Circalunar and ultralunar periodicities in women with partial seizures


  • Progesterone Trial Study Group Members: Donald Schomer MD, Edward Bromfield MD, Barbara Dworetzky MD, Sonia Replansky BS, Cynthia L. Harden MD, Blagovast Nikolov MD, Alison Pack MD, Alison Randle BS, Barbara Jobst MD, Gregory Holmes MD, Emily Clough BS, Page Pennell MD, Melanee Newman RN, Gregory Krauss MD, Peter Kaplan MD, Faith Muigai RN, Teresa Tran MD, Sabina Gapany PharmD, Eva Andermann MD, Frederick Andermann MD, Suha Mercho MD, Joyce Liporace MD, Michael Sperling MD, Gwendolyn Taylor BS, Laura Kalayjian MD, Christianne Heck MD and Sandra Oviedo BS.

Address correspondence to Mark Quigg, M.D., M.Sc., University of Virginia, Department of Neurology 394, Charlottesville VA 22908, U.S.A. E-mail:


Purpose: Little consensus exists for the definition of catamenial epilepsy. Few studies have evaluated the periodicity of seizures to test the hypothesis that seizures in women have periodic patterns of occurrence independent of a priori hormonal considerations. In the present study, we determined seizure periodicity according to a “menstrual clock” provided by a common phase marker of the onset of menses.

Methods: Seizure and menstrual diaries of ∼3 months duration were obtained from women enrolled in a trial of hormonal therapy for localization-related epilepsy. Midluteal progesterone levels identified ovulatory (≥5 ng/ml, OC) from anovulatory cycles (AC). Individual cycles were normalized to a common phase and period (day 0 = menses onset, intervening days = 28 bins). Periodicity of combined data was estimated with cosinor-nonlinear least squares analysis. Best-fit rhythms were estimated with 95% confidence limits.

Results: 100 patients provided 3344 seizures within 293 cycles (77% OC, 20% AC, indeterminate 3%). OC seizures displayed a circalunar rhythm with peak phase of occurrence at onset of menses. AC seizures also featured a circalunar rhythm that peaked at menses onset but also had ultralunar rhythms of ∼14 and ∼9 days.

Discussion: Seizures in women with epilepsy occur in circalunar rhythms. OC and AC seizures differ in seizure timing with the latter occurring in ultralunar rhythms in addition to the predominant circalunar rhythm. This finding supports the existence of catamenial epilepsy and differences in patterns of seizure occurrence between OC and AC.

Past investigations on catamenial epilepsy have focused on the relationship between seizure occurrence and either menstrual phase or reproductive hormone levels (Murri et al., 1986; Herkes et al., 1993; Herzog et al., 1997; Bauer et al., 1998; Herzog et al., 2004). Indeed, the vagaries of self-reporting, variations in menstrual cycle length or regularity, or occurrences of seizures in phases not predicted by menses, lead to controversies in definitions and in calculation of incidence of catamenial epilepsy (French, 2005).

In contrast, few studies have evaluated the periodicity of seizures in women to test the hypothesis that they have circalunar (period length ∼29 days) or ultralunar (<29 days) patterns of occurrence independent of a priori hormonal considerations. In other words, we address the question: is there a significant rhythm of seizure occurrence in women with medically intractable epilepsy that is linked to the “clock” of the menstrual cycle?

In the present study, we determined seizure periodicity according to a menstrual clock provided by a common phase marker of the onset of menstrual bleeding.



These data were collected from the first 100 consecutive women with localization-related epilepsy who participated in the baseline phase of an ongoing multicenter, prospective, double-blind investigation of the use of hormonal treatment for medically intractable seizures (Herzog et al., 2004). Ages ranged from 13 to 45 years (mean ± SD: 33.1 ± 6.9). The women had intractable seizures despite trials of at least two antiepileptic drugs. Antiepileptic drug use was as follows: Monotherapy—37 (carbamazepine—9, phenytoin—8, lamotrigine—7, topiramate—2, oxcarbazepine—2, single cases on other monotherapies—9), combination therapy with two antiepileptic drugs—50, polytherapy—10, no therapy—3. Subjects had at least two seizures each month. Catamenial epilepsy was not a selection criterion. None of the subjects took any form of hormonal supplement nor was on a hormonal form of contraceptive. Data evaluated in the current study focus specifically on pretreatment seizure and menstrual cycle dairies provided by each patient.

