Stress and compulsive buying-shopping disorder: A scoping review

Introduction: Theoretical frameworks of


Introduction
Compulsive buying-shopping disorder (CBSD) is characterized by preoccupations/urges to buy, impaired control over buying, not using consumer goods at all/appropriately, buying/shopping as mean to control emotions, persistence to buy despite negative consequences, and experiencing negative emotional and/or cognitive states when reducing or quitting with buying/shopping [1].CBSD constitutes a rather prevalent phenomenon with a point prevalence of about 5% [2].It is listed in the coding tool of the International Statistical Classification of Diseases and Related Health Problems (ICD-11) as an example of "other specified impulse control disorder" (6C7Y) but not as a standalone diagnosis [3].Some researchers argue that CBSD would be better classified as a behavioral addiction.This is proposed due to similarities with other addictive behaviors, such as cue reactivity and craving, impaired control over the behavior as well as reward and relief motivations [4][5][6].
CBSD was firstly mentioned in psychiatry textbooks more than 100 years ago (for an overview see: [7]).To that time, termini like oniomania or impulsive mania were used for CBSD-like phenomena [7].From the 1980s and 1990s on, different terminology, e.g., compulsive buying, pathological buying, shopping addiction, was established [7].In this article, we refer to the ICD-11 terminology compulsive buying-shopping disorder.The above-mentioned terms refer to the same phenomenon although with different conceptualization/nosology (either impulsecontrol disorder, obsessive-compulsive disorder or behavioral addiction) [7].Impulsive buying on the contrary that is often investigated in the field of economy psychology or marketing research rather considers temporary overshopping that often lacks clinically relevant impairment and thus does not represent a form of CBSD [4].Concepts like excessive buying or dysfunctional buying are related to certain aspects of CBSD (i.e., loss of control or emotion regulation) but do not necessarily correspond to the full picture of CBSD [1].
An emerging field constitutes online CBSD which can be viewed as the virtual equivalent of traditional CBSD or a potential specific internetuse disorder [8,9].Research on affective and cognitive mechanisms of CBSD demonstrates a profound level of evidence for cue reactivity and craving and mostly indicates inhibitory control deficits in individuals with CBSD [10].Research on cognitive and affective biases provided mixed results and research on habit formation is not yet far enough advanced to be able to derive clear conclusions [10].Precursors of CBSD include high impulsivity, strong materialistic values orientation, selfdiscrepancies, reward sensitivity and low self-esteem (e.g., [11][12][13]).Another factor that has been related to addictive behavior is stress.However, stress has been hardly investigated as an antecedent or correlate of CBSD.
Stress induces a neuroendocrine response which activates the sympathetic nervous system and the hypothalamic pituitary-adrenal (HPA) axis to cope with the demands of a homeostatic challenge (i.e., stress) and to re-establish homeostasis [14,15].In individuals with addictive behaviors and substance-related disorders, stress may boost seemingly habitual (or even compulsive) behaviors at the expense of goal-directed, cognitively controlled behavior [16][17][18].Stress is thought to increase the risk of relapse in addictions [17,19].As outlined in the Interaction of Person-Affect-Cognition-Execution (I-PACE) model of addictive behaviors [20,21], stress may trigger affective responses or craving towards addiction-related stimuli [20].
To our knowledge, theoretical models of CBSD have not explicitly taken stress into account [22][23][24][25][26] (for details see Table 1, for further description of the models and in how far they included stress see Supplementary Material Text S1).All five models at least indirectly describe a relief from negative mood as a motive for engaging in CBSD.Stress, which has been repeatedly linked to the development of mental disorders in general [27], including substance use disorders [19,28] and behavioral addictions [20,[29][30][31], has only been integrated in the model by Trotzke et al. [23].
Stress is however considered important in generic models of behavioral addictions (overview in table S1 in the supplement).In the components model of addiction [32], stress is mentioned in the context of mood modification as a "destressing" (p.193) effect of the addiction.In an earlier version of the I-PACE model [21], stress is referred to in the context of stress vulnerability as a person-related factor favoring specific internet-use disorders.Also, stress (e.g., interpersonal stress) was considered to influence the situation in which for instance craving and the urge for emotion regulation occur [21].Furthermore, stress is indirectly included as adverse childhood experiences in the initial and updated version of the I-PACE model [20,21].Two pathways can be considered in which stress contributes to CBSD: First, stress vulnerability, general perceived stress or chronic stress constitute a personrelated predisposing variable for CBSD (pathway 1) [20].According to this proposed mechanism, elevated (chronic) stress levels may be associated with more CBSD symptoms if coinciding with other predisposing variables such as genetic predispositions and specific values (e.g., high materialistic value orientation) [20].Second, perceived stress during the day might act as internal trigger leading to affective or craving responses which "activates" the inner circle of the I-PACE model (pathway 2) [20].CBSD episodes may occur when perceived stress during the day increases to an extraordinary extent that cannot be coped with functionally.Blaszczynski et al. [33] mention life stress as an example of emotional variabilities in their pathway model for problematic gambling which assigns stress a rather subordinate role.Dong and Potenza [31] refer to reward sensation and stress relief as one of the three domains indirectly fostering gaming behavior.Their model for gaming disorder thus places stress as a central contributing factor to gaming disorder.Two other models that are less frequently mentioned within the context of behavioral addictions are the general strain theory [34] in its adoption for internet addiction [35] and the diathesis-stress model applied to gaming disorder [36].Jun et al. [35] postulate that stress leads to the experience of negative mood which then leads to problematic internet use.Li et al. [36] investigated the well-known concept of a diathesisstress model in gaming disorder.The diathesis-stress model refers to an interaction of stress (e.g., life events) and personality factors or genetic aspects.Concretely, problematic behaviors may occur when stressful events meet favoring conditions such as temperamental or genetic factors representing fertile soil for the development of e.g., gaming disorder.This theory plausibly regards the occurrence of stressful events as a central factor as it causes the "barrel overflow".
Given the assumed relevance of stress in the development and maintenance of CBSD on the one hand [20] and the relatively little focus on this link in CBSD-specific models on the other hand, the aim of the current scoping review is to provide an overview of empirical results on the association between stress and CBSD and to draw conclusions with respect to theoretical CBSD models and clinical implications.Based on the I-PACE model which has been successfully applied to CBSD with respect to cognitive mechanisms [6,10], we have the following hypotheses: 1) General perceived stress is linked to symptom severity of CBSD.
2) The occurrence of momentary or daily stress is associated with an increase in CBSD symptoms/episodes.
Hypothesis 1 builds on pathway 1 of the I-PACE model and refers to the person-related stress vulnerability or non-situational stress.Hypothesis 2 refers to pathway 2 and thus to stress as situational factor which triggers CBSD episodes.

