1
Using heart rate, brainwaves, and
stress hormone levels to determine
whether waiting causes stress
Bachelor Thesis Biomedical Sciences
Research proposal based on a pilot study
Amber Hof
| 11675616 | 30-06-2020
Supervised by dr. A.B. Mulder & dr. J.C. van Hoof
2
General Structure
1. Basic Details
2
2. Research Topic
4 – 6
3. Pilot Study Methods
6 – 9
4. Pilot Study Results
9 – 13
5. Pilot Study Discussion
13 – 15
6. Project Proposal
15 – 19
7. References
19 – 21
3
1. Basic Details
1.1 Abstract
Stress can be experienced during several of aspects of life, one of those is the use of public transport, especially waiting for a delayed train. This study will focus on multiple physiological parameters to determine stress in subjects. Research shows that stress is characterized by increased heart rate, decreased HRV, decreased theta power, decreased alpha power, increased high beta power and/or decreased overall beta power, and additionally, an increased cortisol level. Our pilot study concluded that subjects that had to wait long without a time indication experienced more stress than subjects with an indication, signified by an increased heart rate, and lower high alpha and beta power during a CPT after waiting. The objective of the current proposal is to expand and improve the design of the pilot study, in order to gain better insight into how waiting affects travellers. To investigate this, EEG and ECG recordings will be used, in combination with saliva tests to determine cortisol. The outcome of this study can provide a new perspective on what waiting does to the brain, and combined with earlier studies it may serve as a basis for further research into the potentially stressful effects of waiting.
1.2 Summary
Our everyday life is filled with small, stressful events, one of those is the uncertainty of when public transport is going to arrive when it is delayed. The objective of the current proposal is to gain better insight into how waiting affects travellers, and to ultimately alleviate any associated stress. Stress can be measured by examining different parameters; heart rate, brainwaves and the level of stress hormone. Results of this study can be used to implement changes in delay reporting that are beneficial for both public transport providers and passengers.
1.3 Keywords
4
2. Research topic
2.1 General Background
Our everyday life is filled with many small, stressful events, such as when we use public transport; trying to navigate a confusingly arranged train station or searching for the right bus to get to your destination. But one of the most dominant factors that causes stress is the uncertainty of when public transport is going to arrive, for example due to a delay.
The objective of the current proposal is to gain better insight into how waiting affects travellers, with the ultimate goal to alleviate any stress associated with it. The question we aim to answer is whether waiting causes stress, and if an indication of the remaining waiting time can attenuate this stress. The hypothesis is that especially during longer waiting, without a wait time indication, stress will be experienced. Based on this hypothesis, we expect to find an increased heart rate, a decreased heart rate variability, decreased low frequency EEG activity, but increased high frequency EEG activity, and an increased eye blinking rate during waiting or directly afterwards in those participants who have to wait for a longer period of time and without them knowing when the wait will be over.
To research if waiting causes stress and to explain our predictions, we first have to understand how stress is defined. Stress has multiple definitions and remains a relatively vague term in literature. But, it is often described in terms of subjective and objective stress, corresponding respectively to; the psychological and physiological response to a stressor (Föhr et al., 2015). Alternatively, it is described as a threat or disruption of homeostasis, which leads to a stress response and promotes adaptation to new conditions (Morgado & Cerqueira, 2018). Although waiting mostly induces acute stress, it can also contribute to chronic stress which does not only provide a psychological burden, but also affects physical health (Yaribeygi et al., 2017). Moreover, stress and its consequences are a tremendous social and financial burden, for example through absenteeism at work (EU-OSHA, 2014). It is therefore important to understand what happens to the body during everyday stressful events, like anxiously waiting for a delayed train to arrive. These small events may seem insignificant, but multiple small effects still amount to significant consequences. Summarized, stress is both a mental and physical state, posing risks to the person experiencing the stress.
Now we know how stress is defined in literature, we can go into more detail about the effects of stress. Cognition is one of the functions affected by stress, as has been illustrated by multiple studies researching attentional processes, decision-making and memory (Lindau et al., 2016; Moran, 2016; Simonovic et al., 2018; Qi et al., 2017). These processes may seem unimportant for a traveller at a station, but appearances are deceptive, as attention and decision making are certainly important when choosing the right train or deciding what you can do during transfer time. Stress can lead to poorer attentional processes (Moran, 2016) and impairs decision-making processes (Simonovic et al., 2018). In addition, stress induces cognitive vulnerability, which leads to unhealthy behaviours like decreased consumption of healthy foods (Doom & Haeffel, 2013), and can contribute to the negative health effects. However, not all the effects of stress are negative, as (mild) acute stress seems beneficial to learning and memory (Lindau et al., 2016), and seems to shorten reaction times (Qi et al., 2017). To summarize, (mild) acute stress may impair aspects of the cognitive functioning important for people using public transport.
Before we delve deeper into what stress does in specific situations, it is important to understand how stress is diagnosed. In a clinical setting stress is assessed using the presence of specific hormones and characteristics from electrocardiograms, all of which will be used in this study, and thus are explained below.
The first useful parameter to study the effect of waiting on the stress level of subjects is heart rate variability (HRV) (Jönsson, 2007), which is measured by the variation in the beat-to-beat interval in an electrocardiogram (ECG), and allows for an estimation of autonomic nervous system input to the heart
5 (Bandelow et al., 2017). The study of Qi & Gao (2020) found that stress reduced high-frequency HRV, and elevated heart rate (beats per minute, or BPM), which indicates lower activity of the parasympathetic nervous system (Bandelow et al., 2017). Moreover, cardiac activity seems to parallel self-reported stress in test subjects, as higher reported stress levels also correlate with increased BPM (Pakarinen et al., 2018).
Another physiological parameter in addition to HRV is the concentration of stress hormones in bodily fluids, of which the most widely used is cortisol. Cortisol is a good indicator of the presence of a stress response in subjects, and can easily and non-invasively be determined from saliva samples (Bozovic et al., 2013). The cortisol response lags 5-20 minutes after initial stress, and peak blood levels are reached after 10-30 minutes. Since we want to measure cortisol in saliva it is important to know that the transfer from blood to saliva takes place rather quickly, within 2-3 minutes (Bozovic et al., 2013). Therefore, we can use the levels of cortisol to objectively measure the bodily stress response circa 20 minutes after waiting.
