Introduction

Eating disorders (EDs), including anorexia nervosa (AN) and bulimia nervosa (BN), are mental disorders that have high mortality and a broad impact on the quality of life of affected patients [1]. Shoplifting is a serious behavioral problem among patients with EDs [2, 3] and is of great forensic concern due to a recent report that among 116 female ED patients in prison, 72% (n = 84) were convicted of repeated shoplifting [4]. Furthermore, a prospective cohort study recently reported that female patients with EDs are at an increased risk of convictions for shoplifting [5]. Although behavioral therapy may be effective in treating shoplifting in ED patients [6], neither established treatment programs for shoplifting nor control trials exist.

Previous studies have shown that shoplifting was more prevalent in patients with bulimic symptoms than those with restrictive symptoms [7,8,9,10]. Some studies reported certain characteristics associated with shoplifting ED patients including high impulsivity [3, 7, 11, 12], depression [9, 13], and low self-esteem [13]. Therefore, some researchers have postulated that shoplifting is strongly associated with bulimic symptoms [7, 9, 10, 13] and serves as a marker of ED severity [2, 3, 7, 9, 10, 14]. Thus, some ED specialists have insisted that ED patients prosecuted for repeated shoplifting should be treated rather than punished as shoplifting could be a progressive symptom of the disorder [15, 16]. Surprisingly, there is no evidence that supports this hypothesis. Whether shoplifting occurs before or after the onset of ED is unconfirmed and little is known about its prevalence, etiology, and consequences. Previous studies of shoplifting among ED patients [2, 3, 9, 10, 12, 13] have had several limitations including small sample sizes, the absence of clinical interviews performed by experts, and lack of multivariable analyses with adequate control of possible confounding factors such as body weight or impulsive behaviors. Based on the body of literature establishing the co-occurrence of shoplifting and ED, the aims of this study were to evaluate the temporal onset of shoplifting in relation to ED as well as to assess the overall prevalence and correlates of shoplifting in ED patients.

Methods

Participants

Participants were recruited from consecutive referrals to the ED clinic at the Department of Neuropsychiatry of Osaka City University Hospital in Japan from July 2012 to February 2014. Inclusion criteria comprised females aged 13 years or older, Japanese speaking, and diagnosed with AN or BN by the Diagnostic and Statistical Manual of Mental Disorders-5 (DSM-5) [17], as well as the ability and willingness to provide informed written consent. Males (n = 5) with a diagnosis of avoidant/restrictive food intake disorder, binge-eating disorder, or other specified feeding or eating disorder (n = 19) were excluded considering their small number. Those who presented with severe physical conditions that could have impaired the interview process (n = 10) were excluded from the sample; however, no significant differences in age, duration of illness, and current body mass index (BMI) were found between excluded patients and the final sample of participants. Those who refused to participate in the study (n = 6) were also excluded. The final sample included 284 female patients who were between 13 and 46 years of age with AN (n = 171; 60.2%), which consisted 99 (34.9%) restricting type and 72 (25.4%) binge-eating/purging type, or BN (n = 113; 39.8%) met the inclusion criteria.

All participants and/or their parents gave written informed consent, and the study was approved by the institutional review board of Osaka City University Graduate School of Medicine.

Procedures and measures

This study used a cross-sectional design. Clinical, psychopathological, and sociodemographic data were collected through a face-to-face interview conducted by two psychiatrists, who specialized in the treatment of EDs. Psychiatric diagnoses were made according to DSM-5. As the DSM-5 was published in 2013, not all of the participants were initially diagnosed with the newly established criteria. However, for 82 patients, we were able to recode the previous DSM-IV diagnoses based on the careful examination of each diagnostic criterion, taking into all relevant information, such as case notes and medical records. Afterward, participants completed a series of self-administered questionnaires to assess the variables described below. Finally, the multidisciplinary team, which comprised three to five psychiatrists, a clinical psychologist, and psychiatric nurses, evaluated the diagnosis and patient behaviors and reached a consensus using the best-estimate procedure from available resources including medical records and information provided by parents.

