TYPE Original Research PUBLISHED 09 January 2023 DOI 10.3389/fphys.2022.960118 Psychosomatic response to OPEN ACCESS acute emotional stress in healthy EDITED BY Colin K. Drummond, Case Western Reserve University, students United States REVIEWED BY Esther García Pagès 1*, Adriana Arza 2, Jesús Lazaro 3, Rajay Bharshankar, All India Institute of Medical Sciences Carlos Puig 1, Thais Castro 1,4, Manuel Ottaviano 5, Bhopal, India Maria Teresa Arredondo 5, Maria Luisa Bernal 3, Youngsun Kong, University of Connecticut, United States Raúl López-Antón 3, Concepción De La Cámara 3, Eduardo Gil 3, *CORRESPONDENCE Pablo Laguna 3,4, Raquel Bailón 3,4, Jordi Aguiló 3,4 and Esther García Pagès, Esther.Garcia.Pages@uab.cat Jorge Mario Garzón-Rey 6 1 SPECIALTY SECTION Universitat Autònoma de Barcelona, UAB, Barcelona, Spain, 2École Polytechnique Fédérale de This article was submitted to Physio- Lausanne, EPFL, Lausanne, Switzerland, 3Universidad de Zaragoza, UZ, Zaragoza, Spain, 4Centro de logging, Investigación Biomédica en Red de Bioingeniería, Biomateriales y Nanomedicina, CIBER-BBN, Madrid, a section of the journal Spain, 5Life Supporting Technologies, Universidad Politécnica de Madrid, UPM, Madrid, Spain, 6Omya Frontiers in Physiology Clariana S.L.U., OMYA, Barcelona, Spain RECEIVED 02 June 2022 ACCEPTED 29 November 2022 PUBLISHED 09 January 2023 CITATION The multidimensionality of the stress response has shown the complexity of this García Pagès E, Arza A, Lazaro J, Puig C, phenomenon and therefore the impossibility of finding a unique biomarker Castro T, Ottaviano M, Arredondo MT, Bernal ML, López-Antón R, among the physiological variables related to stress. An experimental study was Cámara CDL, Gil E, Laguna P, Bailón R, designed and performed to guarantee the correct synchronous and concurrent Aguiló J and Garzón-Rey JM (2023), measure of psychometric tests, biochemical variables and physiological Psychosomatic response to acute emotional stress in healthy students. features related to acute emotional stress. The population studied Front. Physiol. 13:960118. corresponds to a group of 120 university students between 20 and 30 years doi: 10.3389/fphys.2022.960118 of age, with healthy habits and without a diagnosis of chronic or psychiatric COPYRIGHT illnesses. Following the protocol of the experimental pilot, each participant © 2023 García Pagès, Arza, Lazaro, Puig, Castro, Ottaviano, Arredondo, Bernal, reached a relaxing state and a stress state in two sessions of measurement for López-Antón, Cámara, Gil, Laguna, equivalent periods. Both states are correctly achieved evidenced by the Bailón, Aguiló and Garzón-Rey. This is psychometric test results and the biochemical variables. A Stress Reference an open-access article distributed under the terms of the Creative Scale is proposed based on these two sets of variables. Then, aiming for a non- Commons Attribution License (CC BY). invasive and continuous approach, the Acute Stress Model correlated to the The use, distribution or reproduction in other forums is permitted, provided the previous scale is also proposed, supported only by physiological signals. original author(s) and the copyright Preliminary results support the feasibility of measuring/quantifying the stress owner(s) are credited and that the level. Although the results are limited to the population and stimulus type, the original publication in this journal is cited, in accordance with accepted procedure and methodological analysis used for the assessment of acute stress academic practice. No use, distribution in young people can be extrapolated to other populations and types of stress. or reproduction is permitted which does not comply with these terms. KEYWORDS stress, trier social stress test, cortisol, heart rate variability, multimodal biomarkers, pulse arrival time (PAT) Frontiers in Physiology 01 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 1 Introduction caused by the autonomic nervous system (ANS) and hypothalamic-pituitary-adrenal axis (HPA) activation. This The definition of stress remains ambiguous despite the efforts response can be measured using anatomical, endocrine, made in its study. In fact, post-traumatic stress disorder (PTSD) haematological, electrical, thermal, immune and genetic is the only one described in the medical diagnostic manuals: markers. Although the physiological approach seems to be the DSM-V (American Psychiatric Association, 2013) and CIE-10 most objective, it can be invasive, not very specific and may leave (Brämer, 1988). Beyond PTSD, the most cited definition of stress the subjective perception of the stress aside. was provided by H. Selye within the framework of what he called In any case, in order to characterise the stress response, a the “General Adaptation Syndrome”. Selye defines stress as “The stress stimulus is needed. It could be based on physical elements, non-specific response of the body to the request for change” or as social interaction, emotional reactions or cognitive tasks. For “a complex pattern of reactions that usually has psychological, example, sinking the hand or foot in cold or hot water are types of cognitive and behavioural components” (Selye, 1950). stimuli known as the Cold Pressor Test and the Warm Water Stress was declared a global epidemic by the World Health Test. Similarly, Project ES3 (Aguilo et al., 2015) has studied the Organization in the 21st century (Bakker and Holenderski, response to stress caused by high ambient temperatures with high 2012). In addition, the prevalence of stress has been linked to humidity simulated in a climatic chamber (Garzon-Rey et al., negative effects on both psychic and physical health (Cohen et al., 2016). Exposure to videos of corneal transplant surgeries has 1997). However, the level and/or duration required for its made it possible to characterise cortisol and alpha-amylase as transformation into a pathogenic agent is unknown. Chronic stress biomarkers (Takai et al., 2004). stress arises when the stress response is prolonged and a The Trier Social Stress Test (TSST) is a widely accepted rearrangement of the homeostatic equilibrium may take place. protocol for robustly inducing moderate acute stress in human These new equilibrium conditions can affect the subject’s populations using different stimuli. TSST has made it possible to wellbeing and even negatively affect their health. Again, study the relationship between psychological and physiological depending on the stressor, the time and the subject himself, responses in recent decades (Kudielka et al., 2007; Sequeira et al., the equilibrium may never be reached again and the subject’s 2021). Perhaps its usefulness lies in the simulation of a health may be severely affected. In this case, we talk about post- combination of a memory task, a social interaction event traumatic stress disorder. On the opposite side, acute stress arises followed by an arithmetic task. when the stress response remains for minutes or a few hours An increase in salivary, plasma and serum cortisol levels has been affecting the homeostatic equilibrium. reported on TSST application, as well as in the androgen precursor Among the strategies reported in the literature for stress dehydroepiandrosterone (DHEA) and its metabolite measurement, three different approaches can be highlighted dehydroepiandrosterone-sulphate (DHEA-S). Higher levels of (Hellhammer et al., 2010): the assessment of the stress adrenaline (Gold et al., 2004) and salivary α-amylase (Het et al., stimulus, the measurement of psychological perception of the 2009) have also been reported following the TSST. Furthermore, the stress stimulus and the measurement of the physiological changes immune system is sensitive to stress and has been linked to the HPA under a stress stimulus. axis activation. Pro-inflammatory cytokines are able to activate the The first strategy is based on the information gathering of HPA axis and alter glucocorticoid secretion, as well as glucocorticoids events that occur in an individual’s lifetime, assessing the effect can affect cytokine production. Pro-inflammatory cytokine they produce and understanding that the accumulation of interleukin-6 (IL-6) levels can increase by 50% in response to TSST stressful events generates a higher affectation level. This and remain elevated 20 min after stress exposure. Elevated levels of IL- strategy assumes that any event similarly affects all 6 following the TSST are associated with a higher heart rate during the individuals. The Holmes and Rahe Stress Scale (Holmes and arithmetic task and a lower cortisol response (Allen et al., 2014). Rahe, 1967) is a well-known instrument based on this strategy, Within the chain of secretions caused by the activation of where a list of possible events is associated with a score. the HPA axis are arginine vasopressin (AVP) and prolactin. The second approach focuses on the perception that an AVP has a small molecular size and a short average lifetime that observer has of a subject exposed to a stimulus. This makes it difficult to measure. However, the concentration of approach usually uses questionnaires. Clearly, this method is Copeptin derived from the same AVP precursor has been subjective and comparing subjects from different observers is proposed as a more stable surrogate, although its difficult. Another alternative within this strategy is the self- physiological interpretation is unknown (Urwyler et al., perception of the subject, on behalf of the own psychosocial 2015). On the other hand, it is well established that prolactin environment. For example, the Derogatis Stress profile secretion and plasma glucose levels are affected by stress. An (Derogatis, 1987) has 77 questions rating the stress level based elevation of their levels in response to different types of on personality traits and the emotional response. psychological stimuli in humans has been described and the The third approach is based on the measurement of the intensity of the response depends on the intensity of the physiological response of the subject to a stress stimulus, namely stressful stimulus (Armario et al., 1996). Frontiers in Physiology 02 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 Additionally, for a non-invasive measure of the ANS resolution is limited and it is invasive, our second objective is response to stress, Heart Rate Variability (HRV) derived from to establish a methodology that would allow the objective, non- the electrocardiogram (ECG) signal is commonly used (Task invasive and continuous measurement of stress, generating a Force, 1996). The psychophysiological factors that influence the model from physiological features alone that could lead to a more stress level can be reflected in the heart rate (HR) and its practical and extensive application. variations, HRV (HR variability). These are both mediated by For this purpose, the present paper reports an experimental the ANS in its sympathetic and parasympathetic branches, which study performed on healthy university students, designed to is reflected in the low-frequency (LF) and high-frequency (HF) guarantee a synchronous and concurrent measure of components, depending on whether one or the other branch was psychometric tests, biochemical variables and predominant, respectively. A meta-analysis shows that there is a electrophysiological features related to acute stress using a reduction in the HRV, an increase in the LF band parallel to a modification of the TSST to induce levels of stress based on reduction in the HF band, and/or an increase in the LF/HF different stimuli. quotient, thereby reflecting an increase in sympathetic stimulation and parasympathetic inhibition during a stressful event (Kim et al., 2018). 