The last parameter that will be used to determine if waiting causes stress are the effects on the brain, as stress is related to changes in electroencephalography (EEG) measurements. EEG is an electrophysiological measuring method used to record the electrical activity of the brain, which is the result of voltage fluctuations caused by ionic currents within neurons (Niedermeyer & da Silva, 2005). This electrical activity is often recorded over a period of time, most commonly via multiple non-invasive electrodes on the scalp (Niedermeyer & da Silva, 2005). The most studied waveforms in EEG research are associated with different states of the brain and are defined based on frequency. They include theta, alpha and beta waves, as shown in table 1 (Abhang et al., 2016). Studies will often refer to the power of one of these frequency bands, which is a single computed number that summarizes the contribution of a band to the overall power of the signal (Vallat, 2018). In this study we are going to focus on the power of the above mentioned brainwaves.
Table 1. Characteristics of brainwaves in EEG measurements. Drafted using the paper of Abhang et al. (2016).
Now we know which brainwaves can be observed it is important to know how we can recognize the effects of acute stress, since there are multiple EEG features that seem to be influenced by it. The studies mentioned below used different tasks to measure effects of stress; a Flanker task (Qi & Gao, 2020), a Stroop test (Alonso et al., 2015), and a Sternberg paradigm and N-back paradigm (Marshall et al., 2015). First, the study of Qi & Gao (2020) observed an increased arousal level, although not exactly the same as stress, arousal is related to a stress reaction (Pfaff et al., 2007). Second, Alonso et al. (2015) found a decrease of the high alpha power (11-12 Hz) and an increase in the high beta band (23-36 Hz) after stress induction. Furthermore, acute stress caused power differences in the theta, alpha, and beta bands; all of these were decreased in a stressed state (Giannakakis et al., 2015). Additionally, stress decreased short-lasting amplitude enhancements in the alpha band (Marshall et al., 2015). Thus if we want to know if travellers are experiencing stress, we can measure this by looking at changes in theta, alpha and beta power.
6 In addition to brain activity, another physiological characteristic can also be observed using EEG, namely eye blinks. The frequency of eye blinks correlates with experienced stress in test subjects, as a higher frequency of eye blinks was observed in stressful situations (Haak et al., 2009). Therefore, in addition to the frequency bands stress can be measured by looking at the frequency of eye blinking.
2.2 Pilot study
Recently, a pilot study was carried out by UvA Psychobiology and VU Medicine students, who conducted experiments to research how train travellers experience waiting. This NS Stations Lab took place at Amsterdam Central Station from October 4th to October 8th 2017. The students were guided by dr. Tonny Mulder and Prof. Erik Scherder, and received help from NS researchers Mark van Hagen and Do van Elferen. The goal of the experiments was to start understanding how travellers react to waiting, and if this induces unnecessary stress. In total EEG data was collected from 81 subjects, whom were divided into three distinct groups; 1. control, a short wait (n=30), 2. long wait without time indication (n=26), and 3. long wait with time indication (n=25). These subjects completed a task with a build-in waiting period and continuous performance task (or CPT for short), with which the effect of stress was investigated. The CPT is one of several types of neuropsychological tests that measure sustained and selective attention. Since this pilot study provides the starting point for our proposal, it will be thoroughly dealt with.
This current paper used the Stations Lab data to look at the how waiting affects commuters. Data from the waiting period, and the consecutive first 75 sec. of the CPT were analysed, to determine whether stress occurs during and/or just after waiting. Analysis consisted of processing ECG data to determine HRV and BPM, and processing EEG data to determine high alpha, low & high beta EEG power. The alpha and beta ranges have been indicated by previous studies as frequency bands in which power differences occur during stress (Alonso et al., 2015; Giannakakis et al., 2015), and were therefore analysed in this study. In addition, reaction time during CPT is analysed, since previous research found that acute stress shortens reaction time (Qi et al., 2017). The question addressed in the pilot study is whether having test subjects wait causes stress, and furthermore, if a waiting time indication can mediate this stress. The hypothesis was that during the time that corresponds between the short wait and long wait groups, the groups will show no difference, but that during further waiting the group that has to wait a long time without indication of how long the period is will experience more stress than the group that does have an indication. What we expect to see in that case is a shortened reaction time, an increased heart rate, a decreased HRV, a decreased high alpha power, increased high beta power and decreased overall beta power in the group without indication, when compared to the group that did have a time indication.
3. Pilot Study Methods
Now that the important concepts and our pilot study are introduced, we will discuss the methods and techniques used for recording and processing of the data in the pilot study.
3.1 Experimental design
The experiment consisted of an intake with a questionnaire, where after blood pressure and heartbeat values were measured. This was followed by the waiting experiment and a continuous performance task (CPT), which is detailed below, a second short questionnaire, again measuring blood pressure and heartbeat, and finally, a debriefing in which subjects received a printout of their own EEG data. The flow chart (fig. 2A) shows the consecutive sequence of these steps.
7 The questionnaires included questions about gender, age, weight and mood, and the full questionnaires can be found in the appendix. Questionnaires were completed before and after the EEG task.
The task was designed and presented to the subjects in Stimulus Presenter running in a MATLAB environment and consisted of an introduction with explanation, a practice round, a waiting period, and the CPT. The general structure of the EEG task is shown in figure 2B. The instructions were given before the practice round, and before and after the waiting period, subjects could take as long as they wanted for the instructions, and continue with the rest of the experiment by pressing the spacebar.
Figure 1. Flow chart of the experimental design. (A) Overview of the entire experiment from start to finish, the large arrow shows the part of the total experiment that consists of the EEG task. (B) Chart that shows the different parts of the EEG task, the arrow indicates the course of the experiment, the time taken for each part is indicated above the blocks.
The main goal of the experiment is to test if waiting induces stress, but ideally without the subjects being aware that this was the case. Therefore, the waiting block in the task was presented as an equipment calibrating block. This part of the task is of special interest for the research question, since we are investigating whether waiting causes stress. To be able to distinguish if waiting can induce stress three groups were made; in the short wait group the waiting time lasted about 25 seconds without a time indicating bar, in the group with the long wait without indication this period lasted about 100 seconds without a time indicating bar, and in the group with the long wait with a time indicating bar this period lasted about 90 seconds. Waiting periods could differ slightly between individuals within groups. The groups will henceforward be referred to as SW (short wait), LW (long wait without indication) and LWi (long wait with indication).
The subsequent CPT was done as to further investigate stress. In the CPT the subject is shown pictures of several different means of transportation, followed by a traffic light with a specific colour. The goal of task is for subjects to press the spacebar as fast as possible when two almost colliding trains are shown, followed by a green traffic light. Performance on this task was measured by reaction time.
3.2 EEG & ECG recording
The EEG electrode was placed on the centre of the subjects forehead, position Fpz according to the 10-20 system (Niedermeyer & da Silva, 2004) (fig. 2), along with a ground electrode on the forehead and a reference on the earlobe. The ECG was measured by placing an electrode on the wrist that the subject would not use to press keys during the task.