Method
A scoping review of the literature including the search engines PubMed, Scopus and Web of Science was performed.The search was carried out in November 2022, was updated in November 2023 and was not restricted regarding publication year.A broad search string was used covering multiple terms for CBSD (e.g., CBSD, shopping addiction) and the most relevant terms concerning stress (i.e., stress, strain, cortisol).The detailed search strings for PubMed, Scopus and Web of Science can be found in the supplementary material table S2.
Inclusion criteria were original research articles on CBSD (or at least problematic levels of buying-shopping behavior) and (perceived or objective markers of) stress that also examined the association between stress and CBSD or the influence of stress on CBSD or vice versa as these were the main questions of interest.We included all articles that considered CBSD, regardless of whether the shopping environment was specified or not (i.e., offline, online, mixed, or not specified).Exclusion criteria were non-original articles (e.g., reviews), non-quantitative articles, case studies, articles that did not examine stress or CBSD, or the link between those two/the influence of the one on the other and articles that investigated persons with CBSD being the result of a medication/ other mental disorder.Also, articles in a language other than German or English were excluded.Screening process by application of those criteria to title/abstracts of the articles was carried out by the first (TAT) and second author (AMS).An agreement could be achieved concerning the few discrepancies in title/abstract screening.This resulted in 31 articles being selected for full text screening which reduced the final number of eligible articles to 16.For details on screening and selection process of articles see fig.S1.
It was further decided to group the articles by methods employed in these studies as they are content-wise extremely heterogeneous.This decision led to the following sections: Questionnaire-based crosssectional evidence that is either correlative ('Correlative evidence') or relies on between-groups comparisons ('Between-group comparisons'), questionnaire-based cross-sectional evidence that involves (theorydriven) model testing/generating (e.g., regression or structure equation models; 'Model-based investigations'), ecological momentary assessments (EMA) with a high number of measurement points ('Ecological momentary assessments') and studies that experimentally induced acute stress/ negative mood and observed reactions on CBSD behavior ('Experimental investigation').
Interpretation of correlation sizes was based on the conventions implemented by Cohen [37].For between-group comparisons, effect sizes for pairwise comparisons in the between-group comparisons section were, in case not already given in the original article, calculated by use of one web-based tool [38] and IBM SPSS version 28/29.Conventions for effect sizes established by Cohen [37] were used.Note that in case of between-group comparisons with more than two groups, the respective comparison of the CBSD group(s) with the control group was calculated based on the indicated means and standard deviations (see supplementary material 'Calculation of t-tests' and 'Calculation of effect sizes').

Correlative evidence
Nine studies were identified that examined correlations between general perceived stress and CBSD symptom severity mostly using Pearson correlations.Table 2 provides an overview of these studies, including the instruments used to assess CBSD symptom severity and general perceived stress.All studies found significant associations between general perceived stress and CBSD symptom severity (except for the association between general perceived stress and CBSD measured with the subscale 'Feelings about shopping' of the Edwards Compulsive Buying Scale in [39]), but with varying size of correlations.Singh et al. [40] reported significant but very small correlations between CBSD symptom severity and general perceived stress.The studies by Gallagher et al. [39] and Ridgway et al. [41] found small to moderate correlations.A moderate to large correlation was observed in two studies [42,43].Large correlations were found in the other four studies [44][45][46][47].As shown in Table 2, these correlations were found in both adolescents and older persons.The stress subscale of the Depression Anxiety Stress Scale-21 (DASS-21 [48,49]) was used in six studies to capture general perceived stress [39,[41][42][43]46,47].Vogel et al. [45] used the Perceived Stress Questionnaire [50] and Maraz et al. [44] and Singh et al. [40] (at least partly) used the Perceived Stress Scale (PSS [51]).A variety of different scales to capture CBSD symptom severity was used: For general CBSD symptom severity, the Compulsive Buying Scale ( [52], used in [39,41,46]), the Richmond Compulsive Buying Scale respectively a Polish or revised version of it ( [41], used in [41]; Polish version: [43]; revised version: [47]]), the Edwards Compulsive Buying Scale ( [53], used in [39]), the Bergen Shopping Addiction Scale ( [54], used in [44]), a six-item scale based on the Compulsive Buying Scale [52] and, according to the authors, the Adolescent Compulsive Buying Scale [55] used in [40] and the Compulsive Acquisition Scale ( [56], used in [46]) were used.For online CBSD symptom severity, the Compulsive Online Shopping Scale ( [57], used in [44]), the short Internet Addiction Test modified version for shopping ( [58], adapted for shopping as in [59], Notes.CBSD = compulsive buying-shopping disorder.a Original version of the CBSD questionnaire containing a stress-related item or stress-related item included in the version of the scale reported in this study.Singh et al. [40] reported having measured CBSD with items based on the Compulsive Buying Scale and the Adolescent Compulsive Buying Scale.However, the items reported in the article seemed to be based rather on the Compulsive Buying Scale and did not contain the stress-related item of the Adolescent Compulsive Buying Scale.CBSD-NAFI = compulsive buying-shopping disorder without acquisition of free items, CP = control participants, PG = pathological gambling, SNUD = social networksuse disorder.1 d was not provided in the original articles and was thus calculated by the authors (see supplementary material).2 t-test was not provided in the original articles and was thus calculated by the authors (see supplementary material).a Original version of the CBSD questionnaire containing a stress-related item or stressrelated item included in the version of the scale reported in this study.
used in [45]) and the Online Compulsive Buying Scale ( [60,61], used in [42]) were employed.The studies by Vogel et al. [45] and Zheng et al. [42] explicitly considered online CBSD symptom severity only, whereas most studies examined offline CBSD symptom severity or had no explicit distinguishing characteristics.Only the study by Maraz et al. [44] examined both online and offline CBSD symptom severity.Overall, three studies investigated online CBSD symptom severity [42,44,45], including two studies [44,45] that reported among the strongest correlations of all included studies.The study with the youngest participant group found the weakest correlation between CBSD symptom severity and general perceived stress [40].Samples varied from relatively small samples in laboratory settings [45] to large online surveys [44].Convenience samples were used in all studies except for the study by Williams et al. [46] (see Table 2).

