https://orcid.org/0000-0002-2304-1783
ABSTRACT
Introduction: Following a mass casualty event, such as the Paris terrorist attacks of 13 November 2015, first responders need to identify individuals at risk of PTSD. Physical peritraumatic symptoms involving the autonomic nervous system may be useful in this task.
Objective: We sought to determine the trajectory of physical response intensity in individuals exposed to the Paris terrorist attacks using repeated measures, and to examine its associations with PTSD. Using network modelling, we examined whether peritraumatic physical symptom associations differed by PTSD status.
Methods: Physical reactions were assessed using the Subjective Physical Reactions Scale at three time points: peritraumatic by retrospective recall, then current at one year (8–18 months) and three years (30–42 months) after the attacks. Interaction networks between peritraumatic physical reactions were compared according to PTSD status.
Results: On the one hand, the reported intensity of physical reactions was significantly higher in the PTSD group at all time points. On the other hand, using the dynamic approach, more robust positive interactions between peritraumatic physical reactions were found in the PTSD group one and three years after the attacks. Negative interactions were found in the no-PTSD group at one year. Peritraumatic physical numbness was found to be the most central network symptom in the PTSD group, whereas it was least central in the no-PTSD group.
Discussion: Network analysis of the interaction between peritraumatic physical subjective responses, particularly physical numbness, may provide insight into the clinical course of PTSD. Our knowledge of the brain regions involved in dissociation supports the hypothesis that the periaqueductal grey may contribute to the process leading to physical numbing.
Conclusions: Our findings highlight the role of peritraumatic somatic symptoms in the course of PTSD. Peritraumatic physical numbness appears to be a key marker of PTSD and its identification may help to improve early triage.
HIGHLIGHTS
Physical numbness was found to be a central symptom in people developing PTSD in our study examining peritraumatic physical symptoms related to the 2015 Paris terrorist attacks.
Introducción: Después de un evento con víctimas masivas, como los ataques terroristas de París del 13 de noviembre de 2015, los primeros socorristas necesitan identificar a los individuos en riesgo de TEPT. Los síntomas físicos peri traumáticos que implican al sistema nervioso autónomo pueden ser útiles en esta tarea.
Objetivo: Buscamos determinar la trayectoria de la intensidad de la respuesta física en individuos expuestos a los ataques terroristas de París utilizando medidas repetidas y examinar sus asociaciones con el TEPT. Mediante el uso de modelos de red, examinamos si las asociaciones de síntomas físicos peri traumáticos diferían según el estado del TEPT.
Métodos: Se evaluaron las reacciones físicas mediante la Escala de Reacciones Físicas Subjetivas en tres momentos: peritraumático mediante recuerdo retrospectivo, luego actual al año (8–18 meses) y a los tres años (30–42 meses) después de los ataques. Se compararon las redes de interacción entre las reacciones físicas peritraumáticas según el estado del TEPT.
Resultados: Por un lado, la intensidad declarada de las reacciones físicas fue significativamente mayor en el grupo de TEPT en todos los puntos temporales. Por otro lado, utilizando el enfoque dinámico, se encontraron interacciones positivas más sólidas entre las reacciones físicas peritraumáticas en el grupo con TEPT uno y tres años después de los ataques. Se encontraron interacciones negativas en el grupo sin TEPT al año. Se observó que la sensación de anestesia física peri traumática era el síntoma de red más central en el grupo con TEPT, mientras que era el menos central en el grupo sin TEPT.
Discusión: El análisis en red de la interacción entre las respuestas subjetivas físicas peritraumáticas, en particular la anestesia física, puede proporcionar entendimiento sobre el curso clínico del TEPT. Nuestro conocimiento de las regiones cerebrales implicadas en la disociación apoya la hipótesis de que la sustancia gris periacueductal puede contribuir al proceso que conduce a la sensación de anestesia física.
Conclusiones: Nuestros hallazgos destacan el papel de los síntomas somáticos peri traumáticos en el curso del TEPT. La anestesia física peritraumática parece ser un marcador clave del TEPT y su identificación puede ayudar a mejorar la evaluación precoz.
简介：在发生大规模伤亡事件后，例如 2015 年 11 月 13 日的巴黎恐怖袭击，急救人员需要识别有患 PTSD 风险的个人。涉及自主神经系统的围创伤性躯体症状可能在这项任务有用。
目的：我们试图通过重复测量识别巴黎恐怖袭击暴露个体的躯体反应强度轨迹，并检查其与 PTSD 的关联。使用网络模型，我们考查了围创伤性躯体症状关联是否因 PTSD 状态而异。
方法：在三个时间点使用主观躯体反应量表评估躯体反应：通过回顾性回忆的围创伤性，然后是攻击后一年（8–18个月）和三年（30–42个月）的当前反应。根据PTSD状态比较了围创伤性躯体反应之间的相互作用网络。
结果：一方面，PTSD 组报告的躯体反应强度在所有时间点都显著更高。 另一方面，使用动态方法，在PTSD组中，在袭击发生后一年和三年内发现了围创伤性躯体反应之间更强的正性相互作用。一年后，在无 PTSD 组中发现了负性交互。围创伤性躯体麻木被发现是 PTSD 组中最核心的网络症状，而在非 PTSD 组中它是最不重要的。
讨论：对围创伤性躯体主观反应（尤其是躯体麻木）之间相互作用的网络分析，可能有助于深入了解 PTSD 的临床过程。我们对涉及解离脑区的了解支持这样的假设，即导水管周围灰质可能有助于导致躯体麻木的过程。
结论：我们的研究结果强调了围创伤性躯体症状在 PTSD 过程中的作用。围创伤性躯体麻木似乎是 PTSD 的关键标志，其识别可能有助于改善早期分类。
1. Introduction
The analysis of adaptation mechanisms immediately following a life-threatening event is essential to our understanding of the subsequent development of psychopathological states. The same traumatic event may give rise to an adaptive, acute stress reaction in one individual, but prove pathological and maladaptive for another, and many factors (personality, psychological defenses, etc.) can modulate whether the response is effective or not (Crocq, Citation1999; DiGangi et al., Citation2013).
Very early studies established that a threatening confrontation leads to an adaptive physiological stress response, characterized by an increase in autonomic nervous system (ANS) activity, and an increase in hormones that drive the fight or flight response (Cannon, Citation1929; Hess, Citation1957; Selye, Citation1956). However, prolonged exposure to stress, or the inhibition of the fight or flight response, can result in maladaptive behaviour (Adams et al., Citation1969; Lacey, Citation1967). Animal experiments have shown that when confronted with a predator, in some cases, the reaction of the prey follows a predictable sequence. First, the threatened animal stops moving (freezing) and the sympathetic nervous system is stimulated (manifested as trembling, or an increased heart rate); here, the aim is to quickly analyse the danger. In the second stage, it tries to run away or escape. If it cannot escape, it turns to face the predator, to try to regain a dominant position (fight or flight). If neither fight nor flight actions stop the threat, the animal goes into tonic immobility, a state of involuntary paralysis. The purpose of this simulated death is to ‘disinterest’ the predator (Marx et al., Citation2008); the heart rate slows, along with the animal’s overall activity. The immediate physical response is sideration (Volchan et al., Citation2011). Physiologically, this suggests a shift from intense activation of the sympathetic nervous system to an equal and opposite activation of the parasympathetic system (Ford, Citation2009), and research has shown that tonic immobility is associated with decreased autonomic arousal (Gentle et al., Citation1989; Reese et al., Citation1982). These early findings are supported by current studies that postulate there are phylogenetic changes to the ANS, notably based on Porges’ polyvagal theory (Porges, Citation1995, Citation2001, Citation2007). Polyvagal theory argues that peritraumatic physical reactions are indicative of neurobiological changes related to ANS activation, when an individual faces a stressor. The ANS is therefore recognized as a key player in the response to a stressor (Godoy et al., Citation2018), which makes it relevant when studying PTSD severity (Park et al., Citation2017; Williamson et al., Citation2015).
