Migraine aura, a predictor of near-death experiences in a crowdsourced study

Study design

Our objective was to investigate whether people with a history of migraine aura are more likely to have NDE, and vice versa, than people without migraine aura. We used an online platform, Prolific Academic (https://prolific.ac/), to recruit an international sample of laypeople. Like Amazon’s Mechanical Turk, Prolific Academic is a crowdsourcing online platform to recruit human subjects that can be used for research purposes (Kondziella, Dreier & Olsen, 2019; Kondziella, Cheung & Dutta, 2019) and that compares favorably in terms of data quality, including honesty and diversity of participants (Peer et al., 2017). Participants were recruited without any filters except for English language and age ≥18 years, and we excluded participants who had been enrolled in our previous study on NDE and REM intrusion (Kondziella, Dreier & Olsen, 2019). The study was announced under the headline “Survey on NDE and headache” using the following text: “We wish to explore the frequency with which NDE occur in the public. This should take no more than 1.5 min on average (a little bit longer, if you have had such an experience, and a little bit less, if you haven’t). You will be paid 0.20$ after completing the survey. Please note that we might use your anonymous answers when writing a paper.”

From all participants, we collected information about age, gender, place of residence and employment status (data provided automatically by Prolific Academic); if they had frequent headaches; if yes, if these headaches could last longer than 4 h and were associated with visual or non-visual aura (Kaiser et al., 2019); if participants ever had an NDE; if yes, if this experience occurred in a truly life-threatening situation or in a situation that just felt so; if the experience was neutral, pleasant or unpleasant; and all participants with an NDE were asked to provide information about all 16 items of the GNDES, the most widely used standardized tool to identify, confirm and characterize NDE in research (Greyson, 1983). Like in our previous study (Kondziella, Dreier & Olsen, 2019), NDE was defined by a GNDES score ≥7. We evaluated all reports of NDE, irrespective of whether they occurred in truly life-threatening situations or in situations that just felt so. Participants with an NDE (and those who claimed an NDE but scored six or less points on the GNDES) were also given the opportunity to describe this in their own words (optional). We did not inquire about REM sleep intrusion, which we assessed in our previous study (Kondziella, Dreier & Olsen, 2019). Table 1 provides details.

Statistics

Using a very high population size (300,000,000), a confidence level of 95% and a margin of error of 5%, we estimated the required sample size to be 384 participants. However, since previous studies have estimated the frequency with which NDE occur in the public to be 5–10%, including our own on NDE and REM intrusion (Kondziella, Dreier & Olsen, 2019), we decided to enroll approximately 1,000 participants to identify an estimated number of 100 individuals with an NDE.

In univariate analysis, associations between potential predictors (age, gender, migraine aura) for NDE were examined using Chi-square test and t-test for independent samples. Additionally, we used multivariable logistic regression to analyze the association between migraine aura and NDE adjusted for age and gender. The level of significance was 0.05 (two-sided) for all statistical tests. Statistical analysis was performed with SPSS 23.0 (IBM, Armonk, NY, USA).

Ethics

Participants gave consent for publication of their anonymous data. Participation was voluntary, anonymous and restricted to those aged 18 years or older. Participants received a monetary reimbursement after completing the survey, in accordance with the Prolific Academic’s ethical rewards principle (≥$6.50/h). The Ethics Committee of the Capital Region of Denmark waives approval for online surveys (Section 14 (1) of the Committee Act. 2; http://www.nvk.dk/english).

Data availability statement

The de-identified raw data are provided in the Supplemental Files.

Near-death experiences: frequency and phenomenology

A total of 286 participants (28% (95% CI [25–30%])) claimed an NDE. The most frequent symptoms were abnormal time perception (faster or slower than normal; reported by 257 participants; 90%); extraordinary speed of thoughts (n = 169; 59%); exceptional vivid senses (n = 165; 58%); and feeling separated from one’s body, including out-of-body experiences (n = 113; 40%). Participants perceived the situation in which they made their experience slightly more often as truly life-threatening (n = 165; 58%) than not (n = 121; 42%).

