|Home | About | Journals | Submit | Contact Us | Français|
The definition of the stressor criterion for posttraumatic stress disorder (“Criterion A1”) is hotly debated with major revisions being considered for DSM-V. We examine whether symptoms, course, and consequences of PTSD vary predictably with the type of stressful event that precipitates symptoms.
We used data from the 2009 PTSD diagnostic subsample (N=3,013) of the Nurses Health Study II. We asked respondents about exposure to stressful events qualifying under 1) DSM-III, 2) DSM-IV, or 3) not qualifying under DSM Criterion A1. Respondents selected the event they considered worst and reported subsequent PTSD symptoms. Among participants who met all other DSM-IV PTSD criteria, we compared distress, symptom severity, duration, impairment, receipt of professional help, and nine physical, behavioral, and psychiatric sequelae (e.g. physical functioning, unemployment, depression) by precipitating event group. Various assessment tools were used to determine fulfillment of PTSD Criteria B through F and to assess these 14 outcomes.
Participants with PTSD from DSM-III events reported on average 1 more symptom (DSM-III mean=11.8 symptoms, DSM-IV=10.7, non-DSM=10.9) and more often reported symptoms lasted one year or longer compared to participants with PTSD from other groups. However, sequelae of PTSD did not vary systematically with precipitating event type.
Results indicate the stressor criterion as defined by the DSM may not be informative in characterizing PTSD symptoms and sequelae. In the context of ongoing DSM-V revision, these results suggest that Criterion A1 could be expanded in DSM-V without much consequence for our understanding of PTSD phenomenology. Events not considered qualifying stressors under the DSM produced PTSD as consequential as PTSD following DSM-III events, suggesting PTSD may be an aberrantly severe but nonspecific stress response syndrome.
The diagnosis of posttraumatic stress disorder (PTSD) requires exposure to a potentially-traumatic event, referred to as “Criterion A.” Since the construction of diagnostic criteria for PTSD, the requirement and definition of Criterion A has been controversial. Critics have pointed out that precisely defining events that qualify as highly stressful is challenging(1), that individuals can experience a PTSD-like syndrome in response to a wide variety of stressors (reviewed in (2)), and that there has been no support for particular stressors producing distinct variants of other disorders, such as postpartum depression(3).
Partly in response to these criticisms, the definition of Criterion A has changed substantially since DSM-III(4). Initially, DSM-III mandated that PTSD be diagnosed only subsequent to a life-threatening event so intense that it was outside the range of normal human experience, for example, active combat, natural disaster, or rape. In DSM-IV, Criterion A was split into two parts. Criterion A1, the event itself, was no longer required to be outside the range of normal experiences: witnessing or being confronted with events that merely threatened serious injury was sufficient. A new Criterion A2 was also added, which required events to elicit horror, intense fear, or helplessness(5). However, recent research has found Criterion A2 to be redundant with other diagnostic criteria, therefore we focus this paper on Criterion A1(6). Criterion A1 is being considered for revision again in DSM-V, provoking intense debate about whether to leave the definition as is(7), narrow it (8), or drop Criterion A1 altogether(9, 10).
The definition of Criterion A1 has implications for nosology, theory, and practice. For nosology, Criterion A1 determines who is eligible for the diagnosis of PTSD and whether people with PTSD-like symptoms are diagnosed with PTSD, with adjustment disorder, or with another anxiety or mood disorder. For theory, fear conditioning has been proposed as a central mechanism of PTSD(11). If a similar stress syndrome arises from both fear-inducing events, such as violence, and from events that typically do not involve intense fear, such as job loss or divorce, then the fear conditioning model is no longer central to the diagnosis(5). For practice, patients diagnosed with PTSD may be more likely to qualify for insurance coverage for their treatment or be compensated for their traumatic experience in court(12). Treatments developed for PTSD related to DSM-III stressors may not be effective for PTSD from ordinary stressors(13).
