| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLE |
From the Eating Disorders Unit, Department of Psychiatry, "La Fe" University Hospital, School of Medicine, University of Valencia, Spain (L.R., L.L.); Psychiatric Service, Hospital of Sagunto, Spain (L.C.); Vanderbilt University School of Medicine, Nashville, Tennessee (O.B.).
Address correspondence and reprint requests to Luis Rojo, MD, UTCA, Planta 11, Hospital Maternal, H.U. La Fe, Avda de Campanar 21, 46009, Valencia, Spain. E-mail: luis.rojo{at}uv.es
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Methods: The Life Events and Difficulties Schedule interview was applied to a sample with 32 cases and 32 controls selected from a two-phase epidemiologic study among a representative population of adolescents, followed by a decay model to assess acute and chronic stress in the year preceding the onset of ED. Psychiatric comorbidity was assessed using the SCAN interview.
Results: Cases (46.9%) and 9.4% of controls were found to have associated psychiatric comorbidity (
2 = 11.74, p = .001). Of cases, 6.3% and none of the controls had at least one severe stressful event (N.S). Of cases, 18.8% and 3.1% of controls had at least one major difficulty (Fisher exact test = 0.05). Of cases, 25% and 3.1% of controls had a provoking stressful agent (Fisher exact test = 0.026). Psychiatric comorbidity partially mediated the relationship between stress and EDs. The Structural Equation Modeling Analysis shows that chronic stress is strongly associated with the onset of EDs, both directly (r2 = 0.38) and indirectly, through psychiatric comorbidity (r2 = 0.56).
Conclusion: Chronic stress and psychiatric comorbidity are strongly associated with the onset of EDs. Psychiatric comorbidity is a partial mediating factor in the association of stress with eating disorders.
Key Words: eating disorders • anorexia nervosa • stress • life events • chronic stress • comorbidity
Abbreviations: ED = eating disorders; AN = anorexia nervosa; LEDS = Life Events and Difficulties Schedule; BN = bulimia nervosa; BED = binge eating disorder; SCAN = Schedules for Clinical Assessment in Neuropsychiatry; EAT = Eating Attitudes Test; SEM = structural equation modeling; DSM-IV = Diagnostic and Statistical Manual of Mental Disorders; SPSS = Statistical Package for Social Science; AMOS = Analysis of Moment Structures.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Research on stress and eating pathologies has been mostly retrospective and on clinical populations. Strober (3) used a semistructured interview in a clinical sample to gather information on the "life events" experienced during the 18 months preceding the onset of the ED. Subjects with purging types of anorexia nervosa (AN) had significantly more events than those with restricting types. Horesh et al. (6,7) did two case-control studies based on adolescents hospitalized for AN. The control groups were adolescents admitted for other psychiatric disorders and healthy adolescents. Results supported the identification of family stress as a factor related to EDs.
Schmidt and colleagues (5) used a different methodologic approach to evaluate stress with Brown and Harris's (8) Life Events and Difficulties Schedule (LEDS). They selected a sample of patients with AN and bulimia nervosa (BN), using a Camberwell community cohort from Brown and Harris as the control group. There were no differences between the sample and controls as to having had at least one severe stressful event during the previous year. There was a significantly higher incidence of major difficulties (severe chronic stress) in the sample. The most common stressful events were those related to family and friends. Limitations of this study were the use of a clinical sample and a control group that was not chosen from the same community as the subjects.
Gowers et al. (9) used a semistructured interview to assess stress in the year before onset of ED in a sample of adolescents with AN, compared with a control group with other psychopathologies and another of healthy controls from a primary care registry. No significant differences were detected, although there was a trend toward significance when extremely negative events were considered.
There are two case-control studies with nonclinical, community-based samples. Rastam and Gillberg (4) evaluated the role of stress in a sample partly made up of a community-based population and detected an accumulation of stressful events during the 3 months before the onset of AN but provided no further details regarding the instruments used or the stress characteristics. Welch et al. (10) analyzed the relationship between stress and BN. Among the sample group, there was a higher life events rate, especially involving disruption of family or social relationships or threat to physical safety, when compared with same age controls.