However, only 81 of 286 individuals who claimed an NDE reached the threshold of ≥7 points on the GNDES (28% (95% CI [23–34%])). Hence, confirmed NDE were reported by 81 of 1,037 participants (8% (95% CI [6.3–9.7%])) (Figs. 2 and 3). Confirmed NDE were perceived much more often as pleasant (n = 29; 49%) than experiences that did not qualify as NDE according to the GNDES (n = 21; 13%; p < 0.0001; Chi-square test; neutral experiences excluded). Table 2 provides selected written reports from participants with an NDE of ≥7 GNDES points and Table 3 from participants with <7 GNDES points.

Headache and migraine aura

Seven-hundred-twenty of 1,037 individuals (69%) answered “yes” to the following question about a primary headache disorder: “Do you get headaches that are NOT caused by a head injury, hangover, or an illness such as the cold or the flu?” The male-to-female ratio of people who responded “yes” to this question was 1:1.3. Two-hundred-fifty-four of 1,037 individuals (24%) fulfilled criteria (Kaiser et al., 2019) of having experienced a migraine aura at any point during their lifetime. Individuals could have different types of migraine aura. Two-hundred-thirty of 254 (91%) individuals reported having had visual auras, 60 (24%) somato-sensory auras, 49 (19%) motor auras and 21 (8%) aphasic/dysarthric auras. Hundred-seventy-four of 531 women (33%) had a migraine aura and 77 of 500 men (15%). This difference was statistically significant (p < 0.001, Chi-square test; male-to-female ratio: 1:2.2). People with migraine aura were slightly older than people without migraine aura (median age: 30 (IQR: 24–38) years vs median age: 28 (IQR: 22–36) years, p = 0.005, Mann–Whitney Rank Sum Test).

Near-death experiences and evidence of migraine aura

There were no significant associations between confirmed NDE and age (p > 0.6, t-test independent samples) or gender (p > 0.9, Chi-square test). The only significant association was between confirmed NDE and migraine aura: 48 (6.1%) of 783 subjects without migraine aura and 33 (13.0%) of 254 subjects with migraine aura had experienced an NDE (p < 0.001, odds ratio (OR) = 2.29). In multiple logistic regression analysis with age, gender and the interaction of age and gender, none of these potential predictors was significant. However, migraine aura remained significant after adjustment for age (p < 0.001, OR = 2.31), gender (p < 0.001, OR = 2.33), and both age and gender (p < 0.001, OR = 2.33).

The prevalence of NDE

The prevalence of individuals with an NDE is estimated at about 4–8% in the general population (Gallup & Proctor, 1982; Knoblauch, Schmied & Schnettler, 2001; Perera, Padmasekara & Belanti, 2005; Facco & Agrillo, 2012; Chandradasa et al., 2018). In our survey, it was 8%. We found a prevalence of 10% using the same criteria in our previous crowdsourcing online survey on NDE and REM intrusion (Kondziella, Dreier & Olsen, 2019), indicating that this prevalence is quite robust. Unlike most previous reports in which NDE were almost always associated with peace and well-being (Thonnard et al., 2013; Charland-Verville et al., 2014; Martial et al., 2017, 2018; Cassol et al., 2018), we confirmed recent findings that many people find their NDE unpleasant (Charland-Verville et al., 2015; Kondziella, Dreier & Olsen, 2019; Cassol et al., 2019). However, experiences with the cut-off score of ≥7 GNDES points were reported more often as pleasant (49%) than experiences with a lower score (13%).

