Cortisol and Catecholamines in Posttraumatic Stress Disorder: Title and subTitle BreakAn Epidemiologic Community Study

Stress has been viewed as a precipitant for a number of psychiatricdiseases. A unique form of stress, psychologic trauma is viewed in the DSM as linked in an essential way with a specific syndrome,that of posttraumatic stress disorder (PTSD).¹ Ithas been proposed that PTSD demonstrates a paradoxical cortisol profile incomparison with the expected response to stress, with evidence of low basalcortisol levels in urine and plasma and enhanced negative feedback to dexamethasone.^{2 - 4} The low cortisol levelswere accompanied by increased catecholamine secretion in urine, leading tothe claim that there is a disconnect between basal cortisol and catecholaminein individuals with PTSD.^{2 - 3} Whilethere was some (but incomplete⁵ ) consistencyamong early studies with regards to the urinary cortisol findings in malecombat veterans,³ recent reports suggest thateven these findings are uncertain.⁶ The generalizabilityof these findings to the general community, where PTSD occurs predominantlyin women, is questionable.

Studies in women with PTSD, generally recruited by advertising, havefocused on childhood sexual abuse, which may not reflect the effects of traumain adulthood. The results of these studies are inconsistent, with one studysuggesting increased cortisol levels,⁷ andanother, low cortisol levels,⁸ while stillanother, no difference from control subjects.⁹ Studiesof individuals who have experienced natural disasters have focused on symptomsof anxiety or distress but not on DSM-defined PTSD.^{10 - 14} Anexception is a study by Maes et al,¹³ whichreported a higher mean urinary free cortisol (UFC) level in subjects withPTSD following natural disasters. However, in that study, subjects were volunteersand a comparison with persons exposed to trauma without PTSD was not included.Studies of individuals who have experienced disasters^{10 - 14} havereported elevated mean cortisol levels in urine, saliva, or plasma. Boscarino⁴ studied a representative sample of male veterans,but there have been no reports from representative community samples on basalcortisol levels in individuals exposed to trauma with PTSD.

Three studies have examined 24-hour urinary catecholamine levels inveterans with PTSD, yielding conflicting results. Kosten et al¹⁵ andYehuda et al¹⁶ reported an increase in urinaryepinephrine and norepinephrine levels in veterans with PTSD, compared withpatients with other psychiatric disorders¹⁵ orhealthy control subjects.¹⁶ Pittman and Orr⁵ reported no difference in catecholamine levels betweencombat veterans with PTSD and combat veterans without PTSD. The nature ofthe control group might be critical because exposure to trauma may alter urinaryepinephrine and norepinephrine levels. This possibility is supported by evidenceof increased urinary norepinephrine levels in a civilian population exposedto the Three Mile Island nuclear accident, compared with individuals residing80 miles away¹⁷ and increased urinary catecholaminelevels in American hostages shortly after they were freed from Iran.¹⁸ Urinary catecholamine levels in persons exposed totrauma with and without PTSD compared with nonexposed persons have not beenexamined.

Because PTSD is highly comorbid with major depressive disorder (MDD),^{19 - 20} the effect of MDD, which shows hypothalamic-pituitary-adrenalaxis activation, on the neuroendocrine picture of PTSD is an important question.Results have been inconsistent. In earlier studies, the presence of comorbidMDD did not appear to alter the low cortisol findings in PTSD.³ Incontrast, a recent study reported that MDD and PTSD influenced 24-hour UFClevels in opposite directions, with no interaction, so that the comorbid groupshowed an intermediate cortisol excretion.²¹

In this report, we examine urinary catecholamine and cortisol levelsin relation to PTSD and exposure to traumatic events. The study was conductedin the context of an ongoing longitudinal community study in the Detroit,Mich, metropolitan area. We present data on 24-hour UFC and catecholaminelevels, collected from a subset of the sample in a sleep research center.The key questions addressed concern the association of exposure to traumaand PTSD with urinary cortisol and catecholamine levels and the role of comorbidMDD in the relationship of PTSD with these hormonal measures.

The neurobiologic investigation was nested in a large-scale longitudinalcommunity study of young adults. The study was described previously.^{19 ,22} In brief, a sample of 1200 individualswas randomly selected from all 21- to 30-year-old members of a large healthmaintenance organization in southeast Michigan. The membership of the healthmaintenance organization was representative of the population of the geographicarea as depicted in the 1990 US census, with the exception of the extremesof the socioeconomic status range. Personal interviews were conducted in 1989with 1007 members (84%) of the sample. Follow-up personal interviews wereconducted in 1992, 1994, and 1999 through 2001. In each wave, more than 90%follow-up completion was achieved. In the 10-year follow-up in 1999 through2001, 913 members (91.1%) of the initial sample completed interviews. Biologicmeasures were collected on a subset of these respondents.

