Seasonality as a risk factor for sudden unexpected death in epilepsy: A study in a large cohort
Bell GS, Peacock JL, and Sander JW (2009) Seasonality as a risk factor for sudden unexpected death in epilepsy: A study in a large cohort. Epilepsia. 2010 May;51(5):773-6
Abstract: PURPOSE: We attempted to establish whether month and season, or lunar phase, are risk factors for sudden unexpected death in epilepsy (SUDEP), as these have all been suggested as potential risk factors. METHODS: We used the dataset of the National Sentinel Clinical Audit of Epilepsy-Related Death; all death certificates of people who died in England and Wales between September 1999 and August 2000 and in which epilepsy was mentioned were reviewed. We identified people who probably died of SUDEP during that time. We compared the number of observed SUDEP cases with those expected according to month of death, season of death, minimum external temperature, mean external temperature, and lunar phase. RESULTS: Of 2,412 deaths, we identified 409 probable SUDEPs. We found no evidence for an association between either mean temperature group or minimum temperature group and SUDEP, although there appeared to be a slight excess of sudden deaths in the coldest (mean temperature) groups. We also found no association between SUDEP and month, season, or lunar phase. DISCUSSION: Some causes of death appear to be related to temperature, but we found no convincing evidence to support a link between temperatures or seasons and sudden unexpected death in epilepsy. It is important to identify risk factors for SUDEP, but it is also important that evidence is established before hypotheses are accepted.
Article
The cause of sudden unexpected death in epilepsy (SUDEP) is currently, by definition, unknown. SUDEP is one of the most common causes of premature death in people with epilepsy, yet it occurs rarely; there are therefore difficulties in elucidating the cause of SUDEP. Over the years, studies have investigated risk factors for SUDEP, and the following have been suggested as consistent risk factors: young adult age, early age of onset of seizures, presence of generalized tonic–clonic seizures, higher seizure frequency, and possibly antiepileptic drug polytherapy (Bell & Sander, 2006). It has been suggested that SUDEP is a seizure-related event. There are several hypotheses about the cause of SUDEP, including cardiac problems, respiratory compromise, cerebral shutdown (Bell & Sander, 2006), and possibly genetic susceptibility (Nashef et al., 2007). It may be that different causes lead to death in different people. It is important that the risk factors for SUDEP be established, and that suggested risk factors are substantiated.
Recently it has been suggested that winter temperatures may be a risk factor for SUDEP (Scorza et al., 2009, 2007). The authors provide interesting hypotheses as to why SUDEP may be more common in the winter, but there is, to the best of our knowledge, no evidence to support that this is indeed the case. It is accepted that there is an excess of deaths overall in winter months in the United Kingdom (Johnson & Griffiths, 2003). It should be noted, however, that the excess of deaths was greater in the winter of 1999–2000 due to “epidemic flu activity” (UK Statistics Authority, 2008).
As well as the hypotheses suggested by the original authors, there are other interesting theories as to why SUDEP may occur more frequently in the winter. Vitamin D levels are lower in northern hemispheres from January to April, and some studies have shown a possible increased risk for developing epilepsy in people born in winter months (Procopio et al., 1997), suggesting that, in some way, vitamin D levels might be relevant to neurologic development and function.
More recently, based on a series of 10 patients, it has also been suggested that SUDEP may be more common around the time of the full moon (Terra-Bustamante et al., 2009). There is less convincing evidence that this is likely to be the case and fewer explanations for this being a risk factor.
Any investigation into risk factors for SUDEP is hampered by the small number of cases recorded as such. SUDEP is frequently coded as “epilepsy” or “status epilepticus” on death certificates. Nevertheless, it is likely that a substantial proportion of deaths recorded as “epilepsy” in young people, without other causes of death listed, are due to SUDEP.
