Journal of Affective Disorders 118 (2009) 166–172

Contents lists available at ScienceDirect

Journal of Affective Disorders

j ourna l homepage: www.e lsev ie r.com/ locate / j ad

Research report

A longitudinal analysis of alcohol consumption and the risk ofposttraumatic symptoms

A.C. McFarlane a,⁎, D. Browne a, R.A. Bryant b, M. O'Donnell c, D. Silove d, M. Creamer c, K. Horsley a

a Centre for Military and Veterans' Health, University of Adelaide, South Australia, Australiab School of Psychology, University of New South Wales, NSW, Australiac Australian Centre For Posttraumatic Mental Health and Department of Psychiatry, University of Melbourne, Victoria, Australiad School of Psychiatry, University of New South Wales, NSW, Australia

a r t i c l e i n f o

⁎ Corresponding author. CMVH Level 2/122 FromAustralia, Australia. Tel.: +61 08 83035200.

E-mail address: alexander.mcfarlane@adelaide.edu

0165-0327/$ – see front matter © 2009 Elsevier B.V.doi:10.1016/j.jad.2009.01.017

a b s t r a c t

Article history:Received 10 September 2008Received in revised form 21 January 2009Accepted 21 January 2009Available online 23 February 2009

Background: Previous studies investigating the impact of alcohol ingestion on the emergence ofposttraumatic psychological symptoms have generated contradictory findings.Methods: One thousand forty-five patients, admitted to hospital following traumatic injurywere assessed during hospitalisation for patterns of alcohol consumption prior to the injury andalso during the month prior to reassessment at 3 months. Anxiety, depression andposttraumatic stress disorder (PTSD) were assessed post accident and at 3 months. In a subsample (n=167), blood alcohol levels were measured at the time of admission to emergencydepartments.Results: Moderate alcohol consumption prior to and following the accident predicted lowerlevels of psychological distress at 1 week and 3 months. No significant relationship was foundbetween the blood alcohol level and psychiatric outcomes. PTSD predicted the emergence ofalcohol abuse following the accident, suggesting self-medication in a subgroup of survivors.Limitations: The impact of alcohol consumption upon injury severity and the nature of injurywas not controlled for and some non-participation may have been related to patterns of alcoholconsumption. We relied on retrospective reports of alcohol use obtained shortly after thetraumatic injury to index prior alcohol use and these reports may have been influenced bymood states at the time of recall. Our follow-up was limited to 3 months and there is a need forlonger-term assessment of the relationship between prior alcohol use and subsequentposttraumatic adjustment.Conclusion: Given the potential impact of alcohol use on traumatic injury and post-injuryrecovery, we advocate active screening and early intervention strategies that focus onmoderatealcohol usage.

© 2009 Elsevier B.V. All rights reserved.

Keywords:Posttraumatic stress disorderAlcohol

There has been a long-standing interest in the influence ofalcohol use on the development of psychiatric disorders aftertrauma. During WWI Sir Frederick Mott and Dr Edith Green(Mott, 1919), working at the Maudsley Hospital, comparedalcohol consumption and use of the rum ration in a group ofpatients with shell shock or war neurosis with surgicalpatients that had neither shell shock nor war neurosis,

e St Adelaide, South

.au (A.C. McFarlane).

All rights reserved.

making their work an early example of the case–controlmethod. A re-analysis of these data found that there was astatistically protective effect for the consumption of alcohol(OR .37 95% CI .18, .77, pb .001) (Horsley et al., 2006). Using asimilar methodology, at the same time, Wolfsohn (1918)found fewer non-drinkers amongst soldiers who were beingtreated for physical wounds than amongst soldiers beingtreated for psychoneurosis.

More recently, a series of studies have reported conflictingevidence about the role of alcohol use on the emergence ofposttraumatic symptomatology. Several studies have indexed

167A.C. McFarlane et al. / Journal of Affective Disorders 118 (2009) 166–172

the influence of alcohol intoxication at the time of the traumaby calculating the relationship between blood alcohol level(BAL) and subsequent PTSD development. BAL has beenlinked to decreased rates of PTSD (Mellman et al., 1998),increased rates of PTSD (Richmond and Kauder, 2000), or nochange in PTSD rates (Zatzick et al., 2002). One study thatassessed alcohol use by self-report found that it was linked toreduced PTSD (Maes et al., 2001). Some studies have alsoinvestigated the role of alcohol use in the period prior totrauma exposure and found no relationship between alcoholuse and subsequent PTSD (Mason, 2006).

There are several methodological issues that may con-tribute to these disparate findings. First, most of these studiesinvolved small sample sizes (Mellman et al., 1998; Zatzick etal., 2002; Mason, 2006). Second, some studies retrospectivelyassessed alcohol consumption many months after the trauma(Roy-Byrne et al., 2004; Peele and Brodsky, 2000). Third,studies that used self-report did not employ validatedmeasures of alcohol use (Roy-Byrne et al., 2004) and othersdefined intoxication as the presence of any alcohol on a bloodscreen (Zatzick et al., 2002; Mason, 2006). Fourth, theconsumption of alcohol is directly related to mental healthstatus, with a number of studies suggesting that individualswho do not drink at all or regularly drink excessively are lesspsychologically healthy than individuals who are moderatedrinkers (Peele and Brodsky, 2000). Therefore, abstinence orextreme alcohol consumption may be a proxy marker in it'sown right for high risk to mental disorder, hence potentiallyexplaining the apparent link between alcohol and post-traumatic morbidity. As a consequence, previous studies thathave analysed the relationship between alcohol use and PTSDsymptoms in a linear manner may not have identified askewed relationship.