8
Figure 2. Placement of the EEG electrode according to the 10-20 system. (A) Top view of the 10-20 system, the electrode used in this study is shown in yellow. (B) Side view of the 10-20 system, the electrode used in this study would be next to Fp1 shown here.
3.3 Inclusion of subjects
According to the inclusion criteria, subjects were excluded at different stages of data processing. The main criterium was that a subject must have participated in every aspect of the task without errors in the running of this task or having major disturbances, as well as have a complete data set. For a complete list of criteria we refer to the appendix.
3.4 Subjects
After exclusion the data consists of a total of 37 subjects, all with the intake and outtake questionnaires, an EEG measurement and an ECG measurement. The characteristics of the subjects groups are displayed in table 2.
Table 2. Characteristics of experimental groups.
3.5 Data processing
3.5.1 Questionnaire data processing
Intake and outtake questionnaire data were merged for all subjects, to which the mean reaction times extracted from the task result files were added.
3.5.2 EEG processing
EEG analysis was performed using the EEG recorder software in MATLAB. EEG files were first filtered using a bandpass filter from 0.1 to 48 Hz, to remove the interference caused by the mains. Then the first 75 seconds of EEG from the waiting period and the CPT task were isolated and separated into 3 blocks of 25 seconds for each period, with exception of the short wait data, of which there is only 25 seconds of waiting. For each block the average power (V2/Hz) was calculated for the high alpha (10-12 Hz), low beta (12.5-16 Hz), beta (16.6-20 Hz) and high beta (20.5-28 Hz) frequencies.
9
3.5.3 ECG processing
3.4.3.1 HRV
ECG analysis was performed using the EEG recorder & ECG Tool software in MATLAB, following the same procedure as described above. In short, the waiting period and CPT was divided into 25 sec. blocks. Within the 25 second blocks the timings of the T peaks were determined and RR-interval calculated. To determine HRV the SDRR method was used, which is a HRV time-domain measure, calculated by taking the standard deviation (SD) of all the RR-intervals from a subject over a given period of time (Shaffer & Ginsberg, 2017).
3.4.3.2 BPM
From the RR-interval, the beats per minute (BPM) could be calculated for each point in time. The first 10 seconds of the first block for each period were set at 100% (baseline) and the BPM for remainder of the period was normalized to it, averaged over short periods of 5 seconds. This resulted in 5 data points per block for each subject, which could be used for averaging and further statistical analysis.
3.6 Statistical analysis
Statistical analysis was performed using R, a P-value <0.05 was considered significant if a non-parametric alternative was available for a test. If a non-non-parametric alternative was not available for a test, the parametric test was used and a P-value <0.01 was considered significant, since a non-parametric alternative would have been more conservative. A Shapiro-Wilk test was used to determine if the data was distributed normally. Outliers were determined using a outlier analysis with Z-values (P<0.05) and removed from the data set. Most comparisons were made between ≥ 3 groups, therefore ANOVA was used to determine significant differences, and a Kruskal Wallis test was used as a non-parametric alternative. Comparisons between two groups were made using a two-sample t-test, using the Welsh df modification and the Mann-Whitney U test as a non-parametric alternative. A 2x3 or 3x3 mixed ANOVA was used to determine if there was an interaction between group and time slot, which will be discussed in more detail in the results.
4. Results Pilot Study
The pilot study results follow from the data processing, these results and their statistical significance are discussed in this section. For each mentioned insignificant result plots and tables are available in the appendix.
4.1 Experiment duration and reaction time
To see if there was a significant difference (p<0.05) between the three groups in the reaction time during the CPT, an one-way ANOVA was performed. An outlier analysis excluded 3 subjects, and no significant differences were found between groups or between male and female subjects within groups. ANOVA did show a significant difference (F(6,31) =6.3361, p<0.001) between age groups (fig. 3). The 20-29 age group was significantly faster than the 60-69 (p<0.05) and 70-79 groups (p<0.05), and the 18-19 age group was significantly faster than the 60-69 age group (p<0.05).
10
Figure 3. Mean response time during CPT. Time is given in seconds, subjects were divided into age range groups independent of experimental group. Significance p<0.05 is indicated by *.
4.2 Power changes over time and differences between groups
Foremost, statistical testing showed no significant (p<0.01) differences between groups or interactions during the entire waiting period. However, a 3x3 mixed ANOVA with group as between subjects factor and time slot as within subject factor determined that there were significant differences (p<0.01), but no interactions during the CPT. Bar plots for each bandwidth are shown in figure 4. For high alpha (fig. 4A) there was a significant difference between groups (F(2,105)=6.19, p<0.001), where the average power of the LW group was lower than the power of SW group (p<0.01). The average low beta power (fig. 4B) showed a significant difference between groups (F(2,105)=6.001, p<0.01), the LW group had a lower average low beta power than the LWi group (P<0.01). For beta oscillations (fig. 4C) there was a significant power difference between groups (F(2,105)=7.223, p<0.01), where the lack of a time indication leads to lower beta power compared to the LWi group (p<0.01). High beta power (fig. 4D) also showed a significant difference between groups (F(2,105)=6.332, p<0.01), the LW group had a lower high beta power compared to the LWi group (p<0.01). Overall, these results, with the exception of the high beta frequency, indicate that the long wait without indication group experienced more stress during these 75 sec. of the CPT.
12
Figure 4. Bar plot of mean average power over time for all three groups during 75 sec of CPT. The plots show the course of the normalized average power during 75 sec. of waiting for high alpha (A), low beta (B), beta (C) and high beta (D) oscillations. The bars represent the mean for each group ±SD. The x-axis displays the different time slots; the first 25 seconds, the second 25 seconds and the third 25 seconds, and the y-axis shows the normalized average power in V2/Hz. Significance of p<0.01 is depicted as **. Mixed 3x3 ANOVA showed no significant
differences in average power overtime within groups for any of the oscillations.
4.3 HRV changes over time and differences between groups
Although the EEG results did indicate a difference in stress levels, statistical testing of HRV showed that there were no significant (p>0.01) differences between groups or interactions over time, during both waiting and the 75 sec. of the CPT task analysed.
4.4 Changes in heart rate
The analysis of BPM showed no significant (p<0.05) differences between groups during waiting. However, one-way ANOVA did determine significant differences in BPM between groups during 75 sec. of the CPT (fig. 5). All differences found were between the SW and LW groups at 15 (F(2,35)=6.1663, p<0.01), 40 (F(2,35)=5.3771, p<0.01), 45 (F(2,35)=3.5184, p<0.05), 50 (F(2,35)=5.574, p<0.01), 55 (F(2,35)=4.5252, p<0.05), 60 (F(2,35)=3.5847, p<0.05) and 65 (F(2,35)=3.5262, p<0.05) seconds. At all these points in time the LW group had a significantly higher mean BPM, suggesting that longer waiting without an indication is more stressful than shorter waiting.