Data analysis

One problem in documenting the association between seizures and the menstrual cycle is the relative paucity of seizures and short lengths of reliably documented menstrual cycles. One way to surmount these limitations is to pool data among many subjects, but the variability in the duration of menstrual cycles among subjects may obscure underlying rhythms.

To avoid these limitations, we normalized data in reference to a common menstrual phase and period (Fig. 1). The starting phase for each menstrual cycle was identified by the first day of menstrual bleeding. Intervening days between phase markers were then converted to 28 menstrual cycle days, creating a normalized 28 bin circalunar cycle with day 0 designating the onset of menstrual bleeding. Normalized cycles among all subjects were then combined. The resulting time series consisted of the mean number of seizures per normalized day. This method not only allowed combining data with the use of a common menstrual phase marker and period, it also accurately accounted for multiple daily seizures in those patients who experienced catamenial seizure clustering.

Figure 1.

Conversion of a representative menstrual cycle and seizure diary data to normalized 28 “day” bins with phase markers of onset of menstrual bleeding (black). Days with seizures are marked grey.

To determine periodicity of rhythms within the resulting normalized serial data, we used cosinor-nonlinear least squares analysis (Quigg et al., 1998; Quigg & Straume, 2000). The algorithm (Michael Johnson, Pulse, University of Virginia (Johnson et al., 2004)) iteratively summarizes time series data to a series of cosine functions described by period, amplitude, phase, and mean within specified confidence limits. Rhythms were designated significant when amplitude exceeded its lower confidence limit by a single-tailed p-value <0.05.

Midluteal progesterone levels were used to identify ovulatory cycles (OC) (≥5 ng/ml) from anovulatory cycles (AC).


A total of 100 consecutive patients were available for study. Two subjects had at least one full menstrual cycle, three subjects had at least two full cycles, and 95 had three full cycles. The mean duration ± standard deviation of cycles before normalization ranged from 20 to 37 days (mean ± standard deviation, 27.92 ± 2.89). A total of 3470 seizures occurred among all patients. Average daily seizure frequency was 0.07 ± 0.16 for secondary generalized motor seizures, 0.34 ± 0.57 for complex partial seizures, and 0.39 ± 0.64 for simple partial seizures. After normalization to 28-day cycles, 3344 seizures remained after elimination of partial cycles. Of these remaining seizures, 77% occurred during OC (2587/3344) and 20% during AC (673/3344). Ovulatory status was not known in 3% of seizures (84/3344).

Table 1 and Fig. 2 contain best-fit parameters of rhythms of seizures. When all seizures were considered regardless of ovulatory status, a significant circalunar cycle was measured with the peak of seizure occurrence corresponding to the onset of menses. Similar circalunar dominant rhythms were seen in the subgroup of seizures occurring during ovulatory menstrual cycles (31.68 ± 2.41 days) and during AC (28.58 ± 1.09). Rhythms with ultralunar periods were detected in both groups, but in the case of OC were not significant. In the case of the anovulatory subgroup, ultralunar rhythms of 13.99 ± 0.18 days and 8.99 ± 0.05 days were statistically significant.

Table 1.  Menstrual rhythms of group seizure occurrence as described by cosinor-nonlinear least squares analysis
  1. Values are presented for seizures occurring during ovulatory and anovulatory menstrual cycles.*, significant rhythm; SD, standard deviation; p, value from single-tailed z-score of best-fit amplitude; sz/day, mean rate of seizures per day for each group.

Ovulatory seizures n = 2587
Anovulatory seizures n = 673
Figure 2.