The following persons from among those who completed the 10-year follow-upwere eligible for the sleep and urine collection study: (1) all persons whowere exposed to traumatic events in the 5-year interval between 1994 and thelast assessment, (2) other individuals (not from group 1) who met PTSD criteriain previous assessments, and (3) other individuals (not from groups 1 or 2)from a randomly preselected subsample of the total sample. Persons who movedaway from the Detroit metropolitan area were not eligible. Persons with recentsubstance abuse, use of psychotropic medications, and smokers who reportedthat they could not abstain for the 8-hour sleep period were not eligiblefor this study. However, in this general population study, no subjects wereexcluded based on these criteria. Eligible respondents were invited to spend2 consecutive nights and the intermediate day at the Sleep Center of HenryFord Hospital, Detroit. Bedtimes were determined by matching the respondents'regular bedtimes at home. Respondents stayed in bed for 8 hours each night.Of a total 439 eligible individuals, 292 (66.5%) participated.

Urine collection during the 32-hour stay at the sleep center was madein 4 parts, with each covering an 8-hour period. The first collection coveredthe first night, the second collection covered 8 hours from rising to earlyafternoon, the third collection covered the late afternoon and evening, andthe fourth, the second night. The procedure provided total 24-hour measuresand had the added capacity to provide separate timed measures that capturedthe morning surge (in the 8 AM hour) and the evening low (inthe 8 PM hour) for an examination of effects related to diurnalphase. At the time of collection, urine was split into 2 aliquots, with theurine for catecholamine level testing acidified with 7 mL of 6 N hydrochloricacid as a preservative, and the urine for cortisol level testing was frozenwithout further processing. Data on 13 persons were deleted because of lowcreatinine levels, defined as less than 1 g/24 h, leaving a sample of 275persons for the analysis.

The National Institute of Mental Health—Diagnostic Interview Schedule(NIMH-DIS)²³ for DSM-III-R was used to diagnose psychiatric disorders. The baseline interviewin 1989 inquired about lifetime history of disorders, and each follow-up assessmentinquired about disorders occurring during the interval period since the previousassessment. The diagnosis of PTSD in DSM-III-R requiresexposure to a qualifying traumatic event and the presence of PTSD criterionsymptoms that are linked to the traumatic event. Two earlier studies reportedhigh concordance between NIHM-DIS–diagnosed PTSD and independent clinicalreinterviews.^{20 ,24} The latter²⁴ used the clinician-administered PTSD scale^{25 - 26} and reported sensitivity of 76% andspecificity of 97%.²⁴

We assayed UFC levels with DPC Coat-a-Count cortisol kits (DPC, LosAngeles, Calif). Urine was extracted using dichloromethane following the instructionsfrom the kit. Interassay variation for cortisol levels was 6%. Urine catecholaminelevels were assayed by high-pressure liquid chromatography using a commerciallaboratory (Warde Medical Laboratories, Ann Arbor, Mich).

All statistical analyses were performed on log-transformed data, whichnormalized distributions of the hormone measures. Comparisons of nonsmokerswho had never smoked (n = 156), current smokers (n = 58), and past smokers(n = 61) showed no significant differences in total 24-hour UFC levels (P = .44). We found no significant effect of oral contraceptives(n = 84) on 24-hour UFC levels (P = .77).

Two series of analyses were conducted for cortisol, epinephrine, norepinephrine,and dopamine levels. The first consisted of comparisons across 3 groups: (1)PTSD (n = 69; 16 men, 53 women, 10 of whom met criteria for current PTSD atthe time of the interview), (2) exposed to trauma without PTSD (n = 105; 38men, 67 women), and (3) no exposure to trauma (n = 101; 40 men, 61 women).This provided 2 control groups with which PTSD could be compared, includinga group of persons who were never exposed to trauma, according to the baselineinterview and the 3 follow-up assessments. The second series examined comorbidityof PTSD with MDD and consisted of comparisons across 4 groups defined by historyof PTSD and MDD: (1) neither disorder (n = 139; 61 men, 78 women), (2) MDDonly (n = 67; 17 men, 50 women), (3) PTSD only (n = 23; 5 men, 18 women),and (4) both disorders (n = 46; 11 men, 35 women). All persons in the MDDgroup had reported at least 1 exposure in their lifetime, either precedingor following MDD onset. In each series of analyses and for each hormone, wefirst examined total 24-hour urine hormone levels and then we examined thenight, morning (AM), and evening (PM) urine hormone levels, using an analyticalapproach that took into account diurnal variation.