13
Figure 5. BPM changes from baseline during 75 sec. of the CPT. The first 10 sec. of the measurements were set as a 100% artificial baseline from which changes were calculated. The thin lines show individual subjects, whereas the thick lines represent the group mean. The vertical black dotted lines show the demarcation of the three 25 sec. time intervals. All differences found were between the short wait and long wait groups at 15 sec. (p<0.01), at 40 sec. (p<0.01), at 45 sec. (p<0.05), at 50 sec. (p<0.01), at 55 sec. (p<0.05), at 60 sec. (p<0.05) and at 65 sec. (p<0.05).
5. Interpretation Pilot Study Results & Discussion
This section will first provide a summary and interpretation of the pilot study results discussed above in a scientific context, after which various methodological discussion points are addressed.
5.1 Summary & interpretation of results
5.1.2 Reaction time
Previous research showed that reaction time shortened in individuals exposed to acute stress (Qi et al., 2017), therefore it was expected that individuals stressed due to the waiting period would have a shorter reaction time. However, no significant difference between the different experimental groups was observed, suggesting that either the subjects were not stressed, or that acute stress does not affect reaction times, which is something that has been observed by Dierolf et al. (2018). The observed significant increase in reaction time with age is not related to stress, but a known phenomenon, a large study by Thompson et al. (2014) showed that reaction times start to increase after the age of 24.
5.2.2 Power differences between groups during waiting & CPT
Stress during the waiting period was expected to be detected by an increased heart rate (Pakarinen et al., 2018), a decreased HRV (Qi & Gao, 2020), a decreased high alpha power (Alonso et al., 2015; Giannakakis et al., 2015), increased high beta power (Alonso et al., 2015 ) and/or decreased overall beta power (Giannakakis et al., 2015). However, during waiting no significant differences between groups were found for any of these markers. This would suggest that the subjects were not stressed during waiting, though this is somewhat difficult to say due to the lack of a baseline measurement, as will be discussed in more detail later on.
14 Significant differences were observed during the CPT directly after waiting, the LW group had lower high alpha (10 -12 Hz) compared to the SW group, and decreased low beta (12.5 -16 Hz), beta (16.6 - 20 Hz) and high beta (20.5 -28 Hz) power compared to the LWi group. This is in line with previous findings about stress, except for the decrease in high beta oscillations, and would suggest that the LW group is more stressed as was expected. However, the anticipation of the in-task trigger could also have influenced differences, since alpha oscillations have long been proposed to play a crucial role in anticipatory attention (Palva & Palva, 2007). In addition, the decrease in average power could also be an effect of the in-task stressors, such as the anticipation of a spacebar click. Moreover, observed differences could also be the result of the CPT, since decreases in alpha and beta power have been observed (Karamacoska et al., 2019). However, since this decline is expected to be similar in all groups, the significant differences may be an indicator that the long wait group became more stressed after the waiting period.
5.2.3 Heart rate during waiting & CPT
Lastly, heart rate and HRV showed no differences between groups during waiting, suggesting that during this time the groups did not experience stress.
During the CPT the LW group did show a significant higher heart rate (BMP) for a prolonged amount of time compared to the SW group, but not compared to the LWi group. This higher heart rate can be an indication of stress in this group, but this is not supported by the HRV results. Additionally, the increase in heart rate could also be an effect of the in-task stressors. Notably, the differences in power were mostly related to the two long waiting groups, whereas the differences in BPM are between the SW and LW groups.
In conclusion, there is no indication that subjects were more stressed during waiting without a time indication, compared to subjects that did get a time indication. However, the results suggest that the subjects in the long wait without indication group were more stressed during the first 75 seconds of the continuous performance task.
5.2 Discussion of methodological shortcomings
Multiple obstacles emerged during data processing, which will be discussed shortly. One of the main concerns is the lack of a baseline measurement to which the data can be compared. With a baseline showing the default, relaxed, state of the subject, more can be said about the alpha and beta decreases during the experiment, which can then be compared between groups. Additionally, the pilot study did not include a more objective measure for stress, like a cortisol level measurement. Moreover, the waiting long with and without indication periods were supposed to be the same length, however, the task result files showed that they were not. This makes it difficult to compare these groups, since any observed effect in the CPT can also be the result of the time difference, and does not have to be the result of the time indication. Lastly, the practice round was assigned to the task before waiting, with no period of rest in between, which could also affect brain waves and heart rate.
The EEG recordings were taken using one electrode at Fpz, in a non-controlled environment. Recording with one electrode limits the measurement to only very local and frontal effects, and is therefore likely to be a limitation in observing the effects of stress on brain activity. In addition, frontal electrodes are sensitive to eye blinks, since the muscles in the forehead also generate measurable electric activity (Iwasaki et al., 2004), this causes notable and hard to remove artifacts in the EEG measurements, which could not be corrected in this study. Another notable factor is the wide variation of age within the groups, a range from 18 to 70, as multiple studies have shown that EEG oscillation patterns change when people age (Liu et al., 2012; Christov & Dushanova, 2016), changes include reduction in the amplitude of alpha activity, a slowing of the background activity (dominant alpha rhythm), and a global increase of delta and theta power (Ishii et al., 2017). Therefore, it would be more difficult to observe general changes in the subject population.
15 To summarise, previous research has shown that acute stress is characterized by an increased BPM (Pakarinen et al., 2018), a decreased HRV (Qi & Gao, 2020), and changes in EEG: a decreased alpha power (Alonso et al., 2015; Giannakakis et al., 2015), increased high beta power (Alonso et al., 2015 ) and/or decreased overall beta power (Giannakakis et al., 2015). Using these stress indicators this pilot study showed that our long waiting period, about 2 minutes, did not cause acute stress in small subject groups, however, differences in stress level could be detected directly afterwards during the CPT. These differences suggested that subjects who had to wait long without a waiting time indication were more stressed, as they had a higher heart rate, and lower high alpha and beta power. This is an interesting finding, especially since a short delay of a train is often not indicated on notification screens.