(Left) Actual and curve-fitted rhythms of seizure occurrence in reference to normalized menstrual cycles of 28 days. Separate plots show seizures occurring during ovulatory cycles (top), and seizures during anovulatory cycles (bottom). (Right) Percent variance of best-fit accounted for each rhythm, designated by period of the rhythm along the x-axis, listed in Table 1. Arrows designate significant rhythms. Percent variance can be thought of as the “strength” of the determined rhythm. Whereas seizures that occurred during ovulatory cycles have a prominent rhythmic occurrence at a circalunar frequency, seizures during anovulatory cycles occur in a polyrhythmic fashion with mixed ultralunar rhythms.


This study determined that seizures in women with medically intractable localization-related epilepsy occur in a circalunar rhythm during both ovulatory and anovulatory menstrual cycles. Peak incidence of seizures is associated with onset of menstrual bleeding. OC and AC differ in seizure timing with the seizures in the latter occurring in ultralunar rhythms in addition to the predominant circalunar rhythm. This finding supports the existence of catamenial epilepsy and differences in patterns of seizure occurrence between OC and AC.

Because the techniques presented here rely on relatively short durations of group data to calculate circalunar seizure patterns, group measures of rhythmicity do not address adequately the important question of which individual patient can be considered to have catamenial epilepsy; neither can the method calculate the incidence of catamenial epilepsy in a population. Nevertheless, it can calculate the degree to which a population expresses seizures rhythmically, and thus compare patterns of seizure expression among groups. Accordingly, this study is the first to use methods of chronobiology to estimate patterns of seizure occurrence occurring to a “menstrual clock,” thus accounting for variations in menstrual cycle length. The technique demonstrates, independently from any a priori hormonal status, that seizures tend to occur in rhythmic patterns associated with the menstrual cycle. These findings argue against the opinion that catamenial epilepsy does not exist or is rare. We acknowledge that the women who participated in a hormone-treatment trial could bias recruitment in favor of women with catamenial epilepsy.

The findings of the present study are supported by previous work from the same patient cohort (Herzog et al., 2004) as well as from a different patient sample (Herzog et al., 1997) that, with the use of standard statistical methods, showed that seizures are nonuniformly distributed by the four menstrual phases (menstrual, follicular, ovulatory, and luteal) and ovulatory status (ovulatory or anovulatory). Although methodology is not directly transferable, the findings of ultralunar rhythms of seizure occurrence in the present study, especially in the anovulatory group, have analogues in earlier studies that broke catamenial seizures into three groups defined by menstrual phase; these three groupings probably account for the “polyrhythmic” patterns found in the present study (Herzog et al., 1997). The presence of an ultralunar rhythm of ∼14 days of seizure occurrence that is significant in AC and is close to significant in OC has analogues in midcycle catamenial seizures (Herzog Type 2) that occur in association with the preovulatory surge in estrogen (Herzog et al., 1997).

Few studies have evaluated circalunar patterns of seizure occurrence with similar methods. In an analysis of seizure patterns spanning a decade in a patient with partial epilepsy who was postmenopausal, rhythms of seizure occurrence occurred with a period of ∼18 days (Quigg & Straume, 2000) which did not correspond to any rhythms in the current study of premenopausal women. The absence of circalunar rhythms of seizures in the postmenopausal female suggests that other, unknown modulatory factors may contribute to periodic seizure clustering when the circalunar menstrual cycle is absent. Other studies of long-term seizure patterns confirm that clustering of seizures in patients with and without menstrual cycles, is common in intractable epilepsy (Taubøll & Lundervold, 1991); clustering, in the case of the current study, is presumed to be tied to the menstrual cycle, but modulating factors to account for rhythms of different periods are unknown.

The significance of the present study is that it offers additional perspective on modulating factors of seizure occurrence that may be countered by treatment directed at rhythmic influences. Reproductive steroids have influences on brain electrophysiological properties (Scharfman & MacLusky, 2006). The investigation of progesterone supplement in the treatment of epilepsy is currently underway; analyses of rhythmic seizure occurrence will be explored as prognostic factors upon study completion.


This work was supported by NINDS NIH (NIH RO1 NS39466) and an NIH General Clinical Research Center grant (MO1RR01032). We thank Michael Johnson PhD (University of Virginia, Department of Pharmacology) for use of Pulse software.

Conflict of interest: We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. The authors have no conflicts of interest in the reporting of this study.