2 Methodology Photoplethysmography (PPG) is an uncomplicated and inexpensive optical measurement alternatively used for heart 2.1 Participants rate monitoring purposes among others, more and more present in wearables nowadays. This signal can be analysed in Students from the University of Zaragoza, the Technical order to extract similar information to the HRV, that is, the Pulse University of Madrid and the Autonomous University of Rate Variability (PRV). Moreover, the PPG signal together with Barcelona, Spain were recruited consecutively using the ECG signal provides useful information about the time lapse advertisements on websites. Through the web portal of the between the two, the Pulse Arrival Time (PAT), which is project, a participation survey was carried out among inversely proportional to the pulse wave velocity, which is in interested students. In the survey, the objective of the project turn associated with arterial stiffness and cardiovascular output. and the conditions of participation were explained in a general Therefore, a decrease in PAT is related to an increase in blood way, as well as the inclusion and exclusion criteria. vessel resistance and cardiovascular output, as well as being The conditions for inclusion in the study were to be a healthy inversely related to blood pressure (Wong et al., 2011). student aged between 19 and 30 years old and a BMI lower than There are also reports of an increase in the respiratory rate, 30, while any participant with psychopathology, regular RR, and its variability, RRV (Vlemincx et al., 2011; Milagro et al., consumption of psychotropic substances, or when the baseline 2017), and even an alteration affecting the synchronisation stress level was greater than 70% on the visual analogue scale, was between them (Lackner et al., 2011) as a response to the excluded. Moreover, they could not consume tobacco, alcohol or action of a stressor. Increased electrodermal activity (EDA) stimulants of the Nervous System (caffeine, theine, taurine, etc) has also been observed during TSST (Jezova et al., 2004). for at least 8 h prior to the study. Finally, in other studies a relationship has been established Those interested who meet the inclusion criteria were between stress and blood pressure (BP), the electrical selected, up to a maximum of 60 participants per site and characteristics of the skin, and skin temperature (T) (De they were contacted to indicate the date and place of the first Santos-Sierra et al., 2011; Vinkers et al., 2013; Allen et al., session. All participants signed an informed consent form 2014; Haufe et al., 2014). approved by the ethics committee of each institution. To sum up, the stress response is mainly generated through the hypothalamic centres and the hypophysis in order to simultaneously activate the ANS and the HPA axis, although 2.2 Experimental protocol with different timings, being the former faster than the latter. Therefore, the main goal of this study is to establish a The proposed experimental design is an observational and methodology to facilitate the objective, non-invasive and transversal study, in which each individual is compared with continuous measurement of stress through the study of himself or herself in a relaxed condition and in an acute stress changes induced by these two pathways. To that end, two condition in two measurement sessions. Each session took place steps were needed which constitute the two specific objectives on different days one or 2 weeks from each other, at the same of the present paper. time of day normally 2 h after waking up. Both sessions started The first objective is to generate a Stress Reference Scale with a relaxation period for 10 min with the help of an (SRS) as a linear combination of psychometric tests and autogenous relaxing audio to prevent the arriving state from biochemical variables allowing to better differentiate between interfering with the study, named baseline relax (BR) and stressed and relax states. However, since the SRS temporal baseline stress (BS). Frontiers in Physiology 03 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 FIGURE 1 Experimental design sessions. BR: Baseline relax. BS: Baseline stress. ST: Story Telling. MT: Memory test. SA: Stress anticipation. VD: Public video display. AT: Arithmetic Task. At the end of each session, a nurse took a blood sample. In After 10 min, the interviewer and an audience of at least three addition, two samples of saliva were extracted per session at the people entered the room to make the public presentation of the end and at the beginning, which for the stress stage, it was 25 min video segments (VD). The recorded videos of the subject were before the end of the session, which corresponds to the end of the interspersed with videos from the control segments, which are baseline stage. equivalent video segments where a dummy participant responds to the same stories correctly. This unknown subject was matched 2.2.1 Relax session in age and gender for each participant. During the first session, the relevance of the study and the Finally, during the arithmetic task (AT) the subject was asked protocol were explained to the participant and the informed to count down from 1022 to 0 in steps of 13. If he/she made an consent was signed. Then, sensors were placed and clinical and error, he/she had to start over from the beginning (i.e., 1022). The socio-demographic data were documented. Afterwards, the subject had 5 min to complete the task, even though the subject subject was left alone in the room, with dimmed lights, in a was not expected to finish the task. After that, the second saliva comfortable position and with an audio for autogenous sample and blood draw took place along with the psychometric relaxation for 25 min following Schultz’s method (Stetter tests. et al., 2002). After the BR stage, the first saliva sample was Figure 1 summarises the process followed during the two taken. The next 25 min were designated the Relax (R) stage, and, sessions. after that, saliva and blood samples were collected and the psychometric tests administered. 2.3 Outcome measures 2.2.2 Stress session During this second session, after the 10 min BS stage, the first 2.3.1 Psychometric tests saliva sample was collected and the stress protocol began, which The psychometric tests that were self-administered are the included three short stories, a memory test, a stress anticipation Perceived Stress Scale, the Visual Analog Scale, and the State- period, the video presentation and an arithmetic task. Trait Anxiety Inventory. All tests are validated, supported, well The memory test was carried out with an interviewer and the documented by the medical world and have a contrasted and subject alone in a room and consists of two stages. Within the accepted Spanish language version. The Symptomatic Stress first stage, referred to as Story Telling (ST), the subject was asked Scale, designed at the Clinical Hospital of Zaragoza in the to listen carefully to three stories, which he/she will have to Project ES3 framework, was also administered. reproduce with the greatest number of details during the second The two sessions took place on different days to prevent the stage, the Memory Test (MT), with a 30 s time limit for each learning effect or memorisation of the questionnaires, which story. The MT stage was video recorded. could invalidate the psychometric test results. Under the pretext that the video will be evaluated by experts The State-Trait Anxiety Inventory (STAI) was built by C. to assess memory performance, the subject remains alone in the Spielberger, R.L. Gorsuch, and R.E. Lushene, and consists of room without knowing how long he/she will be there. This stage 40 questions related to the subject’s introspection about his aims to increase the state of anxiety before the next stage, and it feelings of discomfort, worry, tension and stress (Takai et al., was labelled as Stress Anticipation (SA). 2004). The scale aims to measure two components: a trait Frontiers in Physiology 04 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 component (STAIt, Trait Anxiety Inventory) that would quantify after waking up. Moreover, it helps to have control of the the relatively stable individual differences in responding to foods consumed by the subjects and in the awakening, two situations perceived as threatening or the tendency to perceive factors that could invalidate the biochemical variables. All situations as more threatening. The second component refers to subjects were asked to avoid coffee and tea before the session the state (STAIs, State Anxiety Inventory) in a transitional period and refrain from exercising or drinking alcohol in the last 24 h. characterised by a feeling of tension, apprehension and increased The saliva collection was performed in Salivette tubes activity of the ANS (Spielberger, 2010). Each of the components following the manufacturer’s recommendations (Sarstedt AG is measured through a subset of 20 questions out of 40 on the & Co., Nümbrecht, Germany). Chewing the cotton swab scale. The STAIt and STAIs have total ranges from 0 to 60, with a provided with the tube and avoiding swallowing saliva for higher score corresponding to higher detected anxiety. The 1 min to obtain an adequate amount of sample. Cronbach’s alpha of the scale, both in its trait and state Subsequently, the sample was preserved on ice until it was component, is greater than 0.9 (Guillen-Riquelme & Buela- centrifuged in the laboratory (15 min at 1315 g, 4°C) to separate Casal, 2011). the saliva from the cotton swab. Then it was aliquoted in The Visual Analog Scale (VAS) is a simple method to detect duplicate in Eppendorf tubes with the subject identifying data, changes in mood and subjective perceptions (Monk, 1989). It type of session and date. The tubes were kept frozen at −20°C consists of a single question: In general, where would you say until processing. your stress level is at this moment? Being 0 = no stress at all and The concentration of cortisol and α-amylase in saliva were 100 = absolutely stressed. The use of the VAS has shown high quantified through commercial kits. A commercial immunoassay internal consistency in the measurement of transient and technique (Salimetrics, State College, PA, United States) was used subjective psychological states (Cronbach’s alpha range: for cortisol and a kinetic enzyme assay was used for α-amylase 0.84–0.96) (Abend et al., 2014). (Salimetrics, State College, PA, United States), capable of The Perceived Stress Scale (PSS) was proposed in the late measuring the activity of the enzyme in international units/ml 20th century by Sheldon Cohen and his colleagues with a set of of saliva (U/ml). 