Methods
We used the dataset and the same methodology as in the moderation of the National Sentinel Clinical Audit into Epilepsy-Related Deaths (Hanna et al., 2002) to find probable cases of SUDEP. Specifically, we reviewed the death certificates of people who died in England and Wales between September 1999 and August 2000 and in whom epilepsy was recorded on the death certificate. We searched for deaths in people for whom “epilepsy” was given as the only cause of death (or for whom “epilepsy” was the underlying cause, where the primary cause registered was considered to be the terminal event). We excluded those individuals in whom the cause of death was given as “status epilepticus,” and any in whom there was any other pathology listed that could have caused death. We also excluded any deaths occurring in people aged 60 years or older at death. After a shortlist was drawn up with these criteria, as in the original audit, one author (JWS) examined each case individually to determine whether it should be included in the study as probable SUDEP. During the National Sentinel Clinical Audit, we also examined primary care, secondary care, and postmortem examination for a sample of these cases, and this provided support for the diagnosis in our methodology (Hanna et al., 2002).
The deaths considered all occurred in England and Wales, but we do not know the exact place of death. Daily minimum and maximum air temperatures for the year were retrieved from published data for Maidenhead, a town in southern England (Brugge, 2009). Mean temperatures were calculated as the average of the minimum and maximum temperatures. Moon phases for each date during the relevant time period were taken from the database of “Sky and Telescope” (Sky and Telescope, 2009). For the purposes of this study each phase was taken as the date of the phase ±3 days, as provided by the website.
Because the number of deaths considered as due to SUDEP was small, few deaths would be expected on each day. Therefore, we divided the mean and minimum temperatures each into 10 percentiles. We divided the year into four seasons (December–February, March–May, June–August, and September–November), into three groups based on vitamin D levels in the population during those months (low vitamin D, January–April; high vitamin D, July–October; intermediate vitamin D, May, June, November, December) and into winter (December–March) and summer (April–November) according to the definition used by UK statistics authority.
Statistics
We compared the number of observed probable SUDEPs with those expected using the following groupings: month of death, four seasons, winter and summer using UK statistics authority definition, and seasons based on vitamin D levels. For each analysis the number of deaths to be expected was calculated by assuming the null position that death was not affected by month, season, or temperature. For the analysis by month, this meant using the number of days in each month—for example, the expected number of deaths for January was calculated as (31/366) × total deaths. For the analysis by temperature, we compared the number of probable SUDEPs occurring in each temperature group with those expected according to the number of days in which that range of temperatures occurred. This approach was used for mean temperature and for minimum temperature. Similarly, we compared the number of SUDEPs occurring in each lunar phase with those expected based on the number of days in each lunar phase between September 1999 and August 2000.
The chi-square goodness of fit test was used to test the null hypothesis that SUDEP is not associated with month of year, season, lunar phase, or temperature against a general alternative hypothesis of any association. For each analysis the chi-square goodness of fit statistic was calculated in Excel (Microsoft Office Excel 2003, Redmond, WA, U.S.A.), and the relevant p-value was obtained from Stata (version 10.1, StataCorp, College Station, TX, U.S.A.).
Results
Two thousand, four hundred twelve death certificates were included in the National Sentinel Clinical Audit dataset, of which 409 cases were probable SUDEP. Table 1 shows the distribution by month and season of death. Although there was some variation by month, this difference was not statistically significant. Similarly for season, there was no evidence of an association when season was categorized into four groups, into three groups according to vitamin D levels, or into two groups based on the classification used by the UK statistics authority.
Table 2 shows the distribution by mean and minimum temperature groups, and by lunar phase. There appeared to be a slight excess of sudden deaths in the coldest two mean temperature groups, with 56 deaths compared with 42.5 expected in the coldest group and 53 compared with 39.1 expected in the second coldest group. Nevertheless, overall there was no evidence of an association between mean temperature group and sudden death. There was no clear association between the minimum temperature in 10 groups and SUDEP (chi-square goodness of fit = 6.86, df = 9, p = 0.65).
In the period under review, there were 88 days in the first quarter of the moon, 89 days of full moon, 96 days in the last quarter, and 93 days of new moon. There was no evidence for a relationship between the number of SUDEPs and lunar phase.