6. Project proposal
6.1 Objectives and expectations of the proposed study
The objective of the current proposal is to expand and improve the design of the pilot study in order to gain better insight into how waiting affects travellers. The question we aim to answer is if waiting causes stress, and if an accurate time indication would attenuate this stress. The hypothesis is that during longer waiting the group that has to wait a long time without indication will experience more stress than the group that does have a time indication. Based on this hypothesis, we expect to find an increased heart rate, a decreased HRV, decreased theta and alpha power, an increased high beta power and/or decreased overall beta power, and an increased eye blink frequency during waiting or directly afterwards. All in the group that does not have a time indication compared to the group that does get a time indication. Moreover, we expect the group without time indication to have an increased cortisol level in their saliva.
6.2 Approach
This section will be divided into three different segments, the first will discuss how the methods and experimental design are altered from the pilot study, the second the techniques that will be used, including specific EEG recordings and analysis tools, and the third the planning and duration of the study.
6.2.1 Methods
This study will have a similar overall structure to the pilot study, but there are a few important changes that we want to make due to the methodological shortcomings that have been discussed.
6.2.1.1 Subjects and recruitment
Based on the pilot study and literature (Lenth, 2001) 40 subjects per group will be recruited in the age range of 20 to 30. Subjects will be randomly placed in one of the two groups, without knowing in which group they are placed. Subjects and their data will be treated according to the Netherlands privacy guidelines and the UvA ethical guidelines.
6.2.1.2 Setting
Instead of at a busy station, this study will be conducted in a controlled laboratory. The labs at Science Park 904 in Amsterdam are easily accessible via public transport and offer a multitude of resources, and are therefore favoured.
16 6.2.1.3 Experimental Design
In this study we will keep the basic framework of the task in the pilot study, with some adaptations (fig. 6A). The practice round is moved to a separate short task before the actual complete task. No EEG is recorded during the practice round. In addition, a baseline recording of 1 minute will be made, so during processing the recordings during waiting and the continuous performance task (CPT) can be compared with this baseline. This baseline block will be preceded and followed up by a short introduction, where after the 3 minute waiting block starts, again formulated as a necessary equipment calibration period (fig. 6B). We expect that longer waiting gives better insight into
whether waiting actually causes stress and how it manifests itself. Additionally, waiting time must be identical between the group that waits with indication and the group that waits without. Waiting is again followed by a short introduction and then the same CPT as in the pilot experiment. Both the waiting and CPT block will start with a trigger. Before the intake and after the outtake questionnaire a saliva sample will be taken to determine the cortisol level, the latter coincides with the cortisol peak.
Figure 6. Flowchart of the experiment. (A) Overview of the entire experiment from start to finish, the large arrow shows the part of the total experiment that consists of the EEG task. (B) Chart that shows the different parts of the EEG task, the arrow indicates the course of the experiment, the time taken for each part is indicated above the blocks.
6.2.2 Techniques
This study uses two methods to measure and three methods to analyse data. The results of these techniques are independent of each other and do not depend on the outcome of other measurements. Changes from the pilot study in techniques and analysis methodology will be discussed below.
17 6.2.2.1 EEG set-up
In contrast to the pilot study, EEG recording will be made using a EEG cap and multiple electrode dispersed over the scalp, as was done in cited articles (table 3). Electrodes will be placed frontal, central and parietal, at Fp1, Fp2, F7, F3, Fz, F4, F8, T3, C3, Cz, C4, T4, T5, P3, Pz, P4, T6, O1 and O2, and a reference electrode, as was done by Alonso et al. (2015). Our consideration of not using the total 64 10-20 system electrodes is practical, as more electrodes could make the subjects more nervous, and the analysis of a larger number of electrodes will take substantially more time. In addition, an electrode will be placed at the wrist of the hand that is not used for the task, to record heart rate.
Table 3. Placement of EEG electrodes in the references of this article.
The EEG set up consists of two laptops, one running MATLAB Stimulus Presenter software and the other running MATLAB EEG recorder software. The setup components are; a MOBIlab Amplifier, GAMMA connector box, digital input/output (DIO) box, connector cables, 19 EEG electrodes, ground and reference electrode, EEG cap, electrode gel, and gel syringe. Setups will be assembled according to the user manual, and are easily applicable and accessible. Test subjects will experience very minor discomfort, as putting on the EEG cap is a very low impact procedure.
6.2.2.2 EEG recordings
The goal of the study is to analyse the theta (6-10 Hz), the alpha (8-12 Hz), low beta (12.5-16 Hz), beta (16.6-20 Hz) and high beta (20.5-28 Hz) frequencies. No special adjustments have to be made before processing to acquire these frequencies. It is important that disruptions, such as the movement of test subjects during the recording, are prevented, and that noise is minimized.
6.2.2.3 EEG processing & analysis
EEG analysis will be performed using the method described in the pilot study with five major changes. Firstly, the baseline and waiting period will start with a trigger, allowing cutting of the recordings to be performed accurately. Second, artifacts, like excessive blinking, will be removed and counted after cutting using special algorithms (Golmohammadi et al., 2015). Third, after artifact removal, the baseline block and waiting or CPT block will be added together, resulting in one recording. Fourth, the waiting and CPT periods will be divided into blocks of 30 seconds to determine average power. Lastly, power averages will first be compared to the baseline, before they are compared between groups, to accurately observe increases and decreases in brain waves as described in the literature.
6.2.2.4 ECG processing & analysis 6.2.2.4.1 HRV
HRV analysis will be performed using the method described in the pilot study with two major changes. The first change is that the waiting and CPT periods will be divided into blocks of 30 seconds, and the second that the HRV is first compared to the baseline.
18 6.2.2.4.2 BPM
BPM analysis will be performed using the method described in the pilot study with two major changes. First, the actual baseline recording is set at 100% and second, the 30 second blocks are divided resulting in 6 points per block for each subject.
6.2.2.5 Cortisol measurement & analysis
Cortisol levels will be determined using saliva samples collected from subjects and are analysed using the Salimetrics Salivary Cortisol ELISA kit (Assay #1-3002). These cortisol levels can be used to reliably estimate serum cortisol levels. Although cortisol levels fluctuate during the day, this should not matter for our samples, as levels rise independently of this circadian rhythm in response to stress (Kalman & Grahn, 2004). This technique is a simple, painless, non-invasive sampling procedure, and all measurements will be duplicated. The measurements will be used to confirm if test subjects had a physiological stress response.
6.2.2.6 Statistical analysis
Statistical analysis will be performed using R, via the same methods as the pilot study.
6.2.3 Planning & Duration
Figure 7 shows the estimated duration of each experiment part according to the structure of the experiment as discussed in section 6.3.1.3. The maximum duration of the experiment is 1 hour, which translates to measuring 4-6 subjects per day.
Figure 7. Planning of the individual experiment with a time indication for each part. The green squares represent the part of the task, estimated duration is shown in brackets.