14 items that were later reduced to 10. This scale assesses the The extraction of blood was carried out in two tubes: the first extent to which life situations are assessed as stressful based on with EDTA anticoagulant and the second with gel separator. feelings and thoughts over the past month. Intrinsically, it Both were preserved on ice until centrifuged in the laboratory at considers the influence of how well the subject can handle a 3000 rpm for 10 min. The plasma and serum were then aliquoted stressful situation given his or her coping resources, which separately in tubes with the identifying data of the subject, type of ultimately offers a characterization of a person’s trait (Cohen session and date. The tubes were kept frozen at −20 °C until et al., 1983). A higher score corresponds to a higher level of processing at the Biomedical Diagnostic Center of the Hospital perceived stress. The short European Spanish version of 10 items Clínico de Barcelona by molecular absorption and immunoassay demonstrated adequate reliability with Cronbach’s alpha 0.82 spectrometry techniques to determine the concentration of (Remor, 2006). prolactin, copeptin and glucose. The Symptomatic Stress Scale (SSS) is a psychometric test designed on behalf of the ES3 Project (Aguilo et al., 2015) by the 2.3.3 Physiological signals psychiatry research group of the Hospital Clínico de Zaragoza. It Physiological signals such as electrocardiogram (ECG), tries to evaluate the effect of the stressful stimulus on the subject photopletismography (PPG), electrodermal activity (EDA), from the somatic and psycho-cognitive symptoms that the respiration (Resp) and skin temperature (T) change with subject perceives of himself (Table A1). The scale is of the stress. Many physiological features extracted from them are Likert type and consists of 20 questions. reported to vary with stress in the literature, but all of them are non-specific. We simultaneously recorded all these signals 2.3.2 Biochemical variables and processed them separately. The biochemical variables considered in the present study are The final set of signals selected due to their relation to the stress the concentrations of hormones, enzymes or molecules, analysed response are the following: ECG of three orthogonal leads in the in blood or saliva samples that offer information about the chest (1000 Hz), PPG in the middle finger of the non-dominant hypothalamic-pituitary-adrenal axis (HPA) and the autonomic hand and the temple (250 Hz), EDA between the second phalanx nervous system (ANS). of the ring finger and the index finger of the non-dominant hand The biochemical variables measured in saliva were cortisol (250 Hz), ST in the little finger and non-dominant cheek and α-amylase, whereas the biochemical variables measured in (sampling frequency 250 Hz) and the RR with a chest band blood were prolactin, copeptin, and glucose. (250 Hz). All the signals were recorded using the Medicom It was proposed to hold the morning sessions because of the 83 system, ABP-10 module (Medicom MTD Ltd, Russia). circadian rhythm of cortisol and α-amylase, specifically the For the signal processing and feature extraction in 1-min sessions were scheduled in the morning (09:45–10:45 h), 2 h windows, the MATLAB software was used. The beat detection in Frontiers in Physiology 05 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 the ECG signal was performed using the wavelet transform-based algorithm described in (Gil et al., 2009). Finally, the detection algorithm reported in (Martinez, et al., 2004). After beat of beats or pulses in the signal without artifacts was performed detection, the existence of ectopic or non-normal beats was using the strategy reported in (Lázaro et al., 2014). The pulse verified using the algorithm reported in (Mateo et al., 2003). detector algorithm consists of two phases: a linear filtering The algorithm looks for sudden changes in the time between transformation, and an adaptive thresholding operation. beats and proposer an interpolated beat where non-normal beats Finally, the generation of pulse rate signal for the PRV are detected. analysis, the fiducial point of each pulse in PPG was The conditioning of the photoplethysmography signal was defined as the one in which the amplitude has reached the carried out with a low-pass FIR filter with cut-off frequency 50% of its maximum. Once the pulses were detected, the 35 Hz and order 50 and a high-pass FIR filter with cut-off existence of ectopic or non-normal beats was also verified. frequency of 0.3 Hz and order 5000. In order to detect artifacts Subsequently, the pulse arrival time was estimated as the time in the signal, the average power was estimated from the between the beats detected on the ECG and the pulse detected by filtered signal in windows of 1 s displaced every 0.5 s. The the PPG on the finger. power was calculated as the variance of the signal every 0.5 s in The skin conductance signal was visually inspected to delete the previous second and it was linearly interpolated over the motion artefacts and linearly interpolated. The 1-min windows time vector of the filtered signal. An artifact was defined as a with interpolated segments larger than 25% were discarded and signal point whose associated average power was greater than the signal was then resampled at 4 Hz. For the EDA processing the mean plus three standard deviations of the average power and decomposition, we applied the convex optimization model associated with the signal of the immediately previous minute. (cvxEDA). In parallel, the signal was filtered with a band pass FIR Given that the PPG signal records present abrupt changes in filter with order 500 and a cut-off frequency of 0.01 Hz and amplitude of long duration or that they can be considered 0.9 Hz for the frequency-domain analysis. sections of low signal quality, the detection of this type of Finally, for the temperature signal, visual inspection showed artifacts was carried out based on Hjorth parameters in that the recorded signals did not contain significant artifacts, windows of 2 seconds displaced every half-second using the therefore no pre-processing was needed. TABLE 1 List of extracted features from physiological signals. Physiological signal Extracted Definition features Electrocardiogram (ECG) or SDNN (bpm) Standard deviation of intervals between beats considered normal (NN) Photoplethysmography (PPG) RMSSD (bpm) Root mean square of successive differences between adjacent NN intervals mHR (bpm) Mean heart rate −2 PHF (s ) HRV High frequency band power (0.15–0.4 Hz) PHFex (s−2) HRV Power in extended high frequency band (0.15 Hz-mHR/2) Pulse Arrival Time (PAT) mPAT (ms) Mean pulse arrival time stdPAT (ms) Standard deviation of the pulse arrival time Electrodermal Activity (EDA) mTonic (μS) Average value of the tonic component, i.e. Skin Conductance Level (SCL) stdTonic (μS) Standard deviation of the tonic component mPhasic (μS) Average value over time of the phasic component stdPhasic (μS) Standard deviation of the phasic component maxPhasic (μS) Maximum peak value of the curve of the phasic component aucPhasic (μS·s) Area under the curve value of the phasic component normalised by the length of the session, i.e. Skin Conductance Responses (SCRs) EDASymp (µS2) Power spectral density of EDA signal in the frequency range of 0.045–0.25 Hz Skin Temperature (T) mT (°C) Mean temperature mGrad (°C) Mean temperature gradient medGrad (°C) Median temperature gradient Frontiers in Physiology 06 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 TABLE 2 Sociodemographic data. factor. All analyses were run in SAS 9.4 software. To fulfil the assumption of normality required by the model, the Box-Cox Variables Categories N = 120 transform with λ = 0 was used when it was necessary. Principal Age (m, std) 21,992 (2,865) components analysis (PCA) was performed to find the so-called BMI (m, std) 22,145 (2,760) Stress Reference Scale weights, and the most relevant features extracted from the physiological signals. Afterwards, we used Gender (n,%) Female 60 (50) linear regression with these physiological features to generate the Male 60 (50) Acute Stress Model. Before performing the PCA analysis, the Smoker (n,%) No 100 (83) features to be included in the model were re-scaled to a range of 0–100 in order to avoid unfair weighting given the differences in the Yes 20 (16) order of magnitude of the values of each feature. Coffee (n,%) No 59 (49) Yes 61 (50) Alcohol (n,%) None 22 (18) 3 Results Occasional 88 (73) 3.1 Participants Moderate 10 (8) The study included 120 healthy young students, recruited Drugs (n,%) No 104 (86) from the University of Zaragoza, the Technical University of Occasional 13 (10) Madrid and the Autonomous University of Barcelona, Spain. Habitual 3 (2) After individual checking for criteria fulfilment, 40 candidates per site were able to participate in the two different sessions. Chronic diseases (n,%) No 108 (90) The population studied corresponds to a group of 60 men Yes 12 (10) and 60 women. Their average age is 22 ± 3 years and in normal Pharmaceutical drugs (n,%) No 94 (78) ranges according to the WHO based on their body mass index (BMI) (Barba et al., 2004). 10% of the sample subjects suffered Yes 26 (21) from asthma, respiratory allergies, migraine or intestinal reflux, Relationship (n,%) No 33 (27) but they were treated and it did not represent an inconvenience to Yes 87 (72) carry out the measurement sessions. Most of the subjects had extracurricular activities, mainly sports, language courses or Origin (n,%) Rural 22 (18) some type of artistic activities. Urban 98 (81) The perceived stress measured prior to the inclusion of the Cohabitation (n,%) Alone 6 (5) study did not exceed 47 units on average where from a scale from 0 to 100 it would be considered from no to mild stress. The self- Family 76 (63) perception of stress in subjects who were not drug consumers was Shared_flat 38 (31) higher (VAS = 46.21) than in those who consumed at least once a Sport (n,%) No 21 (17) day (VAS = 15.3) but lower than in those who consumed less than once a day on average (VAS = 52.08). Occasional 40 (33) The consumption habits were mainly non-toxic in the Habitual 59 (49) measured sample. In other words, non-use of tobacco or drugs and occasional consumption of liquor prevailed. The consumption or not of coffee was evenly distributed between Further details concerning feature extraction are described men and women. Regarding the health status of the subjects in our previous works (Greco et al., 2016; Posada-Quintero based on Table 2, it can be defined as good or healthy. Given that et al., 2016; Garzón-Rey et al., 2017; Arza et al., 2018) and most practice sports regularly or occasionally, they do not have Table 1 summarises the final list of features used in the diagnosed degenerative diseases and the consumption of analysis. pharmaceutical drugs was limited to the use of oral contraceptives. Finally, the majority of the subjects had a partner or emotional relationship, lived in family homes or 2.4 Statistical analysis shared a flat which suggested that there was no anomaly in the capacity for social interaction. In conclusion, the 120 subjects The results of the psychometric tests and biochemical variables who participated in the measurement sessions met the proposed were analysed using mixed models with the subject as a random inclusion/exclusion criteria. Frontiers in Physiology 07 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 TABLE 3 Psychometric test results. Total Female Male RS SS RS SS RS SS PSS 20.68±(2.91) 20.00±(3.25) 20.65±(3.16) 20.41±(3.28) 20.70±(2.67) 19.59±(3.21) STAIt 19.93±(9.53) 19.52±(9.53) 20.83±(9.80) 19.71±(9.84) 19.02±(9.24) 19.33±(9.29) STAIs 15.99±(8.41)* 23.45±(11.37)* 16.03±(8.75)* 25.05±(11.10)* 15.95±(8.13)* 21.83±(11.52)* VAS 32.68±(21.36)* 51.02±(21.18)* 34.85±(23.58)* 52.56±(19.48)* 30.50±(18.84)* 49.45±(22.85)* SSS 21.21±(13.39)* 28.84±(15.05)* 21.92±(14.41)* 30.73±(14.64)* 20.50±(12.37)* 26.91±(15.36)* Data are Mean and Standard deviation. RS: Relax State. SS: Stress State. PSS: Perceived Stress Scale. STAIs: State Anxiety Inventory. STAIt: Trait Anxiety Inventory. SSS: Symptomatic Stress Scale. VAS: Visual Analog Scale. * Statistically significant differences between RS, and SS (p < .05). TABLE 4 Biochemical variable results. Total Female Male RS SS RS SS RS SS ΔCortisol ⚤ −0.06±(0.09)* −0.02±(0.11)* −0.04±(0.07)* −0.02±(0.08)* −0.08±(0.11)* −0.01±(0.13)* Δα-Amylase 9.22±(56.48)* 49.71±(77.76)* 10.50±(50.09)* 46.00±(78.31)* 7.92±(62.73)* 53.48±(77.69)* ⚤ Copeptin 6.28±(3.64)* 7.28±(3.97)* 4.74±(2.71)* 5.67±(3.46)* 7.75±(3.81)* 8.88±(3.82)* Glucose 87.14±(14.77) 88.22±(9.80) 85.36±(14.95) 87.03±(8.28) 88.87±(14.51) 89.39±(11.05) ⚤ Prolactin 7.95±(3.62)* 9.16±(4.54)* 8.92±(4.27)* 10.45±(4.93)* 7.02±(2.58)* 7.92±(3.77)* Data are Mean and Standard deviation. RS: Relax State. SS: Stress State. * Statistically significant differences (p < 0.05). ⚤ Statistically significant differences between gender. 