Patton et al. (1) come closest to having optimal methodologic parameters by using a two-phase, case-controlled, general population design. Its results showed no significant relationship between acute or chronic previous stress and the posterior development of ED. Among its limitations was that they only evaluated stress in certain social fields and that it lacked an evaluation of contextual threat of the stressful events.
Finally, Rosen et al. (11) did a prospective study in which, for a 4-month period, they followed the interaction among stress and psychological and eating symptoms via questionnaires. They found a reciprocal association between stress and food-related symptoms in that recent stress was associated with changes in eating behaviors, but also eating symptoms predicted the occurrence of psychological stress. Limitations of this study were the use of a nonrepresentative sample, self-administered questionnaires, and the lack of a clinical assessment of the subjects.
There is also evidence that there is an association between development of eating-related pathology and preexisting psychopathology, especially depressive disorders. Patton et al. (12) followed a cohort of 2032 adolescents for 3 years and found that psychiatric morbidity was the most important predictor for the development of EDs. Johnson et al. (13) found an association between depressive disorders and the onset of EDs in a longitudinal study from a random community sample of adolescents. Given that the association of stress and psychopathology, specifically between stress and depression, has been substantiated (14,15), it is crucial to control for the presence of psychopathology when investigating the association between stress and EDs.
Ample evidence supports the dimensional perspective of EDs. Full and partial syndromes share sociocultural, familial, psychometric, behavioral, medical, and hereditary variables. Dancyger and Garfinkel (16) found that full syndromes showed higher levels of depression, lack of interoceptive awareness, and ineffectiveness in the Eating Disorder Inventory than partial syndromes, although the latter also showed higher scores than the control group without EDs. Patton and King (17) suggest that, in view of the psychopathology they share with AN, partial syndromes should be considered as part of the same spectrum. Other shared aspects of full and partial syndromes are comorbidity with other psychiatric disorders, high rates of previous psychiatric treatment (18), and psychosocial characteristics (19). Strober et al. (20) studied the hereditary characteristics of partial syndromes and found that both full and partial forms are familial, although there appears to be greater familial predisposition in full syndromes, which supports the ED vulnerability continuum hypothesis. These authors consider that there is a significant heritable component to the etiology of broad forms of AN and BN and that there may be a continuum of liability to developing an ED. Authors who support the dimensional approach to EDs agree that incomplete syndromes are less severe forms of these disorders and that they should not be omitted in epidemiologic studies.
Research on the role of stress in EDs has yielded varied results, given the different methodologies used. The LEDS (8) includes a series of methodologic advantages that reduce certain biases and methodologic limitations noted in previous studies. First, it allows the subject's main areas of life to be assessed (home, family, friends, work, partner relationships, and health). Second, the classification of the individual's answers (i.e., the threat of the life events) is not dependent on the interviewer's subjective assessment but on a group of experts who are unaware of the subject's opinion and base themselves on the interviewer's report. Third, since the period covered is 1 year, loss of information is as low as 2.9% annually for severe threat events (14). Forth, great effort is taken to stimulate event recall with accurate dating. Fifth, it includes an independence scale that allows selection of stressful events and difficulties that are independent from the studied disorder. The LEDS offers the best assurance that the majority of the stressful events from the period studied are included and that the assessment is objective and valid (8).
A novelty in Brown and Harris's (14) work was their way of measuring life events whose impact on the individual is reflected through the contextual measurement of threat. The LEDS taught them that only certain relatively infrequent events that scored as "severe" had an etiologic role and that prolonged major difficulties had a causal role, although of lesser importance. To study the impact of events or difficulties, the concept of "provoking agent" was introduced. It was considered to operate when the patient experienced at least one "severe event" or one "major difficulty" during the 12 months assessed. To avoid confusion with severe events caused by psychiatric disorders, only those events that were outside the patient's control ("independent") or appeared unrelated to the psychiatric disorder ("possibly independent") were considered (8). These findings on the etiologic role of these specific types of stress are based on research with depression, and it is unclear how applicable they are to other disorders.