Figure 8 shows the planning and duration for the complete study, the complete estimated duration of the study is 12 months. For a risk assessment of this study we refer to the appendix.
Figure 8. GANTT diagram of the duration of the study. The parts of the study are listed on the left, the duration in months of each part is represented by the coloured rectangles. Rectangles that overlap mean that parts can be executed parallel to each other. The first 2 months are the setup and recruitment, the 3rd and 4th months the data collection, data processing starts in month 4 and continuous until month 6, data analysis is planned for months 5-6, and the writing of the research paper is a task that overlaps with the whole study and includes reviewing.
6.3 Scientific and societal impact
In neuroscience, there are relatively few studies investigating whether waiting causes stress, although there are certainly studies that research the effects of waiting. As a result, this study is innovative to some extent, and the outcome will provide a new perspective on what waiting does to the brain, as
19 the used methods have been validated in countless studies. Our results can be combined with studies that have investigated other effects of waiting, and serve as a basis for further research into the diverse effects of waiting.
We think it is important that scientific research benefits society wherever possible and as quickly as possible. This requires active cooperation from all parties, in this case our research team and parties that offer public transport. Therefore, we will make sure that our data becomes freely available as soon as possible, after the research had been finished. Moreover, we will actively involve the NS in order to make a direct link between our study, the pilot study and the societal benefit of our results. The results of our study can be directly applied to reporting in public transport, for example by suggesting adjustments regarding how delays are communicated. Making the travel experience more pleasant is not only an advantage for the passenger, but for the NS, or any other public transport service, it also increases the chance that travellers will use their service again. It therefore has a two-sided societal benefit, one to reduce stress among a large group of people who use public transport, and second to increase the use of public transport.
7. References
Abhang, P. A., Gawali, B. W., & Mehrotra, S. C. (2016). Chapter 2 - Technological Basics of EEG Recording and Operation of Apparatus. In, Introduction to EEG- and speech-based emotion recognition. Amsterdam: Elsevier. ISBN: 978-0-12-804490-2.
Alonso, J. F., Romero, S., Ballester, M. R., Antonijoan, R. M., & Mañanas, M. A. (2015). Stress assessment based on EEG univariate features and functional connectivity measures. Physiological Measurement, 36(7), 1351–1365. doi: https://doi.org/10.1088/0967-3334/36/7/1351
Bandelow, B., Baldwin, D., Abelli, M., Bolea-Alamanac, B., Bourin, M., Chamberlain, S. R., Cinosi, E., Davies, S., Domschke, K., Fineberg, N., Grünblatt, E., Jarema, M., Kim, Y. K., Maron, E., Masdrakis, V., Mikova, O., Nutt, D., Pallanti, S., Pini, S., Ströhle, A., … Riederer, P. (2017). Biological markers for anxiety disorders, OCD and PTSD: A consensus statement. Part II: Neurochemistry, neurophysiology and neurocognition. The world journal of biological psychiatry: the official journal of the World Federation of Societies of Biological Psychiatry, 18(3), 162–214. doi:
https://doi.org/10.1080/15622975.2016.1190867
Bozovic, D., Racic, M., & Ivkovic, N. (2013). Salivary Cortisol Levels as a Biological Marker of Stress Reaction. Medical Archives, 67, 374. doi: https://doi.org/10.5455/medarh.2013.67.374-377
Christov, M., & Dushanova, J. (2016). Functional Correlates of the Aging Brain: Beta Frequency Band Responses to Age-related Cortical Changes. International Journal of Neurorehabilitation, 03(01). doi: https://doi.org/10.4172/2376-0281.1000194
Dierolf, A. M., Schoofs, D., Hessas, E.-M., Falkenstein, M., Otto, T., Paul, M., … Wolf, O. T. (2018). Good to be stressed? Improved response inhibition and error processing after acute stress in young and older men. Neuropsychologia, 119, 434–447. doi:
https://doi.org/10.1016/j.neuropsychologia.2018.08.020
Doom, J. R., & Haeffel, G. J. (2013). Teasing apart the effects of cognition, stress, and depression on health. American journal of health behavior, 37(5), 610–619.
https://doi.org/10.5993/AJHB.37.5.4
European Agency for Safety and Health at Work (EU-OSHA) (2014). Calculating the cost of work-related stress and psychosocial risks. ISBN: 978-92-9240-420-8 doi:
https://doi.org/10.2802/20493
Föhr, T., Tolvanen, A., Myllymäki, T., Järvelä-Reijonen, E., Rantala, S., Korpela, R., Peuhkuri, K., Kolehmainen, M., Puttonen, S., Lappalainen, R., Rusko, H., & Kujala, U. M. (2015). Subjective stress, objective heart rate variability-based stress, and recovery on workdays among overweight and psychologically distressed individuals: a cross-sectional study. Journal of occupational medicine and toxicology, 10, 39. https://doi.org/10.1186/s12995-015-0081-6
20 Giannakakis, G., Grigoriadis, D., & Tsiknakis, M. (2015). Detection of stress/anxiety state from EEG features during video watching. 2015 37th Annual International Conference of the IEEE
Engineering in Medicine and Biology Society (EMBC). doi:
https://doi.org/10.1109/embc.2015.7319767
Golmohammadi, M., Lopez, S., Obeid, I., & Picone, J. (2015). Improved EEG event classification using differential energy. 2015 IEEE Signal Processing in Medicine and Biology Symposium (SPMB). doi: https://doi.org/10.1109/spmb.2015.7405421
Haak, M., Bos, S., Panic, S., & Rothkrantz, L.J. (2009). Detecting stress using eye blinks and brain activity from EEG signals.
https://www.stevenbos.com/dl/publications/Detecting_Stress_Using_Eye_Blinks_And_Brain_Activit y_From_EEG_Signals.pdf
Iwasaki, M., Kellinghaus, C., Alexopoulos, A.V., Burgess, R.C., Kumar, A.N., Han, Y.H., Lüders, H.O. and Leigh, R.J. (2004). Effects of eyelid closure, blinks, and eye movements on the
electroencephalogram. Clinical Neurophysiology, 116(4), 878-885. doi:
https://doi.org/10.1016/j.clinph.2004.11.001
Jönsson, P. (2007). Respiratory sinus arrhythmia as a function of state anxiety in healthy individuals. International Journal of Psychophysiology. 63 (1): 48–54.
doi: https://doi.org/10.1016/j.ijpsycho.2006.08.002.
Kalman, B. A., & Grahn, R. E. (2004). Measuring salivary cortisol in the behavioral
neuroscience laboratory. Journal of undergraduate neuroscience education : JUNE : a publication of Faculty for Undergraduate Neuroscience, 2(2), A41–A49.