This paper examines whether the definition of Criterion A1 matters. By ‘matters,’ we mean do symptoms, course, and consequences of PTSD vary predictably with the type of stressful event that precipitates PTSD. Although many studies have shown that more intense event exposures increase the likelihood of developing PTSD, few have examined the relationship between event type and symptoms, chronicity(14), distress(15), impairment(16), or sequelae of PTSD. Specifically, we consider three questions: 1) Are DSM-III-qualifying events typically ranked as worse than other stressful events? 2) Are people more likely to develop PTSD from DSM-III worst events than from DSM-IV or non-DSM worst events? 3) Do DSM-III worst events, compared with DSM-IV and non-DSM worst events, lead to a more serious form of PTSD with more sequelae?
We used data from the PTSD diagnostic subsample (N=3,013) of the Nurses Health Study II (NHSII). The NHSII is a cohort of 116,678 female nurses from the 14 most populous U.S. states, established in 1989 and followed up with biennial questionnaires. Briefly, in 2008 we mailed the Trauma and PTSD Screening Questionnaire(17, 18) to 60,804 NHSII participants who had completed recent questionnaires (to retain participation in the longitudinal study, participants who have not responded to the most recent biennial questionnaire are not asked to participate in supplemental studies). The response rate was 89% (N = 54,282). Of these 54,282 respondents, 43,413 reported exposure to at least one traumatic event on the Screening Questionnaire. Of these, 23,104 respondents agreed to be interviewed (53%). We then identified probable PTSD cases and trauma-exposed controls using Breslau’s lifetime PTSD screen(19), which classifies PTSD cases with 85% sensitivity, 93% specificity, 68% positive predictive value, and 98% negative predictive value.
We randomly selected 2,112 probable PTSD cases and 2,001 probable controls for diagnostic interviews. Among those select, 3013 people (73%) completed interviews, including 1510 people with probable PTSD (71% completion rate) and 1503 people without probable PTSD (75% completion rate). Compared with the whole NHS2 cohort of 116,678, our sample was somewhat more likely to be white (98.1% versus 95.5%), was slightly younger (mean age=53.3 years versus 54.5 years) and was more likely to be married (81.2% versus 77.4%, p<0.001 for all comparisons) at the establishment of the cohort in 1989. The Partners Human Research Committee approved this study; the protocol has been published(20).
We interviewed participants via telephone using a highly structured interview. The PTSD Checklist (PCL-C), a 17-item self-report measure of DSM-IV PTSD symptoms(21) was used to assess re-experiencing symptoms (Criterion B), avoidance/numbing symptoms (Criterion C), and arousal symptoms (Criterion D). Participants were cued to think of the period following the event during which symptoms were most frequent and intense, and were then asked whether they had ever been bothered by each of the 17 symptoms. Participants rated each symptom on a scale indicating how much they had been bothered by a particular symptom as a result of the event, from “not at all” to “extremely”(22). To be a PTSD case, respondents must have reported experiencing one or more of the 5 re-experiencing symptoms, 3 or more of the 7 avoidance/numbing symptoms, and 2 or more of the 5 arousal symptoms at least “moderately.” Additional questions assessed the other three DSM-IV criteria: intense fear, horror, or helplessness in response to the event (Criterion A2), symptom duration of at least one month (Criterion E), and clinically significant impairment in functioning due to symptoms (Criterion F)(23). The PCL-C had excellent internal consistency (Cronbach’s α=0.87).
Participants were initially asked to identify stressful events they had experienced from a list of 25 events, including both events standard in diagnostic interviews of PTSD and events included in measures of life stressors following event lists used in diagnostic interviews in epidemiologic samples including the World Health Organization (WHO) Composite International Diagnostic Interview (CIDI)(23, 24) and the NIMH Diagnostic Interview Schedule(25–27)(eTable 1). They were further asked an open-response question, whether they had “experienced any other very stressful situation or event.” Events described in response were recorded verbatim. PTSD symptoms were then assessed by referencing the event that the participant regarded as the worst. Respondents were considered affected by lifetime PTSD if all six DSM-IV criteria were met in reference to the worst event. To assess reliability, a blind review of audiotapes from 50 interviews was conducted by the senior author, a licensed clinical psychologist who is an experienced diagnostician specializing in PTSD treatment. Reliability was assessed by comparing this diagnosis with that made via computer algorithm from the structured interviews using Cohen’s kappa statistic. The kappa was 1.0 (perfect reliability).