Shoplifting

Lifetime and current shoplifting were rated categorically (yes/no) using an interview which was developed for the study. Shoplifting was defined as taking an item intentionally from a store without permission or payment. Current shoplifting was defined as shoplifting that had occurred within the past 6 months to compare with the previous studies [3, 13]. The Shoplifting Interview comprised five main questions to assess the following: (1) whether the participant has ever shoplifted; (2) the age at which the participant first shoplifted; (3) the type of items that the participant shoplifted; (4) presence or absence of current shoplifting; and (5) whether the participant has ever been caught by the police (without regard to prosecution). We could not verify the inter-rater and test–retest reliability of individual items because people tend to deny shoplifting due to social stigma [18]. Repeated assessment may make it difficult to encourage honest answers and motivate participants for treatment.

Impulsive behaviors

Lifetime impulsive behaviors were rated categorically using an interview consisting of the following questions: “Have you ever cut yourself on purpose?” “burned yourself?” “hit yourself?” “scratched yourself?” “banged your head?” and “Have you ever tried to commit suicide?” Impulsive behaviors included self-injury (the act of deliberately harming oneself by cutting, burning, hitting, scratching, or banging), attempting suicide, sexual promiscuity (engaging in sexual intercourse with three or more partners within 1 year), alcohol abuse (drinking alcohol four or more days a week or passing out drunk at least 1 day a week), and illicit drug use.

Depression

Depression was rated categorically and defined as having been diagnosed with major depressive disorder according to the DSM-5 criteria [17].

Severity of eating disorder psychopathology

The eating disorder inventory (EDI) is a widely used multidimensional inventory consisting of a 64-item self-report questionnaire measuring psychological and behavioral traits common among ED patients [19]. Previous studies reported the reliability and validity of the EDI Japanese version [20, 21].

Self-esteem

To evaluate self-esteem, a scale of 0–40 was used. Participants completed all ten items of the Rosenberg Self-Esteem Scale (RSES), which had a good internal consistency [22]. The psychometric properties of the RSES Japanese version were satisfactorily documented [23]. A lower score indicated more pathology (lower self-esteem).

Statistical analyses

The Chi square test was used for categorical comparisons of the data. For continuous variables, the Mann–Whitney U test was used to examine intergroup differences due to the non-normal distribution of data. We examined variables between participants with AN and those with BN, in addition to variables between participants with a history of shoplifting and those without a history. Univariate and multivariate logistic regression analyses were performed to reveal potential risk factors (independent variables) associated with a history of shoplifting (a dependent variable). As independent variables, we included age, current BMI, duration of illness, ED subtype (AN vs. BN), sociodemographic factors [education, job status (full-time vs. part-time)], and impulsive behaviors (self-injury, suicide attempt, sexual promiscuity, alcohol, and illicit drug use). We added depression, RSES scores, and the severity of ED symptoms (total EDI scores) as independent variables, which were suspected to be associated with shoplifting in the previous reports [9, 13]. The Spearman’s rank correlation coefficients between independent variables were satisfactory (less than 0.7) regarding the multicollinearity.

Descriptive results were presented as means and standard deviations, medians, and ranges, or frequencies and percentages. The results of regression models were reported as crude odds ratios (OR) and adjusted odds ratios (AOR) with 95% confidence intervals (Cl). Statistical significance was defined as p < 0.05. Adjustment for multiple testing was not performed [24]. All statistical analyses were performed using SPSS ver. 20.0 (IBM, Tokyo, Japan).