Karamacoska, D., Barry, R. J., Blasio, F. M. D., & Steiner, G. Z. (2019). EEG-ERP dynamics in a visual Continuous Performance Test. International Journal of Psychophysiology, 146, 249–260. doi:
https://doi.org/10.1016/j.ijpsycho.2019.08.013
Lindau, M., Almkvist, O., & Mohammed, A. H. (2016). Effects of stress on learning and memory. In G. Fink (Ed.), Handbook of stress: Vol. 1. Stress: Concepts, cognition, emotion, and behavior (p. 153–160). Elsevier Academic Press.
Liu, C. J., Huang, C. F., Chou, C. Y., Kuo, W. J., Lin, Y. T., Hung, C. M., Chen, T. C., & Ho, M. C. (2012). Age- and disease-related features of task-related brain oscillations by using mutual
information. Brain and behavior, 2(6), 754–762. doi: https://doi.org/10.1002/brb3.93
Marshall, A. C., Cooper, N. R., Segrave, R., & Geeraert, N. (2015). The effects of long-term stress exposure on aging cognition: a behavioral and EEG investigation. Neurobiology of Aging, 36(6), 2136–2144. doi: https://doi.org/10.1016/j.neurobiolaging.2015.02.026
Moran, T. P. (2016). Anxiety and working memory capacity: A meta-analysis and narrative review. Psychological Bulletin, 142(8), 831–864. doi: https://doi.org/10.1037/bul0000051
Morgado, P., & Cerqueira, J. J. (2018). Editorial: The Impact of Stress on Cognition and Motivation. Frontiers in Behavioral Neuroscience, 12. doi:
https://doi.org/10.3389/fnbeh.2018.00326
Niedermeyer, E. & da Silva, F.L. (2005). Electroencephalography: Basic Principles, Clinical Applications, and Related Fields. Lippincott Williams & Wilkins. ISBN 978-0-7817-5126-1.
Pakarinen, S., Korpela, J., Torniainen, J., Laarni, J., & Karvonen, H. (2018). Cardiac measures of nuclear power plant operator stress during simulated incident and accident scenarios.
Psychophysiology, 55(7). doi: https://doi.org/10.1111/psyp.13071
Palva, S., & Palva, J. M. (2007). New vistas for α-frequency band oscillations. Trends in Neurosciences, 30(4), 150–158. doi: https://doi.org/10.1016/j.tins.2007.02.001
Pfaff, D. W., Martin, E. M., & Ribeiro, A. C. (2007). Relations between mechanisms of CNS arousal and mechanisms of stress. Stress, 10(4), 316–325. doi:
https://doi.org/10.1080/10253890701638030
Qi, M., & Gao, H. (2020). Acute psychological stress promotes general alertness and attentional control processes: An ERP study. Psychophysiology, 57(4). doi:
21 Qi, M., Gao, H., & Liu, G. (2017). Effect of acute psychological stress on response inhibition: An event-related potential study. Behavioural Brain Research, 323, 32–37. doi:
https://doi.org/10.1016/j.bbr.2017.01.036
Shaffer, F., & Ginsberg, J. P. (2017). An Overview of Heart Rate Variability Metrics and Norms. Frontiers in public health, 5, 258. doi: https://doi.org/10.3389/fpubh.2017.00258
Simonovic, B., Stupple, E. J. N., Gale, M., & Sheffield, D. (2018). Performance Under Stress: An Eye-Tracking Investigation of the Iowa Gambling Task (IGT). Frontiers in Behavioral Neuroscience, 12. doi: https://doi.org/10.3389/fnbeh.2018.00217
Thompson, J. J., Blair, M. R., & Henrey, A. J. (2014). Over the Hill at 24: Persistent Age-Related Cognitive-Motor Decline in Reaction Times in an Ecologically Valid Video Game Task Begins in Early Adulthood. PLoS ONE, 9(4). doi: https://doi.org/10.1371/journal.pone.0094215
Yaribeygi, H., Panahi, Y., Sahraei, H., Johnston, T. P. & Sahebkar, A. (2017). The impact of stress on body function: a review. EXCLI Journal, 16, 1057-1072. doi:
http://dx.doi.org/10.17179/excli2017-480
Vallat, R. (2018, May). Compute the average bandpower of an EEG signal. Retrieved June 15, 2020, from https://raphaelvallat.com/bandpower.html
8. Appendix
8.1 Questionnaires
8.1.1 Intake questionnaire
Q1. User Language
Q2. What is your participant number? Q3. What is your gender?
Q4. What is your age in years?
Q5. What is your weight in kilograms? (please use whole numbers) Q6. How tall are you in centimetres?
Q7. Are you colour-blind?
Q8. Why are you in the StationLab? (you can choose multiple) a. I am waiting on my train.
b. I just came from my train.
c. I came here specifically for the StationLab. d. The promotion team has approached me. e. Other
Q9. If you answered "other" in the previous question, can you specify your answer? Q10. How are you feeling at this specific moment?
Q10_1. Irritated Q10_2. Relaxed Q10_3. Tired Q10_4. Agitated Q10_5. Calm Q10_6. Impatient Q10_7. Stressed
22
8.1.2 Outtake questionnaire
Q1. User Language
Q2. What is your participant number?
Q3. How long do you think the experiment took? (Time in minutes) Q4. How much do you agree with these statements?
Q4_1. I thought the experiment took a short amount of time. Q4_2. I thought the time during the experiment was pleasant. Q4_3. I thought the experiment was enjoyable.
Q4_4. I was tense during the experiment. Q4_5. I was bored during the experiment. Q4_6. I thought this was a cool experiment.
Q4_7. I could immerse myself well into the situation of the experiment. Q4_8. I usually leave very early to make sure I am on time for an appointment.
MC:a. Strongly disagree, b. Disagree, c. Neutral, d. Agree, e. Strongly Agree
8.2 Inclusion criteria
• The subject should have no missing intake or outtake questionnaire data.
• The subject should not have an EEG recording.
• The subject should not be an outlier with regards to reaction time in the continuous performance task, or have only wrong answers or no answers during CPT.
• The subject should not have an error in the results of the EEG task, this includes a missing waiting period, or a waiting period that was uncharacteristically long.
• The subject should not have an error in the trigger in de EEG recording, for example no trigger for the waiting period.
• The subject should not have too much noise the EEG/ECG recording.