3.2 Outcome measures On the other hand, there was an increase in the scores for STAIs in the stress session compared to the relaxation session, 3.2.1 Psychometrics tests regardless of gender. This result supports that the stressful The psychometric scores were calculated following the stimulus applied in the stress session had an effect on guidelines available in the CIBER-SAM bank of instruments modifying the state of the subjects. Women had a significantly and methodologies in mental health. The following tables lower trait (p-value = .0009) than the score reported in (Guillen- summarise the results obtained from the psychometric test Riquelme and Buela-Casal, 2011), but an equivalent score administered: Table 3. to men. The Perceived Stress Scale scores were calculated as the sum The scores of the Visual Analog Scale and the Symptomatic of all the items with the reverse scores for questions 4, 5, 7, and 8. Stress Scale had statistically significant increases that were The results for each of the sessions and by gender were not consistent with the State Anxiety Inventory. Again, this result significant. The studied sample had approximately three more provides evidence of the stressful effect of the stress session and units in the total score (2.76 ± 3, p-value < .0001) than the the correct application of the stimulus. Likewise, there were no normative data of the PSS reported (Remor, 2006). This indicates differences related to sex. that the studied population was more likely to perceive a The Symptomatic Stress Scale (SSS) was validated with a situation as stressful or was more sensitive to it. There was no sample of 71 individuals by JM Garzón and published in his PhD statistical evidence that this feature changed from one session to thesis (Garzón-Rey J.M., 2017). The score is the sum of all the the next or that it was influenced by gender. items with no reverse scoring. Internal validation of the SSS scale For the STAI, scores were calculated as the sum of all the was carried out as an instrument for measuring acute emotional items with the reverse scores for questions 1, 2, 5, 8, 10, 11, 15, 16, stress through a descriptive and inferential statistical analysis 19, 20, 21, 23, 26, 27, 30, 33, 34, 36, and 39. In the case of the following the guidelines and procedures described in (Cook et al., STAIt, there were no statistically significant changes observed, 2006). Significant differences in the means of the relaxation and which is consistent with the result of the Perceived Stress Scale. stress sessions were observed through a t-test for repeated Frontiers in Physiology 08 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 measures with a p < .0001 in all state tests and no difference in the TABLE 5 Physiological features. trait tests as expected. Cronbach’s Alpha Coefficient was Electrocardiogram determined for each of the two sessions from the descriptive — — Mean Std No Diff statistics for each of the questions on the scale. No negative correlations or Cronbach’s Alpha Coefficient less than 0.8 were mHR [BMP] BR 73,6 11,55 BS SA VD found, that is, the scale has internal consistency or reliability in R 72,85 11,22 BS VD both measurement states for the entire set of questions. BS 72,99 11,85 BR R VD 3.2.2 Biochemical variables ST 82,81 15,84 MT The differences between the second sample and first sample MT 83,74 16,41 ST of cortisol and α-amylase concentrations are shown in Table 4. SA 73,85 12,92 BR VD Concerning cortisol, there was a lower decrease in concentration in the stress session compared to the relax session, that is, in both VD 73,22 12,84 BR R BS SA cases, between the two measurements of each session, the AT 89,74 17,71 — circadian cycle is observed, a decrease in cortisol and an −2 increase in α-amylase. However, the lower decrease in SDNN [s ] BR 0,13 1,2 R BS ST AT cortisol in the stress session shows a greater release of R 0,11 0,87 BR BS ST cortisol opposite to the variation due to the circadian cycle. BS 0,07 0,08 BR R ST AT And the same occurs with α-amylase, there is a greater increase in the stress session than in the relaxation session. Both results ST 0,22 1,01 BR R BS SA VD AT showed greater activity in the HPA axis and in the sympathetic MT 0,35 1,62 system during stress. Moreover, men had a higher concentration SA 0,31 6,38 ST VD AT of cortisol than women while no differences were found in α- amylase concentrations. VD 0,59 3,85 ST SA AT As expected, there was a statistically higher value for AT 0,18 0,8 BR BS ST SA VD prolactin and copeptin in the stress session with respect to the RMS [s−2] BR 0,11 1,71 R BS ST relaxation session. However, glucose levels were not significantly altered by the stressful stimulus applied. R 0,07 1,24 BR BS ST Likewise, it is noteworthy that the values of all the variables BS 0,03 0,11 BR R ST were within the normal range of clinical use. It was also observed ST 0,25 1,42 BR R BS SA AT that the values were affected by sex, finding a higher concentration of copeptin in men than in women. On the MT 0,38 2,26 VD contrary, the prolactin concentration in women was higher SA 0,36 9,05 ST AT than in men. In fact, the concentration of prolactin in men in VD 0,77 5,45 MT the stress session is equivalent to that of women in the relaxation session. AT 0,19 1,16 ST SA −2 PLF [s ] BR 0,61 0,81 ST SA 3.2.3 Physiological signals R 0,51 1,22 BS ST The results on the physiological features extracted during each protocol stage are presented on Table 5. BS 0,5 0,85 R ST SA EDA values multiplied by 100. DL: Dimensionless. BR: Basal ST 0,83 4,13 BR R BS SA relax. R: Relax. BS: Basal stress. ST: Storytelling. MT: Memory MT 1,22 2,43 AT test. SA: Stress Anticipation. VD:Public video display. AT: Arithmetic Task. No Diff: No significant difference with the SA 0,5 0,61 BR BS ST stage listed (p-value < .05). VD 0,68 2,5 For the HRV, regarding the temporal indices, none of AT 0,86 2,24 MT them differed between the basal moments but they were significantly different compared to the AT stage, which is PHF [s−2] BR 0,24 0,27 R BS ST SA VD the moment of the highest stress level. However, it is R 0,23 0,25 BR ST SA VD surprising that the moments of relaxation, VD and SA did not differ. The frequency indices in both the classic and the BS 0,28 0,42 BR ST SA VD extended bands differed between the two moments BR and (Continued on following page) Frontiers in Physiology 09 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 TABLE 5 (Continued) Physiological features. TABLE 5 (Continued) Physiological features. Electrocardiogram Photoplethysmography ST 0,57 3,34 BR R BS SA VD VD 0,02 0,02 ST AT MT 0,75 2,02 AT 0,02 0,01 ST VD SA 0,26 0,35 BR R BS ST VD RMS [BPM] BR 1,23 0,19 BS SA VD VD 0,34 0,82 BR R BS ST SA R 1,22 0,22 BS SA VD AT 0,48 0,94 BS 1,23 0,2 BR R SA VD −2 PHFex [s ] BR 0,27 0,29 R BS ST SA VD ST 1,39 0,28 MT R 0,26 0,38 BR ST VD MT 1,43 0,29 H BS 0,31 0,46 BR ST SA VD SA 1,21 0,21 BR R BS VD ST 0,62 3,41 BR R BS SA VD VD 1,22 0,21 BR R BS SA MT 0,87 2,26 AT 1,48 0,32 — SA 0,3 0,41 BR BS ST VD Electrodermal Activity VD 0,38 0,9 BR R BS ST SA mTonic [DL] BR −0,18 0,68 R AT 0,58 1,18 R −0,26 0,69 BR LF_HF [DL] BR 380,56 489,64 ST MT SA AT BS −1,09 0,57 — R 299,32 423,79 ST MT SA ST 0,10 0,55 SA MT VD BS 297,94 391,1 VD MT 0,12 0,62 SA ST VD ST 284,94 253,19 BR R MT SA AT SA −0,03 0,52 ST MT MT 284,73 382,89 BR R H SA AT VD 0,18 0,58 ST MT SA 295,95 352,11 BR R ST MT AT AT 0,40 0,79 — VD 209,97 310,99 BS stdTonic [DL] BR 0,11 0,12 BS AT 311,26 492,72 BR ST MT SA R 0,15 0,19 ST Photoplethysmography BS 0,10 0,16 BR mHR [BPM] BR 73,68 11,36 BS SA VD ST 0,16 0,15 S MT R AT VD R 73,08 12,93 BS SA VD MT 0,20 0,22 SA ST AT VD BS 73,37 11,74 BR R SA VD SA 0,19 0,16 ST MT VD ST 83,12 16,72 MT VD 0,18 0,19 SA ST MT AT MT 85,4 17,27 H AT 0,22 0,23 ST MT VD SA 72,04 12,43 BR R BS VD EDASymp [DL] BR 0,38 1,17 BS R VD 72,78 12,55 BR R BS SA R 0,90 3,05 BR BS AT 88,68 19,17 — BS 0,44 4,66 BR R SDNN [BPM] BR 0,02 0,01 R BS H ST 5,96 50,54 MT AT R 0,02 0,01 BR BS H MT 1,73 4,01 ST AT BS 0,02 0,01 BR R H SA 1,08 3,63 — ST 0,02 0,01 BR R BS VD AT VD 1,65 3,80 — MT 0,02 0,02 — AT 1,85 4,94 ST MT SA 0,08 1,3 — mPhasic [DL] BR 0,07 0,14 BS (Continued in next column) (Continued on following page) Frontiers in Physiology 10 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 TABLE 5 (Continued) Physiological features. TABLE 5 (Continued) Physiological features. Electrodermal Activity Skin Temperature R 0,10 0,19 BS MT 32,5 2,74 SA BS 0,07 0,17 BR R SA 32,3 3,49 MT ST 0,38 0,45 SA MT AT VD VD 31,68 3,69 — MT 0,39 0,39 ST AT VD AT 30,79 3,52 — SA 0,22 0,25 ST VD meanGrad [°C] BR 0,03 0,07 BS VD 0,28 0,30 SA ST MT AT R 0 0,04 SA AT 0,35 0,35 ST MT VD BS 0,03 0,07 BR stdPhasic [DL] BR 0,07 0,12 BS ST −0,07 0,07 MT R 0,10 0,16 — MT −0,07 0,07 H BS 0,06 0,14 BR SA 0 0,08 R ST 0,17 0,28 MT AT VD VD −0,04 0,06 AT MT 0,21 0,23 ST AT VD AT −0,04 0,06 VD SA 0,17 0,17 — medGrad [°C] BR 0,03 0,07 BS VD 0,19 0,20 ST MT AT R 0 0,04 SA AT 0,19 0,21 ST MT VD BS 0,03 0,07 BR maxPhasic [DL] BR 0,29 0,48 BS R ST −0,07 0,07 MT R 0,42 0,66 BR BS MT −0,07 0,07 H BS 0,25 0,51 BR R SA 0 0,09 R ST 0,82 1,03 MT AT VD −0,04 0,07 AT MT 0,93 0,88 ST AT AT −0,04 0,06 VD SA 0,70 0,68 — Pulse Arrival Time VD 0,80 0,76 — mPAT [ms] BR 281,69 69,46 R BS ST MT SA AT 0,88 0,86 ST MT R 277,41 65,41 BR BS ST MT SA aucPhasic [DL] BR 4,10 8,05 BS MT R AT SA VD BS 270,06 66,77 BR R ST MT SA VD AT R 5,99 11,31 SA BR BS MT AT VD ST 259,48 77,5 BR R BS MT SA VD AT BS 4,08 9,96 SA BR MT R AT VD MT 277,9 90,85 BR R BS ST SA VD AT ST 21,84 25,49 SA 268,59 69,31 BR R BS ST MT VD AT MT 22,67 23,15 SA BR BS R AR VD VD 256,59 68,02 BS ST MT SA AT SA 12,88 14,15 BR BS MT R AT VD AT 258,3 88,29 BS ST MT SA VD VD 16,59 17,49 SA BR BS MT R AT stdPAT [ms] BR 15,12 29,98 R BS ST SA VD AT 20,36 20,73 SA BR BS MT R VD R 13,69 29,43 BR BS ST SA VD Skin Temperature BS 14,64 39,88 BR R ST SA VD MeanT [°C] BR 33,71 3,33 — ST 19,88 52,31 BR R BS MT SA AT R 34,7 2,83 H MT 30,11 49,36 ST AT BS 34,31 3,28 H SA 18,41 56,6 BR R BS H H 34,03 2,63 R BS VD 12,76 40,03 BR R BS (Continued in next column) AT 22,66 42,88 ST MT Frontiers in Physiology 11 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 BS, failing to meet one of the selection criteria. The only average of the psychometric test scores and biochemical frequency indices that met this criterion were the indices in variables. The weights for each element were assigned in the HF band, although this was only calculated when the accordance with a principal components analysis (PCA) being subject was quiet. In conclusion, the heart rate variability proportional to the scoring coefficient and the variances of the features that were considered adequate for estimating the component. To define the SRS, the components with an stress level were: the mean value (mHR), the standard eigenvalue higher than 0.9 that together explained at least deviation (SDNN) and the mean square value of the heart 75% of the total variance were selected. rate (RMS) and in those cases where the subjects were not As shown in Table 6, the first five components fulfilled speaking, the indices on the HF band. Similar results were the selection criteria of an eigenvalue higher than 0.9 and found in the PRV analysis. accumulated explained variance at least 75% of the total When combining both signals, no significant differences variance. The resulting scoring coefficients for the five were found between the two baselines BR and BS for both components are shown in Table 7 and the highest values mPAT and stdPAT. In the stages where the stressful stimulus are highlighted. The first and the fourth component had the was applied, a decrease in pulse arrival time and greater highest scores for the psychometric tests. Components 2 and variability were observed. 