The human body is a collection of thousands of clocks, each cycling in accordance with their own pacemaker.1 Cells such as neurons can track time with millisecond accuracy.2 Hormonal cycles might have longer periods than cells of 24 h (eg, cortisol), 28 days (eg, oestrogen), or more. The presence of many metabolic cycles with their various peaks and troughs has a fundamental effect on human health. For example, sleep cycles are linked to healthy brain function.3 Understanding the cyclic nature of disease progression is vital for treating diseases that continuously fluctuate through stages of severity, rather than steadily worsening.
Cycles of disease severity have been extensively studied in epilepsy, with data accumulating over hundreds of years. The rate of seizures has long been shown to oscillate regularly over the course of days,4 months,5, 6 and years.7, 8 Modern research has provided extensive documentation of circadian patterns of seizures.9, 10, 11 Longer cycles of seizure frequency are less easily studied, although they have been recorded.12, 13 Monthly seizure rhythms are often interpreted as catamenial seizures that coordinate with hormonal cycles.14, 15 Notably, multiple studies have identified approximate monthly cycles that did not appear to be more prevalent in women.5, 16, 17 Individuals vary in circadian rhythms of epilepsy, with different patients having distinct times of peak seizure onset.16, 17
Although the existence of seizure cycles is clear, their cause is somewhat controversial. Peak seizure times might arise purely from behavioural changes, such as different stress levels at the weekend, or seasonal variation in sleep quality. Alternatively, innate and persistent biological drivers, such as those that govern sleep, menstruation, hibernation, or breeding cycles (eg, in mammals) might affect seizure cycles. Increasing evidence shows that seizures are co-modulated with subclinical epileptic activity that also adheres to patient-specific circadian and multiday cycles. In 2018, Baud and colleagues16 reported that seizures preferentially occurred within a particular phase of the underlying cycles of epileptic activity (referred to as phase locking). Karoly and colleagues17 had previously hypothesised that circadian rhythms of seizure and spike-wave discharges occur with aligned phases. Taken together, these results suggest that subclinical epileptic activity and seizures share similar regulatory factors. Underlying oscillation in metabolic or regulatory factors would modulate both seizure susceptibility and the rate of interictal discharges. A better understanding of seizure cycles might provide new targets for treatment.
Research in context
Evidence before this study
We searched for studies on epilepsy and cycles published in MEDLINE from Jan 1, 1946, to Nov 1, 2016, and Embase from Jan 1, 1974, to Nov 1, 2016, using comprehensive electronic search strategies combining terms “epilepsy”, “seizures”, “convulsions”, “cycles”, “circadian”, “diurnal”, “patterns”, “circaseptan”, and “catamenial” with no language restrictions. Historical publications relating to the same search terms were identified through Google Books. We identified a small number of studies in between 1946 and Aug 23, 2018, and some older historical texts relating to these terms, mostly in relation to cycles of the moon and the postulated connection with seizures. Although many authors have appreciated that cycles in the patterns of epileptic seizures exist, these have been poorly defined, chiefly because no accurate databases of seizure activity over sufficiently long timeframes were available.
Added value of this study
We used two unique databases of seizure activity: the NeuroVista study, which captured continuous EEG recordings from intracranial electrodes for up to 3 years, and the SeizureTracker study, which has data from more than 12 000 patients for periods of up to 8 years. The NeuroVista study captured much higher-resolution data over a shorter period than SeizureTracker, though SeizureTracker contains a much larger sample of seizures from a greater number of patients. This allowed hypotheses regarding the underlying patterns of seizure activity identified in the NeuroVista study to be validated in a much larger population.
Implications of all the available evidence
Marked and highly individual patterns of seizure activity were seen in patients over multiple time scales. Circadian rhythms were most common, but a notable minority of patients had 7-day cycles (circaseptan), as well as much longer cycles (>3 weeks). These cycles were similarly prominent in men and women. The identification of these patterns might allow the development of personalised chronotherapies, more accurate seizure prediction algorithms, and provide an insight into the biological basis of culturally ubiquitous calendar measures such as the week.
Even without fully understanding the mechanisms of seizure cycles, temporal patterns can be incorporated into patient management plans. Chronotherapy, or scheduling medication so that drug concentrations coincide with times of peak seizure propensity has been explored in the treatment of epilepsy.18, 19 Daily cycles of seizures might be caused by peaks and troughs in drug effectiveness due to metabolic cycles. Hence, trying to treat these seizure cycles by drug scheduling might simply shift the timing of patient's peak seizure risk, unless long-acting drugs with stable blood concentrations are used. By contrast, longer monthly cycles are less likely to be caused by medication. There is evidence that longer cycles of epileptic activity might be as strong as circadian rhythms;16 however, the possibility of titrating therapy over weekly or monthly cycles has not yet been explored.
A better understanding of seizure cycles will help advance our understanding and treatment of epilepsy. However, scarcity of reliable, long-term records of patients' seizure times has restricted the ability to investigate temporal rhythms. The study of patterns greater than 24 h is particularly challenging because of the requirement of very long recording periods. Documentation of monthly and annual cycles has generally been limited to small cohorts of tens of patients.15 Other, larger studies have not taken patient-specific patterns into account. Given the individual nature of epilepsy, where each person's seizures might be uniquely caused by the specific wiring of their brain, population approaches will probably never reveal all facets of epilepsy progression. In this retrospective cohort study, we investigated whether repeating (cyclic) patterns of seizure onset aligned with a particular phase of (ie, are phase locked to) an underlying periodic signal in patients with epilepsy.