Our diagnostic interview also assessed lifetime depression via a modified version of the Patient Health Questionnaire (PHQ-9)(28) using the DSM-IV coding criteria. The PHQ-9 had excellent internal consistency (α=0.87).
We assessed the validity of our identification of PTSD and depression in a separate cohort, the Detroit Neighborhood Health Study, via clinical interviews among a random subsample of 51 participants(29, 30). A licensed psychologist conducted one-hour in-person clinical interviews, using the Clinician-Administered PTSD Scale for DSM-IV to evaluate PTSD(31) and the Structured Clinical Interview for DSM-IV Disorders to evaluate depression(32). The psychologist was blinded to the information obtained from the main study. Comparison of the clinical interviews with the PCL-C and PHQ-9 from the main study showed excellent concordance for both PTSD and depression(29, 30).
For analysis, events were categorized in one of three mutually-exclusive groups. DSM-III-qualifying events included events that threatened life or physical integrity, such as disasters, interpersonal violence, and serious accidents(4). Events that qualifying only under DSM-IIIR/DSM-IV and not DSM-III criteria were serious illness, witnessing violence, and serious events to someone close, including sudden death(21). Events not qualifying under any DSM criteria included, for example, serious financial or legal problems, and sexual harassment at work. Added to this group were other experiences women described in the open-ended question, for example, marital problems, divorce, and job loss. Answers to the open-ended question were grouped into 7 categories by a research assistant and the first author, and were reviewed by the senior author (Table 1).
For this analysis, we classified people as PTSD cases if they met the B, C, D, E, and F criteria for PTSD related to their worst event via structured interview as described above, whether or not it fit under DSM-IV criteria A(1) or A(2).
We asked participants the age at which their worst event occurred. We counted the number of Criteria B, C and D symptoms which the participant endorsed as moderately or severely distressing. We assessed symptom duration with one question about the longest time period symptoms were experienced. The 4 response options were: less than one month, between 1 and 3 months, between 3 and 6 months, between 6 months and 1 year, or more than 1 year. Participants rated distress related to symptoms from “not at all distressing” to “severely distressing,” and rated difficulties at work, home and in social interactions due to symptoms from “not difficult at all” to “extremely difficult.”
We used age-of-onset information from the diagnostic interview described above to determine if participants experienced a first-onset depression following their worst event. For other possible sequelae age-of-onset information was not available, therefore, we assessed these behaviors and health indicators at a single time point following the worst event. In addition to responses provided during our interview, responses provided to 2 questionnaires given to the whole cohort in 2001 and 2005 were used to examine possible PTSD sequelae. The 2005 questionnaire queried body mass index, current smoking, binge drinking, exercise frequency, and phobic anxiety. Binge drinking was assessed with a single question, “In a typical month during the past year, what was the largest number of drinks of beer, wine and/or liquor you may have had in one day?” Six or more drinks in a day on a typical month was considered binge drinking(33). Phobic anxiety was assessed with the Crown-Crisp index(34); the total score was divided into 4 groups indicating no anxiety to highest level of anxiety(35). Physical functioning was assessed in 2001 with the Short Form 36 Health Survey (SF-36)(36). The PTSD diagnostic interview included questions about unemployment and divorced or separated status among participants ever married.
We ranked events based on the frequency with which each event was selected as worst among women exposed to it. To adjust for the number of other events people were exposed to, we also calculated likelihood of being chosen as worst in pairwise comparisons. For example, a respondent who had experienced mugging, financial crisis, and a serious accident and selected mugging as worst would contribute information to the mugging/financial crisis and mugging/accident comparisons, and mugging would be ranked worse in each comparison. Since we do not know whether the respondent considered the financial crisis worse than the accident, this pair would not contribute to the event rankings. We then calculated the percent of times each event was chosen as worst in these comparisons, across all participants, and ranked events accordingly. Finally, we looked at participants who had been exposed to events from all three groups, and reported the percent choosing a worst event from each group.
We calculated the odds ratios of developing PTSD from non-DSM and DSM-IV worst events compared with DSM-III worst events.
To assess whether PTSD resulting from DSM-III events was experienced as more deleterious than that resulting from events in the other two groups, we compared distress, symptom duration, impairment and receipt of professional help among participants who met PTSD criteria from the three event groups using χ2 tests.