3 can be associated with the HPA activation since the All the features from the electrodermal activity except the cortisol, copeptin and prolactin scores are the most skin conductance level met the selection criteria of being the relevant among these components. Finally, the last same at the baseline moments of both sessions and showed a component, the fifth, has the peak value for the α-amylase significant difference between the moments of relaxation and the that gives this component a relation to the sympathetic moments in which the stressful stimulus was applied. It is nervous system. important to highlight that, for the EDA analysis, 40 subjects The resulted Stress Reference Scale (SRS) computed for a of the total sample of 120 were discarded due to configuration population of 120 young people was: problems with the measurement equipment, therefore the results SRS = 0.14*PSS + .11*SSS + 0.11*STAIs + .10*VAS + presented correspond to only 80 subjects. 0.08*STAIt + .16*Copeptin + .13*⍺Amylase + .09*Cortisol Finally, for the skin temperature features, only the +0.08*Prolactin. temperature gradient in the finger (mGrad and medGrad) met Table 8 shows that there were no significant differences the selection criteria since there were no differences at the two between male and female SRS scores, although, as expected, baseline moments (BS and RB). A temperature decrease was the SRS was significantly higher in the Stress Session compared to observed during moments of stress, which was more intense in the relax session. stages ST and MT than VD and AT. 3.4 Physiological stress model 3.3 Stress Reference Scale To generate the acute stress model, first of all, feature values Following the methodology suggested in (Garzon-Rey et al., that were significantly different at the baseline stages of both 2017), a Stress Reference Scale (SRS) was computed as a weighted sessions were discarded considering that their values were TABLE 6 Eigenvalues and explained variances for components of PCA. Comp Eigenvalue Dif % VAR VAR Acum (%) 1 2.78 1.56 31 31 2 1.22 0.12 14 44 3 1.10 0.12 12 57 4 0.99 0.04 11 68 5 0.95 0.16 11 78 6 0.78 0.20 9 87 7 0.58 0.22 6 93 8 0.36 0.11 4 97 9 0.24 3 100 Comp: component. % VAR: Percentage of explained variance. VAR, Acum: Percentage of accumulated explained variance. Frontiers in Physiology 12 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 TABLE 7 Scoring coefficients for computing component scores. PPG the PRV_SDNN, PRV_RMS and PRV_mHR, from the PAT the mPAT and stdPAT; from the EDA the stdTonic, Component 1 2 3 4 5 stdPhasic, maxPhasic, mPhasic, aucPhasic and EDASymp; STAIs 90 10 −1 0 4 and from the ST the mGrad and medGrad. It was SSS 87 2 13 1 -4 considered that this number of features could be easily handled to give results in real-time without requiring large VAS 79 12 −13 −3 23 computing capabilities. STAIt 69 -6 5 31 -20 Table 9 shows the eigenvalues and explained variances for PSS 9 0 −11 93 5 the components resulting from the PCA analysis on the set of features. It is observed that the first five components are the Cortisol 5 77 29 20 0 only ones that have an eigenvalue greater than one and with α-Amylase 3 3 3 4 97 them, 87% of the variance of the data is explained, which is Copeptin 3 1 92 −11 3 sufficient for the type of analysis that was carried out. For the five components selected, Table 10 shows the Prolactin 7 75 −27 −21 5 coefficients or associated weights of each feature. It is % VAR 40% 17% 16% 14% 13% observed that each feature loads in a single component VAR, Prop: Percentage of Total Variance Explained by Component. with quite a difference compared to the others. This fact Bold text indicates values >60. can be understood as a great collinearity between the features that make up each group and therefore an evident reflecting conditions other than the stress. Moreover, features redundancy in the information they offer. To select the affected by speech such as respiratory rate were also excluded to features included for each of the components, only those avoid intrinsic noise. that had a coefficient greater than 60 were selected. Unlike the Stress Reference Scale, where values can be The first component was associated with all the features associated only with a state of stress (SS) or a state of extracted from the electrodermal activity, which in turn is relaxation (RS), the physiological features selected are, in a associated with the state of the sympathetic system. The way, “continuous.” Although, due to their method of phasic component from the EDA signal revealed to calculation, they are associated with different moments of the provide greater information among the EDA parameters session, and, therefore, offer information on the different dimension. We selected the standard deviation of the intensities of stress reached at this time. phasic component because it showed better correlation To select the relevant features, we associated the values of the with the SRS. SRS for the state RS to the features calculated at the moment of R The group of features associated with the second component of the first session measurement, and the values of the SRS in the was composed of the heart rate and pulse rate variability indices SS state to the last moment of the second measurement session, related to the power in the HF band. The features HRV_PHFex that is, to the AT stage. and HRV_PHF had the highest coefficients associated with this Two sequential selections were done in order to simplify component, but HRV_PHFex was selected since its weight was further analysis. Firstly, from each set of features showing slightly higher. statistically significant linear dependencies, only one was The mean values of heart rate estimated from the ECG and selected. From this set of features, only those having a PPG were in the third component group. The differences in significant correlation with the SRS score were selected in a weight between the two features were very small and although second step. Only 18 out of the initial 80 physiological features the weight corresponding to the one coming from the PPG were finally selected: from the ECG the HRV_SDNN, HRV_ (PRV_mHR) was strictly greater, in this case, other reasons RMS, HRV_PHFex, HRV_PHF and HRV_mHR; from the were considered. On the one hand, taking the HRV_mHR TABLE 8 Stress Reference Scale results. SRS N Mean Std Min Q1 Median Q3 Max Session* RS 114 39,40 6,70 27,28 35,09 38,01 42,70 62,34 SS 113 45,40 7,78 29,81 40,26 45,34 51,01 71,31 Gender F 112 42,89 7,89 28,59 37,56 41,08 47,30 71,31 M 115 41,90 7,80 27,28 36,40 40,43 47,94 60,13 *p < .05. Frontiers in Physiology 13 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 TABLE 9 Eigenvalues and explained variances for components of PCA. Comp Eigenvalue Dif % VAR VAR Acum 1 5.79 1.27 32 32 2 4.52 2.35 25 57 3 2.17 0.27 12 69 4 1.90 0.54 11 80 5 1.36 0.72 8 87 6 0.63 0.17 4 91 7 0.46 0.03 3 93 8 0.43 0.11 2 96 9 0.31 0.13 2 98 10 0.19 0.06 1 99 11 0.12 0.08 1 99 12 0.04 0.01 0 100 Comp: component. % VAR: Percentage of explained variance. VAR, Acum: Percentage of accumulated explained variance. TABLE 10 Scoring coefficients for computing component scores. into consideration since the previous group already took the C1 C2 C3 C4 C5 ECG. On the other hand, using the PPG represents greater comfort for the subject, since the signal can be extracted from HRV_mHR 21 −7 91 −4 15 a watch or bracelet and adhered electrodes would not be HRV_SDNN 11 77 −40 0 6 necessary. However, this option makes no sense if we keep HRV_RMS 1 79 −34 −1 3 the choice of HRV_PHFex in the previous group (which already requires having the ECG). On the contrary, HRV_PHF 3 94 18 0 0 selecting PPG in both groups would definitely mean greater HRV_PHFex 5 94 21 −1 2 comfort for the subject at the cost of losing precision in the stdTonic 79 3 1 −5 5 second component. The mean and median of the hand temperature gradient mPhasic 98 5 9 −5 1 were the features that form the fourth group. The coefficient stdPhasic 98 2 7 −3 1 associated with the fourth component was identical for both maxPhasic 98 2 8 −4 1 features. mGrad was selected since calculating the mean is computationally less expensive than the median. aucPhasic 98 5 9 −5 1 Finally, the last group, associated with the fifth EDASymp 94 7 10 −7 3 component, corresponded to the pulse arrival time features. Among these, we selected the mean value PRV_mHR 19 −6 93 −3 14 (mPAT) for having the highest coefficient associated with PRV_SDNN 9 65 −56 4 7 component five. From the selected features and using a linear PRV_RMS 0 65 −57 1 0 regression model, the Stress Reference Scale was estimated, which resulted in the following equation with a mPAT −2 −10 7 1 90 determination coefficient R2 of 0.97: stdPAT 10 21 14 −4 86 PSM = .5 · mGrad + .4 · PRV_mHR + .2 · HRV_PHFex + .06 · meanGrad −10 0 −3 99 −1 mPAT + .01 · stdPhasic. Table 11 shows the descriptive statistics of the physiological medGrad −9 1 −4 99 −2 stress model. A net difference between the stress states SS and VAR, Prop: Percentage of Total Variance Explained by Component. Bold text indicates values >60. relaxation states RS is preserved. Similar to the SRS results, there were no significant feature (from the ECG) would avoid having to deal with the differences between male and female PSM scores, although, greater number of artefacts that PPG undoubtedly generates as expected, they were significantly higher (p < 0.05) in the and, in turn, would eliminate the need to take the PPG signal Stress Session compared to the relax session. Frontiers in Physiology 14 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 TABLE 11 Physiological Stress Model results. SRS N Mean Std Min Q1 Median Q3 Max Session* RS 114 40.48 3.62 26.84 38.45 40.7 42.54 52.98 SS 113 44.64 4.74 30.71 41.87 44.55 48.03 52.83 Gender F 112 41.35 4.01 30.71 38.69 41.18 43.56 52.83 M 115 40.87 4.14 26.84 38.81 40.83 43.18 52.98 *p < .05. 4 Discussion task, a social interaction event and an arithmetic task, all without previous warning. Stress is a phenomenon that we can perceive, but hardly The reasoning behind the two separate sessions in the design quantified. The aim of this study was to establish a quantitative, of the present study considered the effect that could resemble the continuous, objective and non-invasive measure of the level of commonly known as “white coat syndrome”, in which acute emotional stress in healthy young subjects. From a practical participants are altered due to the clinical or experimental point of view, this tool will help clinicians improve a more setting, although they do not exhibit it in other settings. On accurate diagnosis and during the follow-up of therapeutic the one hand, participants knew what to expect during the second interventions while facilitating a common ground among session regarding the sensor placement and the experimental professionals. setting and on the other hand, the features that were significantly different in both basal states (BR and BS) were discarded considering their variations were not directly related to stress. 