Results

Demographics

Of the 284 participants, the mean age at evaluation was 24.6 ± 7.0 years, duration of illness was 5.3 ± 5.6 years, mean age at ED onset was 19.3 ± 4.9 years, mean current BMI was 16.6 ± 4.1 kg/m2, mean past highest BMI was 22.3 ± 3.8 kg/m2, mean past lowest BMI was 14.8 ± 3.0 kg/m2, mean school education was 12.9 ± 2.4 years, 22% (n = 61) worked full-time, and 85% (n = 241) were single. All participants were Japanese females. There were no significant differences in the main demographics (age, duration of illness, BMI, education, and job status) between participants included in our study and those excluded.

Demographic characteristics of AN and BN participants are shown in Table 1. Duration of illness, current BMI, past highest BMI, past lowest BMI, and school education were lower in participants with AN than in those with BN (Table 1).

Table 1 Sociodemographic characteristics of the participants
Full size table

Shoplifting

The lifetime prevalence of shoplifting in participants was 28.5% (n = 81; 95% Cl, 22.3–33.8%), which was 12.1% in AN restricting type, 26.3% in AN binge-eating/purging type, and 44.2% in BN. The mean age at the initial shoplifting incident was 13.7 ± 7.0 years. Among 81 participants with a history of shoplifting, 70.4% (n = 57) initially shoplifted (at least 6 months) prior to the onset of ED, 13.6% (n = 11) shoplifted at the same age of (less than 6 months before) the onset of ED, and the remaining 16% (n = 13) shoplifted following the onset of ED. The main stolen items were classified into four categories: food, cosmetic items, clothes, and others. 38% (n = 31) of participants mainly shoplifted food. The remaining 62% (n = 50) mainly shoplifted other items (cosmetic items in 17%, clothes in 10%, and others in 35%). Among 81 participants with a history of shoplifting, 53% (n = 43) shoplifted within the past 6 months, and 17% (n = 14) were caught by the police irrespective of prosecution. Results indicated that the prevalence of lifetime and current shoplifting were significantly lower in patients with AN than in those with BN (18.1% vs. 44.2%, 9.4% vs. 23.9%, respectively; Table 2).

Table 2 Shoplifting and clinical features
Full size table

Impulsive behaviors and clinical features

High prevalence of impulsive behaviors and depression were identified in all participants with self-injury, attempted suicide, sexual promiscuity, alcohol abuse, illicit drug use, and depression occurring in 24% (n = 68), 17% (n = 47), 8% (n = 23), 4% (n = 11), and 57% (n = 161), respectively. Impulsive behaviors and clinical features in patients with AN and BN are shown in Table 2. Results showed that the prevalence of impulsive behaviors and depression were significantly lower in participants with AN compared to those with BN. RSES scores were significantly higher in participants with AN than in those with BN indicating lower self-esteem in BN patients. Regarding the severity of participants’ ED symptoms, all EDI scores were significantly lower in participants with AN than in those with BN except for maturity fears.

Risk factors associated with shoplifting

The results of the logistic regression (Table 3) revealed that in the unadjusted model, current BMI, duration of illness, BN, impulsive behaviors, depression, RSES, and EDI were associated with a history of shoplifting. In the adjusted model, in which all covariates were entered in the logistic regression, alcohol abuse, illicit drug use, depression, and RSES were associated with a history of shoplifting. Participants with alcohol abuse, illicit drug use, and depression were, respectively, 3.91 (95% Cl, 1.34–11.38), 14.42 (95% Cl, 1.65–125.86), and 2.63 (95% Cl, 1.24–5.60) times more likely to have shoplifted compared to those without each.

Table 3 Logistic regressions examining depression, impulsive behaviors, EDI, and RSES scores with and without lifetime shoplifting (n = 81 vs. 203)
Full size table

Discussion

This study demonstrated a high prevalence of shoplifting in ED patients. Shoplifting preceded the onset of EDs in most patients with a history of shoplifting. Contrary to previous studies, shoplifting was associated with impulsive behaviors such as alcohol abuse and illicit drug use in addition to depression and low self-esteem after controlling for possible confounding factors. These findings did not support the previous assumptions that shoplifting was strongly associated with bulimic symptoms [7, 9, 10] or was a marker of ED severity [2, 3, 7, 9, 10, 14]. However, the causal relationship between shoplifting and EDs remains inconclusive because of the cross-sectional design of this study.