This method described by Brown and Harris (14) uses a system of categories to classify stressful agents, the "category model." Surtees (21) then proposed an alternative conceptualization of life-stress scores called the "decay model," which contains two hypotheses: a) stressful effects of events decay over time at a constant rate, which is the same for all events; b) stressful effects of independent events are cumulative. Thus, the residual adversity that a person suffers at a given moment can be determined by combining the effect of the last event before that moment, with the residual adversity of all the previous events.
Based themselves on the LEDS method, which assigns special importance to long-term threats, the severity of events (severe, moderate, some, none), and the persons affected (focus on the subject or on others), were attributed decreasing weights to the eight different possible combinations of these two variables. These weights then reveal the extinction period of events (21). Using this model in depressive women, the authors found a relationship between residual adversity and the subjects' subsequent evolution. They drew special attention to the fact that low-threat life events may also contribute to the statistical association between residual adversity and symptom severity. Surtees (21) claimed that clinical experience has shown that the best period for calculating extinction is 26 weeks, although on occasion a longer (52 weeks) extinction period is used (22).
The current study was designed to assess the relationships between stress and EDs while overcoming the majority of the prevailing methodological limitations. Our cases and controls were selected from a representative community population, and we used a semistructured interview to evaluate for stress that incorporates contextual measures and controls for stressful events due to the subject's behaviors, controlling for psychiatric comorbidity through this semistructured interview technique. Our study subscribes to the dimensional perspective of EDs by including both partial syndromes and subclinical forms of EDs in the sample.
Research Questions
The first question was to assess the incidence of life stress events and major difficulties in the year before the onset of an ED in a community simple, compared with a control group matched in age, sex, and schooling.
The second was to evaluate the role of psychiatric comorbidity in the association between stress as a precipitating factor and eating pathology.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
First Stage
The sample population was screened using the Spanish version of Garner and Garfinkel's EAT-40 (Eating Attitudes Test) to detect the at-risk population (24,25).
We selected the sample through conglomerates extracted from a probability design proportional to size. A conglomerate sample provides greater precision than a simple sample when conglomerates of low homogeneity (with respect to the variable studied) are obtained. Simple random samples were taken within each segment, and they were then united to obtain the total desired sample with the classroom as the sampling unit.
During the first trimester of 1999, we surveyed the total sample population of 544 youths, 273 males, selected in three stages: a) random sampling of townships in each demographic-geographic area, b) random sampling of the teaching centers in each selected township, c) stratification by age and sex of each teaching center and random classroom selection.
Second Stage
In interviews with the case group (with scores of 
30 on the EAT) and the control group (with scores of <30 on the EAT and randomly selected using individual pairing stratified by age and sex), 57 out of 60 "possible cases" (scoring higher than the EAT cutoff point), 10.47% of the sample, were interviewed, 44 of them female (77.2%) and 13 males (22.8%). Experimental mortality was 5% (two people refused the interview, and one gave false answers to the questionnaires). A control group of 57 subjects was chosen by random sampling among those individuals with an EAT score below 30, using individual stratified pairing by age and sex.
Procedures
Initial Screening for EDs Using the EAT
The EAT (24) is a self-report questionnaire that measures cognitions, emotions, and behaviors associated with AN and BN. It has been the most widely used instrument for the evaluation of EDs in adults and adolescents in diverse cultures and populations (26). A version of the EAT validated for our cultural context with a cutoff point of 30 was used (25). Participants responded to each item on a 6-point scale scored as 0 (never, rarely, sometimes), 1 (often), 2 (usually), and 3 (always). Crombach's 
for this scale in the study sample was 0.857.