Between-group comparisons
Three studies examined the average differences in general perceived stress between individuals with problematic buying-shopping behavior/ CBSD and control participants (see Table 3).All studies found that individuals with problematic buying-shopping behavior/CBSD reported higher general perceived stress than control participants.It cannot be explicitly stated if these results refer to offline CBSD, online CBSD or a mixed form as shopping environment was not further specified.Harvanko et al. [62] collected data as part of a large computer survey of college students and assessed CBSD with the Minnesota Impulsive Disorders Interview [63].General perceived stress was measured with the PSS.The results showed a high effect size (d = 0.98; for computation, please see supplementary material) for the significant difference.Persons with CBSD had higher general perceived stress levels than the control participants.In the study by Williams et al. [64], participants were screened by telephone, and those who reported symptoms suggestive of CBSD (symptom severity measured with Yale Brown Obsessive Compulsive Scale -Shopping Version [65], clinical interview: SCID-IV TR for impulse-control disorders not elsewhere classified) were allocated to the CBSD group.This group was compared to a group of patients with pathological gambling and a control group without problematic gambling or buying-shopping.These groups were matched regarding education, ethnicity and marital status.Significant differences, as measured by the stress subscale of the DASS-21, were found between the three groups in an analysis of variance.The comparison between the individuals with CBSD and the control participants revealed a large effect (d = 1.10).No group differences were found between the CBSD and the pathological gambling group (d = 0.01).The comparisons indicated higher values for general perceived stress in the patient groups compared to the control participants.Another study by Williams et al. [46] distinguished control participants and two groups of persons with CBSD (i.e., with or without excessive acquisition of free items).The results showed that there are significant differences in general perceived stress between control participants and both CBSD groups.The comparisons between the two CBSD groups (without or with acquisition of free goods) and the control participants also indicated large effect sizes (d = 1.23 respectively d = 1.48).Again, the DASS-21 was the measurement instrument for general perceived stress and for CBSD, the Compulsive Acquisition Scale and the Compulsive Buying Scale were used.Persons with CBSD fulfilled the McElroy et al. [66] criteria for CBSD.The CBSD groups had higher stress levels than the control participants.
Recently, Wegmann et al. [67] compared women with CBSD and women with social-networks-use disorder with respect to perceived chronic stress using the Trier Inventory for Chronic Stress (TICS [68,69]).CBSD and social-networks-use disorder symptom severity was measured with the Assessment of criteria for specific internet-use disorders (ACSID-11 [70]) and an adapted version of the Assessment of internet and computer game addiction, Structured clinical interview on Internet-use disorders [Strukturiertes klinisches Interview zu Internetbezogenen Störungen] (AICA-SKI:IBS [71]) was used as clinical interview for CBSD and social-networks-use disorder.The effect size (d = 0.22) indicated that persons with CBSD tended to report more perceived chronic stress than those with social-networks-use disorder.
The comparison did however not reach significance.