Correlation of the first and second night cortisol levels was 0.43.The correlations of dopamine, epinephrine, and norepinephrine levels werehigher, 0.67, 0.62, and 0.65, respectively. Differences between the 2 nightsoverall or across diagnostic groups were not significant, as tested by paired t tests, and consequently, hormone values of the 2 nightswere averaged. Averaging across 2 nights yields a more reliable measure. Analysisof the 4 time periods, with each night treated separately, did not alter theresults. For the total 24-hour measures, we combined the average of the 2nights with the 8-hour AM and the 8-hour PM values.

We tested whether current PTSD or MDD cases (ie, with symptoms continuingin the preceding 12 months) differed from remitted cases but found no significantdifferences in cortisol or catecholamine levels. We therefore used lifetimedisorders in the analyses. The neither-disorder group comprised 38 persons(27.3%) who were exposed to trauma and 101 who were never exposed. Comparisonof these 2 subgroups in the neither-disorder group revealed significant differencesonly in dopamine levels, with persons who were exposed to trauma having alower mean value than those never exposed. In a separate analysis, we used2 no-disorder groups, classified by the presence vs the absence of exposureto trauma. The analysis with 2 no-disorder groups yielded the same resultsas the 4-group analysis.

Analysis of variance (ANOVA) was used to compare total 24-hour urinarymeasures across groups. Significant results were followed up with post hocpaired comparisons using the Tukey test. We tested the effects of sex, race,education, and interactions between group membership and these variables.When a significant covariate or an interaction was detected (using α= .05), models that included these variables were displayed. Sample size wasslightly smaller in the total 24-hour analyses because persons without 3 consecutive8-hour urine collections (n = 19) were excluded.

Multiple regression analysis was used, applying generalized estimatingequations (GEEs),^{27 - 29} totest and estimate associations between group membership and each of the 4hormones across 3 time periods of the 24-hour diurnal cycle—night, AM,and PM. The GEE approach permits simultaneous modeling of the relationshipbetween group classification and hormone measures at the 3 time periods. TheGEE approach takes into account correlations within persons across the multiplemeasures and uses information on persons with incomplete data. Interactionterms were used to examine whether differences across groups varied by timeand by covariates (eg, sex). No significant interactions between time andgroup membership were detected; post hoc tests of differences by group membershipacross night, AM, and PM were not performed. The estimated log-transformedmeans in the GEE results were averaged across the night, AM, and PM measures.In the analyses of cortisol levels, we found a significant interaction betweendiagnostic groups and sex. The final model for the cortisol analysis is illustratedin the equation

Y = α + β₁ (group) + β₂ (sex) + β₃ (group × sex) + β₄ (time)

where log cortisol measures at 3 times (night, AM, and PM) are the outcomes (Y).

No significant differences in the biologic subsample were observed insex, race, age, education, and substance use disorders in comparison withthe entire sample (Table 1). The4 diagnostic groups defined by history of PTSD and MDD did not differ on thesecharacteristics from the total sample.

The raw mean ± SD of total 24-hour urinary cortisol levels forthe entire sample with complete data (n = 256) was 52.0 ± 32.2 µg/24h (143.5 ± 88.8 nmol/d). The ANOVA of total 24-hour urinary cortisollevels in persons with PTSD, exposure to trauma without PTSD, and no exposureto trauma revealed neither a significant group difference nor a sex ×group interaction. The GEE analysis on night, AM, and PM data (n = 275) revealeda significant diurnal variation (P<.001) and asignificant group × sex interaction (P = .04).However, sex-specific analyses detected no significant effects of group membershipin either men (P = .21) or women (P = .09).

We explored the potential effect of type of trauma (assaultive vs nonassaultive)on cortisol levels and found no support for such an effect. We found no supportfor effects of early trauma (≤16 years of age) vs late trauma or historyof prior exposure to trauma vs no prior exposure on cortisol levels. Recencyof exposure (ie, within the preceding year) showed an effect on cortisol levelsin men only, with those who were recently exposed exceeding significantlythose who experienced a trauma more than 1 year ago, as well as those withno trauma exposure in their lifetime.