Assessment for EDs
The interviewers were psychiatrists well experienced with EDs. The clinical interviews were conducted in a semistructured fashion using a list of questions that covered the totality of the DSM-IV diagnostic criteria for AN, BN, and binge eating disorder (BED) (27). This allowed us to recognize full syndromes (meeting all AN (307.1) or BN (307.51) criteria), partial syndromes (meeting AN criteria B and C but not A or D; or meeting BN criteria A1, B, D and E but not A2 or C), and subclinical forms (meeting AN criteria B and C but neither A nor D; or meeting BN criteria A1, B, D, and E but neither A2 nor C) (28).
Detection of Comorbid Psychiatric Pathologies
As an additional diagnostic measure of psychiatric pathologies, we administered the Schedules for Clinical Assessment in Neuropsychiatry interview (SCAN) (29). The SCAN is a set of instruments designed for analysis, measurement, and classification of psychopathology and behaviors associated with main adult psychiatric disorders. Our study sample first received the general module and the Depression and Anxiety modules when necessary. The SCAN was designed to identify diagnosis using ICD-10 and DSM-IIIR criteria. We reassigned the diagnosis to the corresponding categories of the DSM-IV (27). The psychiatrists administering the SCAN were specifically trained.
Assessing Stress
The LEDS of Brown and Harris (8) was applied, the period covered being the 12 months preceding onset of EDs in case subjects. In control subjects, the presence of life events was assessed in the year preceding the appearance of their matched ED case.
The cross-cultural validity of the LEDS has been evidenced (14). In Spain, Gaminde et al. (30) applied the Spanish version of the LEDS to nonclinical representative samples in three different communities and were able to duplicate the findings obtained in Anglo-Saxon populations in relation to depressive disorders. Because it was also applied to depressive patients and matched controls in our clinical setting before this study, our team was experienced using it (15). The LEDS interview was administered in all cases by one of the coauthors, who was blinded to the SCAN results obtained by the participants.
Measurement of LEDS
Following Brown and Harris's (8) procedures, the interviewer presented the results of the LEDS to at least two other team members to reach consensus. The distress caused by the different events in the majority of people sharing similar biographical circumstances was taken as the basis. Researchers were blind to the subjective reaction of the test subject and their psychiatric status. The same method was used to score the "contextual threat." Severe events were defined as those with an important, long-lasting threat (with the highest score on a threat or distress scale of 4 points or with a moderate score if they were centered on the individual). Moderate and light events were not included. "Major difficulties" were defined as those chronic adverse events capable of causing distress in the individual (with a severity of 1 to 4 on a 6-point scale), lasted at least 2 years, and were not health related.
Extinction or Decay Model
To determine the optimal extinction times, we calculated variable extinction times, from the maximum permitted by our study (52 weeks) to 1 week and then calculated the respective residual adversity for patients and controls. The optimum extinction time selected was the one with maximum statistical differences between the groups by t tests (15). There is no empiric rationale to select 52 weeks, but it seemed to be the consensus period in prior studies. The data appear to support the notion that life events that occurred in the 20 weeks before the onset of an ED have greater importance as a provoking agent. Even so, assessing in a longer period is reassuring in that we did not fail to evaluate prior events.
Data Processing and Statistical Analysis
The study data were entered in an MS Access database and then exported to an SPSS file for statistical analysis. The LEDS data were directly entered into the SPSS database. The threat extinction calculations were done with the algorithms proposed by Surtees (21), which we used in a program of our own design.
Binary logistic regression analysis was used to study the relationship between stress, comorbidity, and EDs (31). We also performed a structural equation modeling (SEM) analysis to test the best fit model. The best-fitting model was chosen using the Specification Search option of AMOS (v.5.0). The SEM as a model of multiple regression analysis allowed us to compare competitive models to discern if the variables had a direct effect on the onset of EDs or if they were mediated by other variables. The number of final subjects in our study was not conducive to a rigorous SEM analysis, but the model used illustrates well the results of the logistic regression analysis.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Comorbidity With Other Psychiatric Diagnoses
We found 18 subjects (28.1%) of the total interviewed to have an additional psychiatric disorder by DSM-IV criteria. Among the cases, 15 (46.9%) had an additional diagnosis of comorbid psychiatric disorder: 12 depressive disorders (3 major depressive disorders, 4 adjustment disorders with depressed mood, 5 depressive disorders not otherwise specified) and 3 anxiety disorders (1 generalized anxiety disorder and 2 adjustment disorders with anxious mood). Among the controls, three (9.4%) were found to have comorbid psychiatric diagnosis
There was one adjustment disorder with depressed mood and 2 anxiety disorders (1 panic disorder and 1 adjustment disorder with anxious mood); the presence of comorbid psychiatric psychopathology was significantly greater among the cases (2 = 11.74, df = 1, p = .001).