Model-based investigations
Eight studies were identified that used either regression models or structural equation models to investigate the effects of general perceived stress and other variables on CBSD symptom severity (see Table 4).Moon et al. [47] investigated precursors and consequences of CBSD in shopping mall visitors in Pakistan using a structural equation model.Self-esteem, materialism, negative feelings as well as depression, anxiety and general perceived stress (as subscales of the DASS-21) were included as predictors in the model.CBSD symptom severity was measured with a revised form of the Richmond Compulsive Buying Scale [41] that was adjusted to non-western societies and that was reduced to a final version of four items [47].According to the authors, CBSD symptom severity should predict positive feelings and hiding behaviors.Results confirmed the above-mentioned hypotheses as each of the predictors significantly predicted CBSD symptom severity leading to 97% of variance in CBSD symptom severity being explained by the set of predictors.Moon et al. [47] referred to general perceived stress as "the strongest antecedent of compulsive buying" (p.561) as it showed the largest regression coefficient (see Table 4).
Aydin et al. [72] explored the impact of chronotype (morning-, evening-or neither type), impulsivity and fear of missing out on CBSD symptom severity in a sample of university students in Türkiye.They used the DASS-21 subscales together with sociodemographic factors, tobacco and caffeine use, impulsivity, sleep hygiene, fear of missing out, life satisfaction and chronotype in their hierarchical regression to predict CBSD symptom severity (measured with the Turkish version of Compulsive Buying Measurement Scale [73,74]).The model explained 30.4% of variance in CBSD symptom severity.General perceived stress was not a significant predictor of CBSD symptom severity.
Gallagher et al. [39] studied the effect of sex, anxiety sensitivity (i.e., "fear of arousal-related somatic sensations" [39], p. 38), and negative affect (depression, anxiety and general perceived stress as measured by the DASS-21 subscales) on CBSD symptom severity in a sample of Canadian university students.CBSD symptom severity was measured with the Compulsive Buying Scale [52] and with the three subscales 'Tendency or compulsion to spend', 'Feelings about shopping' and 'Postpurchase guilt' of the Edwards Compulsive Buying Scale [53] that resulted from an exploratory factor analysis the authors conducted.Four different hierarchical regression models were calculated ranging with variance explained from 9 to 17%.General perceived stress failed to predict CBSD symptom severity as measured with Compulsive Buying Scale.General perceived stress further failed to predict the Edwards Compulsive Buying Scale subscales 'Tendency or compulsion to spend', 'Feelings about shopping', but significantly predicted the subscale 'Postpurchase guilt'.
Tarka et al. [43] validated a Polish adaptation of the Richmond Compulsive Buying Scale [41] and intended to estimate the prevalence of CBSD in Polish young consumers.Besides this, they investigated if known precursors of CBSD found in previous literature in Western Europe/USA were valid also in Eastern Europe.They used a sample of students (high school, professional school and university students) aged 17-25 years and found that anxiety, depression, general perceived stress (as measured with the DASS-21 subscales), materialism, negative feelings and self-esteem were able to predict CBSD symptom severity.General perceived stress was the third most important predictor of CBSD symptom severity.Predictive ability for general perceived stress was stronger for female participants (γ = 0.59) than for male participants (γ = 0.40).
Singh et al. [40] conducted a study that examined familial and non-  [52] and, according to the authors, the Adolescent Compulsive Buying Scale [55]) and the influence of CBSD on post-purchase regret.Only five items of PSS-10 were used because "[t] […]he five items that did not particularly relate to younger people were excluded" (p.259).The authors used an Indian sample of 15-to 18-year-old adolescents.They included adolescents with "higher value" (p.258) on the PSS without stating how higher values were operationalized.Their analyses involved a structural equation model.General perceived stress was found to significantly predict CBSD symptom severity.Furthermore, the authors tested whether gender moderated the association between general perceived stress and CBSD symptom severity.They reported that gender did not significantly moderate the association between general perceived stress and CBSD symptom severity and concluded that "at higher level of general perceived stress, both boys and girls reported higher level of compulsivity" (p.264).This study differed from most other studies as general perceived stress was the only predictor of CBSD symptom severity in this model whereas in most other studies, stress was one predictor among an entire set of predictors.Maraz et al. [44] performed a study in which they assessed the impact of the first six months of COVID-19 pandemic (March to October 2020) on offline (measured with the Bergen Shopping Addiction Scale) and online CBSD behavior (measured with the Compulsive Online Shopping Scale).A second focus was placed on the impact of general perceived stress "above and beyond the association of income and [economic position] […]" (p.92) on CBSD symptom severity.They used subsequent cohorts of 25 persons so that measurement repeated data of individuals were not gathered.The authors used the PSS-14 as general perceived stress measure but restricted the reference period of the questionnaire to the last 7 days due to the dynamic conditions in times of COVID-19 pandemic.It was found that general perceived stress significantly predicted offline as well as online CBSD symptom severity but the effect of general perceived stress was considered "weak" (p.92).Furthermore, the association between general perceived stress and CBSD symptom severity seemed to be particularly relevant in a highincome group compared to a rather low-income group.The authors concluded that it is rather income/economic position that determines CBSD behavior during COVID-19 pandemic but not general perceived stress itself [44].
Wegmann et al. [67] investigated potential predictors for symptom severity of online CBSD (measured with the ACSID-11) in a clinical sample.They computed a regression model in which online CBSD symptom severity should be predicted by a set of potential vulnerability factors of CBSD (i.e., materialistic values orientation, impulsivity, chronic stress).Perceived chronic stress, measured with the TICS, failed to significantly predict online CBSD symptom severity.
Zheng et al. [42] investigated the relationship between online CBSD symptom severity (measured with the Online Compulsive Buying Scale), general perceived stress (measured the Chinese version of the DASS-21), negative coping and self-esteem in a sample of young female consumers from China.The authors proposed a moderated mediation model with a direct path from general perceived stress to online CBSD symptom severity and a mediation from general perceived stress to online CBSD symptom severity via negative coping.According to Zheng and colleagues [42], both paths were assumed to be moderated by self-esteem.They found a significant direct relationship between general perceived stress and online CBSD symptom severity that was partially mediated by negative coping.Furthermore, the direct path between online CBSD symptom severity and general perceived stress as well as the indirect path via negative coping were moderated by self-esteem with stronger effects for consumers with low self-esteem.Besides the study of Singh et al. [40], this study was one of the two studies that (also) assessed a direct path between general perceived stress and CBSD symptom severity without the influence of further predictors.

Ecological momentary assessments
Two studies used EMA to explore the relationship between perceived stress during the day and CBSD episodes in the natural environment of clinical samples [76,77].In addition, both studies were interested in the interplay of mood and CBSD episodes, whereas only results on stress will be reported here given the scope of this scoping review.In each study, the EMA was conducted over a time period of two weeks and signalcontingent, event-contingent as well as interval-contingent recordings were used to track CBSD episodes, (impact of) minor stressful events, and mood states during the day.For signal-contingent recordings, a sound reminded the participants at three semi-random time points during the day to answer a set of questions, while for the eventcontingent recordings, the participants should provide information directly after a CBSD episode occurred.Finally, interval-contingent recordings were scheduled at the end of the day.While Silbermann et al. [77] assessed perceived stress during the day only during the signalcontingent recordings, Müller et al. [76] had participants report stressful events and perceived stress also after each CBSD episode.To assess minor stressful events, both studies used a subset of items from the Daily Stress Inventory [78].To be eligible for participation, individuals had to meet the McElroy et al. [66] criteria andonly in the Silbermann et al. [77] studythe criteria for impulse-control disorder of the Structural Clinical Interview for DSM-IV [79].Müller et al. [76] additionally relied on the Compulsive Buying Scale (52).
Participants experienced on average two [77], respectively six [76] CBSD episodes during the two weeks.The proportion of participants who did not report any CBSD episodes was 12% in the study of Müller et al. [76] and 31% in the study of Silbermann et al. [77].Concerning the link between perceived stress during the day and CBSD episodes, Silbermann et al. [77] found that on days with CBSD episodes, participants reported significantly more stressful events.Additionally, the participants felt more stressed by these events, although this difference slightly missed significance.In contrast, Müller et al. [76] did not find a difference in perceived stress during the day between days with and without a CBSD episode.The authors additionally conducted within-day analyses, but did not observe a change in the perceived stress levels reported before or after a CBSD episode [76].