Other work has shown that the intensity of peritraumatic physical reactions is a powerful predictive factor in the development of PTSD (Greene et al., Citation2020). However, analyses of the role of peritraumatic physical reactions in the PTSD clinical trajectory have been based on simple linear regression methods. The classic model underestimates or neglects the fact that symptoms may interact with each other and then, once activated, may be self-sustaining (Borsboom & Cramer, Citation2013). The latter perspective is at the core of the psychopathological network model, which is currently attracting much scientific interest (Nelson et al., Citation2017).
In psychopathology, network modelling is a dynamic approach that is based on network theory. It aims to illustrate the multifactorial complex links between a mental disorder and its symptoms (Borsboom, Citation2017). More recently, the idea of a dynamic, interacting psychopathological network has emerged. The latter assumes that mental disorders are stable states of closely connected, self-sustaining symptom networks that have been activated by a common cause (Fried et al., Citation2017). Networks can be estimated using cross-sectional or longitudinal time series data and analysed at the group or the individual level (Rhemtulla et al., Citation2016). In these networks, nodes represent various psychological variables (e.g. symptoms, behaviours), while connections between nodes represent unknown statistical relationships (e.g. correlations, predictive relationships) that can be estimated from psychopathological data (Borsboom, Citation2017). There are two types of connections: (i) directed nodes are connected by an arrow, indicating a unidirectional effect, or (ii) undirected nodes are connected with a line that indicates a mutual relationship, but there is no arrow to indicate the direction of the effect (Bringmann et al., Citation2016). The analysis of these networks can be divided into three fundamental steps. First, the structure of the network is estimated, based on a statistical model that reflects empirical patterns of relationships between variables. Second, this structure is analysed, and third, the accuracy of network parameters and measurements is evaluated.
On 13 November 2015, Paris (France) was the target of a multi-site terrorist attacks involving three kamikaze bombs around the ‘Stade de France’, a football stadium in the northern suburbs of Paris, four different shootings and bombings places in the 10th and 11th arrondissements of Paris and the Bataclan theatre, located in central Paris. The total number of victims was 130 dead and 354 injured in hospital, including 94 absolute emergencies and 250 relative emergencies (Hirsch et al., Citation2015). In a web-based study of 454 people exposed to the Paris attacks conducted 8–11 months after the events, almost 37% of the study sample had probable PTSD according to the PCL-5 (Pirard et al., Citation2020). In another web-based study of 698 rescue and law enforcement responders who were mobilized on 13 November 2015, the prevalence of PTSD according to the PCL-5 ranged from 3.5% to 9.9% depending on the type of responders (Motreff et al., Citation2018). In the context of the Paris attacks, the high prevalence of PTSD among exposed individuals underlines the need to better-understand the disorder (Pirard et al., Citation2020). The need is twofold: at the clinical level, we need to better-determine the trajectory of PTSD, based on the peritraumatic response (O’Donnell et al., Citation2008), and, at the practical level, we need to help front-line practitioners identify individuals who are most at risk of developing PTSD (Bossini et al., Citation2016).
The first aim of our study is to compare immediate (peritraumatic) and delayed physical reactions according to PTSD status assessed one year and three years after the attacks.
We hypothesize that the intensity of the subjective peritraumatic physical reaction is significantly higher for PTSD subjects and that this difference could be observed whatever the period of assessment. The second aim is to use network theory to explore the structure of the associations between peritraumatic physical reactions, taking into account the individual's PTSD status and the period of assessment.
2. Methods
2.1. Participants and procedures
This monocentric longitudinal study is part of the REMEMBER (REsilience and Modification of brain control network following novEMBER 13) biomedical research project (Mary et al., Citation2020; Postel et al., Citation2021), which received prior approval from the Nord Ouest III Personal Protection Committee (12/2016; ID RCB: 2016-A00661-50). Exposed participants were recruited through the Programme 13-Novembre cross-disciplinary and longitudinal research initiative (http://www.memoire13novembre.fr/), funded by the French General Secretariat for Investment (SGPI) through the National Research Agency (ANR) and the ‘Programme d’investissement pour l’Avenir’ (PIA ANR-10-EQPX-0021-01). REMEMBER study is a component of this Program.
All exposed participants met the requirements of DSM-5 Criterion A (they were all directly or indirectly involved in the Paris attacks). Trauma-exposed participants (see for demographic and clinical characteristics) were divided into two subgroups: those with PTSD symptoms according to the DSM-5 criterion of PTSD, and those without.
Table 1. Socio-demographic statistics for the study population as a function of PTSD according to the two phases of the study (time T1 and time T2).
Previous studies realized in this research program also define the notion of partial PTSD (or subthreshold or sub-syndromal) to define exposed individuals (criterion A) who have reexperiencing symptoms (criterion B), persisting for more than one month (criterion F), that caused significant distress and functional impairment (criterion G), without yet meeting all the DSM-5 diagnostic criteria (Leone et al., Citation2022; Mary et al., Citation2020; Postel et al., Citation2021). In the specific context of this work, we decided not to consider this population of partial PTSD because the psychopathological characterization of this entity will be the subject of a later specific analysis.
All subjects received information on the protocol and gave written consent prior to participation.
2.2. Instruments
PTSD status, and peritraumatic physical reactions were measured using validated questionnaires in French and English, chosen for their psychometric and clinical quality. Inclusion, and initial psychological testing took place between 13 June 2016 and 7 June 2017 (Time 1, one year (8–18 months) post-trauma). A second round of testing took place between 6 July 2018 and 29 June 2019 (Time 2, three years (30–42 months) post-trauma).
2.2.1. Measurement of PTSD
The Structured Clinical Interview for DSM-5 (American Psychiatric Association, Citation2013) was used to diagnose possible disorders related to exposure to the attacks. Any participant who met Criterion A was considered to have PTSD if he or she fully met DSM-5 specifications. Exposed participants who did not meet all criteria were considered to have no PTSD.
2.2.2. Measurement of physical reactions
Physical reactions were assessed using the Initial Subjective Reaction Physical Scale, which is part of the Potential Stressful Events Interview (Falsetti et al., Citation1994). The 10-item self-administered questionnaire assesses the subjective, physical reactions of individuals at the time of the event or at its recall. Ten reactions are evaluated: feeling short of breath, choking (1); feeling dizzy, unsteady or faint (2); heart palpitations or rapid heart rate (3); trembling (4); sweating (5); nausea or abdominal discomfort (6); numbness or a tingling sensation (7); hot flushes and chills (8); a feeling of strangulation (9); and chest pain or discomfort (10). Responses are reported on a four-point Likert-type scale (1 = not at all, 2 = a little bit, 3 = moderately, and 4 = extremely). The total score therefore ranges from 10 to 40. The instrument has good internal consistency (Cronbach’s α = 0.86).
2.3. Data analysis
The R software package (version 4.0.3) was used in all analyses.