Stress: Comparison of Case and Control Subjects
Life Events
The average number of events was 3.81 ± 2.66 among cases and 3.97 ± 2.6 among controls. Of the cases, 6.3% (2/32) had at least one severe event compared with 0% (0/32) of noncases. Severe events occurred exclusively among females. Cases with severe events averaged 15.3 years of age. We did the same analysis on the 12 weeks before onset, never finding statistical differences between cases and controls.
Difficulties
Case subjects showed a significantly higher average number of difficulties (2.88 ± 1.99) than controls (1.26 ± 1.05) (t = 4.31; df = 62, p < .001). We detected 90 difficulties among cases, mostly in "relations with parents" (15.9%) and "social relations" (13.63%). Controls showed 36 difficulties, predominantly school related (20%). Of the total number of cases, 40.6% had at least one difficulty in "relations with parents" (versus 18.75% in controls), 28.1% in "relations between parents" (3.1% in controls), and 28.1% in "social relations" (15.6% in controls).
The incidence of major nonhealth difficulties (Table 1) among cases was 18.8% (6/32), compared with 3.1% (1/32) in controls. These differences tended to significance (Fisher exact test unilateral = 0.05).
|
Provoking Agents
Twenty-five percent (8/32) of case subjects had at least one provoking agent, compared with 3.1% (1/32) of controls (Table 1). These differences were significant (Fisher exact test = 0.026). The risk estimate that an individual exposed to a provoking agent developed an ED was 10.3 (95% CI, 1.2–88.4).
Decay Model
Figure 1 shows the threat evolution caused by acute stressful events over the year preceding ED onset. Case group tends to show higher accumulated or residual stress than the controls during weeks 10 to 18 before onset. Case group also shows accumulation of acute stress in the 3 weeks before onset. This may be due to both the impact of new events in that time period and the residual threat of older events. The differences with the control group only reached significance in week 1. Figure 2 shows the combined evolution of events and difficulties during the period studied. There are significant differences between case and control subjects in all 52 weeks before ED onset.
|
|
Analysis of the Relationship Between Stress, ED, and Other Comorbid Psychiatric Disorders
As can be seen in Figure 3, 62.5% (5/8) of case subjects with prior stressful provoking agent had an accompanying psychiatric disorder. Of the case subjects, only 9.4% (3/32) without psychiatric comorbidity had been exposed to a provoking agent; 31.3% (10/32) had psychiatric comorbidity in the absence of stressful provoking agent. Another interesting finding was that 43.8% (14/32) of the cases did not report a stressful provoking agent or were found to have psychiatric comorbidity. Therefore, the relationship between stress and EDs does not appear to be totally independent of psychiatric comorbidity.
|
We used binary logistic regression analysis to study the relationships and interactions between these variables. The dependent variable was the presence of ED, the other two variables being the predictor variables. The regressions with stress-ED and comorbidity-ED were both significant (Table 2). When we entered both factors into the equation simultaneously, thereby assessing their covariation, the relationship stress-ED lost its significance, whereas comorbidity and ED remained significant. No changes occurred in any of the relationships analyzed when controlling for the variables age or gender.
|
Figure 4 shows the best model explaining the relationships between stress, comorbidity and EDs using SEM analysis. The numbers next to the arrows represent regression weight, whereas those above the variables show the squared multiple correlation and can be interpreted as the variance proportion of endogenous variables explained by the model. SEM 2 should be interpreted as measure of "badness of fit," which proves that the model fits especially well. With a 2 = 0.005 and a p = .945, the model cannot be rejected.