Experimental investigation
Only one study used an experimental procedure to investigate the influence of induced negative mood which -according to the authorswas defined as stress induction on CBSD symptoms [80].Besides the impact of induced stress/negative mood on CBSD symptoms in adolescents, Roberts et al. [80] were also interested in the moderating effect of gender.Their convenience sample consisted of 82 seventh-graders aged 12-13 years, with females being slightly overrepresented (n = 46).To induce stress (i.e.negative mood), the authors used text stimuli describing an academic stress scenario.Participants in the high stress condition received a scenario in which a student received poor results in the final exams and was punished therefore.In the low stress scenario, the student succeeded in the final exams and was rewarded.In both conditions, participants had to imagine how worried and upset they would feel in such a scenario.After the stress induction, CBSD symptoms were assessed with a ten-item scale developed by the authors for the purpose of the study (in the following, referred to as Compulsive Buying Scale by Roberts and colleagues).The items of the scale mainly covered aspects like craving, impulsive buying and buying to regulate one's mood (e.g., "When I go to the store, I always feel the urge to buy something" or "When I feel bad, I like to buy things") [80].
The manipulation check confirmed that the participants in the high stress scenario reported significantly higher levels of perceived stress than the participants in the low stress scenario.For the further analyses, however, the authors did not compare both experimental conditions, but grouped participants based on their stress ratings.Participants who reported perceived stress levels above the scale mean formed the high stress group, and participants with perceived stress ratings below the scale mean were allocated to the low stress group.The results revealed a significant main effect of stress, meaning that students in the high stress group displayed higher scores on the Compulsive Buying Scale by Roberts and colleagues compared to the low stress group.In addition, a main effect of gender was observed, meaning females scored higher on the Compulsive Buying Scale by Roberts and colleagues than males.A stress X gender interaction failed to reach significance.Based on their results, the authors concluded that compulsive buying is used, regardless of gender, as means to cope with stress [80].

Discussion
This scoping review aimed to evaluate research on the association between stress and CBSD.Drawing on the I-PACE model for behavioral addictions, two pathways in which stress might influence CBSD (symptoms) were examined.First, general perceived stress as personrelated vulnerability factor (pathway 1) was investigated.This pathway of the stress-CBSD association refers to correlative research, mean comparisons and regression/structural equation models.Second, stress as a situational factor representing an internal trigger (pathway 2) was considered.This pathway comprises studies with ecological momentary assessments and experimental manipulations of stress.Hypothesis 1.General perceived stress/chronic stress as stress vulnerability factor (pathway 1).
Concerning the first pathway, general perceived stress was shown to be positively associated with CBSD symptom severity as shown by nine studies of correlative examination.These correlations were found for both online and offline CBSD in different countries and different age groups, and the magnitude of the correlations varied.These results confirm hypothesis 1 but do not enable inferences regarding the direction of the association.
Furthermore, three studies involving mean comparisons showed large differences and higher general perceived stress values in persons with problematic CBSD behavior/CBSD compared to a control group.Of interest is also the lack of differences in general perceived stress/chronic stress between individuals with CBSD and individuals with other behavioral addictions [64,67].This indicates a relationship between general perceived stress/chronic stress and addictive behaviors regardless of the type of behavioral addiction, which is in line with theoretical models of addiction [20].Again, the direction of the stress-CBSD relation cannot be derived from these analyses.
Mixed evidence in the field of general perceived stress originated from regression/structural equation models that studied the impact of general perceived stress on CBSD symptoms severity.Five studies found that general perceived stress was among the significant predictors in their predictive set of variables for CBSD symptom severity [40,[42][43][44]47] whereas three studies (mostly) did not report significant prediction [39,67,72].The most obvious reason for these discrepancies lies within the methodological nature of the statistical approaches used in these studies: Regression coefficients of multiple regression models or structural equation models are specific to the particular set of predictors and results might change enormously if only one predictor is removed or added [81].To obtain accurate regression coefficients a "perfectly specified model" ([82], p. 2; i.e., model containing all relevant predictors and that thus is (almost) optimal) is required.However, "[t] […] he problem is that the true model is rarely, if ever, known" ([82], p. 2) which also applies to CBSD research as theoretical models in this field are rather sparse.Relying on single evidence can lead to ambiguous conclusions e.g., when it comes to including gender as a factor that might account for differences in CBSD.Singh et al. [40] for instance studied the sole effect of general perceived stress on CBSD symptom severity whereas Aydin et al. [72] incorporated general perceived stress in a set together with eleven other predictors.Furthermore, these models do not show the actual direct contribution of general perceived stress to CBSD symptom severity beyond significance of a predictor and no conventions exist for interpreting beta weight as the predictor set is highly specific.Theoretical models of behavioral addictions such as the I-PACE model [20,21] assume a complex interaction of several personrelated factors (e.g., stress vulnerability, impulsivity, genetic aspects, motives, values) as the basis for the development of behavioral addictions with affective, cognitive and executive factors.Given this theoretical basis, it does not seem surprising that direct pathways between stress and CBSD symptom severity are only partly of statistical significance.Thus, the stress-CBSD relationship might be more complex than just a linear relationship and might be influenced by several other mechanisms due to the interplay of several factors.Moreover, different measures were used for operationalization of general perceived stress and CBSD symptom severity in regression/structural equation models.This might be particularly relevant for general perceived stress as those studies using the DASS-21 [48,49] measured general perceived stress only with a subscale of the actual scale and used depression and anxiety as further predictors.This implies a certain risk for multicollinearity as depression and anxiety are usually strongly correlated with general perceived stress and studies should therefore conduct multicollinearity assessments like Moon et al. [47] did.Computing structural coefficients would have enabled further assessment of suppression or multicollinearity effects as differences in the magnitude of regression coefficients and structural coefficients might be indicative of these effect [81,83].This was unfortunately not possible as not all authors provided bivariate correlations between CBSD symptom severity and general perceived stress that are necessary for structural coefficient computation [84,85].
Taken together, evidence attributed to the first pathway (stress vulnerability, general perceived stress and chronic stress as personrelated factors) suggest an association between stress and CBSD (see Fig. 1).Results from different models however rather point towards a complex interaction of stress with other factors rather than a sole, strong and direct influence of stress on CBSD.Hypothesis 2. Acute stress and perceived stress during the day as situational triggers (pathway 2).
Regarding the second pathway, perceived stress during the day was captured in two studies using EMA.The diverging findings of the two EMA studies may be explained by different statistical approaches [76].In the study of Müller et al. [76], perceived stress on days with and without CBSD episodes was compared with a methodologically superior approach (mixed model analysis) to the Wilcoxon test for dependent samples used by Silbermann et al. [77].Apart from this methodological consideration, there might be another explanation for the different results.The participants in the study of Silbermann et al. [77] had started a group therapy for their pathological buying behavior, simultaneously with the beginning of the EMA study.Hence, they may already have begun to change their pathological buying behavior, an assumption that is supported by the lower number of CBSD episodes compared to the study of Müller et al. [76].On days without perceived stressful events, they might have been more successful to restrain from compulsive buying than on days with a higher number of or more intense stressors.On stressful days, they might have fallen back into their pathological buying habits, since stress can override goal-directed behavior [18] and promote relapse in addictions [86].In contrast, the participants in the study of Müller et al. [76], who were not in treatment at the time of the study, might have been less aware of or intended to change their pathological shopping behavior less.Consequently, they may have habitually engaged in CBSD episodes, even on less stressful days.Following the I-PACE model, perceived stress during the day might act as an internal trigger [20].This trigger then leads to an urge to shop which activates the inner circle of the I-PACE model and finally results in CBSD episodes as a dysfunctional coping strategy to reduce stress.For hypothesis 2 that built upon the role of perceived stress during the day as an internal trigger, mixed evidence was reported.This might be explained by methodological shortcomings of the studies but may also question the ubiquity in which stress leads to CBSD episodes.This process might only run under specific circumstances (e.g., unavailability of other, more functional coping strategies) that need to be further investigated, as some stressful situations seemed to lead to CBSD episodes whereas Fig. 1.Evidence for stress acting as vulnerability factor (predisposing variable) or situational factor for CBSD supporting (+) or not supporting (− ) the assumptions of the I-PACE model for addictive behaviors (Brand et al., 2019).Note.DV = dependent variable, CBSD = compulsive buying-shopping disorder.others did not.
The work of Roberts et al. [80] was, to the best of our knowledge, the only study that tried to experimentally manipulate stress and to study its impact on CBSD symptoms.While the authors' approach to experimentally manipulate stress is a valuable and much-needed addition to the mostly correlational questionnaire studies in this research area, the study has some shortcomings that challenge the interpretations made by the authors.First, participants had to indicate how worried and upset (i.e., stress according to the authors) they would feel if they were in this scenario.It can be debated if reading the academic failure scenario induced rather bad mood than stress and if reading the academic success scenario was rather a good mood scenario than a neutral condition.Second, asking participants to indicate how they would feel if they were in this scenario does not necessarily imply that the participants experienced more stress when reading those scenarios.Third, analyzing the effect of stress by comparing participants based on their stress ratings instead of comparing the two experimental conditions, weakens the possibility to make causal conclusions.Hence, the interpretation that stress triggered compulsive buying may be questioned.Alternatively, participants with higher CBSD symptom severity may have provided higher stress ratings, given that stress vulnerability has been suggested as predisposing factor for behavioral addictions [21].Consequently, the high-stress group might have consisted mainly of individuals with higher symptom severity of CBSD, leading to the observed link between stress and CBSD symptom severity.Adding to this line of argumentation, the Compulsive Buying Scale by Roberts and colleagues [80] the authors used does not measure momentary CBSD symptoms, cognitions or craving.To capture the effect of stress on CBSD symptom severity in an experimental setting, a measure that assessed affective and cognitive responses, for example craving or the desire for mood regulation, would have been more suitable.Besides its limited appropriateness to measure stress-induced CBSD tendencies, another shortcoming of the questionnaire used was the sparse information regarding its psychometric properties, given that the scale was developed for this very study and had not been validated before.Although the authors provided evidence for its internal reliability, which was with 0.79 close to a good reliability [87], evidence for its (diagnostic) validity is missing.
Overall, evidence for the second pathway in which stress acts as an internal trigger and situational factor is rather mixed (see Fig. 1).Further investigations concerning this pathway are needed.