4.1 Artificial stress environment protocol Furthermore, this protocol allowed all subjects to be their own control reducing possible biases. First of all, from the literature, we have seen that there have coexisted three distinct approaches to the stress assessment. The first one relies on the subject’s lifetime stressful 4.2 Reference scale for stress experiences, which generate a higher affectation level, the measurement second one evaluates the psychological perception of a stressful stimulus and the third one measures some variables Once the experimental setting was designed, our initial scope of the physiological response of the subject (Hellhammer et al., was to find a comprehensive scale for stress measurement that 2010). Gathering information from these three approaches, we could be later used as reference for more practical use cases have designed a protocol that includes two different sessions. offering synthesised information on both the response The relax session in which a state of relaxation is achieved with corresponding to the «cogno » or of the psyche as well as of music. The stress session begins with a brief relaxation time to the somatic response to a stressful stimulus. reach similar conditions to the previous session. The stress state To that end, there are two sets of highly correlated systems is then achieved using the TSST as a moderate stressor. At the that activate simultaneously when faced with a stressful stimulus, end of both sessions, various psychometric tests and generating a single response. The first is associated with changes biochemical markers are administered as accepted gold in the psyche or « cogno», i.e. changes in thinking, the generation standards for stress assessment in humans and animal of emotions, the perception of reality, behaviour and social models respectively. The psychometric tests evaluate the interaction, among others. The second set is associated with psychological affectation due to the sequence of events while somatic changes from the system or organ level to the molecular the biochemical variables evaluate the response of the level. This in turn is activated by two separate although endocrine system. Additionally, several electrophysiological interrelated and cooperative pathways: the hypothalamic- variables, reported in the literature as stress markers, are pituitary-adrenal (HPA) axis and the autonomic nervous also measured synchronously throughout the sessions to system (ANS). assess ANS activation. From the cognitive point of view, our study results clearly The Trier Social Stress Test is an experimental protocol that showed the stress reactivity caused by the artificial stressor; has made it possible to study the relationship between the questionnaires evaluating the imminent stressful psychological and physiological responses in recent decades situation, i.e., STAIs, VAS and SSS, showed at least a 10% (Kudielka, B.M. et al., 2007). Its usefulness lies in sequentially increase in the stress session, whereas the trait questionnaires, including various different common types of stressors: a memory i.e., STAIt and PSS, remained the same. These results are in Frontiers in Physiology 15 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 accordance with the reported literature (Kudielka, et al., 2007; To elaborate the Stress Reference Scale, all the above variables Bae et al., 2018). that respond to the activity of the hypothalamic-pituitary-adrenal Regarding the HPA axis, results show a lower decrease in axis and the sympathetic nervous system that significantly changed cortisol concentration during the stress session compared to the as a result of the TSST were considered in the principal component relax session. Previously documented patterns of higher analysis and weighted according to their relevance. Each coefficient anticipatory salivary cortisol responses during the TSST in of each variable is proportional to the scores in the component and men compared to women have also been found (Liu et al., explained variances of the component. The inclusion of the trait 2017). Cortisol release also increases the glucose availability psychometric test (i.e. PSS and STAIt), allowed the Scale to have the for a fight or flight response although our results do not show information of predisposition (trait) of the individual to respond to a significant differences. stressful stimulus. The synthesis and release of cortisol in humans are subject to The SRS scale shows that 54% of the score corresponds to a circadian rhythm, with high levels in the early morning that Psychometric tests, rather than 60% proposed in (Garzon-Rey decrease as the day goes by. Additionally, the secretion of cortisol et al., 2017). The difference can be explained due to the inclusion is stimulated as a result of the activation of the HPA axis triggered of the Symptomatic Stress Scale (SSS) test and α-amylase since by a stressful stimulus. Even though the morning cortisol may be the increase up to 120 participants showed significant difficult to compare during the cortisol awakening response, concentration differences between Relax and Stress states, stress reactivity in healthy subjects has been proven to show unlike the subset of 40 participants used in (Garzon-Rey greater response compared to the evening results (Yamanaka et al., 2017). The SRS was validated for the type of population et al., 2019). and stimulus of the present study and is independent of gender. Outside the adrenal gland, cortisol is mainly bound to Once validated, the proposed SRS represented a quantitative corticosteroid-binding globulin (CBG). The remaining part, assessment of stress on a scale from 0 to 100 in which SRS = called free cortisol, represents the biologically active part of 0 means the total absence of stress (no response to stimuli) and the hormone. The cortisol found in saliva corresponds to the SRS = 100 means the highest level of distress. free cortisol that diffuses passively into the salivary glands and its concentration is associated with the levels of free cortisol in plasma (Foley and Kirschbaum, 2010). Although salivary cortisol 4.3 Non-invasive stress measurement can be altered as a consequence of a variation in the CBG concentration (Hellhammer et al., 2009), we analysed it since Finally, our protocol included the synchronous monitoring of it still has numerous advantages over its measurement in plasma, several non-invasive stress biomarkers. We extracted several stress- as it does not require any equipment or specialised personnel for related physiological features found in the literature to study their its collection. evolution under different stress intensities and to find the most The activation of the HPA axis also gives rise to arginine relevant ones for a continuous acute stress measurement. vasopressin (AVP) and prolactin hormones secretion. AVP has a In (Arza et al., 2018) an approach to estimate stress level small molecular size and a short average lifetime that makes it every 1 min from physiological signals was evaluated on a subset difficult to measure. Therefore, we determined the concentration of the database. Results supported the feasibility of quantifying of a co-secreted glycopeptide, known as copeptin, as a surrogate the stress level using a multiparametric measure of the biomarker, although physiological interpretation is unknown physiological stress response through biomarkers derived from (Carrillo-Esper et al., 2013; Urwyler et al., 2015). Both the processing of ECG, PPG and ST signals. Moreover, the prolactin and copeptin blood concentrations were elevated relationship between the features extracted from the signals after the stressor. and the SRS showed higher correlations than with either the Regarding the ANS, the sympathetic activation of the salivary psychometric variables or the biochemical variables alone. This glands increases the secretion of α-amylase while that of the suggested the feasibility of the SRS used to assess the stress levels. parasympathetic increases the volume of secreted saliva (Garrett In the present study, the conditioning and treatment of et al., 1999). Furthermore, the concentration of α-amylase also physiological signals was performed on ECG, PPG and ST, changes depending on the circadian rhythm of the subject and EDA was also added. From them, more than 80 features inversely to cortisol (Rohleder et al., 2004). widely reported in the literature or that are associated with The greater increase in the stress session compared to the relax physiological processes related to stress were extracted. shows the effect of the TSST, which has been associated with The heart rate variability features that were considered increased plasma catecholamines (adrenaline, noradrenaline) due adequate for estimating the stress level were the mHR, the to the presence of stressful stimuli. Therefore, there is a SDNN, RMS and HF band, in line with previous analysis in presumption that α-amylase in saliva could be considered a the data subset (Arza et al., 2015) where they significantly non-invasive substitute for the measurement of catecholamines changed between highest stress stages (MT and AT) and all and even cortisol (Rohleder et al., 2004). other stages of the session. Frontiers in Physiology 16 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 Regarding pulse rate variability (PRV), in (Garzón-Rey et al., From the PCA analysis, the selected final set of five features 2017) the complementary or additional information of PRV with explaining 87% of the variance of the data were: respect to HRV was investigated for stress assessment. While PSM = .5 · mGrad + .4 · PRV_mHR + .2 · HRV_PHFex + .06 · significant differences were found between PRV and HRV at mPAT + .01 · stdPhasic. baseline, no such differences were observed during stress stages, This model takes into account all the before mentioned suggesting that pulse arrival time variability is higher during relax features emphasising changes in the increase of the mean than during stress. Changes in the morphology of the PPG signal heart rate and the temperature decrease. Contrary to what the induced by stress were studied using a Gaussian mixture literature suggests, the power in the extended high frequency modelling technique, revealing significant differences in band of the HRV analysis increases with stress, which is features related to the position and width of the systolic and usually related to parasympathetic activity (as reviewed by reflected waves during stress stages with respect to baseline Kim et al., 2018). This could be explained because the power of (Banerjee et al., 2017). the extended band of the HF goes from 0.15 Hz to half the For the joint analysis of the ECG and PPG signals, we heart rate. Therefore, if the heart rate increases during the extracted the PAT features which reduced during the stressor. stress stage, the band is extended and consequently the power These are reported to be a proxy for vasoconstriction and hence can increase. Moreover, speech should be taken into account higher blood pressure. to increase the reliability of HRV as a marker of stress as The PAT is the time difference between the peak of the frequency features of HRV are influenced by respiration R-wave in the ECG signal (corresponding to left ventricular (Hernando et al., 2016). During the relax state, respiratory depolarisation) and a fiducial point in the PPG waveform (as frequency is within the LF band, so the HF band is not measured by a pulse oximeter attached to the fingertip). measuring respiratory sinus arrhythmia but just “noise”. However, it is often erroneously described in literature as On the contrary, during the stress tasks, respiratory pulse transit time (PTT) (Finnegan et al., 2021). The PAT frequency increases and falls within the HF band, thus includes not only PTT, but also the time delay between the increasing the power in the HF band. However, this can be electrical depolarisation of the heart’s left ventricle and the misinterpreted as an