number of days than previously?). At 36 months after randomization, the migraine substudy participants were sent a nearly identical follow-up questionnaire asking about changes during the three years since randomization. We defined improvement at 12 and 36 months as any response indicating fewer migraines, less severity, shorter duration, and less incapacitation compared with baseline; we determined no improvement as any response indicating the same number or more migraines, the same or greater severity, the same or longer duration, and the same or more incapacitation.
Using responses to the follow-up migraine substudy questionnaires, we compared migraine frequency, severity, duration, and degree of incapacitation among women allocated to aspirin with those allocated to placebo. We used logistic regression analyses to calculate the odds ratios of improvement in migraine symptoms associated with aspirin vs. placebo. Analyses were carried out according to intention to treat and were adjusted for age (single years) and vitamin E assignment. For each odds ratio, we calculated the 95% confidence interval (27).
Migraine substudy diaries
Participants in the migraine substudy received monthly diaries with grids that allowed them to report for each day whether a headache had occurred, its severity (mild, moderate, or severe), and duration (less than 4 h, 4–12 h, or 13–24 h). The daily grids also contained checklists for symptoms associated with each headache, including aura, nausea, sensitivity to light, and others. In analysing data from the diaries, we classified as migraines any headaches of moderate or severe intensity accompanied by one or more of the following symptoms: aura, sensitivity to light, or nausea.
Participants kept diaries during the run-in phase of the WHS trial (pre-randomization phase) and after randomization (post-randomization phase). We excluded information from the first calendar month of each woman's pre-randomization diary because women frequently received the diary after the start of the month and thus the data for that month were incomplete.
In analysing the diaries, we determined the number of headaches reported on each monthly diary, standardized to a 30-day month. For descriptive purposes, we calculated the average number of headaches of any type per month for each woman during the pre-randomization period and the post-randomization period, as well as the average number of migraines. Using a t-statistic, we compared the mean number of headaches and the mean number of migraines per month averaged over all women in the pre- and post-randomization periods, as well the changes in these means from the pre-randomization to the post-randomization period among those randomized to aspirin and to aspirin placebo. In addition, we used the general estimating equation (GEE) approach to allow for the varying number of diaries and the correlation among diaries for each woman. Poisson regression using PROC GENMOD of SAS was used to model the number of headaches of any type and migraines per month as a function of randomization period, aspirin assignment, and the interaction between them. The interaction term was used to estimate the change in the log number of headaches from the pre-randomization period to the post-randomization period in the aspirin group vs. the placebo group. The exponent function of this term is an estimate of the ratio of number of headaches in the aspirin vs. the placebo group in the post-randomization period, adjusted for the baseline number.
We classified all migraines reported in the diaries as either severe or moderate in intensity, and lasting either 13 h or more or 12 h or less. We used a GEE model to compare intensity and duration of migraine, and modelled the proportions of migraines that were severe or of longer duration using a binomial distribution with a logit link. To estimate the relative odds of having more serious migraine attacks in the post-randomization period than in the pre-randomization period in the aspirin group vs. the placebo group, the intensity and duration of migraine attacks were modelled as a function of randomization period, aspirin assignment, and the interaction between them. We again used the exponent function of the interaction term to estimate the odds ratio for migraines of greater intensity or longer duration in the aspirin vs. placebo groups.
Results
Analysis of entire randomized group
For all 39 876 women randomized in the Women's Health Study, 2418 self-reported occurrence of one or more migraine attacks during the first 12 months of the trial – 1177 among those assigned to aspirin and 1241 among those assigned to aspirin placebo. The odds ratio for migraine among women randomized to aspirin compared with women randomized to aspirin placebo was 0.95 (95% CI, 0.87–1.03) after 12 months of treatment (Table 1). When we analysed the four subgroups of women defined by their menopausal status and current use of hormone replacement therapy, there were no significant differences in the odds ratios between women allocated to aspirin and women allocated to aspirin placebo in any of the subgroups.
Odds ratio and 95% confidence interval (CI) of self-reported migraine on the 12-month follow-up questionnaire among all 39 876 women in the Women's Health Study, according to menopause status and current use of hormone replacement therapy∗
Menopausal status and use of hormone replacement therapy could be accurately classified in 28 913 women.
Analysis of the migraine substudy group using questionnaires
Table 2 presents the baseline characteristics of the 1001 women in the aspirin and placebo groups of the migraine substudy. The two groups were similar with respect to age; body mass index; history of hypertension, diabetes, and parental myocardial infarction before age 60; smoking; alcohol intake; menopausal status; use of hormone replacement therapy; and randomization to vitamin E. However, women allocated to aspirin were more likely to report high cholesterol levels and exercise than women allocated to aspirin placebo.