The first stage of our analyses compared the intensity of physical reactions over time between exposed individuals with and without PTSD. We retrospectively assessed the self-reported intensity of peri-traumatic symptoms during time of Session 1 (T1), which was held 8–18 months after the trauma (2016–2017); current symptoms at the time of Session 1; and, finally, current symptoms three years (30–42 months) after the trauma during the time of Session 2 (T2; 2018–2019). An ANOVA was followed by the Tukey post hoc test, if the Shapiro–Wilk test determined that the distribution of the data was normal.
The second stage compared interaction networks between physical reactions at the time of the trauma for individuals with and without PTSD at time T1 (one year (8–18 months) post-trauma) then at time T2 (three years (30–42 months) post trauma).
2.3.1. Estimation and visualization of networks
When estimating interaction networks, it is common to find false connections due to unmeasured, confounding variables. Typically, this problem is overcome by using partial correlations to create relationships between variables. The method estimates the strength of relationships between variables by controlling for the effects of other measured variables in the model. Two nodes are connected if there is a covariance between them that cannot be explained by any other variable in the network (Epskamp et al., Citation2018). Coefficients range from −1 to +1 and determine the interaction between two nodes. In general, connections are visualized using red lines to indicate negative partial correlations, and green lines to indicate positive partial correlations (Borsboom & Cramer, Citation2013). The wider the line, the greater the number of connections. The more nodes there are, the more connections need to be estimated: in a 5-node network, 10 connections are estimated; in a 10-node network, 45 connections are estimated; and in a 20-node network, 190 connections are estimated (Epskamp et al., Citation2018). In order to successfully control for false connections between variables, their significance is examined. The most commonly used approach is the LASSO (Least Absolute Shrinkage and Selection Operator) method, together with the Extended Bayesian Information Criteria (EBIC) (Epskamp et al., Citation2018). Networks are estimated using polychoric partial correlations for ordinal data, and then modelled using the LASSO method, with the EBIC, using the qgraph package (Epskamp et al., Citation2012). In this approach, association networks are undirected and weighted with partial correlations, and estimate association parameters between all nodes using a Gaussian graphical model (Epskamp et al., Citation2018). The model uses the Fruchterman–Reingold algorithm (Fruchterman & Reingold, Citation1991), which produces easy-to-visualize networks where the edges are of similar length and overlapping edges do not impede the visualization. However, this method may not be appropriate for small sample sizes since it reduces weak connections at zero, which could result in the removal of moderately strong connections. This can lead to an inaccurate estimation of networks in samples with low statistical power, ultimately reducing the reliability of the analysis (Epskamp et al., Citation2018). To estimate networks, centrality, and connection strengths in small samples, the Information Filtering Networks (IFN) approach can be used (Christensen et al., Citation2018). IFNs estimate a fixed number of connections based on the formula ‘3 * nodes – 6,’ and the Triangulated Maximally Filtered Graph (TMFG) is a variant of IFN that filters the network by generating a chordal network (Massara et al., Citation2016). For more details of the methodology, please see the Supplementary Material.
2.3.2. Analysis of the network structure
Not all of the nodes in a network are equally important in determining its structure. Centrality indices provide an insight into the relative importance of a node with respect to other nodes (Borgatti, Citation2005; Freeman, Citation1978). For example, a central symptom is a symptom with a large number of connections, which can propagate the activation of the symptom network. Different centrality indices are developed to provide insight into different dimensions. These indices are presented as standardized z-scores that provide information about the relative importance and centrality of nodes. They are based on the connection model, in which the node of interest plays a role, and they can be used to model or predict several network processes, such as the amount of flow through a node or the tolerance of the network to the deletion of selected nodes (Borgatti, Citation2005). In the present study, we used the NetworkComparisonTest (NCT) package in R (Van Borkulo et al., Citation2016) to directly compare node centralities, with a single invariance measure, expected influence, as the measure of centrality.
2.3.3. Assessment of the model’s accuracy and robustness
In the present study, we used the bootnet R package to determine robustness (Epskamp et al., Citation2018). Specifically, non-parametric bootstrapping was used to calculate the accuracy of edge weights and node centralities. Edge stability was evaluated by calculating the mean correlation with the original sample. Node stability was determined by calculating a correlation stability (CS) coefficient. CS coefficients were calculated for the centrality measure, and for each of the two networks. To compare edge weights and centralities within networks, we used the difference test function with a p-level of .05. Each bootstrap was run 10,000 times for each network.
2.3.4. Network comparison test
To compare the interaction network for the two independent groups, a permutation test was developed to be able to make direct comparisons estimated in different subpopulations (Van Borkulo, Citation2018). A permutation test compares network structures that contain relationships between variables that are estimated from the data (Van Borkulo et al., Citation2016). In the present study, we used the NetworkComparisonTest (NCT) package in R (Van Borkulo et al., Citation2016) to estimate between-group network differences, and compare node centralities and edge weights. The significance level was set at p < .05.
3. Results
3.1. Participants
Of the 80 participants at time T1, 37 were women (46.3%) and 43 were men (53.7%). Their average age was 36 years. Furthermore, 95% were educated to high school level (the French baccalaureate) and 41.2% had received a further five years (or more) of education. Finally, 53.8% were married or in a common-law relationship, and 86.2% were in employment.
50 of the 80 included subjects (62.5% of the total sample) had no PTSD symptoms, and 30 (37.5% of the total sample) did.
At time T2, only 11 individuals were lost to follow up, and of the remainder, 31 were women (44.9%) and 38 were men (55.1%), with an average age of 37 years. Furthermore, 94% were educated to high school level (the French baccalaureate), and 43% had received a further five years (or more) of education. Finally, 55% were married or in a common-law relationship, and 60% were in employment.
Among the individuals who suffered from PTSD symptoms one year (8–18 months) post-trauma, 11 were clinically free of PTSD, and 13 still had symptoms three years (30–42 months) later. Among those who had no symptoms of PTSD one year (8–18 months) post-trauma, two had developed symptoms, and 43 were still asymptomatic three years (30–42 months) later.
Descriptive statistics regarding the population are detailed in , for the two sessions. shows the clinical trajectory of subjects who attended both sessions.
Figure 1. The clinical PTSD trajectory of exposed individuals between the time T1 (one year (8–18 months) post-trauma) and time T2 (3 years (30–42 months) post-trauma), assessed using DSM-5 diagnostic criteria.
3.2. Univariate longitudinal analysis of the physical reactions' intensity
Significant differences were found in the intensity of physical reactions (total score) reported: at the time of trauma, retrospectively reported at time T1 (F(78,1) = 47.10, p < .001); at time T1 (F(76,1) = 41.38, p < .001); and at time T2 (F(67,1) = 66.27, p < .001), between individuals with, and with no PTSD symptoms ().
Figure 2. Analysis of variance (ANOVA) results of the comparison of the intensity of physical reactions between individuals with, and with no PTSD symptoms for peritraumatic reactions retrospectively reported at time Tl (left); at the time Tl, held 8–18 months later (middle) and at the time T2, held three years (30–42 months) later (right).
3.3. Interaction networks for peritraumatic physical reactions, and the presence (n = 30) or absence (n = 50) of PTSD at time T1 (one year (8–18 months) post-trauma)
3.3.1. Network estimation
Based on the reported peritraumatic physical reactions, recorded retrospectively 8–18 months post-trauma and PTSD status at the same time, two networks were obtained.