We explored whether restricting the analysis to current (active) PTSDwould suggest different conclusions. There were no men with current PTSD;thus, the analysis was restricted to women. Current and past PTSD in womendid not differ significantly (P = .46). Log mean± SDs of no exposure to trauma and current PTSD in women were virtuallyidentical, 2.50 ± 0.06 and 2.49 ± 0.18, respectively (P = .97).

We evaluated the role of comorbidity of PTSD and MDD in total 24-hourUFC levels. Four groups were compared: MDD alone, PTSD alone, PTSD comorbidwith MDD, and neither disorder. The ANOVA revealed a significant group ×sex interaction (P = .02). However, sex-specificmodels revealed a nonsignificant group effect in men (P = .12) and women (P = .07).

The GEE analysis on night, AM, and PM UFC levels revealed a significantdiurnal variation (P<.001) and group × sexinteraction (P = .003). Group effects were significantin women (P = .03) and men (P =.04) (Table 2). In women, thePTSD/MDD comorbid group had significantly higher cortisol levels, comparedwith each of the other 3 groups (ie, MDD alone, PTSD alone, and no disorder).No other comparisons were significant. In men, the MDD group had a significantlyhigher mean cortisol level than the no-disorder and the comorbid PTSD/MDDgroups. No other comparisons were significant. To illustrate the results,data on cortisol levels at night, AM, and PM from these analyses are displayedin Figure 1 and Figure 2.

The sample raw mean ± SD for 24-hour dopamine levels was 381.5± 334.8 µg/24 h; epinephrine, 11.0 ± 9.1 µg/24 h[60.1 ± 49.7 nmol/d]; and norepinephrine, 50.3 ± 8.2 µg/24h [297.3 ± 48.5 nmol/d]. Table 3 presents comparisons of the PTSD, exposed to trauma without PTSD,and no exposure to trauma groups. There were significant effects of groupmembership on the 3 catecholamine levels. The group with PTSD had higher meanlevels of all 3 urinary catecholamines, compared with the exposed to traumawithout PTSD and no exposure groups. (See Table 3 for significant pairwise comparisons.) Pairwise comparisonsalso revealed significant differences between the exposed to trauma withoutPTSD and no exposure groups in dopamine (P = .02)and epinephrine (P = .02) levels, although in theopposite direction from the expected (ie, lower in the exposed to trauma withoutPTSD group than in the no exposure group). Subjects with past vs current PTSDdid not differ significantly on any catecholamine measure (P>.25). The GEE models yielded similar results and, in addition, revealedsignificant diurnal variations (P<.002).

Group means of total 24-hour urine dopamine, epinephrine, and norepinephrinelevels across the 4 diagnostic groups are displayed in Table 4. A significant race difference was detected in the analysisof dopamine levels, with African American individuals' levels exceeding thoseof white individuals; the mean ± SEs of dopamine levels in Table 4 are race-adjusted. For all 3 outcomes,there were significant effects of group membership. Pairwise comparisons showedthat, across the 3 catecholamine measures, the comorbid group of PTSD plusMDD had significantly higher means than the groups with neither disorder andwith MDD only. The group with PTSD only also exceeded the groups with neitherdisorder and with MDD on all 3 measures; however, the excess did not alwaysreach significance (Table 4).A consistent finding across the 3 catecholamine measures was the absence ofa significant difference between 2 PTSD groups (ie, the PTSD/MDD comorbidgroup and the PTSD-only group). The GEE analysis yielded similar results and,in addition, detected a significant diurnal variation (P<.002). The data for catecholamine levels across night, AM, andPM are shown in Figure 3, Figure 4, and Figure 5.

Both urinary cortisol and catecholamine levels are indices of stresshormonal systems, but the extent to which these 2 systems are correlated isunclear. Using Spearman rank correlations, we examined whether there was evidenceof a disconnect between the 2 systems in PTSD and, if observed, whether sucha disconnect was unique to PTSD. We found that in the total sample, dopamine,epinephrine, and norepinephrine are robustly intercorrelated; Spearman rankcorrelations ranged from 0.47 to 0.69, with a mean of 0.60. However, noneof the catecholamine levels had a significant correlation with cortisol levels(ρ = 0.04-0.08). We further examined correlations between catecholamineand cortisol levels within subgroups (ie, exposed to trauma without PTSD,no trauma exposure, PTSD, MDD, and PTSD plus MDD), but, with one exception,correlations were less than 0.15 and not significant. There was a low butsignificant correlation (0.20) between cortisol and epinephrine levels inthe group not exposed to trauma. We conclude there is no evidence of a disconnectbetween cortisol and catecholamine levels as a feature of PTSD.