8.3 Detailed description of data processing and analysis
8.3.1 EEG
EEG analysis was performed using the EEG recorder software running in a MATLAB environment. EEG files were first filtered using a bandpass filter from 0.1 to 48 Hz, to remove the interference caused by the mains. After filtering the EEG recordings were cut so that only the waiting period remained. This waiting period started with a trigger, allowing cutting to be performed accurately; the files were cut at the start of the trigger and the waiting duration time after the trigger. The EEG recordings of the different test subjects were combined into a single MATLAB file. The same was done for files that were cut at the start of the CPT. The combined files were loaded into the Time-Frequency (TF) tool, were the power was calculated for normalised data. The waiting and CPT periods were divided into three blocks of 25 seconds, with exception of the short wait data, of which there is only 25 seconds of waiting. For each block the average power was calculated for the high alpha (10-12 Hz), low beta (12.5-16 Hz), beta ((12.5-16.6-20 Hz) and high beta (20.5-28 Hz) frequencies.
8.3.2 HRV
The filtering and cutting was done using the EEG recorder software running in a MATLAB environment in the same manner as the EEG data, with the change that the second channel of the recording contained data from the ECG electrode. ECG analysis was performed using the ECG Tool software running in a MATLAB environment. The isolated samples were loaded into the ECG Tool, in which a manual threshold was set to make sure that all T tops of the ECG could be detected, additionally a heart rate guidance was set to avoid detecting U peaks. The sample numbers of the T tops were exported to Excel, where they were converted to time in seconds by dividing the sample number by the sample rate, in this case 256. The RR-interval in seconds was calculated by subtracting the preceding T top from the current T top. To determine HRV the SDRR method was used, this is a HRV
23 time-domain measure, calculated by taking the standard deviation (SD) of all the RR-intervals from a subject (Shaffer & Ginsberg, 2017). The waiting and CPT periods were divided into three blocks of 25 seconds, with exception of the short wait data, of which there is only 25 seconds of waiting. The data processing as mentioned above was carried out in these 3 blocks for each subject individually.
8.3.3 BPM
From the RR-interval, in seconds, obtained in the processing above, the beats per minute (BPM) could be calculated for each point in time by dividing 60 with this RR-interval. An artificial baseline was created to ensure that changes could be visualized over time. This artificial baseline is the mean of the BPM points during the first 10 seconds of the measurements. The baseline was set at 100% and the changes were determined by dividing the BMP points by the baseline and multiplying by 100. In order to get a good picture of overall changes, it must be possible to average the course of the heartbeat, but this is only possible if the time points between subjects are equal, which is not the case naturally. Therefore, the 25 second blocks were further divided into blocks of 5 seconds, over which the mean BPM change was calculated. This resulted in 5 data points per block for each subject, which could be used for averaging and further statistical analysis.
8.4 Additional figures
8.4.1 Power changes over time and differences between groups
One way ANOVA determined that there were no significant differences between the three groups during the first 25 sec. of waiting. A 2x3 mixed ANOVA with group as between subjects factor and time slot as within subject factor and determined that there were no significant differences between groups or interactions (table A1). Figure A1 shows bar plots of the mean average power for each bandwidth.
Table A1. ANOVA results power during waiting
24
25
Figure A1. Bar plot of mean average power over time for all three groups during waiting. The plots show the course of the normalized average power during 75 sec. of waiting for high alpha (A), low beta (B), beta (C) and high beta (D) oscillations. The bars represent the mean for each group ±SD, and since the short wait group had only ~25 sec. of waiting, a bar is only shown for the first 25 sec. interval. The x-axis shows the different time slots; the first 25 seconds, the second 25 seconds and the third 25 seconds, and the y-axis the normalized average power in V2/Hz. Mixed 3x3 ANOVA showed no significant differences in average power overtime within groups for any of the oscillations.
8.4.2 HRV changes over time and differences between groups
One way ANOVA determined that there were no significant differences between the three groups during the first 25 sec. of waiting. A 2x3 mixed ANOVA with group as between subjects factor and time slot as within subject factor and determined that there were no significant (p>0.01) differences between groups waiting or interactions over time (table A2, left). A 3x3 mixed ANOVA with group as between subjects factor and time slot as within subject factor determined that there were no significant differences or interactions during CPT (table A2, right). Some interactions over time were close to significant, and are noticeable in the bar plots (fig. A2).
26
Figure A2. Bar plot of HRV over time for all groups during waiting and 75s of the CPT. (A) Bar plot of HRV over time during waiting. The bars represent the mean for each group ±SD, since the short wait group had only ~25 sec. of waiting, a bar is only shown for the first 25 sec. interval. The x-axis shows the different time slots; the first 25 seconds, the second 25 seconds and the third 25 seconds, and the y-axis the HRV in SDRR. (B) Bar plot of HRV over time during 75s of the CPT. The bars represent the mean for each group ±SD. The x-axis shows the different time slots; the first 25 seconds, the second 25 seconds and the third 25 seconds, and the y-axis the HRV in SDRR.
8.4.3 Changes in heart rate
ANOVA determined no significant differences in BPM between groups during the first 25 sec. of waiting, T-tests determined no significant differences in BPM between groups during the remaining 50 sec. of waiting. Figure A3 shows the line plots of BPM over time during the waiting period.
27
Figure A3. BPM changes from baseline during waiting. The first 10 sec. of the measurement were set as a 100% artificial baseline from which changes were calculated. The thin lines show individual subjects, whereas the thick lines represent the group mean. The vertical black dotted lines show the demarcation of the three 25 sec. time intervals.
8.5 Risk assessment
There are a few points in this study that could pose a risk to the success of the research, therefore we would like to discuss them here and propose additional measures to reduce these risks. First, there is a change that the student recruitment fails, an alternative approach could be to recruit older test subjects through the NS panel, as was done in the pilot study. This would result in an experimental group with a higher age average, but as long as this group is homogeneous this should not pose a problem. In our opinion, the location choice does not pose any additional risk since, Science Park is easily accessible by car and public transport, and there is sufficient material and space available. Next, we would like to discuss some risks related to the experimental design and techniques used. To start, it could be that the waiting in our task is too short to see an effect, if this seems to be the case when data processing starts in the second month, the group of subjects can be tested with a longer waiting time in the second month of measuring. This ensures that if observations are made, they can be supported by the longer waiting time. Second, the use of multiple EEG electrodes leaves more room for possible errors or more noise in the recordings, but it is a commonly used technique that can be used with the right prior knowledge. Therefore, we will make sure that staff operating the EEG measurements are well educated on the use of this setup. Last, the cortisol measurement uses a kit that has been used in research for a long time, and therefore gives a relatively low level of uncertainty, nevertheless, everything will be done in duplicates. Overall, we expect that this study is relatively low risk.