increase in parasympathetic activity. An opening of the aortic valve, known as pre-ejection period approach to better estimate the sympathovagal balance after (PEP). PEP can vary depending on contractility, sympathetic separating respiratory influences from the heart rate was nervous system activity, preload and afterload. In our study, we developed in (Varon et al., 2019). Finally, a minor did not separate the PEP from the analysis, although the contribution goes to the mean pulse arrival time and the literature suggests that the time elapsed between the R-peak standard deviation of the phasic component of the and the R-wave gives the best systolic blood pressure prediction electrodermal activity. (Wong et al., 2011). Our previous publication on a subset of the database (Arza For the skin temperature features, the experiments reported et al., 2018) where we designed a similar model for the non- in the literature do not converge to the same conclusion. In this invasive measurement of stress, the set of extracted features study, the mean temperature decreases rapidly during the first was more limited and the rescaling of the magnitudes to stages of the TSST, remains at its lowest values until the end of balance weightings given the differences in the order of the whole task and it has been reported that more than an hour is magnitude of the values of each feature was not carried needed post-stress for temperature to return to baseline (Pisanski out. However, both models are able to quantitatively et al., 2018). measure the stress changes over time and show a good Finally, electrodermal activity features measure a person’s correlation with the reference scale designed showing that sweat level in glands since the skin is usually an insulator but its the methodology used is reliable and can be adapted to the conductance changes when there is sweat as a consequence of conditions needed for a specific end-user application. sympathetic activation. Skin conductance has been found to have Moreover, after all the analysis, both models include the a linearly varying property with respect to emotional arousal, and mHR and mPAT. Therefore, these features show robustness apart from being used to classify different states such as anger and should be considered for future stress assessment and fear, it is also to detect the stress level while performing a task methods. (Baig et al., 2019). The EDA can be divided in two independent The purpose of this analysis is to achieve a continuous, components; the first in which rapid, immediate responses are non-invasive measurement that allows quantitatively observed and of short duration, known as phasic component, and assessing different levels of stress in a repeatable manner the second, the tonic component, is associated with the level of and from an approach that groups the different aspects of arousal of the sympathetic nervous system, that is, a sustained or the psychosomatic changes induced due to a stressor. In fact, accumulated response. Our results show this arousal in both the Physiological Stress Model can be calculated throughout components of the time-domain analysis and in the frequency- the entire session. The highest values of the model can be domain analysis, supporting the sympathetic arousal. observed in MT and AT and the lowest in BR, R BS, as Frontiers in Physiology 17 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 expected. It is important to highlight that due to the type of re- Data availability statement scaling that was carried out to “uniform” the ranges of the features, this model will never reach a minimum of 0 or a The raw data supporting the conclusion of this article will be maximum of 100, which is adequate for the type of moderate made available by the authors, without undue reservation. stress in healthy students being studied. Ethics statement 4.4 Limitations The studies involving human participants were reviewed and The main limitation of the current study is the specific target approved by Comisión de Ètica en la Experimentación Animal y subject of study, narrowed to healthy young students and a Humana de la UAB; Comité Ético de Investigación Clínica de perfect acute stress experimental setting although not Aragón; Comité de Ética de la UPM. The patients/participants applicable in a wider population and other types of stress. provided their written informed consent to participate in this Further studies including clinical and sociodemographic data study. such as age, gender, BMI or fitness may enhance the model’s robustness. The ultrashort term analysis for the HRV is a trade-off Author contributions between usability and reliability and should also be more carefully studied. Moreover, respiration and speech inclusion EG, JA, JG-R, and RB were substantial contributions to the into the analysis could greatly improve the reliability of the conception or design of the work and the acquisition, analysis or measurement. interpretation of data for the work; AA, JL, MB, CD, EG, and RL- Finally, only linear relations between the stress response and A helped drafting the work or revising it critically for important the stress reference features were analysed in this approach, intellectual content; CP, TC, MO, MA, and PL provided approval although including non-linear relations could contribute to a for publication of the content; All of the above agree to be more accurate measure of the stress reactivity nature. accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. 5 Conclusion The present study explores the multidimensionality of Funding perceived stress via a carefully designed holistic protocol and analysis. To reach the proposed objectives a wide set of This work was funded by the MINECO (Spanish Ministry of measurable variables related to the psychometric Economy and Competitiveness) and by the Centro de perception, hypothalamic-pituitary-adrenal axis and the Investigación Biomédica en Red sobre Bioingeniería, autonomic nervous system response have been taken into Biomateriales y Nanomedicina (CIBER-BBN) at the Instituto account. de Salud Carlos III de España. We used principal component analysis to find the variables that mostly contribute to differentiate the relax and stress states and weigh them according to the Stress Reference Scale. Acknowledgments Therefore, the SRS gives the quantitative stress level of an individual based on the biochemical variables and the We would like to express our very great appreciation to psychometric scores. The regression analysis with Llorenç Badiella from the Applied Statistics Service of the physiological variables to forecast the SRS level is then used to Autonomous University of Barcelona for his statistical expert elaborate the Physiological Stress Model. Although the model has guidance and support. been developed to match the SRS values in the initial (relaxation) and final (stress) states, this model allows to observe changes in the level of stress throughout all the stages of the TSST Conflict of interest continuously, non-invasively and specifically for acute emotional stress. Author JG-R was employed by Omya Clariana S.L.U. Additionally, the Symptomatic Stress Scale has been The remaining authors declare that the research was validated as a new psychometric test to assess the somatic and conducted in the absence of any commercial or financial psycho-cognitive stress perceived symptoms. Sequeira relationships that could be construed as a potential conflict of et al., 2021. interest. Frontiers in Physiology 18 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 Publisher’s note organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or All claims expressed in this article are solely those of the claim that may be made by its manufacturer, is not guaranteed or authors and do not necessarily represent those of their affiliated endorsed by the publisher. References Abend, R., Dan, O., Maoz, K., Raz, S., and Bar-Haim, Y. (2014). Reliability, Finnegan, E., Davidson, S., Harford, M., Jorge, J., Watkinson, P., Young, D., et al. validity and sensitivity of a computerized visual analog scale measuring state (2021). Pulse arrival time as a surrogate of blood pressure. Sci. Rep 11 (1), 22767. anxiety. Journal of Behavior Therapy and Experimental Psychiatry, 45 (4), doi:10.1038/s41598-021-01358-4 447–453. doi:10.1016/j.jbtep.2014.06.004 Foley, P., and Kirschbaum, C. (2010). Human hypothalamus–pituitary–adrenal Aguilo, J., Ferrer-Salvans, P., Garcia-Rozo, A., Armario, A., Corbi, A., Cambra, F. axis responses to acute psychosocial stress in laboratory settings. Neurosci. J., et al. (2015). Project ES3: Attempting to quantify and measure the level of stress. Biobehav. Rev 35, 91–96. doi:10.1016/j.neubiorev.2010.01.010 Revue neurologique, 61, 405–415. doi:10.33588/rn.6109.2015136 Garrett, J. R., John, R., Ekstro€m, J., and Anderson, L. C. (1999). Neural Allen, A. P., Kennedy, P. J., Cryan, J. F., Dinan, T. G., and Clarke, G. (2014). mechanisms of salivary gland secretion. Karger, Basel, Switzerland, Biological and psychological markers of stress in humans: Focus on the trier social Garzon-Rey, J. M., Arza, A., De-la-Camara, C., Lobo, A., Armario, A., and Aguilo, stress test. Neurosci. Biobehav. Rev 38, 94–124. doi:10.1016/j.neubiorev.2013.11.005 J. (2017). An approach to an acute emotional stress reference scale. Rev. Neurol 64, American Psychiatric Association (2013). Diagnostic and statistical manual of mental disorders. 529–537. doi:10.33588/rn.6412.2016509 Fifth Edition. American Psychiatric Association. doi:10.1176/appi.books.9780890425596 Garzon-Rey, J. M., Arza, A., Salama, A. A. K., Caja, G., and Aguilo, J. (April 2016). Armario, A., Marti, O., Molina, T., de Pablo, J., and Valdes, M. (1996). Acute Environmental temperature changes as stress stimulus, Proceedings of the Global stress markers in humans: Response of plasma glucose, cortisol and prolactin to two medical engineering physics exchanges/Pan American health care exchanges examinations differing in the anxiety they provoke. Psychoneuroendocrinology 21 (GMEPE/PAHCE). Madrid, Spain, IEEE, 1–4. doi:10.1109/GMEPE-PAHCE. (1), 17–24. doi:10.1016/0306-4530(95)00048-8 2016.7504666 Arza, A., Garzón, J. M., Hemando, A., Aguiló, J., and Bailon, R. (2015). Towards Garzón-Rey, J. M., Lázaro, J., Milagro, J., Gil, E., Aguiló, J., and Bailón, R. (2017). an objective measurement of emotional stress: Preliminary analysis based on heart “Respiration-guided analysis of pulse and heart rate variabilities for acute emotional rate variability. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc 2015, 3331–3334. doi:10. stress assessment, in Proceedings of the 2017 Computing in cardiology. IEEE, 1109/EMBC.2015.7319105 Rennes, France, doi:10.22489/CinC.2017.318-093 Arza, A., Garzón-Rey, J. M., Lázaro, J., Gil, E., Lopez-Anton, R., de la Camara, C., Garzón-Rey, J. M. (2017). Psychosomatic approach to stress measurement [PhD et al. (2018). Measuring acute stress response through physiological signals: thesis. Barcelona: Autonomous Univerity of Barcelona. Edith Cowan Online Towards a quantitative assessment of stress. Med. Biol. Eng. Comput 57, RepositoryRetrieved from https://hdl.handle.net/10803/458661. 