In the PTSD group strong positive correlations were found between: feeling short of breath (1) and a feeling of strangulation (9); and between trembling (4) and numbness (7). Moreover, numbness (7) was positively correlated with most of the other symptoms.
In the no-PTSD group, most negative partial correlations related to nausea or abdominal discomfort (6). Strong positive correlations were found between trembling (4) and hot flushes and shivering (8); between trembling (4) and feeling dizzy, unsteady or faint (2); and between chest pain (10) and feeling short of breath (1) and a feeling of strangulation (9). See .
Figure 3. Interaction networks for peritraumatic physical reactions for PTSD (left) and no-PTSD (right) groups with EBIC lasso method. Green lines represent positive partial correlations, red lines represent negative partial correlations, and the thickness of the line indicates the strength of the correlation. Partial correlations between −1 and +1 are indicated in the middle of the lines. The maximum absolute value of the partial correlation is shown at the bottom right of the network. For the comparison of the two networks, the tuning parameter was set to λ = 0.40.
We investigated central indices as the expected influence of the peritraumatic physical reactions in each network. The resulting plot is shown in . The analysis of node centralities found that the expected influence of physical reactions was globally superposable, except for numbness (7). The latter factor is the most central in the network of the PTSD group, whereas it is the least central in the no-PTSD group. In individuals with no PTSD, the key impacts were: trembling, chest pain, and feeling short of breath.
Figure 4. Expected influence in the network structure of peri-traumatic physical reactions interaction and bootstrap 95% confidence intervals for estimated expected influence (n = 10,000) in PTSD group (left) or no-PTSD group (right) at time Tl (8–18 months post-trauma).
3.3.2. Edge weight accuracy and stability
An accuracy analysis (refer to the Supplementary Material for details of the method) revealed that 95% confidence intervals for edge weights mostly overlapped in both networks, which suggests that they are accurate (see Supplementary Material, Figure Aa). However, the large size of bootstrapped CIs implies that interpreting the order of most edges in the network should be done with care. Nevertheless, the bootstrapped significant difference test was used to compare differences in edge weights in each network (see Supplementary Material, Figure Ab).
In the PTSD group, no significant differences were found for edge weights after bootstrapping. Concerning edge stability, the mean correlation with the original sample was mostly positive and ranged from 0.75 to 1 which confirmed the robustness of edge weights (see Supplementary Material, Figure Aa).
In the no-PTSD group, the significant difference test showed that the negative interaction between nausea (6) and numbness (7) was significantly larger than most of the other interactions, after bootstrapping; this was followed by positive interactions between feeling dizzy, unsteady or faint (2) and trembling (4), and between trembling (4) and hot flushes (8). These three edges are reliably the strongest in the no-PTSD network, since their bootstraped CIs do not overlap with the bootstrapped CIs of most of the other edges. Concerning edge stability, the mean correlation with the original sample was between 0.5 and 1, with a decreasing trend, which suggests that robustness falls as resampling proceeds (see Supplementary Material, Figure Aa).
3.3.3. Centrality stability
The case-dropping bootstrap method was used to investigate centrality stability (refer to the Supplementary Material for details of the method). Stability was quantified using the CS coefficient. It should be noted that the CS coefficient should not be below 0.25, and preferably greater than 0.5. The results of the analysis indicated that centrality was stable in both networks, with CS = 0.51 in the no-PTSD group and CS = 0.60 in the PTSD group. These results mean that the order of the expected influence of nodes can be interpreted, although such an interpretation should be done with care.
The bootstrapped difference test was used to compare node centralities in each network (see Supplementary Material, Figure B). In the PTSD group, the expected influence of numbness (7) was significantly stronger than that of most of the other nodes. In the no-PTSD group, the expected influence of numbness (7) was significantly lower than feeling short of breath (1), trembling (4), and chest pain (10).
3.3.4. Network comparisons
We investigated differences in the network structure, and significant differences in edge weights and node expected influence between the PTSD and the no-PTSD group. The results revealed that the expected influence of numbness (7) was significantly different (p < .01). The results revealed no significant difference concerning the maximum difference in edge weights (p = .39) (see Supplementary Material, Tables A and B).
3.4. Interaction networks for peritraumatic physical reactions, and the presence (n = 13) or absence (n = 43) of PTSD at time T2 (three years (30–42 months) post-trauma)
3.4.1. Network estimation
Based on the reported peritraumatic physical reactions, recorded retrospectively 8–18 months post-trauma, and PTSD status three years (30–42 months) post-trauma, two networks were obtained with the triangulated maximally filtered graph (TMFG) method.
In the PTSD group strong positive correlations were found between: feeling short of breath (1) and a feeling of strangulation (9) and numbness (7) and between hot flushes and chills (8) and chest pain or discomfort (10) and numbness (7).
In the no-PTSD group, strong positive correlations were found between trembling (4) and feeling dizzy, unsteady or faint (2) and hot flushes and chills (8); and between chest pain (10) and feeling short of breath (1) and a feeling of strangulation (9). See .
Figure 5. Interaction networks for peritraumatic physical sympton1s for PTSD (left) and no-PTSD (right) groups at time 2 (30–42 months post-trauma) with TMFG method. Green lines represent positive partial correlations, red lines represent negative partial correlations, and the thickness of the line indicates the strength of the correlation. Partial correlations between −1 and +1 are indicated in the middle of the lines. The maximum absolute value of the partial correlation is shown at the bottom right of the network.
We investigated central indices as the expected influence of the peritraumatic physical reactions in each network. The resulting plot is shown in . The analysis of node centralities found that the expected influence of numbness (7) is one of the most central in the network of the PTSD group, as well as feeling short of breath (1), feeling of strangulation (9) and chest pain (10) whereas trembling (4) and nausea (6) were the least central in the network. In the no-PTSD group, trembling (4) and feeling short of breath (1) are the most central whereas numbness (7), sweating (5) and nausea (6) are the least central in the network.
Figure 6. Expected influence in the network structure of peri-traumatic physical reactions interaction and bootstrap 95% confidence intervals for estimated expected influence (n = 10,000) in PTSD group (left) or no PTSD group (right) at time T2 (30–42 months post-trawna).
3.4.2. Edge weight accuracy and stability
An accuracy analysis revealed that 95% confidence intervals for edge weights mostly overlapped in both networks, which suggests that they are accurate (see Supplementary Material, Figure Ca). However, the large size of bootstrapped CIs implies that interpreting the order of most edges in the network should be done with care. Nevertheless, the bootstrapped significant difference test was used to compare differences in edge weights in each network (see Supplementary Material, Figure Cb).
In the PTSD group, the significant difference test showed that the positive interaction between numbness (7) and feeling short of breath (1) and between numbness (7) and feeling of strangulation (9) were significantly larger than most of the other interactions, after bootstrapping. Concerning edge stability, the mean correlation with the original sample was mostly positive and ranged from 0.5 to 1 which confirmed the robustness of edge weights (see Supplementary Material, Figure Ca).
In the no-PTSD group, the significant difference test showed that the positive interactions between feeling short of breath (1), chest pain (10); between feeling of strangulation (9) were significantly larger than most of the other interactions, after bootstrapping. Concerning edge stability, the mean correlation with the original sample was under 0.5 which suggests that robustness falls as resampling proceeds (see Supplementary Material, Figure Ca).