Characteristics of the 1001 women in the migraine substudy of the Women's Health Study according to treatment allocation
Mean±standard deviation.
Defined as hypertension diagnosed by physician, or self-reported systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg or higher, or treatment for hypertension.
Defined as self-reported physician's diagnosis of elevated cholesterol or treatment for high cholesterol.
Of the 1001 women in the migraine substudy, 705 (70.4%) returned the 12-month migraine substudy questionnaire while 893 (89.2%) returned the 36-month migraine substudy questionnaire. Because women entered the WHS in a staggered fashion, 100% of the subjects had reached their 12 month follow-up when this analysis was undertaken, while 96% had attained 36 months of follow-up.
The follow-up questionnaires assessed improvement in features of migraine. Overall, about half of women reported a lower frequency of migraine at 12 months and about 58% at 36 months compared with baseline. On the 12-month follow-up questionnaire, after adjusting for differences in age and randomized assignment to vitamin E, 16% more women allocated to aspirin self-reported less frequent migraine (OR, 1.16; 95% CI, 0.86–1.57), 24% more women reported less severe migraine (OR, 1.24; 95% CI, 0.92–1.68), 20% more women reported shorter migraine attacks (OR, 1.20; 95% CI, 0.88–1.62), and 23% more women reported less migraine-related incapacitation (OR, 1.23; 95% CI, 0.91–1.66) than women allocated to placebo. (Table 3). These differences, however, were not statistically significant.
Odds ratio and 95% confidence interval (CI) for improvement in migraine characteristics, ascertained from questionnaires, among 1001 women in the migraine substudy of the Women's Health Study∗
705 women(70.4%) returned the 12-month migraine substudy questionnaire and 893 women(89.2%) returned the 36-month questionnaire.
adjusted for age and assignment to vitamin E.
Similar results were seen at the 36-month follow-up. Women allocated to aspirin were more likely to report less frequent migraine (OR, 1.13; 95% CI, 0.86–1.48), less severe migraine (OR, 1.06; 95% CI, 0.81–1.39), shorter duration of migraine attacks (OR, 1.11; 95% CI, 0.85–1.45), and less migraine-related incapacitation (OR, 1.12; 95% CI, 0.86–1.47) than women allocated to aspirin placebo. Again, these differences were not statistically significant.
To ascertain the validity of self-reported migraine, we compared women's self-reports of migraine and their descriptions of migraine characteristics with modified IHS criteria that omitted the lifetime frequency requirement, since we did not collect this information. We defined agreement as fulfilling all of the modified IHS criteria (IHS codes 1.1, migraine without aura, and 1.2, migraine with aura), and likely agreement as fulfilling all but one of the modified IHS criteria (IHS code 1.7, migrainous disorder) (13). Among the 1001 participants in the WHS migraine substudy, the self reports of migraine agreed with modified IHS criteria for 589 (58.9%) of women (including migraine with and without aura). In addition, we found a likely agreement in 271 (27.1%) women. When we restricted our analyses to the 860 (85.9%) women who fulfilled the modified IHS criteria for migraine or likely migraine, the 12-month and 36-month findings were unchanged except for a significant difference in incapacitation at 12 months (OR, 1.45; 95% CI, 1.04–2.02) and severity at 12 months (OR, 1.45; 95% CI, 1.04–2.01). The results were also not significantly different for those whose headaches were classified as with or without aura.
Migraine substudy diaries
Diaries were returned by 864 women in the pre-randomization phase and 790 women in the post-randomization phase, with 750 women returning diaries in both phases. We excluded information from the first monthly pre-randomization diary for all women because some women received their first diary after the beginning of the month and thus these first-month data were incomplete. After these exclusions, 724 women returned diaries with complete information in both phases. The median number of pre-randomization diaries was 3 per participant, with a range of 1–21; the median number of post-randomization diaries was 12 per participant, with a range of 1–31.
In the diaries, women reported 54 409 headaches. In the pre-randomization phase, the mean number of monthly headaches was 5.2 ± 4.6 for women ultimately randomized to aspirin and 5.3 ± 4.6 for women randomized to aspirin placebo. In the post-randomization phase, the mean number of monthly headaches was 4.3 ± 4.4 for women randomized to aspirin and 4.8 ± 4.6 for women randomized to aspirin placebo. Change (which is represented by the mean number of monthly headaches in the post-randomization phase minus the mean number in the pre-randomization phase) was −0.80 for women randomized to aspirin and −0.41 for women randomized to aspirin placebo (P ≤ 0.03).