|
Chronic stress is strongly associated with presence of EDs, both directly (r2 = 0.38), and indirectly through psychiatric comorbidity (r2 = 0.56). Acute stress is only moderately and directly associated with EDs (r2 = 0.14) and almost completely unassociated with comorbidity.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Our results tend to support the hypothesis that stress is a precipitating factor for EDs, although psychiatric comorbidity does mediate this effect to a large degree. Using diagnostic criteria, including mild clinical forms (partial and subclinical), the number of cases suffering the impact of a stressful provoking agent before ED onset is significantly higher than in the control group (Table 1). Exposure to at least one stressful provoking agent increases by 10.3-fold the risk of developing ED, including subclinical forms. Our results specifically highlight the importance of severe chronic stress (major difficulties) in the development of EDs. The case group showed significantly higher numbers of chronic stressful events preceding ED onset than the control group. Finally, during the weeks immediately before ED onset, the case group showed an increase of acute stress (life events) as compared with the control group (Figure 1).
The relationship between stress and EDs appears to be fundamentally mediated by psychiatric comorbidity. More than 60% of case subjects with a stressful provoking agent had an accompanying psychiatric disorder. Moreover, the relationship between stress and EDs loses its significance when the variable of comorbidity is introduced into the regression equation. The model of structural equations points to the direct effect of stress on EDs is less than the indirect effect, mediated by psychiatric comorbidity. Psychiatric comorbidity seems to be a major mediator in the association of stress to ED. Psychiatric comorbidity can thus be said to behave as a "vulnerability factor," i.e., a variable that helps to predict the response to a stressful agent or difficulty (8). An alternative interpretation would be that stress induces psychiatric comorbidity. This, when combined with other predisposing factors, facilitates the emergence of an ED. Our study did not exclude this possibility, but it does highlight the need for more multivariate studies that explore the complex etiology of EDs while controlling for the interaction of other risk factors.
The previous studies that specifically analyzed the relationship between stress and EDs yielded highly disparate results for many possible reasons; for instance, biased recall of events, the use or nonuse of standardized procedures for determining the existence and severity of stressful events, or the origin of the samples used. When dealing with clinical samples, the relationship between stress and ED may be spurious and due to the subject seeking treatment. Our community sample eliminates this bias. The methodology we used for stress assessment (LEDS) also eliminates other sources of bias: it facilitates recall of stressful events; these are defined by the interviewer, not the interviewee; there is a concise definition of the different stressful events and difficulties; special attention is paid to accurately dating stressful events and to ensuring they are the result of the disorder and not its cause (using the independence scale).
Some prospective studies have analyzed the relationship between stress and EDs, with inconclusive results (11,33). That of Patton et al. (1) probably comes closest to having optimal methodologic parameters, using a two-phase, case-control, and general-population design. Its results, however, showed no significant relationship between acute or chronic prior stress and the subsequent development of an ED. This divergence may be due to the two studies' different objectives and designs. Whereas Patton et al. (1) fundamentally studied risk factors and limited their investigation to stress in certain social fields, our study examined whether stress precipitates EDs. We used a transversal study design and applied a stress-measuring instrument covering many areas in addition to social stress. Furthermore, in our study, whether an event or chronic difficulty was considered a stressful agent does not depend exclusively on its presence but on an evaluation of its contextual threat.
Few studies are based on representative community samples. Rastam and Gillberg (4) studied a mixed population of community and clinical cases and a control group matched by age, sex, and school. They analyzed "major life events," dividing them into recent (occurring within 3 months before disorder onset) and chronic, showing no differences in the number of chronic life events, although there was a higher incidence of recent stressful events. Their results are also difficult to compare with ours because half of their sample was clinical cases and they used a different definition of chronic stress. In our methodology, difficulties are adversities that remain active over a prolonged period. To Rastam and Gillberg (4), stress was apparently chronic if it took place more than 3 months in the past, regardless of whether it was still active at the time of onset. They did not specify the method used for measuring stress.