Discussion
SUDEP may occur in people with epilepsy who are otherwise healthy, and it is imperative that its risk factor(s) and cause(s) be found as soon as possible so that any preventative measures can be encouraged. It is thus laudable to hypothesize possible risk factors for SUDEP. Even if the risk factors themselves (such as gender, time of birth, or season) cannot be ameliorated, knowledge that they exist may lead to useful information as to the causes of SUDEP.
Some causes of death seem to be related to the temperature. A U.K. study between 1993 and 2003 found a small increase in suicide associated with increase in mean temperature above 18°C (Page et al., 2007). Sudden cardiac death also seems to be related to cold temperatures (Gerber et al., 2006). We performed many different analyses and did not find any convincing evidence to support the hypotheses of an association between either winter temperatures or lunar phase. We did not have enough deaths to investigate a possible direct relationship between temperature and number of deaths on a daily basis. There appears to be a slight increase in deaths in the coldest two centiles but, overall, there was no relationship between temperature centiles and death. It is interesting that any effect, if present, was more pronounced when considering mean temperature than minimum temperature; if an association with temperature is genuine it might be expected to be linked more to minimum temperature. Thus, we could not find any convincing evidence that SUDEP is related to winter temperatures; if there is genuinely any effect it is likely to be only a minor risk factor. A German study investigated the date of death for 46 people who died of SUDEP and also a group of 108 people with epilepsy who died of causes other than SUDEP (Schnabel et al., 2000). They also found no significant effect of month of death overall in the number of SUDEPs or in number of other deaths. Reanalysis of data from the United States (Leestma, 1990 quoted in Persinger & Psych, 1995) also shows no significant effect of month on death on the rate of sudden death.
We found no association whatsoever to support the notion that SUDEP might be related to the lunar cycle. There have been suggestions over the years that seizure frequency may be related to the phase of the moon, but a recent study suggested that it is more likely the brightness of the night sky that contributes to a negative correlation with seizure frequency (Baxendale & Fisher, 2008). Indeed, we were unable to reproduce the effect of the lunar cycle proposed in the article that first suggested this association (Terra-Bustamante et al., 2009) using the dates of death provided and our lunar phase methodology.
Limitations of the Study
This study has several limitations, but a prospective study with a reasonably large number of cases of SUDEP would be a huge logistical undertaking. The initial case ascertainment was done through death certification, which has been shown to be a poor source of case identification in epilepsy (Bell et al., 2004). It is thus likely that we missed many cases of SUDEP. It is, however, unlikely that this introduced a bias into the case ascertainment, as there is no reason that deaths occurring during certain time periods would be more likely to be missed than others. Excluding any deaths with any known underlying pathology and those occurring in people aged 60 and older means that the cause of death is more likely to be SUDEP. In the United Kingdom all cases of sudden death are referred to a coroner and usually subject to an autopsy, so it is unlikely that many deaths certified as being due to epilepsy would, in fact, be due to other conditions visible at postmortem examination.
The deaths investigated occurred in England and Wales, but the ambient air temperatures were taken from a weather station in Maidenhead, southern England. The location and time of death were not known precisely for individual deaths, so we were unable to match accurately each death with an exact measure of ambient temperature. Dividing the mean and minimum temperatures into 10 roughly equal centiles may have reduced the effect of variation of temperature around England and Wales.
In the statistical analysis we tested for a general association for each factor and did not attempt to fit any particular model, such as a sinusoidal pattern. We took this approach because we felt it to be more rigorous to test a general hypothesis first and then to test a particular model if there was evidence for general variability. As it turned out, this was not necessary because no statistical differences were observed. The sample size, although substantial for such a study, is not very large for examining trends with other factors, and so our analysis would not have been able to detect fine variations. Nevertheless, we took great pains to conduct our analyses in an objective and systematic way to avoid bias in our reporting.
Conclusion
We found no evidence to support a link of SUDEP with either winter temperatures or with lunar phase. Nevertheless, the link with winter temperatures may not be entirely false and should, if possible, be investigated with a larger dataset, ideally with accurate data concerning location and time of death.