In further exploration of possible differences in PTSD resulting from different events, we compared physical, behavioral, and psychiatric sequelae in participants with PTSD from the three event groups to participants without PTSD. We examined physical functioning, body mass index, exercise, smoking, binge drinking, phobic anxiety, depression, unemployment, and divorced or separated status after the worst event by event group. Women who experienced their worst event following the year in which a given sequela was assessed were excluded from analyses involving that sequela. We tested for differences in these sequelae in participants who had PTSD from DSV-IV and non-DSM events with the no-PTSD group and compared with participants who had PTSD from a DSM-III event, using Student’s t-test for continuous measures and χ2 tests for ordinal and dichotomous measures. We conducted a Cox survival analysis to assess whether PTSD from each event group increased hazard of first episode of major depression, adjusted for age at interview. All tests were two-sided.
By sample design, nearly every participant (99.7%, n=3005) was exposed to a potentially traumatic event (mean=7.4 events). Individual events varied substantially in their probability of being chosen as the worst event among participants exposed, from sudden death of someone close, chosen as worst by 36.2% of participants exposed, to parents’ serious legal problems, chosen as worst by 0.6% of participants exposed. While some DSM-III events, such as rape and intimate partner violence, ranked among the most often chosen, others, such as disasters and muggings, ranked among the least often chosen (Table 1). Pairwise rankings, which accounted for number of events to which participants were exposed, produced nearly identical results. Three-quarters of participants (76.3%, n=2299) were exposed to events from all three groups. Only 30.1% of these participants selected the DSM-III-qualifying events as their worst, while 41.0% chose a DSM-IIIR/DSM-IV event, and 25.6% chose a non-DSM event.
We found no difference between non-DSM and DSM-III worst events in odds of PTSD (OR=1.0, 95% CI=0.8, 1.2, p=0.7), although odds of PTSD from DSM-IV events were significantly lower than odds from DSM-III events (OR=0.6, 95% CI=0.5, 0.7, p<.0001).
Table 2 presents PTSD phenomenology by event type. Participants with PTSD from DSM-III events reported more symptoms and more often reported that symptoms lasted one year or longer compared to participants with PTSD from the other groups. Distress and impairment from symptoms did not differ significantly across event groups.
Overall, participants with PTSD had poorer health-related behaviors and worse physical and psychiatric sequelae compared to participants without PTSD, with no systematic differences in these outcomes between precipitating event groups. Following their worst event, participants in all three PTSD groups were statistically significantly more likely to report highest levels of phobic anxiety and were at increased risk of depression compared to participants who did not develop PTSD from their worst events. Physical functioning was significantly worse in participants who experienced PTSD from DSM events, compared to participants without PTSD, but did not significantly differ in participants with PTSD from non-DSM events (Table 3).
Prevalence of divorce or separation was approximately double in participants with PTSD from DSM-III (23.6%) or non-qualifying events (20.0%), compared to participants without PTSD (11.1%), but was not significantly elevated in participants with PTSD from DSM-IV events. Unemployment was also significantly more common in participants with DSM-III- and non-DSM-related PTSD compared to participants without PTSD (Table 3).
Compared to participants with PTSD from a DSM-III event, physical, behavioral, and mental health indicators did not statistically differ in the other two groups, except divorce was more common among participants in the DSM-III group than among participants in the DSM-IV group, and risk of depression onset was lower among participants in the DSM-III group than among participants in the DSM-IV group.
Because DSM-III events included both events frequently considered worst and events infrequently considered worst, we conducted supplemental analyses to see if grouping events instead by frequency with which they were selected as worst would reveal a dose-response with phenomenology and sequelae. Mirroring results from the main analyses, we found no systematic differences (eTable 2). Additionally, because presence of more risk factors for PTSD among persons developing symptoms from non-DSM worst events may have accounted for the similarities among PTSD sequelae that we found, we conducted additional analyses examining five risk factors for PTSD by PTSD type: number of other event types endorsed, depression prior to PTSD, exposure to childhood physical and sexual abuse, and age at worst trauma. Contrary to this hypothesis, persons with PTSD from DSM-III events were more likely to have experienced their event before age 11 years and were exposed to more other event types than persons with DSM-IV or non-DSM PTSD. There were no other differences in prevalence of individual risk factors or total number of risk factors by PTSD group.