General limitations of the studies included
Some studies relied on the Compulsive Buying Scale, a valid questionnaire that has been in use for many years [52].This being an advantage also brings the disadvantage with it that some items of the scale are rather outdated as they refer to payment methods that are not in frequent use any more such as checks (item 2c [88,89]).Besides this, the Compulsive Buying Scale does not capture all aspects of the proposed diagnostic criterion "A" for CBSD ("Persistent and/or recurrent dysfunctional buying/shopping-related behaviors, thoughts and related phenomena" [1], p. 217).The questionnaire refers only partly to negative consequences experienced by excessive buying/shopping (financial aspect A5d mostly) and does not refer to buying/shopping items "without utilizing them for their intended purposes" (A3 [1], p. 217) and it uses a rather non-intuitive evaluation procedure which operates with negative values and item-related weights [52].Other shortcomings associated with this questionnaire are a potentially under-exclusive cutoff value, further problems with particular items or the factor structure and signs of missing cultural fairness as a test [90].Regarding the Edwards Compulsive Buying Scale [53], Manolis et al. [90] also assumed psychometric shortcomings without being able to concretize this further.Maraz et al. [91] conducted further research on the Edwards Compulsive Buying Scale and failed to replicate the postulated fivefactor structure [91].They instead created a revised version of the final scale that also contains items of the initial item pool of the Edwards Compulsive Buying Scale and concluded that the Edwards Compulsive Buying Scale particularly refers to an emotional/affective dimension of CBSD [91].
Taken together, this review has to be considered under the prerequisite that it is based on studies with this or these specific questionnaire (s) and comparisons between studies with different operationalizations of CBSD are difficult as CBSD questionnaires differ also with regards to their classification of CBSD (impulse control disorder vs. obsessivecompulsive spectrum vs. behavioral addiction [92]).Fig. 2 shows an overview of the questionnaires used to capture CBSD and positions them within the nosological frameworks of impulse control disorders, obsessive-compulsive disorders and behavioral addictions.As can be seen, the underlying conceptualizations of the CBSD scales used in the included studies are diverse.Still, the results of the correlations and regression/structural equation models seem to be rather robust across different CBSD conceptualizations.Only few studies used the gold standard of clinical interviews to detect persons with CBSD (e.g., [67,77]).
A further limitation concerns the assessment of general perceived stress.The studies that were included in this review mostly used either the PSS that asks for stress in a reference of four weeks [51,93] or the DASS-21 that refers to the last week [48,49].Hence, although these two general perceived stress measures are correlated [94], they still cannot be compared without difficulties.Besides this, both scales rather capture shorter or longer excerpts of stressful experiences but not chronic perceived stress (e.g., as measured by the TICS [68,69]).Chronic perceived stress might be captured with a reference period of three months or longer as some studies even assume that chronic perceived stress would be lasting longer than 12 months [95].Moreover, none of the studies assessed stress biomarkers for chronic (e.g., hair cortisol) or acute stress response (e.g., cortisol, alpha-amylase).This might be problematic as subjective and physiological reactions to (acute) stress might differ.In a review, only 25% of the included studies observed a significant correlation between subjective and physiological stress measures [96].
Another shortcoming of the studies especially applying to regression/structural equation models is their cross-sectional nature.A stronger level of evidence for stress being a predisposing variable/ antecedent or consequence of CBSD would involve longitudinal data such as studies showing that stress can predict or follow CBSD episodes.
A further point that needs to be considered when interpreting the questionnaire-based results on the stress -CBSD association is that some scales capturing CBSD also ask about CBSD-related stress (e.g., 'For me, shopping is a way of facing the stress of my daily life and of relaxing' [73], used in [72]).This is the case in four of the included studies [44,46,64,72].One could argue that the stress items in the CBSD scales might partially explain the correlations or regression coefficients found in the respective studies.However, significant associations of perceived general stress with CBSD were found in all studies reporting correlative resultsindependent from including a stress-related item in the CBSD scale.Furthermore, Aydin et al. [72] also failed to observe a significant prediction of CBSD by perceived general stress despite using a CBSD scale with an item on CBSD-related stress.Of course, stress is often associated with strong negative affects which is captured in most CBSD scales.When responding to this item, participants might have not been able to clearly differentiate between stress and negative affect, so that the above-mentioned artifact cannot be entirely dispelled.At least, it became evident that the association between stress and CBSD seems not to be solely driven by the stress-related item in some CBSD scales.
Unfortunately, not all studies provided a clear outline of methodology/procedure further limiting the ability to draw conclusions from these results.The number of studies with methodological and statistical sound procedure is low.Studies relying on cross-sectional correlative evidence cannot be interpreted in a causal way.According to these results, stress could be both an antecedent and a consequence of CBSD as increasing debts and interpersonal conflicts might enhance stress levels as well.Another option would be a bidirectional relationship between stress and CBSD in which stress triggers CBSD episodes, which in turn result in stress and vice versa.This bidirectional relationship might establish a vicious circle in which shopping is carried out to escape from stress but on the same time elicits stress e.g., by negative consequences such as familial conflicts.This might result in a stronger need to escape negative emotions which may accelerate an already problematic shopping behavior.As there are several potential directions of the relation between stress and CBSD, regression and structural equation models could be criticized for assuming that stress predicts CBSD without a very profound theoretical basis in CBSD literature supporting this assumption.The study by Roberts et al. [80] attempted to manipulate stress and examine its influence on CBSD tendencies.However, the methodological shortcomings of the study mentioned above hinders the conclusions that can be drawn.