Shoplifting is common in patients with EDs and probably precedes the onset of EDs in most patients with a history of shoplifting. Previous studies reported that the lifetime prevalence of shoplifting in ED patients including both AN and BN varied between 28 and 79% [3, 9, 10, 12, 13, 25]. The study reporting the highest prevalence of 79% (27/34) [12] included a small number of participants without restricting symptoms. A major limitation of the study was the lack of a representative sample, which may have inflated the results. We demonstrated in a relatively large and well-defined sample that lifetime prevalence of shoplifting in patients with EDs was approximately 30%, consistent with most studies.

Furthermore, consistent with previous studies [7,8,9,10], our results indicated that shoplifting was more prevalent in patients with BN compared to those with AN. However, we found that in shoplifting that preceded the onset of EDs, there was a prevalence of over 70%, which contradicts three previous studies [3, 10, 12], which reported much lower prevalence of 43% (12/28) [3], 46% (33/71) [10], and 33% (9/27) [12]. This may be explained by the small sample size of these studies that included only patients 19 years and older. These three studies also applied inconsistent definitions for shoplifting, using only a questionnaire; one study [12] included only participants with bulimic symptoms, while another [3] included patients with a higher average age and average duration of illness than those of our study (25 and 5 years, respectively), although the prevalence of shoplifting was consistent with our study. Since ED commonly begins during adolescence or young adulthood and rarely develops before puberty or after the age of 40 [26,27,28], our findings can be justified based on the wide age range (13–46 years old) of participants included.

Moreover, to avoid overestimating the preceding prevalence, we defined shoplifting as “before the onset of ED” when shoplifting occurred at least 6 months prior to the onset of an ED, because of the chance that underlying psychopathology, which can later cause an ED, is initially present with shoplifting. When shoplifting occurs just before an ED manifests, the attribution of shoplifting as a preceding or prodromal symptom of EDs becomes controversial as it is possible that an underlying latent factor, or psychopathology, may be causing both. For this reason, we used a strict definition of shoplifting as “before the onset of ED.” Therefore, we believe our findings to be reliable despite the high prevalence of approximately 70%. To further verify our findings, a longitudinal study should be conducted to confirm how often ED patients without shoplifting at the initial assessment develop shoplifting behavior during a follow-up.

Our results showed that ED patients who shoplifted tended to have severe impulsive behaviors. This study elucidated the association between shoplifting and individual impulsive behaviors such as alcohol abuse and illicit drug use, by controlling for confounding factors. We also showed that shoplifting was associated with depression and low self-esteem, similar to previous reports. In contrast to previous studies, shoplifting was found to be less susceptible to bulimic symptoms and symptoms associated with severe ED after controlling possible confounding factors including age, BMI, duration of illness, and sociodemographic factors. In other words, the logistic regression suggested that impulsive behaviors (such as alcohol use and illicit drug use), low self-esteem, and depression were associated with shoplifting, whereas eating disorder severity was not. Indeed, lack of agreement on what characteristics were associated with shoplifting in patients with EDs [7] and insufficient evidence that EDs can progress to shoplifting [4] have been significant problems. Several studies have reported that a history of shoplifting was associated with high impulsivity [3, 7, 11, 12], although impulsivity is not identical to impulsive behaviors. Other studies suggested that shoplifting was associated with depression [9, 13], low self-esteem [13], or low socioeconomic status [3].

Regarding impulsive behaviors, our results were inconsistent with those of another study [2], although this was arguably likely because their study sample only consisted of prisoners. Inconsistencies were found in the results of depression [10, 13] and socioeconomic status [10, 11]. Possible explanations include differences in the average age of participants and in the instruments used for their evaluation, in addition to the small sample size and lack of conducting multivariable analyses with controlled confounding factors. In this study, univariate logistic regression analysis showed that BN, duration of illness, and the total EDI scores were significantly associated with a history of shoplifting, whereas multivariate logistic regression analysis failed to show an association. This suggested that the previously reported association with shoplifting may have been partially confounded by impulsive behaviors, depression, and low self-esteem, which were not examined.