Our principal finding is that sequelae of PTSD did not vary systematically with the type of stressful event that initiated PTSD symptoms, whether DSM-III-qualifying, DSM-IV-qualifying, or not qualifying under DSM. Although individual events differed substantially in their likelihood of being chosen as a worst event, DSM events as a group were not consistently more likely to be chosen than other events, with some DSM events among the most often chosen and some among the least often chosen. Although the number and duration of symptoms were slightly higher for PTSD resulting from DSM-III-qualifying events, on the whole, the stressor Criterion A1 did not seem to matter for PTSD phenomenology in our study.
Because types of events varied in their likelihood of being chosen as worst, but seriousness of PTSD resulting from worst events did not differ, our results are consistent with at least two possibilities. First, event type may be only an approximate indicator of a specific event’s intensity. Respondents choosing “mild” types as their worst event may have experienced unusually intense instances. This hypothesis could be tested in further research that obtained more information about what happened in these instances(37). Second, the effect of the intensity of the inciting event on symptoms, health and behavior may diminish once the PTSD diagnostic threshold has been reached: the relationship between event intensity and PTSD symptoms and sequelae may be curvilinear rather than dose-response(5).
Our results should be considered in light of several limitations. Because our PTSD subsample was comprised of women in the twentieth year of a longitudinal cohort study, it is possible that the seriousness of PTSD from different events appeared similar because people with more serious cases of PTSD had previously dropped out of the study. However, in supplemental analyses with women of similar ages and race in the National Epidemiologic Survey of Alcohol-Related Conditions, a nationally-representative sample of US adults(38), risk ratios for sequelae from PTSD were similar (eTable 3). All measures were self-reported, which may introduce bias. However, use of data collected at different times reduces the possible influence of current mental health on report of PTSD sequelae. Finally, our sample is primarily white U.S. women ages 44 to 63, therefore results may not apply to other demographic groups. Nevertheless, this study provides the most comprehensive examination to date of Criterion A1 stressors by contrast with other stressors in relation to the PTSD symptom syndrome and sequelae.
Our results suggest that the stressor criterion as defined by the DSM may not be informative in characterizing symptoms and sequelae of PTSD among persons meeting all other PTSD criteria. Three suggestions emerge from this observation. First, in the context of the ongoing DSM-V revision, these results suggest that Criterion A1 could be expanded in DSM-V without much consequence for our understanding of PTSD phenomenology. Second, events not considered qualifying stressors under the DSM produced PTSD as consequential as PTSD following DSM-III events, suggesting PTSD may be an aberrantly severe response to many types of stressors, not just to extreme stressors. Our findings that sudden death of a loved one and serious illness or injury to a loved one were among the stressors most often rated as worst further suggests that the DSM has been overly focused on events with intense physical as opposed to psychological impact. Death or injury to a loved one is very different from DSM-III events that are personally life threatening, occur in circumstances outside of ordinary human experience, and primarily evoke intense fear. Finally, our study does not support the concept of distinct subtypes of stress-response syndromes arising from different stressors. Our findings are consistent with a growing literature that suggests psychopathology following events like financial or legal problems, neither included in even the broadened DSM-IV Criterion A1 and traditionally diagnosed as an adjustment disorder, may be similar phenomenologically to PTSD(39).
Sources of support: This study was supported by the Harvard Training Program in Psychiatric Genetics and Translational Research T32MH017119 (to AR), National Institutes of Health Grants MH070627 and MH078928 (to KK), R01 HL064108 (to RW). This study was also supported by National Institutes of Health Grants DA022720, DA022720-S1, MH088283, MH078152, and MH082729 (to SG & AA) and R01CA50385 to support the Nurses Health Study II cohort follow-up.
We thank Marissa Rewak, M.A., Department of Society, Human Development, and Health, Harvard School of Public Health, Boston, MA, for her careful work in coding answers to the open-ended question and the members of the Nurses Health Study II for their time and loyal participation in this study.
Previous presentation: This work has not been presented previously.
Disclosures: The authors have no disclosures to report.