In general, these studies show a positive relationship between stress and the appearance of EDs. None of them, however, offers any contextual measurements of stress, i.e., measurements based on the impact that events may have on an individual's personal or biographical conditions. One exception is the study by Schmidt et al. (5), who also used the LEDS. They selected a clinical sample and used a Camberwell community cohort from Brown and Harris (8) as control group. They found no significant differences in the incidence of major stressful events between subjects and controls during the year before ED onset. They did find significant differences between cases' and controls' chronic stress, with the case group showing a significant buildup of major difficulties. These results are consistent with those of our study, which, in contrast, used a community case-control sample, thus allowing generalization of the results.
Schmidt et al. (5) observed that most common chronic stressful situations preceding ED onset were about relationships with parents and friends. Horesh et al. (6,7) also found these relationships to be the most frequent problem in severe AN cases. Among our cases, 40.6% had at least one difficulty in their relations with their parents (18.75% among controls), and 28.1%, in their social relations (15.6% among controls). There is a potential risk that these difficulties, rather than explaining the onset of ED, are the result of subjects' behavior in the disorder's prodromic period, when their abnormal eating behavior began to raise concern among family members, causing relational difficulties. This supposed bias cannot be discarded, although it is minimized by this design. Our definition of an ED case allows the diagnostic threshold to include clinical situations that stricter categorizations would have excluded or considered as prodromic. In our study, partial forms of ED comprise almost 94% of the affected sample, and subclinical ones, 53%. In addition, our interviewers paid special attention to the detection of these minor forms and, as the LEDS requires, their correct dating. This reduced the risk of assessing difficulties caused by EDs in a prodromic period because the changes in eating behavior preceding individuals' crossing our case threshold could hardly be important enough to generate difficulties, something which could easily have occurred if we had only considered full syndromes. In any case, the LEDS' independence scale covers this by controlling for and eliminating difficulties related to subjects' behavior and, therefore, dependent on them.
Our study also takes a quantitative approach to the assessment and evolution of stress in the year preceding ED onset. Figure 1 shows the evolution of stress generated by acute stressful events. More exactly, it shows the weekly amount of stress that each subject was exposed to; this stress is composed of the impact of the events occurring during each week, in addition to the residual stress of previous events (15,21). There is a notable increase during the 3 weeks before onset. Chronic stress, shown in Figure 2, is significantly higher in the case group than in the noncase group at all times during the year preceding ED onset.
We want to discuss the limitations of the current study. Regarding the sample itself, the majority of case subjects suffered from AN (full, partial, and subclinical syndromes). There were very few cases of BN (strict or broad criteria), probably due to sample age (34). Our results are, therefore, fundamentally applicable to AN. Because of the low prevalence of EDs, sample size was limited. Verification of our results using similar but larger samples is indicated.
The main limitation with epidemiologic studies of ED, especially if they investigate risk factors, is that EDs have low incidence and prevalence. Unless very large samples are used, one must be very cautious when generalizing the results (1). This must be kept in mind when interpreting our results because our final sample size was small and only included two full ED syndromes. Caution is therefore warranted when generalizing the results to the groups of community cases with full ED.
Another limitation was that we did not measure stress prospectively and were thus not able to assess true risk factors but only correlates. The ideal design would have been that of a prospective cohort study, but this would not only have required a study of long duration but also, because of the low incidence of EDs, a very large cohort.
Finally, the use of a semistructured interview utilizing a list of questions for establishing diagnoses should be considered a methodologic weakness because structured diagnostic interviews would have been more reliable.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Angeles García, Lluis Sanjuán, Marisa Vila, Gloria Rodrigo, Alberto Domínguez, Emilia Chaves, M. J. Abad, and M. J. Hernández participated in the design and/or the clinical/psychometric assessment.
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
This project was financially supported by the EVES, Generalitat Valenciana, Grant 69/1998.
DOI:10.1097/01.psy.0000227749.58726.41
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