Migraine aura is a predictor of NDE

Migraine aura was a predictor of NDE in our sample. This association was very stable. Regardless of whether either no adjustment, an adjustment for age, for sex or for both was performed, the ORs for migraine aura only varied between 2.29 and 2.33. However, a potential limitation of our study is the announcement of the internet query in which we stated that we would investigate for NDE and headache. This might have attracted more people with NDE and headache. The overall prevalence for all types of primary headache, including tension-type headache, was 69% in our survey. Tension-type headache is the most common form of headache (Jensen, 2018). Its aggregate prevalence in the general population across different studies was 38% (Jensen, 2018). Yet, in a population-based study in Denmark, a much higher lifetime prevalence of 78% was found (Lyngberg et al., 2005; Jensen, 2018). The high prevalence of primary headaches in our survey is hence within the realm of possibility but raises the question if we have attracted a disproportionate number of people with headache. This could include people with migraine with aura. The observation that 24% of the participants in our survey met criteria for a migraine aura, while population-based studies have estimated this prevalence at only 4% in the general population, renders this indeed likely (Russel et al., 1995). The young average age, typical of an Internet-based study, could have contributed to over-representation of migraineurs with aura. The way we phrased our headache questions could be another reason, as we did not intend to validate a migraine diagnosis according to established criteria (Kaiser et al., 2019). Instead, we used a more inclusive approach to identify people with a high likelihood of having migraine aura because we were not interested in migraine per se but rather in migraine aura as a possible predictor for an NDE (Kaiser et al., 2019). Since population-based studies suggest that spontaneous migraine aura is four times less common in people without typical migraine headache than in people with typical migraine headache (Russel et al., 1995), it is unlikely that the over-representation of people with migraine aura in our survey resulted from the fact that we also included people with migraine aura without typical migraine headache. However, we did not ask whether the aura symptoms lasted at least 5 min (It should be noted that the threshold of >5 min to classify as migraine aura is arbitrary). Accordingly, in humans it has been shown that SD, the pathophysiological correlate of migraine aura, may occur in spatially very limited fields and that the propagation speed in the cortical tissue ranges between ~2 and nine mm/min (Woitzik et al., 2013). On one hand, this could have contributed to the discrepancy between our data and population-based migraine studies. On the other hand, the male-to-female ratio in individuals with migraine aura was 1:2.2 in our survey, which is well in line with the results of population-based studies and supports that we indeed detected variants of migraine aura (Russel et al., 1995). In contrast, the male-to-female ratio of a primary headache disorder, be it tension-type headache, migraine or a rarer headache, was 1:1.3 overall. This ratio is again well in line with the assumption that the vast majority of primary headache sufferers in our survey had episodic tension-type headache (Jensen, 2018).

The recurrent burden of headache may have increased motivation to participate in our survey, although this remains entirely speculative. The important question, however, is whether the combination of NDE and migraine aura disproportionately increased the motivation of affected people to join our study. Mathematically, we deal with three random factors: migraine aura (yes/no), NDE (yes/no), and participation (yes/no). The twofold dependencies between participation and migraine aura or NDE appear unproblematic. In contrast, a threefold dependency between participation, migraine aura and NDE could have produced a spurious association. However, we consider this unlikely because, for instance, the entire survey was finished during such a short time frame (i.e., within 3 h after posting the survey online) that word-of-mouth communication of the survey’s topic seems very unlikely.

Internet-based surveys and more traditional mail-based questionnaires or laboratory-based studies each have their advantages and disadvantages (Kaiser et al., 2019). We suggest that a combination of the different approaches is more meaningful than using just one method (Kondziella, Dreier & Olsen, 2019). On one side, complex clinical and ethical concepts cannot be fully captured by an online survey (Woods et al., 2015; Peer et al., 2017). For instance, we did not inquire about precipitating factors/contexts in which participants had experienced their NDE (although we did so in our previous study on NDE and REM intrusion (Kondziella, Dreier & Olsen, 2019)). Future non-internet-based studies will therefore be necessary to verify that NDE and migraine aura are indeed associated. On the other side, the anonymous character of a crowdsourcing online survey decreases the influence of psychological bias (Woods et al., 2015; Peer et al., 2017), because there is no incentive to satisfy the investigator by exaggerating or inventing memories. There was no monetary incentive in our survey either, since we instructed participants that their reimbursement would be the same regardless of whether they reported an NDE or headache or not. In addition, we recruited a much larger sample than would have been feasible during a conventional survey. Although participants from Europe and North America made up the largest share, ours was indeed a global sample with people from 35 countries and five continents.