3.4.3. Centrality stability
The case-dropping bootstrap method was used to investigate centrality stability (refer to the Supplementary Material for details of the method). The results of the analysis indicated that centrality was stable in both networks, with CS = 0.41 in the PTSD group and CS = 0.25 in the no-PTSD group at time T2. These results mean that the order of the expected influence of nodes can be interpreted, although such an interpretation should be done with care.
The bootstrapped difference test was used to compare node centralities in each network (see Supplementary Material, Figure D). In the PTSD group, the expected influence of numbness (7) was significantly stronger than that of most of the other nodes. In the no-PTSD group, the expected influence of numbness (7) was significantly lower than feeling short of breath (1), trembling (4), and chest pain (10).
3.4.4. Network comparisons
We investigated differences in the network structure and node expected influence between the PTSD and the no-PTSD group. The results revealed that the expected influence of numbness (7) was significantly different (p < .01) as well as trembling (4) which expected influence was significantly different (p = .01) in the network of the no-PTSD group compared to the network of the PTSD group (see Supplementary Material, Table C).
4. Discussion
Our results first demonstrate a significant and stable difference in the intensity of peritraumatic physical reactions measured by scores on the Initial Subjective Reaction Physical Scale between individuals with PTSD symptoms and those without PTSD, both retrospectively, at 8–18 months after the event, and up to three years (30–42 months) later. The intensity of physical reactions was significantly lower in the group with no PTSD symptoms, at all measurement times. At follow-up, 11 subjects were lost, 11 no longer met the criteria for PTSD diagnosis at time T2 out of the 50 previously diagnosed at time T1, while 2 who did not meet the criteria at time T1 did at time T2 (). Similar patterns of follow-up have been observed for military veterans (Solomon & Mikulincer, Citation2006) and for a terrorist attack (Gibert et al., Citation2021).
In addition, our study identifies two profiles, as a function of PTSD status at 8–18 months post-trauma as at 30–42 months post-trauma. In the no-PTSD group, connection strengths in the interaction network between peritraumatic physical reactions fell after bootstrapping, suggesting a decrease in the stability of the network as we resampled. This network is characterized by strong negative correlations between certain initial physical reactions (unlike the PTSD group, where positive, stable connections remain following bootstrapping). In addition, the no-PTSD network featured strong centrality and positive connections between certain symptoms that are known to evoke a physiological stress response through activation of the ANS. On the other hand, numbness was the central symptom in the PTSD group, and it was the factor that was most strongly connected with other symptoms in the network whatever the time session. Finally, we found an overall difference in significant connection strengths between PTSD and no-PTSD networks.
According to network resilience and vulnerability theory (Borsboom, Citation2017), a resilient network is a network in which connections between symptoms are too weak to be self-sustaining over time, even in the absence of the stressor (Borsboom, Citation2017; Cramer et al., Citation2016; Kalisch et al., Citation2019). In line with this approach, the observed lack of stability between connections in the no-PTSD network suggests that it is resilient. We also observed negative interactions in the no-PTSD network – to the best of our knowledge, there is no theoretical explanation for this finding. It is possible that the presence of negative connections increases the network’s instability, by modulating some self-activation loops. Although future studies are needed to support this hypothesis, it opens up an innovative perspective in the field of network theory. Finally, if we consider our findings from the perspective of adaptive or maladaptive stress responses, it is possible that the initial stress response of individuals in the no-PTSD group activated the ANS, and, consequently, they were not overwhelmed (reflected in characteristics consistent with a resilient network). Conversely, the initial stress response of individuals in the PTSD group was maladaptive, reflected in a state of sideration and manifested in peritraumatic numbness, along with characteristics consistent with a vulnerable network.
Moreover, our results highlight the central role of physical numbness in peritraumatic physical reactions reported by individuals who developed PTSD at time T1 (8–18 months post-trauma) and at time T2 (30–42 months post-trauma). In the context of PTSD, numbness can refer to both emotional and physical numbness. In the literature, numbness is generally considered on its emotional side and often understood as a type of dissociation that can occur in response to trauma (Friedman et al., Citation2007; Najavits et al., Citation2015). Emotional numbness refers to a feeling of detachment or disconnection from one's emotions, as if they are not fully experiencing them (Choi & Seng, Citation2017). A previous network analysis of interactions between posttraumatic stress symptoms and several covariates in individuals exposed to the Oslo terrorist attack showed the centrality of emotional numbness (Birkeland & Heir, Citation2017).
Our results complement these findings, providing evidence for the importance of the role of peritraumatic subjective physical reactions and the centrality of physical numbness in terms of networks analysis. Physical numbness is a loss of sensation in the body (Dworkin et al., Citation2018). This may be a result of the body's natural response to stress, which involves the diversion of blood flow away from non-essential areas of the body to support the response to a threat (Ginzburg et al., Citation2002; Lanius et al., Citation2010) and in which the periaqueducal gray matter may be particularly involved (Brandão & Lovick, Citation2019; Dampney, Citation2019). These conditions can also occur in ischemic or traumatic shock, with a redistribution of vascular flow favouring the brain at the expense of the extremities (Marion, Citation1991). The periaqueductal grey matter, which plays a central role in pain control, is also activated in this response to stroke (Ally et al., Citation2020; Rossi et al., Citation1994). The periaqueductal gray matter (PGM) is a complex structure that coordinates the behavioural and autonomic reactions to stress and injury. Previous research on the neurobiological basis of dissociation has highlighted alterations in functional connectivity in key brain regions such as the prefrontal cortex, amygdala, insula, and periaqueductal gray matter (Terpou et al., Citation2019).
While most of these studies have focused on dissociation as a specific alteration of consciousness (Spiegel et al., Citation2011), our research investigated numbness as a physical symptom, including tingling and physical numbness. Our study suggests that physical numbness may be a central part of the dissociative response which could involve the functional connectivity of key brain regions such as the periaqueductal grey matter. Further studies are needed to integrate these findings into the literature on the role of defensive behaviours in PTSD, and its dissociative subtype (Terpou et al., Citation2019). Although the study of peritraumatic physical symptoms is challenging, it is a key element not only to improve the follow-up of subjects who have experienced a trauma such as a terrorist attack, but also to expand our pathophysiological understanding of PTSD, and to develop effective treatments.
5. Implications
Our study explored subjective peritraumatic physical reactions in the context of a terrorist attack and highlighted some clinical implications. Our findings suggest that the intensity of physical reactions over time is significantly associated with the presence of PTSD. It also suggests that the structure of interactions between peritraumatic physical reactions could be responsible for a form of vulnerability. In practice, our results suggest that screening could be put in place regarding the risk of PTSD development, based on peritraumatic physical reactions in general, and numbness in particular, as the latter appears to be a central element of the development of PTSD. Furthermore, a pathophysiological distinction can be made between no-PTSD and PTSD groups.
PTSD has a major impact, not only on the quality of life of those affected, and their personal or professional life, but also a country’s health system. It is therefore a major public health issue. These issues are not negligible, particularly in terms of co-morbidities (anxiety disorders, depression, suicide, substance abuse), and costs (treatment, hospitalization, inability to work). By specifically targeting the type and intensity of peritraumatic reactions that victims experience, frontline practitioners may be able to rapidly detect whether the individual is able to overcome the stress reaction and indicate the nature and urgency of care the psychically injured person needs. This approach could be decisive in limiting, or even avoiding severe PTSD. Further studies are needed to confirm this postulate in the long term, and to verify whether the implementation of screening tools, based on the collection of peritraumatic physical symptoms can really have a predictive impact on the severity of PTSD expression. In addition, future studies are needed to determine whether there is a link between the structure of an individual’s peritraumatic physical symptoms, and the action of central inhibitory control systems (Mary et al., Citation2020), along with amygdalo hippocampal circuits (Postel et al., Citation2021; Ressler et al., Citation2022), which we know are involved in the expression of classic PTSD symptomatology.