The results for the 4 diagnostic groups, defined by history of PTSDand MDD, can be summarized as follows: (1) with respect to catecholamine levels,we found significantly higher mean levels of dopamine, epinephrine, and norepinephrinein persons with lifetime PTSD (with and without comorbid MDD), compared withpersons with MDD alone or neither disorder; (2) with respect to cortisol levels,we found no distinct pattern associated with PTSD in either sex and no differencewas detected between past and current PTSD; (3) a finding incidental to thePTSD-cortisol relationship was a significantly higher mean cortisol levelin women with lifetime MDD comorbid with PTSD, compared with women with neitherdisorder or with either disorder alone and a significantly higher mean cortisollevel in men with lifetime MDD alone, compared with men with neither disorderor with both disorders. Our analysis provides no support for the hypothesisthat PTSD is associated with lower UFC levels.³ Personswith PTSD did not show lower UFC levels than persons with no history of traumaor than persons with history of trauma but not of PTSD; neither did they showlower UFC levels than persons with a history of MDD or persons with neitherdisorder

Higher mean catecholamine levels have been reported in anxiety disorders,among them PTSD.^{15 ,30 - 32} Thenoradrengergic and dopaminergic systems are implicated in arousal disturbance,³¹ which is one of the hallmark symptoms of PTSD. Ourfindings of higher catecholamine levels in men and women with lifetime PTSD(alone or comorbid with MDD) are consistent with these reports. They extendthe previous reports to lifetime PTSD in the general population.

Higher mean cortisol levels have been observed in patients with MDDcompared with healthy control subjects.^{33 - 35} Ourfinding of a higher mean urinary cortisol level in women with lifetime MDDcomorbid with PTSD and in men with lifetime MDD alone is consistent with thesereports. This finding in lifetime MDD is also consistent with a previous reportthat demonstrated higher mean saliva cortisol levels in women with past MDDcompared with women with no history of MDD.³⁶

Our results, taken together, are consistent with neurobiologic modelsof anxiety and MDD, namely, the involvement of the sympathetic nervous systemin anxiety disorders and the involvement of the hypothalamic-pituitary-adrenalaxis in MDD. Evidence of sympathetic nervous system involvement in PTSD wassimilar in both sexes, whereas evidence of hypothalamic-pituitary-adrenalaxis activation in MDD varied by sex. In men, higher mean cortisol levelswere observed in cases with lifetime MDD alone but not in the small numberof cases with MDD comorbid with PTSD. In women, higher mean cortisol levelswere observed in comorbid cases but not in cases with lifetime MDD alone.Our results suggest that in women, MDD alone might be a less severe disorderthan MDD combined with PTSD. Previous studies have reported that MDD comorbidwith anxiety disorders is more severe than "pure" MDD in terms of depressivesymptoms, course of illness, and treatment response.^{37 - 42} Theinterpretation of the observed sex differences in urinary cortisol levelsis unclear.

Our findings on UFC levels in women with MDD comorbid with PTSD arein agreement with a study by Heim et al,⁴³ whichreported that depressed women with history of early trauma (11 of the 13 metcriteria for PTSD) demonstrated enhanced cortisol secretion in response toa stressor, whereas women with MDD without early trauma showed normal cortisolresponses, compared with healthy control subjects. A report by Rasmusson etal⁹ of women with childhood abuse and PTSDalso is consistent with these findings. The women with PTSD had either activeor past MDD and showed enhanced cortisol responses to corticotropin-releasingfactor and to corticotropin infusion, as well as a trend toward higher 24-hourUFC levels. A study by Lemieux and Coe⁷ ofwomen with childhood sexual abuse found higher 24-hour UFC levels in womenwith PTSD than with no PTSD. While the researchers did not report on comorbidMDD, they indicated that 5 of the 11 women with PTSD were taking antidepressants;the possibility that others may have had a history of depression was not ruledout. Our study and these previous studies converge in suggesting that alterationsin hypothalamic-pituitary-adrenal axis, specifically a higher level of UFC,might be observed in women with MDD comorbid with PTSD. These studies highlightthe importance of diagnosing both MDD and PTSD in research focusing on theneurobiology of either disorder.