NDE and the neurobiology of dying

In the largest prospective multi-center observational trial on AWAreness during Resuscitation (AWARE), 46% of 140 survivors reported memories following their cardiac arrest with seven major cognitive themes (Parnia et al., 2014). Nine percent of the survivors met the criteria for an NDE according to the GNDES. Two percent described awareness with explicit memories of “seeing” or “hearing” real events related to their resuscitation. Importantly, one patient had a verifiable period of conscious awareness during which time cerebral function was not expected (Parnia et al., 2014). Although speculative, it seems likely that there is a neurobiological basis for this observation (Nelson et al., 2006; Parnia et al., 2014; Martial et al., 2019; Peinkhofer, Dreier & Kondziella, 2019). The pathophysiological events that occur during the process of dying are of obvious interest in this regard (Vrselja et al., 2019). The transition from life to death is thus characterized by four major events: loss of circulation, loss of respiration, loss of spontaneous electrocorticography (ECoG) activity, and a terminal SD without repolarization. These four events occur always, but not necessarily in the same order (Dreier et al., 2018, 2019; Carlson et al., 2018). In the most common scenario, arrest of systemic circulation, respiration and ECoG activity develops more or less simultaneously, while terminal SD follows the complete arrest of ECoG activity with a latency of 13–266 s (Dreier et al., 2018). Along this sequence, the invasively recorded direct current (DC)/alternate (AC)-ECoG activity can be roughly divided into four different phases which are illustrated with an original recording from a previous study (Dreier et al., 2018) in Fig. 1B: In phase 1, spontaneous ECoG activity is still measurable; phase 2 is characterized by a complete loss of ECoG activity starting simultaneously in different cortical regions and layers, which is referred to as non-spreading depression of spontaneous activity (Dreier, 2011); in phase 3, the terminal SD starts but, from a phenomenologically point of view, is initially similar to SD spreading in healthy grey brain matter (Fig. 1A) (Dreier & Reiffurth, 2015; Hartings et al., 2017a); and finally, in phase 4 a negative ultraslow potential signals the second phase of terminal SD (Oliveira-Ferreira et al., 2010; Hartings et al., 2017b; Dreier et al., 2018, 2019; Lückl et al., 2018; Carlson et al., 2018).

The pertinent question arising from the AWARE study is whether phase 2 and (the transition to) phase 3 are compatible with a conscious perception by the patient—and hence, might contribute to the pathophysiological mechanisms of an NDE. On closer examination of the experimental data, it is interesting that the non-spreading depression of spontaneous ECoG activity in phase 2 does not result from a loss of synaptic activity, but on the contrary from vesicular release of various transmitters, including GABA and glutamate, leading to an incoherent, massive increase in miniature excitatory and inhibitory postsynaptic potentials that replace the normal postsynaptic potentials (Fleidervish et al., 2001; Allen, Rossi & Attwell, 2004; Revah et al., 2016). This probably leads to gradual depletion of the releasable pool of vesicles in the synaptic terminals, and thereby significantly distorts neuronal interactions (Fleidervish et al., 2001; Revah et al., 2016) (Not only are the miniature potentials small, but the abnormal neuronal desynchronization also prevents these potentials from summing-up, which precludes their measurement using comparatively insensitive methods such as subdural and intracortical ECoG or the even cruder scalp EEG). Initially, neurons are hyperpolarized (Tanaka et al., 1997; Müller & Somjen, 2000). Over time, intracellular calcium and extracellular potassium concentrations gradually increase, while extracellular pH decreases (Kraig, Ferreira-Filho & Nicholson, 1983; Mutch & Hansen, 1984; Nedergaard & Hansen, 1993; Erdemli, Xu & Krnjevic, 1998; Müller & Somjen, 2000; Dreier et al., 2002). Eventually, hyperpolarization turns into neuronal depolarization. When the adenosine triphosphate (ATP) stores are exhausted, ATP-dependent membrane pumps such as the Na, K-ATPase become unable to replenish the leaking ions. Consequently, SD erupts at one or more sites of the cortical tissue and spreads into the environment as a giant wave of depolarization. It is important to understand that this terminal SD marks the onset of the toxic cellular changes that ultimately lead to death, but it is not a marker of death per se, since the SD is reversible—to a certain point—with restoration of the circulation (Hossmann & Sato, 1970; Heiss & Rosner, 1983; Memezawa, Smith & Siesjö, 1992; Ayad, Verity & Rubinstein, 1994; Shen et al., 2005; Pignataro, Simon & Boison, 2007; Nozari et al., 2010; Lückl et al., 2018). Thus, in contrast to what happens during coma or sedation, when the brain dies, it undergoes a massive and unstoppable depolarization process (and hence, a very last state of “activation”) (Dreier, 2011).