Furthermore, given the centrality of numbness in networks of peritraumatic physical symptoms in individuals with PTSD, this symptom appears to be a useful therapeutic target, notably as tonic immobility has been associated with the suppression of autonomic activation of trauma-related recollections. In general, cognitive–behavioural treatments aim to decrease any physiological autonomic activation in response to trauma-related recollections, using strategies such as exposure, relaxation and cognitive restructuring (Foa & Rothbaum, Citation1998), and have been shown to be effective in many cases of PTSD. Conversely, individuals with a diminished physiological autonomic response (as in the case of tonic immobility) may not be the best candidates for these exposure therapies (Hembree et al., Citation2003).
6. Strengths and limitations
Despite some limitations, which we discuss below, our study has a solid methodological basis. It draws upon standardized measurement instruments, which have been scientifically validated as reliable, and interviews were run by professionals trained in the exercise. The sample is relatively homogenous with respect to the time that has passed since the traumatic event, and their characteristics (due to the targeted nature of the attacks with respect to their spatial and temporal dimensions). Although it could be considered a small sample (n = 80), this corpus remains very singular in its dimension given the traumatic event (terrorist attack) and the attrition rate remains low (13.75%). The design of the longitudinal study with repeated measures allowed us to include an early psychopathological analysis of victims, and to assess and confirm the strength of the findings over time. Finally, the network modelling approach that we adopt to analyse interactions between peritraumatic physical symptoms is an original, dynamic advance in the field of trauma semiology.
However, the conclusions that can be drawn from our results are limited, for several reasons. First, the participants were relatively young, highly educated, and not necessarily representative of the general population. Caution must therefore be exercised when generalizing our findings.
Second, the retrospective measurement of peritraumatic physical reactions may lead to reporting and memorization biases, which must be considered when interpreting the results, particularly in the case of dissociative reactions. Although this bias could be limited by running studies in the context of emergency facilities, this poses both ethical and practical problems, due to the difficulty of integrating any assessment into existing procedures.
Third, it is noteworthy that while our findings revealed peri-traumatic numbness to be a central factor in distinguishing the PTSD and no-PTSD groups at both time 1 (8–18 months post-trauma) and time 2 (30–42 months post-trauma), this result should be interpreted with caution. Specifically, it is important to consider that the methodology employed at time T1, the EBIC lasso method, differs from that used at time T2, the TMFG method. Methodologically, the state of the art in modelling psychological interaction networks is based on lasso regularized Markov Random Fields (MRF), which utilize Gaussian graphical models (GGM) for ordinal or continuous data (Epskamp et al., Citation2018). The EBIC lasso function is commonly used to compute Gaussian graphical model using graphical lasso based on extended BIC criterium. However, this method is not suitable for small sample size since it considers weak connections at zero (Epskamp et al., Citation2018). Recently, the Information Filtering Networks (IFN) approach has been proposed as a method for estimating networks, centrality, and connection strengths in small samples (Christensen et al., Citation2018). IFNs estimate a fixed number of connections and the Triangulated Maximally Filtered Graph (TMFG) is a variant of IFN that filters the network by generating a chordal network. The advantage of TMFG is that the number of connections is constant and does not vary with sample size, thus protecting comparability between two samples. However, the limitation of TMFG is that it can add unnecessary connections. Therefore, the consistency of the results across time points may be influenced by the differences in the analytical approaches. Further research is needed to replicate these findings using consistent methodological approaches across both time points.
Fourth, our study was based on psychopathological data regarding the development of PTSD and does not explore physiobiological predictive factors. Further research is clearly needed to clarify the peritraumatic mechanisms involved in the clinical trajectory of individuals exposed to a traumatic event, and to confirm their relevance.
Supplemental Material
Download MS Word (981.6 KB)Supplemental Material
Download MS Word (18.9 KB)Acknowledgements
We thank all participants for volunteering to contribute to this study and the victims’ associations who supported the project. We are grateful to Carine Klein-Peschanski, General Secretary to the 13-November Program and Equipment Excellence MATRICE and the researchers; psychologists B. Marteau, and L. Besnehard; technicians; and administrative staff at U1077 (Caen), at ‘Programme 13-Novembre’ in Paris, at INSERM ‘Délégation Régionale Nord-Ouest’. We also thank E. Seery for English-language editing of the main text.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Data availability statement
Due to the nature of the research, participants in this study did not agree to share their data publicly, consequently, supporting data is not available.
References
- Adams, D. B., Baccelli, G., Mancia, G., & Zanchetti, A. (1969). Cardiovascular changes during naturally elicited fighting behavior in the cat. The American Journal of Physiology, 216(5), 1226–1235. https://doi.org/10.1152/ajplegacy.1969.216.5.1226
- Ally, A., Powell, I., Ally, M. M., Chaitoff, K., & Nauli, S. M. (2020). Role of neuronal nitric oxide synthase on cardiovascular functions in physiological and pathophysiological states. Nitric Oxide, 102, 52–73. https://doi.org/10.1016/j.niox.2020.06.004
- American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). American Psychiatric Publishing.
- Birkeland, M. S., & Heir, T. (2017). Making connections: Exploring the centrality of posttraumatic stress symptoms and covariates after a terrorist attack. European Journal of Psychotraumatology. https://doi.org/10.1080/20008198.2017.1333387
- Borgatti, S. P. (2005). Centrality and network flow. Social Networks, 27(1), 55–71. https://doi.org/10.1016/j.socnet.2004.11.008
- Borsboom, D. (2017). A network theory of mental disorders. World Psychiatry, 16(1), 5–13. https://doi.org/10.1002/wps.20375
- Borsboom, D., & Cramer, A. O. (2013). Network analysis: An integrative approach to the structure of psychopathology. Annual Review of Clinical Psychology, 9(1), 91–121. https://doi.org/10.1146/annurev-clinpsy-050212-185608
- Bossini, L., Ilaria, C., Koukouna, D., Caterini, C., Olivola, M., & Fagiolini, A. (2016). PTSD in victims of terroristic attacks – a comparison with the impact of other traumatic events on patients’ lives. Psychiatria Polska, 50(5), 907–921. https://doi.org/10.12740/PP/65742
- Brandão, M. L., & Lovick, T. A. (2019). Role of the dorsal periaqueductal gray in posttraumatic stress disorder: Mediation by dopamine and neurokinin. Translational Psychiatry, 9(1), 232. https://doi.org/10.1038/s41398-019-0565-8
- Bringmann, L., Pe, M. L., Vissers, N., Ceulemans, E., Borsboom, D., Vanpaemel, W., Francis Tuerlinckx, F., & Kuppens, P. (2016). Assessing temporal emotion dynamics using networks. Assessment, 23(4), 425–435. https://doi.org/10.1177/1073191116645909
- Cannon, W. B. (1929). Bodily changes in pain, hunger, fear and rage (2nd ed.). Appleton.
- Choi, K. R., & Seng, J. S. (2017). Sexual assault and emotional numbness: A study of female veterans with PTSD. Journal of Interpersonal Violence, 32(5), 629–650.