An effect of exposure to trauma per se on urinary catecholamine levelswas observed in both sexes. However, instead of a stress-like effect (ie,higher levels of these stress hormones), we observed lower levels in the exposedto trauma without PTSD group than the no exposure group. It is possible thatthe long-term adaptation to trauma exposure is associated with lower catecholaminelevels. An alternative explanation is that exposure to trauma segregated individualsinto a susceptible group (PTSD) and a nonsusceptible group (exposed to traumawithout PTSD), who had before differed on baseline catecholamine levels beforeexposure to trauma. The nonexposed group (which showed higher mean catecholaminelevels than the exposed group) might include a subset of persons who havea susceptibility to the PTSD-inducing effects of trauma but have never beenexposed. A further possibility is that the subset of unexposed persons witha susceptibility to PTSD have a higher level of basal catecholamines. Thepresence of these persons would push upward the average level of catecholaminesof the unexposed group as a whole. Evidence that preexisting anxiety disorders,which would be accompanied by increased secretion of catecholamines,^{30 - 31} predict PTSD^{19 ,44} supportsthis interpretation.

Four limitations in this study are noted. First, the 66.5% participationrate in the biologic studies, although high, considering respondents' burden,limits the generalizability of the results. However, the exceptionally highfollow-up participation in the epidemiological study across multiple assessmentsallows us to evaluate and confirm the representativeness of the biologic sampleon sociodemographic characteristics and history of substance use disorders.Differences on unmeasured variables cannot be ruled out. Second, althoughwe found no difference between current and past PTSD, the small number ofcases with current PTSD limits our ability to examine this question more definitively.Third, the small number of men with PTSD, either "pure" or comorbid with MDD,limits the reliability of our findings in men. The possibility of a sex differencein UFC levels merits further attention.

The fourth limitation is the use of the NIMH-DIS at baseline and acrossall follow-up assessments rather than a clinical assessment. However, theNIMH-DIS has been found to be highly specific but more conservative than aclinician-administered instrument.²⁴ Thus,cases designated as PTSD are "true" cases, although it is possible that somecases of PTSD undetected by the NIMH-DIS were included in the exposed to traumawithout PTSD group. However, the availability of persons who were never exposed,according to baseline and multiple follow-up reassessments, addresses theconcern that any misclassification of cases of PTSD might have obscured potentialdifferences in cortisol levels associated with PTSD. The analyses comparingPTSD with the 2 control groups (exposed to trauma without PTSD and no exposure)showed no significant differences with regards to cortisol levels betweenPTSD and either of these control groups. In the analyses of comorbidity, theneither-disorder group (no PTSD/no MDD) was separated into (1) exposed totrauma without PTSD and (2) no exposure. No differences in cortisol levelswere observed in a comparison that included only the no exposure group asreference from what was reported when the total group with neither disorderwas used.

Important strengths of the study deserve mention. The most importantstrength is the representative community sample from which the data came.Previous studies on the neurobiology of PTSD in nonveteran populations haveused clinical samples or samples of volunteers, which undoubtedly are biasedin terms of severity of psychopathology, other clinical features, and socialfactors that influence self-selection to treatment and clinical research.In sharp contrast, our neurobiologic study is nested in a longitudinal epidemiologicstudy, with multiple psychiatric assessments during a 10-year period and anexceptionally high follow-up completion rate of the initial cohort. A comparisonof the total sample with the subset on whom data on 32-hour urine samplesare available indicates that we have been successful in achieving a representativesample of the total cohort. The study thus offers an opportunity for an unbiasedevaluation of neurobiologic factors in PTSD. The large sample size is anotherimportant strength. We also introduced a methodologic advance over previousstudies that used total 24-hour urinary hormone secretion levels. It involvedcollecting urine in 8-hour increments linked to the wake-sleep cycle. Thismethod revealed an overall circadian rhythm in cortisol and catecholaminelevels. By taking into account the data from night, AM, and PM, GEE analysisoffered enhanced statistical power for detecting differences across diagnosticgroups, which were obscured in the analysis of total 24-hour UFC levels.

Corresponding author: Elizabeth A. Young, MD, Mental Health ResearchInstitute, University of Michigan, 205 Zina Pitcher Pl, Ann Arbor, MI 48109-0720(e-mail: eayoung@umich.edu).

Submitted for publication February 3, 2003; final revision receivedNovember 10, 2003; accepted November 24, 2003.

This study was supported by grants MH 48802 (Dr Breslau) and MH 01931(Dr Young) from the National Institutes of Health, Bethesda, Md.

We would like to acknowledge the statistical assistance of Lonni Schultz,PhD, and Alissa Kapke, MA.