Of the 54 409 separate headaches reported in the diaries, 16 384 (30.1%) were classified as migraine related (i.e. headaches of moderate or severe intensity accompanied by one or more of the following symptoms: aura, light sensitivity, or nausea). Among completed diaries, the mean number of migraine-related headaches was 1.5 per month.
Table 4 compares changes in the frequency of diary-reported migraine-related headaches among women assigned to aspirin compared with women assigned to aspirin placebo in the pre-randomization and the post-randomization phases of the study, as well as changes in the percentage of severe and long-duration (lasting 13–24 h) migraine. The mean number of migraine-related headaches was lower in women randomized to aspirin compared with women randomized to aspirin placebo in both the pre- and post-randomization phases. The percentage of migraines reported as severe and the percentage reported as lasting 13–24 h are also lower in women randomized to aspirin compared with women randomized to aspirin placebo in the pre- and post-randomization phases. However, changes between the two periods (represented by post-randomization data minus pre-randomization data) were not statistically significant for the mean number of migraine-related headaches, percentage of severe migraines, and percentage of migraines lasting 13–24 h. In addition, the GEE analyses we performed to account for the correlation among diaries indicated no differences in the number of migraine-related headaches per month or the severity or duration of migraines.
Mean number and features of migraine in the pre-randomization and post-randomization phases among 1001 women† in the migraine substudy of the Women's Health Study
Women in aspirin and placebo groups may not total 1001 because of missing data.
Change=mean number of migraines or percentage in post-randomization period minus mean number of migraines or percent in pre-randomization periods.
Ratio of number of migraines per month in the aspirin vs. placebo groups estimated using Poisson regression with generalized estimation equation (GEE) analysis.
Odds ratio for more severe or longer duration migraines in the aspirin vs. placebo groups estimated using logistic regression among migraines with GEE analysis.
Discussion
In the migraine substudy of the Women's Health Study, middle-aged women assigned to 100 mg of aspirin every other day reported on follow-up questionnaires at 12 and 36 months after randomization that they had fewer migraines than women assigned to aspirin placebo. These migraine were also less severe and shorter, and were associated with less incapacitation. Although the trends in the results of the study were always favourable to aspirin, the differences did not achieve statistical significance. However, the confidence intervals were wide, so the data are compatible with a small benefit of low-dose aspirin in migraine prophylaxis. Analyses of information recorded in participants' diaries also showed small but statistically nonsignificant reductions in the severity of migraine between the pre- and post-randomization periods.
Other studies have also evaluated whether aspirin is an effective prophylactic agent for migraine. Most of these have been small studies of less than 50 subjects (3–13) testing aspirin at daily dosages ranging from 80 mg to 1.5 g, either alone or in combination with dipyridamol (4, 5), propranolol (7), flunarizine (11), dihydroergotamine (13), or metoprolol (12). Results from these trials have been inconsistent, with some showing a clear reduction in migraine frequency associated with aspirin use (3, 4, 7, 9, 11–13) and others showing no significant benefit of aspirin therapy (5, 6, 8, 10). In addition, some studies of lower-dose aspirin observed a clear benefit (3, 4, 9, 10) while others using higher-dose aspirin did not (5, 6).
Two large, randomized trials in men have also evaluated low-dose aspirin for migraine prophylaxis. The British Doctors' Trial tested 500 mg of aspirin daily in 5139 healthy male physicians between the ages of 50 and 75; two thirds of the group was allocated to aspirin and one third was asked to avoid aspirin over the 6-year trial. Among those allocated to aspirin there was a statistically significant 29% reduction in the number of self-reported migraine attacks for which medical attention was sought compared with the aspirin avoidance group (15). The Physicians' Health Study was a randomized, double-blind, placebo-controlled trial of 325 mg of aspirin every other day in 22 071 healthy male physicians between the ages of 40 and 84. This study showed a statistically significant 20% decrease (relative risk, 0.80; 95% CI, 0.72–0.88; P = 0.00001) in the number of physicians self-reporting migraine on one or more follow-up questionnaires (at 6, 12, 24, 36, 48, and 60 months after randomization) among those assigned to aspirin compared with those assigned to placebo (14). In neither the British Doctors' Trial or the Physicians' Health Study was randomization done within the headache diagnosis category.