- Christensen, A. P., Kenett, Y. N., Aste, T., Silvia, P. J., & Kwapil, T. R. (2018). Network structure of the Wisconsin Schizotypy scales–short forms: Examining psychometric network filtering approaches. Behavior Research Methods, 50(6), 2531–2550. https://doi.org/10.3758/s13428-018-1032-9
- Cramer, A. O., van Borkulo, C. D., Giltay, E. J., van der Maas, H. L., Kendler, K. S., Scheffer, M., & Borsboom, D. (2016). Major depression as a complex dynamic system. PLoS One, 11(12), e0167490. https://doi.org/10.1371/journal.pone.0167490
- Crocq, L. (1999). Les Traumatismes psychiques de guerre. Odile Jacob.
- Dampney, R. A. L. (2019). Central mechanisms generating cardiovascular and respiratory responses to emotional stress. In J. F. Storz (Ed.), Stress: Physiology, biochemistry, and pathology (pp. 391–402). Academic Press.
- DiGangi, J. A., Gomez, D., Mendoza, L., Jason, L. A., Keys, C. B., & Koenen, K. C. (2013). Pretrauma risk factors for posttraumatic stress disorder: A systematic review of the literature. Clinical Psychology Review, 33(6), 728–744. https://doi.org/10.1016/j.cpr.2013.05.002
- Dworkin, E. R., Wanklyn, S. G., Stasiewicz, P. R., & Coffey, S. F. (2018). Trauma-related physical numbness and dissociative experiences: The roles of anxiety sensitivity and emotion dysregulation. Journal of Trauma & Dissociation, 19(4), 458–472.
- Epskamp, S., Borsboom, D., & Fried, E. I. (2018). Estimating psychological networks and their accuracy: A tutorial paper. Behavior Research Methods, 50(1), 195–212. https://doi.org/10.3758/s13428-017-0862-1
- Epskamp, S., Cramer, A. O. J., Waldorp, L. J., Schmittmann, V. D., & Borsboom, D. (2012). qgraph: Network visualizations of relationships in psychometric data. Journal of Statistical Software, 48(4), 1–18. https://doi.org/10.18637/jss.v048.i04. http://www.jstatsoft.org/v48/i04/
- Falsetti, S. A., Resnick, H. S., Kilpatrick, D. G., & Freedy, J. R. (1994). A review of the "Potential Stressful Events Interview": A comprehensive assessment instrument of high and low magnitude stressors. Behavior Therapy, 17, 66–67.
- Foa, E. B., & Rothbaum, B. O. (1998). Treating the trauma of rape. Guilford Press.
- Ford, J. D. (2009). Chapter five - neurobiology of traumatic stress disorders and their impact on physical health. In J. D. Ford (Ed.), Posttraumatic stress disorder (pp. 117–141). Academic Press. https://doi.org/10.1016/B978-0-12-374462-3.00005-8
- Freeman, L. C. (1978). Centrality in social networks conceptual clarification. Social Networks, 1(3), 215–239. https://doi.org/10.1016/0378-8733(78)90021-7
- Fried, E. I., van Borkulo, C. D., Cramer, A. O., Boschloo, L., Schoevers, R. A., & Borsboom, D. (2017). Mental disorders as networks of problems: A review of recent insights. Social Psychiatry and Psychiatric Epidemiology, 52(1), 1–10. https://doi.org/10.1007/s00127-016-1319-z
- Friedman, M. J., Keane, T. M., & Resick, J. A. (2007). Physical numbing and dissociation in combat-related posttraumatic stress disorder. Journal of Traumatic Stress, 20(4), 601–606. https://doi.org/10.1002/jts.20234
- Fruchterman, T. M. J., & Reingold, E. M. (1991). Graph drawing by force-directed placement. Software: Practice and Experience, 21(11), 1129–1164. https://doi.org/10.1002/spe.4380211102
- Gentle, M. J., Jones, R. B., & Wooley, S. C. (1989). Physiological changes during tonic immobility in Gallus var domesticus. Physiology and Behavior, 46(5), 843–847. https://doi.org/10.1016/0031-9384(89)90046-2
- Gibert, L., Verdonk, C., Tarquinio, C., Falissard, B., El Hage, W., & Trousselard, M. (2021). The protective effect of trait mindfulness on post-traumatic stress disorder: A 4.5-year prospective cohort study. Brain & Behavior, 11(8), e02163. https://doi.org/10.1002/brb3.2163
- Ginzburg, K., Solomon, Z., & Bleich, A. (2002). Repressive coping style, acute stress disorder, and posttraumatic stress disorder after myocardial infarction. Psychosomatic Medicine, 64(5), 748–757. https://doi.org/10.1097/01.psy.0000021949.04969.2e
- Godoy, L. D., Rossignoli, M. T., Delfino-Pereira, P., Garcia-Cairasco, N., & Umeoka, E. H. L. (2018). A comprehensive overview on stress neurobiology: Basic concepts and clinical implications. Frontiers in Behavioral Neuroscience, 12, 127. https://doi.org/10.3389/fnbeh.2018.00127
- Greene, T., Palgi, Y., Dicker-Oren, S. D., & Gilbar, O. (2020). A network analysis of peritraumatic distress reactions and their relation to posttraumatic stress symptoms in people exposed to community fires. Psychiatry, 83(4), 375–389. https://doi.org/10.1080/00332747.2020.1762393
- Hembree, E. A., Rauch, S. A. M., & Foa, E. B. (2003). Beyond the manual: The insider’s guide to exposure for PTSD. Cognitive and Behavioral Practice, 10(1), 22–30. https://doi.org/10.1016/S1077-7229(03)80005-6
- Hess, W. R. (1957). Functional organization of the diencephalons. Grune & Stratton.
- Hirsch, M., Carli, P., Nizard, R., Riou, B., Baroudjian, B., Baubet, T., Chhor, V., Chollet-Xemard, C., Dantchev, N., Fleury, N., Fontaine, J. P., Yordanov, Y., Raphael, M., Burtz, C. P., Lafont, A., & Health Professionals of Assistance Publique-Hôpitaux de Paris (APHP). (2015). The medical response to multisite terrorist attacks in Paris. Lancet, 386(10012), 2535–2538. https://doi.org/10.1016/S0140-6736(15)01063-6
- Kalisch, R., Cramer, A., Binder, H., Fritz, J., Leertouwer, I., Lunansky, G., Meyer, B., Timmer, J., Veer, I. M., & van Harmelen, A. L. (2019). Deconstructing and reconstructing resilience: A dynamic network approach. Perspectives on Psychological Science, 14(5), 765–777. https://doi.org/10.1177/1745691619855637
- Lacey, J. I. (1967). Somatic response patterning and stress: Some revisions of activation theory. In M. H. Appleyo & R. Trumble (Eds.), Psychological stress (pp. 14). Appleton-Century-Crofts.
- Lanius, R. A., Vermetten, E., & Pain, C. (2010). The impact of early life trauma on health and disease: The hidden epidemic. Cambridge University Press.