Specific limitations and strengths of this scoping review
For this scoping review, three databases (PubMed, Scopus, Web of Science) were considered.Articles that are not listed in these databases could thus not be incorporated.However, the databases searched are the most common ones and the potential loss of studies listed otherwise should be rather low.Articles were included if they considered perceived or objective stress levels measured with standardized questionnaires (subscales) or objective markers but not specific stressors (such as academic workload or others).Articles could only be included if written in English or German which again potentially narrows down the number of studies that could be assessed and might lead to a potential loss of articles.Unfortunately, the heterogeneity of the studies' design and statistical analyses and the low number of studies included prevented meta-analytic investigations.The complexity of methodological approaches also hinders the ability to draw a unite conclusion of the relevance for stress in CBSD.Moreover, this scoping review was not preregistered.To the best of our knowledge, this is the first review article on the association between stress and CBSD and no pre-registered but unpublished reviews on this topic were found at an international prospective register of systematic reviews (PROSPERO).Still, having preregistered this article would have minimized the risk of duplication of effort.Also, it may be argued that pre-registration is linked to reduced publication bias and an increase in transparency.For reasons of openness and transparency, extensive information on the search string, the number of results and the decisions to in-or exclude original articles was described in the methods section of this work and was given in the supplementary material.
This review also has several strengths.To the best of our knowledge, it is the first scoping review on an understudied topic.The careful discussion of the advantages and limitations of included works allows an overview of the current state of research on the interplay of stress and CBSD.Most importantly, the findings help to develop further research ideas and implications for clinical practice.