Returning to the association between NDE and REM intrusion, it would be interesting to know if also a link exists between miniature excitatory/inhibitory postsynaptic potentials and REM sleep. Information is scarce, but there is indeed evidence that these potentials occur in the healthy brain and are involved in the sleep-wake cycle and both REM and non-REM sleep (Yang & Brown, 2014; Christensen et al., 2014; Sangare et al., 2016). Yet, the connection between these potentials in healthy people, on one hand, and disordered neuronal processing, including NDE, on the other hand, has never been properly investigated.

Another unsolved question is if terminal SD could produce bright light phenomena and tunnel vision similar to what happens during a migraine aura, when SD spread through healthy cortical tissue. In this context, it is particularly thought-provoking that terminal SD is not always the final event, but data from so far three patients indicate that terminal SD can sometimes indeed precede circulatory arrest and initiate a spreading depression of spontaneous activity like that in migraineurs with aura (Fig. 1C) (Dreier et al., 2018, 2019; Carlson et al., 2018). In contrast to migraine aura, activity then remains depressed at the time of cardiac death.

It is important to bear in mind that virtually all humans (and all animals, including insects (Spong, Dreier & Robertson, 2017)) undergo terminal SD at the end of their life, whereas only a minority of people have a migraine aura during their lifetime. Hence, although terminal SD may play a role in the development of NDE, migraine aura during lifetime is probably not required for having an NDE with a bright light at the end of life. However, people with a propensity for migraine aura may be more likely to experience terminal SD while the brain is still electrically active (Fig. 1C). Thus, if terminal SD facilitates NDE, this would suggest that the event of a terminal SD can still be perceived and remembered.

To substantiate or dismiss these speculations, it would be necessary to fully understand how the changing polarization states of approximately 20 billion neurons in the neocortex (Mortensen et al., 2014) create the conscious awareness of an individual, an area of intense but unsolved research (Owen et al., 2006; Giacino et al., 2014; Kondziella et al., 2016; Paulson et al., 2017; Demertzi et al., 2019). This seems important because of the increasing practice of organ donation after cardio-circulatory death (DCD). In countries where DCD is practiced, physicians have reached consensus that death should occur somewhere between a few seconds and 10 min after loss of circulatory function (Boucek et al., 2008; Stiegler et al., 2012; Dhanani et al., 2012; Van Veen et al., 2018). Thus, a survey on postmortem organ donation in the framework of the CENTER-TBI study recently revealed that as many as 10 out of 64 centers (16%) in Europe and Israel immediately begin organ retrieval from the donor after a “flat line electrocardiogram” is detected on the monitor (Van Veen et al., 2018). Critical voices have been raised, however (Rady & Verheijde, 2016; Youngner & Hyun, 2019). Due to the above-mentioned uncertainties in our understanding of the dying process, we think it is indeed prudent to consider if organ removal should first be permitted when the neurons in the donor’s brain no longer exhibit synaptic transmission and alterations of their polarization state. In other words, organ harvesting should perhaps be postponed until the donor’s entire brain has unmistakably reached the negative ultraslow potential phase of terminal SD. It follows that a better understanding of NDE may be relevant to protect the interests of potential organ donors in the context of DCD.