- Leone, G., Postel, C., Mary, A., Fraisse, F., Vallée, T., Viader, F., de La Sayette, V., Peschanski, D., Dayan, J., Eustache, F., & Gagnepain, P. (2022). Altered predictive control during memory suppression in PTSD. Nature Communications, 13(1), 3300. https://doi.org/10.1038/s41467-022-30855-x
- Marion, D. W. (1991). Traumatic shock: The pathophysiology of hemodynamic collapse. Journal of Trauma, 31(6), 754–765. https://doi.org/10.1097/00005373-199106000-00001
- Marx, B. P., Forsyth, J. P., Gallup, G. G., Fusé, T., & Lexington, J. M. (2008). Tonic immobility as an evolved predator defense: Implications for sexual assault survivors. Clinical Psychology: Science and Practice, 15(1), 74–90. https://doi.org/10.1111/j.1468-2850.2008.00112.x
- Mary, A., Dayan, J., Leone, G., Postel, C., Fraisse, F., Malle, C., Vallée, T., Klein-Peschanski, C., Viader, F., de la Sayette, V., Peschanski, D., Eustache, F., & Gagnepain, P. (2020). Resilience after trauma: The role of memory suppression. Science, 367(6479), eaay8477. https://doi.org/10.1126/science.aay8477
- Massara, G. P., Di Matteo, T., & Aste, T. (2016). Network filtering for big data: Triangulated maximally filtered graph. Journal of Complex Networks, 5, 161–178. https://doi.org/10.1093/comnet/cnw016
- Motreff, Y., Pirard, P., Baubet, T., Chauvin, P., & Vandentorren, S. (2018). Enquête de santé publique post-attentats du 13 novembre 2015 (ESPA 13-Novembre): premiers résultats concernant les intervenants. Bulletin Epidémiologique Hebdomadaire, (38–39), 756–764. http://invs.santepubliquefrance.fr/beh/2018/38-39/2018_38-39_3.html
- Najavits, L. M., Hruska, T. W., & Herman, J. C. (2015). The relationship between dissociative symptoms, PTSD, and physical numbing in veterans. Journal of Trauma & Dissociation, 16(5), 555–570. https://doi.org/10.1080/15299732.2015.1032932
- Nelson, B., McGorry, P. D., Wichers, M., Wigman, J., & Hartmann, J. A. (2017). Moving from static to dynamic models of the onset of mental disorder: A review. JAMA Psychiatry, 74(5), 528–534. https://doi.org/10.1001/jamapsychiatry.2017.0001
- O’Donnell, M. L., Creamer, M. C., Parslow, R., Elliott, P., Holmes, A. C. N., Ellen, S., Judson, R., McFarlane, A. C., Silove, D., & Bryant, R. A. (2008). A predictive screening index for posttraumatic stress disorder and depression following traumatic injury. Journal of Consulting and Clinical Psychology, 76(6), 923–932. https://doi.org/10.1037/a0012918
- Park, J., Marvar, P. J., Liao, P., Kankam, M. L., Norrholm, S. D., Downey, R. M., McCullough, S. A., Le, N. A., & Rothbaum, B. O. (2017). Baroreflex dysfunction and augmented sympathetic nerve responses during mental stress in veterans with post-traumatic stress disorder. Journal of Physiology, 595(14), 4893–4908. https://doi.org/10.1113/JP274269
- Pirard, P., Baubet, T., Motreff, Y., Rabet, G., Marillier, M., Vandentorren, S., Vuillermoz, C., Stene, L. E., & Messiah, A. (2020). Use of mental health supports by civilians exposed to the November 2015 terrorist attacks in Paris. BMC Health Services Research, 20(1), 959. https://doi.org/10.1186/s12913-020-05785-3
- Porges, S. W. (1995). Orienting in a defensive world: Mammalian modifications of our evolutionary heritage. A polyvagal theory. Psychophysiology, 32(4), 301–318. https://doi.org/10.1111/j.1469-8986.1995.tb01213.x
- Porges, S. W. (2001). The polyvagal theory: Phylogenetic substrates of a social nervous system. International Journal of Psychophysiology, 42(2), 123–146. https://doi.org/10.1016/S0167-8760(01)00162-3
- Porges, S. W. (2007). The polyvagal perspective. Biological Psychology, 74(2), 116–143. https://doi.org/10.1016/J.BIOPSYCHO.2006.06.009
- Postel, C., Mary, A., Dayan, J., Fraisse, F., Vallée, T., Guillery-Girard, B., Viader, F., Sayette, V., Peschanski, D., Eustache, F., & Gagnepain, P. (2021). Variations in response to trauma and hippocampal subfield changes. Neurobiology of Stress, 15, Article 100346. https://doi.org/10.1016/j.ynstr.2021.100346
- Reese, W. G., Newton, J. E., & Angel, C. (1982). Induced immobility in nervous and normal pointer dogs. Journal of Nervous & Mental Disease, 170(10), 605–613. https://doi.org/10.1097/00005053-198210000-00004
- Ressler, K. J., Berretta, S., Bolshakov, V. Y., Rosso, I. M., Meloni, E. G., Rauch, S. L., & Carlezon, W. A., Jr. (2022). Post-traumatic stress disorder: Clinical and translational neuroscience from cells to circuits. Nature Reviews Neurology, 18(5), 273–288. https://doi.org/10.1038/s41582-022-00635-8
- Rhemtulla, M., Fried, E. I., Aggen, S. H., Tuerlinckx, F., Kendler, K. S., & Borsboom, D. (2016). Network analysis of substance abuse and dependence symptoms. Drug and Alcohol Dependence, 161, 230–237. https://doi.org/10.1016/j.drugalcdep.2016.02.005
- Rossi, F., Maione, S., & Berrino, L. (1994). Periaqueductal gray area and cardiovascular function. Pharmacological Research, 29(1), 27–36. https://doi.org/10.1016/1043-6618(94)80095-2
- Selye, H. (1956). The stress of life. McGraw-Hill.
- Solomon, Z., & Mikulincer, M. (2006). Trajectories of PTSD: a 20-year longitudinal study. The American journal of psychiatry, 163(4), 659–666. https://doi.org/10.1176/ajp.2006.163.4.659
- Spiegel, D., Loewenstein, R., Lewis-Fernández, R., Sar, V., Simeon, D., Vermetten, E., Cardena, E., & Dell, P. F. (2011). Dissociative disorders in DSM-5. Depression and Anxiety, 28(12), E17–E45. https://doi.org/10.1002/da.20767
- Terpou, B. A., Harricharan, S., McKinnon, M. C., Frewen, P., Jetly, R., & Lanius, R. A. (2019). The effects of trauma on brain and body: A unifying role for the midbrain periaqueductal gray. Journal of Neuroscience Research, 97(9), 1110–1140. https://doi.org/10.1002/jnr.24447
- Van Borkulo, C. D. (2018). Network comparison test: Permutation-based test of differences in strength of networks. https://github.com/cvborkulo/NetworkComparisonTest
- Van Borkulo, C. D., Boschloo, K. J., Tio, P. L., Schoevers, R. A., Borsboom, D., & Waldorp, L. J. (2016). Comparing network structures on three aspects: A permutation test [Manuscript submitted].
- Volchan, E., Souza, G. G., Franklin, C. M., Norte, C. E., Rocha-Rego, V., Oliveira, J. M., David, I. A., Mendlowicz, M. V., Coutinho, E. S., Fiszman, A., Berger, W., Marques-Portella, C., & Figueira, I. (2011). Is there tonic immobility in humans? Biological evidence from victims of traumatic stress. Biological Psychology, 88(1), 13–19. https://doi.org/10.1016/j.biopsycho.2011.06.002
- Williamson, J. B., Porges, E. C., Lamb, D. G., & Porges, S. W. (2015). Maladaptive autonomic regulation in PTSD accelerates physiological aging. Frontiers in Psychology, 5, 1571. https://doi.org/10.3389/fpsyg.2014.01571