271–287. doi:10.1007/s11517-018-1879-z Gil, E., Mendez, M., Vergara, J. M., Cerutti, S., Bianchi, A. M., and Laguna, P. Bae, Y. J., Reinelt, J., Netto, J., Uhlig, M., Willenberg, A., Ceglarek, U., et al. (2019). (2009). Discrimination of sleep-apnea-related decreases in the amplitude Salivary cortisone, as a biomarker for psychosocial stress, is associated with state fluctuations of PPG signal in children by HRV analysis. IEEE Trans. Biomed. anxiety and heart rate. Psychoneuroendocrinology 101, 35–41. doi:10.1016/j. Eng 56 (4), 1005–1014. doi:10.1109/TBME.2008.2009340 psyneuen.2018.10.015 Gold, S. M., Zakowski, S. G., Valdimarsdottir, H. B., and Bovbjerg, D. H. (2004). Baig, M. Z., and Kavakli, M. (2019). A survey on psycho-physiological analysis & Higher Beck depression scores predict delayed epinephrine recovery after acute measurement methods in multimodal systems. Multimodal Technol. Interact 3 (2), psychological stress independent of baseline levels of stress and mood. Biol. Psychol 37. doi:10.3390/mti3020037 67, 261–273. doi:10.1016/j.biopsycho.2003.12.001 Bakker, J., Holenderski, L., Kocielnik, R., Pechenizkiy, M., and Sidorova, N. Greco, A., Valenza, G., Lanata, A., Scilingo, E. P., and Citi, L. (2016). cvxEDA: A (2012). Stess@ work: From measuring stress to its understanding, prediction and convex optimization approach to electrodermal activity processing. IEEE Trans. handling with personalized coaching. Industrial Eng. Chem. Res. IND ENG CHEM Biomed. Eng 63 (4), 797–804. doi:10.1109/TBME.2015.2474131 RES 673-678. doi:10.1145/2110363.2110439 Guillen-Riquelme, A., and Buela-Casal, G. (2011). Psychometric revision and Banerjee, S., Bailon, R., Lazaro, J., Marozas, V., Laguna, P., and Gil, E. (September differential item functioning in the state trait anxiety inventory (STAI). Psicothema 2017). A two step Gaussian modelling to assess PPG morphological variability 23, 510–515. induced by psychological stress. Proceedings of the 2017 Computing in Cardiology Haufe, S., Kim, J. W., Kim, I. H., Sonnleitner, A., Schrauf, M., Curio, G., et al. (CinC), IEEE, Rennes, France doi:10.22489/CinC.2017.270-035 (2014). Electrophysiology-based detection of emergency braking intention in real- Barba, C., Cavalli-Sforza, T., Cutter, J., Darnton-Hill, I., Deurenberg, P., world driving. J. Neural Eng 11, 056011. doi:10.1088/1741-2560/11/5/056011 Deurenberg-Yap, M., et al. (2004). Appropriate body-mass index for Asian Heart rate variability (1996). Standards of measurement, physiological populations and its implications for policy and intervention strategies. Lancet interpretation and clinical use. Task Force of the European society of cardiology 363, 157–163. doi:10.1016/S0140-6736(03)15268-3 and the north American society of pacing and electrophysiology. Circulation 93 (5), Brämer, G. R. (1988). International statistical classification of diseases and related 1043–1065. PMID: 8598068. health problems. Tenth revision. World Health Stat. Q 41, 32–36. Hellhammer, D. H., Stone, A. A., Hellhammer, J., and Broderick, J. (2010). Carrillo-Esper, R., and Copeptina, de la Torre-León T. (2013). U. N. novedoso Measuring stress, in Encyclopedia of behavioral neuroscience. Editors G. F. Koob, interes. biomarcador pronóstico Int Mex 29, 380–387. M. Le Moal, and R. Thompson (Oxford: Academic Press), 186–191. Cohen, S., Kamarck, T., and Mermelstein, R. (1983). A global measure of Hellhammer, D. H., Wüst, S., and Kudielka, B. M. (2009). Salivary cortisol as a perceived stress. J. Health Soc. Behav 24, 385–396. doi:10.2307/2136404 biomarker in stress research. Psychoneuroendocrinology 34, 163–171. doi:10.1016/j. psyneuen.2008.10.026 Cohen, S., Kessler, R. C., and Gordon, L. U. (1997). Measuring stress: A guide for health and social scientists. A project of the Fetzer InstituteOxford University Hernando, A., Lazaro, J., Gil, E., Arza Valdes, A., Garzon-Rey, J., Lopez-Anton, Press. R., et al. (2016). Inclusion of respiratory frequency information in heart rate variability analysis for stress assessment. IEEE J. Biomed. Health Inf 20, Cook, D. A., and Beckman, T. J. (2006). Current concepts in validity and 1016–1025. doi:10.1109/JBHI.2016.2553578 reliability for psychometric instruments: Theory and application. Am. J. Med 119 (2), 166.e7–16. PMID: 16443422. doi:10.1016/j.amjmed.2005.10.036 Het, S., Rohleder, N., Schoofs, D., Kirschbaum, C., and Wolf, O. T. (2009). Neuroendocrine and psychometric evaluation of a placebo version of the “trier social De Santos-Sierra, A., Ávila, C. S., Casanova, J. G., and Del Pozo, G. B. (2011). stress test”. Psychoneuroendocrinology 34, 1075–1086. doi:10.1016/j.psyneuen.2009.02.008 Real-time stress detection by means of physiological signals. Recent Appl. Biometrics 58, 4857–4865. Holmes, T. H., and Rahe, R. H. (1967). The social readjustment rating scale. J. Psychosom. Res 11, 213–218. doi:10.1016/0022-3999(67)90010-4 Derogatis, L. R. (1987). “The Derogatis stress profile (DSP): Quantification of psychological stress,” in Research paradigms in psychosomatic medicine, 17, 30–54. Jezova, D., Makatsori, A., Duncko, R., Moncek, F., and Jakubek, M. (2004). High doi:10.1159/000414005 trait anxiety in healthy subjects is associated with low neuroendocrine activity Frontiers in Physiology 19 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 during psychosocial stress. Prog. Neuropsychopharmacol. Biol. Psychiatry 28, Remor, E. (2006). Psychometric properties of a European Spanish version of the 1331–1336. doi:10.1016/j.pnpbp.2004.08.005 perceived stress scale (PSS). Span. J. Psychol 9, 86–93. doi:10.1017/ s1138741600006004 Kim, H. G., Cheon, E. J., Bai, D. S., Lee, Y. H., and Koo, B. H. (2018). Stress and heart rate variability: A meta-analysis and review of the literature. Psychiatry Rohleder, N., Nater, U. M., Wolf, J. M., Ehlert, U., and Kirschbaum, C. (2004). Investig 15 (3), 235–245. doi:10.30773/PI.2017.08.17 Psychosocial stress-induced activation of salivary alpha-amylase: An indicator of sympathetic activity? Ann. N. Y. Acad. Sci 1032, 258–263. doi:10.1196/annals. Kudielka, B. M., Hellhammer, H., Kirschbaum, C., Hellhammer, D. H., 1314.033 Kirschbaum, C., Harmon-Jones, E., et al. (2007). Ten years of research with the trier social stress test—Revisited, in Social neuroscience (New York, NY: Guilford Selye, H. (1950). Allergy and the general adaptation syndrome. Int. Arch. Allergy Press), 512. Appl. Immunol 3, 267–278. doi:10.1159/000227975 Lackner, H. K., Papousek, I., Batzel, J. J., Roessler, A., Scharfetter, H., and Sequeira, I. K., Longmire, A. S., and McKay, N. J. (2021). Trier social stress test Hinghofer-Szalkay, H. (2011). Phase synchronization of hemodynamic variables elevates blood pressure, heart rate, and anxiety, but a singing test or unsolvable and respiration during mental challenge. Int. J. Psychophysiol 79, 401–409. doi:10. anagrams only elevates heart rate, among healthy young adults. Psych 3 (2), 1016/j.ijpsycho.2011.01.001 171–183. doi:10.3390/psych3020015 Lázaro, J., Gil, E., Vergara, J. M., and Laguna, P. (2014). Pulse rate Spielberger, C. D. (2010). State-trait anxiety inventory, in The corsini encyclopedia variability analysis for discrimination of sleep-apnea-related decreases in of psychology. Editor G. Stricke (Hoboken, NJ, USA: John Wiley & Sons), 4–6. the amplitude fluctuations of pulse photoplethysmographic signal in doi:10.1002/9780470479216.corpsy0943 children. IEEE J. Biomed. Health Inf 18 (1), 240–246. doi:10.1109/JBHI. Stetter, F., and Kupper, S. (2002). Autogenic training: A meta-analysis of clinical 2013.2267096 outcome studies. Appl. Psychophysiol. Biofeedback 27 (1), 45–98. doi:10.1023/A: Liu, J. J. W., Ein, N., Peck, K., Huang, V., Pruessner, J. C., and Vickers, K. (2017). 1014576505223 Sex differences in salivary cortisol reactivity to the trier social stress test (TSST): A Takai, N., Yamaguchi, M., Aragaki, T., Eto, K., Uchihashi, K., and Nishikawa, Y. meta-analysis. Psychoneuroendocrinology 82, 26–37. Epub 2017 Apr 24. PMID: (2004). Effect of psychological stress on the salivary cortisol and amylase levels in 28486178. doi:10.1016/j.psyneuen.2017.04.007 healthy young adults. Arch. Oral Biol 49, 963–968. doi:10.1016/j.archoralbio.2004. Martínez, J. P., Almeida, R., Olmos, S., Rocha, A. P., and Laguna, P. (2004). A 06.007 wavelet-based ECG delineator: Evaluation on standard databases. IEEE Trans. Urwyler, S. A., Schuetz, P., Sailer, C., and Christ-Crain, M. (2015). Copeptin as a Biomed. Eng 51 (4), 570–581. doi:10.1109/TBME.2003.821031 stress marker prior and after a written examination – The CoEXAM study. Stress 18, Mateo, J., and Laguna, P. (2003). Analysis of heart rate variability in the presence 134–137. doi:10.3109/10253890.2014.993966 of ectopic beats using the heart timing signal. IEEE Trans. Biomed. Eng 50 (3), Varon, C., Lazaro, J., Bolea, J., Hernando, A., Aguilo, J., Gil, E., et al. (2019). 334–343. doi:10.1109/TBME.2003.808831 Unconstrained estimation of HRV indices after removing respiratory influences from Milagro, J., Gil, E., Garzón-Rey, J. M., Aguiló, J., and Bailón, R. (September 2017). heart rate. IEEE J. Biomed. Health Inf 23, 2386–2397. doi:10.1109/JBHI.2018.2884644 Inspiration and expiration dynamics in acute emotional stress assessment, in Vinkers, C. H., Penning, R., Hellhammer, J., Verster, J. C., Klaessens, J. H., Olivier, Proceedings of the 2017 Computing in Cardiology (CinC). Rennes, France, B., et al. (2013). The effect of stress on core and peripheral body temperature in IEEE, doi:10.22489/CinC.2017.261-411 humans. Stress 16, 520–530. doi:10.3109/10253890.2013.807243 Monk, T. H. (1989). A Visual Analogue Scale technique to measure global vigor Vlemincx, E., Taelman, J., De Peuter, S., Van Diest, I., and Van den Bergh, O. and affect. Psychiatry Res 27, 89–99. doi:10.1016/0165-1781(89)90013-9 (2011). Sigh rate and respiratory variability during mental load and sustained Pisanski, K., Kobylarek, A., Jakubowska, L., Nowak, J., Walter, A., attention. Psychophysiology 48, 117–120. doi:10.1111/j.1469-8986.2010.01043.x Błaszczyński, K., et al. (2018). Multimodal stress detection: Testing for Wong, M. Y., Pickwell-MacPherson, E., Zhang, Y. T., and Cheng, J. C. (2011). The covariation in vocal, hormonal and physiological responses to Trier Social effects of pre-ejection period on post-exercise systolic blood pressure estimation Stress Test. Horm. Behav 106, 52–61. Epub 2018 Sep 19. PMID: 30189213. using the pulse arrival time technique. Eur. J. Appl. Physiol 111 (1), 135–144. doi:10. doi:10.1016/j.yhbeh.2018.08.014 1007/s00421-010-1626-0 Posada-Quintero, H. F., Florian, J. P., Orjuela-Cañón, A. D., Aljama-Corrales, T., Yamanaka, Y., Motoshima, H., and Uchida, K. (2019). Hypothalamic-pituitary- Charleston-Villalobos, S., and Chon, K. H. (2016). Power spectral density analysis of adrenal axis differentially responses to morning and evening psychological stress in electrodermal activity for sympathetic function assessment. Ann. Biomed. Eng 44 healthy subjects. Neuropsychopharmacol. Rep 39, 41–47. doi:10.1002/npr2.12042 (10), 3124–3135. doi:10.1007/s10439-016-1606-6 Frontiers in Physiology 20 frontiersin.org García Pagès et al. 10.3389/fphys.2022.960118 Appendix A: TABLE A1 Escala de síntomas de estrés (SSS). Señale la opción que mejor Muy de Algo de Ni de acuerdo ni en Algo en Muy en describe cómo se siente en este acuerdo acuerdo desacuerdo desacuerdo desacuerdo momento 1 Siento palpitaciones — — — — — 2 Siento la boca seca — — — — — 3 Siento rigidez en el cuello — — — — — 4 Siento que me falta el aire, suspiro — — — — — con frecuencia 5 Siento opresi n en el pecho — — — — — 6 Siento escalofr os — — — — — 7 Siento urgencia para orinar con — — — — — frecuencia 8 Siento que sudo — — — — — 9 Siento como si tuviera nervios en el — — — — — est mago 10 Siento que mi cara se sonroja — — — — — 11 Me siento mareado — — — — — 12 Cometo muchos errores — — — — — 13 No tengo ganas de hablar — — — — — 14 Me siento enfadado — — — — — 15 Me siento molesto por todo — — — — — 16 Siento que me distraigo f cilmente o — — — — — no me concentro 17 No me siento motivado para hacer — — — — — cosas 18 Me siento al l mite — — — — — 19 Siento que me impaciento f cilmente — — — — — 20 Me siento agitado e/o inquieto — — — — — Frontiers in Physiology 21 frontiersin.org