There were a number of differences in the designs of the present study and the two comparable studies in male physicians. Our substudy of 1001 women was smaller than the two studies in men (5139 participants in the British Doctors' Trial and 22 071 in the Physicians' Health Study) and included only women with frequent migraine, while the two trials in men included those with and without a history of migraine. However, in our analyses of all 39 876 women randomized in the WHS, we also observed no statistically significant effect. In the British Doctors' Trial, physicians in the control group were not given placebo but were merely asked to refrain from using aspirin or aspirin-containing medications, possibly leading to bias in the self-reporting of migraine. In both the Women's Health Study and the Physicians' Health Study, control groups received aspirin placebo. In all three studies, information on migraine was provided by self-report on questionnaires that provided no instructions on how to define migraine. Self-reports of migraine by female health professionals in our study were quite valid, with 88.5% meeting modified IHS criteria for migraine and likely migraine. Restricting our analyses to those participants for whom self-reported migraine diagnosis was confirmed by modified IHS criteria did not materially change the results. It is likely that the physicians in the two trials in men were at least as accurate in their self reports. It is also unlikely that differences in compliance could explain the discrepancy between our results and those of the prospective trials in men, given that treatment compliance was high in all three trials – 87% of the group assigned to active treatment in the WHS migraine substudy reported taking at least two-thirds of the study capsules after two years of follow-up, while compliance was 70% in the British Doctors' Trial and 86% in the Physicians' Health Study at the end of each trial.
It is possible that gender differences could account for the variance between our results and those of the two trials in men. Our study was conducted among women 45 years of age or older (mean age, 51 years), who were either postmenopausal or with no intention to become pregnant. This means that the results of the study may not be generalizable to younger women. However, while rates of migraine change with age and menopausal status, this occurred in both the active agent and placebo groups, thus not affecting in any way the validity of the estimates of effect of aspirin on migraine. To investigate possible gender differences, we examined the association between menopausal status, use of hormone replacement therapy, and prevalence of migraine 12 months after randomization among all participants in the Women's Health Study (Table 2). The odds ratio of migraine observed for postmenopausal women not using hormone replacement therapy (in theory, those who could be considered closest to men of that age in their reproductive hormonal milieu) was no different among women allocated to aspirin than among women allocated to placebo. These data provide some indirect evidence that gender differences may not play a large role in the null findings of the present study.
The aspirin dose used in the present study (100 mg every other day) is substantially lower than the doses used in the two trials in men (500 mg daily in the British Doctors' Trial and 325 mg every other day in the Physicians' Health Study). If platelet activation plays a key role in triggering or maintaining migraine, then low doses of aspirin such as those used in the Women's Health Study could potentially prevent or limit migraine attacks, as aspirin-related platelet inhibition occurs even at doses of 100 mg every other day (28). If, however, migraine prophylaxis is more dependent upon controlling the body's response to serotonin or upon some other as-yet unknown mechanism, then low-dose aspirin may not be effective against migraines. For example, it has been demonstrated that aspirin diminishes the serotonin-mediated firing of neurones in the hypothalamus and brain stem (20). It is possible that the aspirin dose used in the Women's Health Study (100 mg every other day), while sufficient to inhibit platelet activation and while postulated to be effective against cardiovascular disease, may not be sufficient to counter serotonergic or other mechanisms involved in initiating or sustaining migraine, effects that may have been possible with the higher aspirin doses used in the two trials in men.
There are a number of limitations that must be recognized in this study, including the self-reported migraine information, the older age range of the population, and the inability to obtain treatment information. However, with regard to studies of migraine prophylaxis among women, the present study is the largest to date and the first to gather information on migraine severity, duration, and degree of incapacitation. In addition, as an a priori substudy of a large, randomized, placebo-controlled trial, the results are less prone to bias and confounding. As it is unlikely that a large-scale randomized trial would ever be mounted to evaluate specifically the role of aspirin in the prophylaxis of migraine, it is important to utilize opportunities to evaluate this question as part of an ongoing trial. In addition, a meta-analysis of all available trials would be helpful in assessing aspirin's possible role in migraine prophylaxis.
In conclusion, results from this large substudy of a randomized, placebo-controlled trial are compatible with a small but statistically nonsignificant effect of 100 mg of aspirin every other day in the prophylaxis of migraine in middle-aged women.
Footnotes
Acknowledgements
We are grateful to the 39 876 dedicated and conscientious health professionals who are participating in this study. We would like to acknowledge the contributions of Dr Patricia Hebert, Ms. Ann Kitross and Ms. Georgina Friedenberg, as well as the expert editorial assistance of Mr P.J. Skerrett.
Supported by grants CA-47988, HL-43851, and NS-34108 from the National Institutes of Health. Dr Benseñor is the recipient of a fellowship from FAPESP (Fundação de Amparo a Pesquisa do Estado de São Paulo), São Paulo, Brazil (97/4223–0).