Future directions
There is a particular need to conduct studies that surpass the level of questionnaire-based evidence as no EMA or experimental studies on the association between stress and CBSD have been carried out since 2012 (see Fig. 3).Future research might consider conducting research with experimental stress induction in order to allow causal inferences, e.g. with the Trier Social Stress Test [97], socially evaluated cold-pressor test [98] or the Maastricht Acute Stress Test [99].Experiments should be accompanied by the assessment of stress biomarkers such as heart rate variability, breath frequency, blood pressure and stress hormones (e.g.,  [46,76,77] Assessment of criteria for specific internetuse disorders (1 study) [67] Online Compulsive Buying Scale (1 study) [42] (Obsession-) Compulsion
salivary cortisol as a marker of the HPA axis system, salivary alpha amylase as a marker for sympathetic nervous system activity, hair cortisol as a marker for chronic stress).Experimental approaches could be used to study stress reactivity in persons with CBSD, in analogy to studies on other behavioral addictions (for an overview: [Schmid et al., submitted]).Other potential topics could be cue reactivity and craving in persons with CBSD after acute stress induction or after placebo condition and transfer from goaldirected to stimulus-response habits and its modulation by acute stress [100].Research on stress induction and cognition in persons with CBSD could also be enriched by the use of neuroimaging techniques as van Timmeren et al. [101] did in their recent study on the balance between goal-directed and habitual decision making after stress induction.Experimental approaches however do only apply to the effects of acute stress.Future research might also investigate perceived chronic stress and perceived stress during the day.
Chronic stress should be taken into account in the development of CBSD.For gambling, it was shown that stressful life events predicted problem gambling in a life course calendar study [102].The same could be valid for CBSD.Chronic stress could be captured using self-report measures such as the TICS [68,69,103] or objective indicators of chronic stress such as hair cortisol (e.g., [104,105]).Conducting research concerning the development of CBSD and its antecedents by means of a prospective design might enormously benefit the design of prevention programs for CBSD that are currently lacking [Thomas et al., in preparation].
Also, perceived stress during the day represents a facet of stress that might be worth closer investigations.Ecological momentary assessment could be accompanied by objective markers such as saliva cortisol or alpha amylase, and cardiovascular outcomes to objectify subjectively reported stress responses during the day (for an overview: [106]).
Besides this, stress proved to be an important factor in risk of relapse or therapy outcome for both alcohol use disorder [107,108] and gambling disorder [109,110].This could also be valid for cortisol levels that might be able to predict relapses in individuals with alcohol use disorder who are currently abstinent from alcohol [111].Examining the role of stress in the development of CBSD and considering its role in relapse/therapy outcome seem to be important areas of future research fostering tailored prevention, therapy and relapse prevention strategies.
As mentioned above, the questionnaire-based studies included in this review reported general perceived stress that either occurred in the last month or last week.Most studies relied on convenience samples and persons with problematic buying-shopping behavior based on questionnaire-cut-off values.Future research should involve clinical samples and determine CBSD status using clinical interviews.Although the association and concrete mechanisms in which stress contributes to CBSD need further clarification, this scoping review shows associations between stress and CBSD.This link could also be integrated into (theoretical) models for CBSD.

Clinical implications
The findings of the studies reported above at least partly suggest an association between stress and CBSD.This conclusion should be further confirmed in future research and should then be incorporated into psychotherapy, for instance in the following ways: First, therapy itself should also, besides typical cognitive behavioral therapy modules (e.g., psychoeducation, motivational interviewing, problem solving strategies, exposure therapy [112]) include aspects of stress management techniques [113], and should support patients in implementing functional coping strategies that also contribute to stress reduction [114].
Second, patients should be informed about a potential association between stress and CBSD by psychoeducation.In psychoeducation, the diathesis vulnerability-stress model (e.g., [36,115]) that emphasizes the impact of stress in development, maintenance and relapse of CBSD could be used.The role of stress could be illustrated by using a barrel as a metaphor for the vulnerability-stress relationship.First, adverse life events and the associated stress can be regarded a vulnerability factor (contributing to the general "fill level" of the barrel in the diathesisstress metaphor).Second, stress might act as a situational factor which might activate dysfunctional coping strategies such as excessive buying for mood regulation (contributing to "barrel overflow").Research on other disorders such as schizophrenia has shown that this is particularly effective when instruction and psychoeducation are extended to significant others so that the environment is also aware of warning signals [116].This might also be beneficial for CBSD.
Third, relapse prevention should include individual "prodromal" symptoms of CBSD episodes, high risk situations and "warning signs" of a relapse [117].Strategies for relapse prevention could also include a periodical reflection of stress levels, e.g., with the help of a stress protocol/diary or technological systems involving electrocardiogram and respiratory parameters to detect high stress phases [118].A study protocol of a promising micro-randomized trial that used machine learningbased stress detection and prompt stress management intervention might, if results are as expected, also offer an opportunity to ameliorate relapse prevention for CBSD [118].A further aim of relapse prevention should be maintenance of social support and functional coping strategies at a high level and further consolidation of these [114].

Conclusion
Evidence from correlative studies suggested a substantial association between general perceived stress and CBSD symptom severity.General perceived stress seems to be elevated in persons with problematic CBSD behavior/CBSD compared to control participants but not compared to persons with other behavioral addictions.Studies that relied on different statistical models proposing stress as antecedent of CBSD symptom severity showed mixed results.Inconclusive results also originated from EMA studies that investigated stress as situational factor contributing to CBSD episodes.Taken together, these findings suggest an association between general perceived stress and CBSD but the direction and relevance of this association remains subject for future research.It may be that absent links from stress to CBSD in cross-sectional models are due to the complex interplay of factors predicting CBSD symptom severity.Drawing on the I-PACE model, stress rather contributes to interactions involving affective, cognitive and executive aspects as vulnerability or situational factor but might not possess standalone predictive power.Further research should focus on experimental induction of stress, ameliorated procedures to capture real-life stress and objective markers of stress.This knowledge should then be incorporated into the prevention and therapy of CBSD.

Fig. 3 .
Fig. 3. Number of studies by year of publication and country in which the study was conducted and grouped by study type (Questionnaire-based vs. EMA/ empirical study).

Table 1
Theoretical models for CBSD and the role of stress.
-Not explicitly mentioned in the model -Stress considered as negative emotional state that can be reduced by excessive shopping: "Shopping has been recognised to ease anxiety and stress" (p.86)-Stress as C − short ➔ maintenance of online CBSD T.A.Thomas et al.

Table 2
Correlative evidence regarding CBSD and stress.
(continued on next page) T.A.Thomas et al.

Table 3
Cross-sectional studies with mean comparisons.

Table 4
Model-based investigations.familial stressors, their association with general perceived stress as measured with five items of the PSS-10 [75] as well as the association of general perceived stress with CBSD symptom severity (measured with a six item-scale based on the Compulsive Buying Scale

Table 4
[40]tinued ) Assessment of Internet and Computer game Addiction, Strukturiertes klinisches Interview zu Internetbezogenen Störungen [Structured clinical interview on Internet-use disorders], CBSD = compulsive buying-shopping disorder, DASS = Depression Anxiety Stress Scale 21, PSS = Perceived Stress Scale.1Bothfigures were given as Singh et al.[40]refer to different figures on p. 261 and p. 263. a Original version of the CBSD questionnaire containing a stress-related item or stress-related item included in the version of the scale reported in this study.Singh et al.[40]reported having measured CBSD with items based on the Compulsive Buying Scale and the Adolescent Compulsive Buying Scale.However, the items reported in the article seemed to be based rather on the Compulsive Buying Scale and did not contain the stress-related item of the Adolescent Compulsive Buying Scale.
T.A.Thomas et al.