In conclusion, although our study did not support the prospect of EDs progressing to shoplifting, it found that ED patients who shoplifted may have severe underlying impulsivity, irrespective of ED severity. Generally, shoplifting may serve as a marker of a broader impulse control difficulty. We believe that shoplifting frequently occurs earlier than other impulsive behaviors in patients with EDs. Attention should be paid to its unrecognized role as a marker to detect the risk of other impulsive behaviors, including alcohol abuse and illicit drug use.

This study had several limitations common to cross-sectional studies. First, we did not examine the reliability of individual items of shoplifting. Individuals tend to deny impulsive behaviors, especially shoplifting [18]. Repeated assessment for shoplifting possibly made it difficult to obtain honest answers due to social stigma. Recall bias may also have underestimated past shoplifting. Whether a questionnaire or an interview optimally encourages honest answers remains controversial [3, 10], and assessment tools for shoplifting have not been developed. However, since one study suggested that clinical settings could lead patients answering honestly [10], we recruited treatment-seeking patients and used an interview conducted by psychiatrists to maximize sensitivity and accuracy. In addition, we performed the interview during the first visit to reduce the possible bias caused by the therapeutic relationship between the interviewers and the participants. Second, the assessment of shoplifting was restricted to five questions and did not allow for detailed information such as the frequency and severity of shoplifting including the existence of current urges to shoplift. Shoplifting may have occurred for a short period of time or only once. A cross-sectional evaluation meant that it was not possible to differentiate between the state and trait levels of impulsive behaviors including shoplifting. Even with this broad definition, our study suggested that shoplifting was associated with impulsive behaviors such as alcohol abuse and illicit drug use in addition to depression and low self-esteem. Future studies are needed to investigate the similarities and differences between shoplifting in patients with EDs and kleptomania. Third, our sample only included AN or BN patients with a wide-age range that included women age 13 years or older who were referred to a clinic that specialized in EDs, leaving open the possibility that these results may be partially due to the immature impulsivity of adolescent patients. However, to reduce this potential confound, we controlled for participants’ age by entering as an independent variable the multivariate logistic regression. Additionally, information on convicted patients, who may have had severe impulsive behaviors including shoplifting, was unavailable. Our study did not include patients with other types of EDs such as avoidant/restrictive food intake disorder or binge-eating disorder [17] and non-referred individuals. Our interest was in ED patients with AN and BN, which impacts the health and quality of life of those affected [1], and who were referred to the clinic for treatment. Fourth, our study did not have a reference group without EDs or consider the diagnostic fluctuation between AN and BN, which may lead to an underestimated association between shoplifting and BN. Although the sample size was larger than those of previous studies, a multivariate model, including many covariate comparisons, may still limit the statistical power to detect significant associations. Fifth, this study did not assess comorbid mental disorders other than depression that may have been associated with shoplifting and impulsive behaviors. The subjective assessment using EDI for the severity of ED may cause bias. Finally, we were unable to use the EDI-2 or EDI-3 with our participants. The Japanese version of EDI-2 and ED-3 were not available because the reliability and validity of these versions have not been established. Despite the limitations, this study suggested a high prevalence of shoplifting preceding the onset of EDs among ED patients and an association with shoplifting, alcohol abuse, and illicit drug use.

Our findings indicate that clinicians should pay attention to shoplifting in patients with EDs because shoplifting often has consequences that lead to complex legal issues. Future studies are warranted for the development of effective assessment tools for shoplifting to elucidate the longitudinal course of shoplifting, delineate the potential boundaries from other impulsive behaviors, and establish a common treatment strategy for impulsive behaviors in patients with EDs.