There are several potential mechanisms that could explaina relationship between alcohol use and symptom develop-ment. PTSD symptoms and increased alcohol abuse maybelinked via the mechanism of self-medication (Stewart et al.,1998; McFarlane, 1998; Jacobsen et al., 2001). As a conse-quence, high levels of consumption in trauma-exposedindividuals can be indicative of prior trauma (Fear et al.,2007) or prior psychiatric disorder, which are recognised riskfactors for PTSD (Brewin et al., 2000). This possibilityunderscores the need to assess alcohol use prior to traumaexposure, and to control for the potential confoundinginfluence of prior psychiatric disorder. Another possibility isthat alcohol use at the time of the trauma may have aprotective pharmacological effect. Alcohol attenuates nora-drenergic activation in the amygdala and locus coeruleus(Shirao et al., 1988), which underpin the fear conditioning oftrauma memories. Alcohol also modifies the N-methyl-D-aspartate (NMDA)-systemwhich also plays a role in the stressresponse (Grillon et al., 1996). It has been hypothesised thatalcohol may provide some protective function by modulatingthe neurobiological systems in the immediate aftermath ofthe trauma, thereby reducing the likelihood of subsequentPTSD. Accordingly, it is important to consider both alcohol useprior to, at the time of the traumatic event, and in theaftermath.

This study investigated the role of alcohol consumption asa risk or protective factor in the development of psychologicalsymptoms, and in particular PTSD, in a large sample of

survivors of traumatic injury. We focused on the patterns ofalcohol consumption prior to the trauma and again in theperiod of time after the trauma because of the potential forthe pattern of drinking to change in a population that hassustained traumatic injury. We also investigated objectiveblood alcohol levels at the time of the accident to determinewhether intoxication at the time of the trauma protectsagainst developing PTSD. We also assessed prior and currentpsychiatric disorder to disentangle the potential confoundbetween prior alcohol use and psychiatric disorder ascontributors to PTSD development. On the basis of priorevidence that abstinence and excessive alcohol use areassociated with psychological dysfunction (Alati et al., 2005;Rodgers et al., 2000), we hypothesised that no or excessivealcohol use prior to the trauma, rather than moderate alcoholuse, would be associated with PTSD symptom development.Secondly, we proposed that this relationship would remainwhen prior psychiatric history was controlled for, indicating aprotective effect of moderate alcohol consumption. In asmaller subgroup, where there was a legal requirement tohave blood assessed following the motor vehicle accident, itwas hypothesised that PTSD symptoms would be less in thosewho had a detectable blood alcohol level in the immediateaftermath of the accident.

1. Method

1.1. Study design

Participants in this study were recruited from one of fourhospitals in three states of Australia. Patients were included inthe study if they had experienced a traumatic injury thatrequired a hospital admission of greater than 24 h; no braininjury ormild traumatic brain injury (mTBI): (mTBI: defined asa loss of consciousness of approximately 30 min or less, aGlasgow Coma Scale (GCS) score of 13–15 after 30 min, orposttraumatic amnesia (PTA) not greater that 24 h: AmericanCongress of RehabilitationMedicine,1993);were aged between16 and 70 years of age; and had a reasonable comprehension ofEnglish to complete the assessment. Patients were excludedfrom the study if they were currently suicidal or psychotic.

Weekday trauma service admissions were randomlyselected using an automated, random assignment procedure,stratified by length of stay. Patients were recruited over a 22-month period (April 2004 to February 2006) during which3771 met inclusion criteria. A total of 1593 were randomisedinto the study and 1157 (73%) consented to participate in thestudy. At 3 months, 988 (85%) participants completed thefollow-up assessment.

1.2. Participants

The majority of participants were male (n=837, 72%) withan average age of 37.93 years (SD=13.63). Forty two percent ofparticipants experiencedanmTBI (n=488)and themean injuryseverity score (ISS) (Baker et al., 1974) was 11.02 (SD=8.13),which is in the moderate severity range. Participants spent anaverage of 12.34 (SD=12.87) days in hospital and 14% ofparticipants had an intensive care unit (ICU) admission(n=157). The principal mechanism of injury was transportaccidents (63%, n=734); 15% experienced falls (n=168), 6%

168 A.C. McFarlane et al. / Journal of Affective Disorders 118 (2009) 166–172

were assaults (n=70; including stabbings and firearm), 7%experienced work related accidents not specified in the abovecategories (n=77), and 8% were other injuries (n=87). Themajority of participants were discharged home (77%, n=895),with the remainder being discharged to a rehabilitation facility.

At the 3-month follow-up assessment, 136 patients couldnot be contacted or declined to participate; 990 wereinterviewed by telephone, representing 86% of the initialsample. Patients at the follow-up assessment did not differfrom thosewho did not participate in terms of gender (c2=.13,df=1, p=.71), length of hospital admission (t(1151)=1.52,p=.13), or injury severity score (t(1112)=1.30, p=.19).Participants lost to follow-up were younger than those whowere retained (35.71±13.34 vs 38.25±13.63) t(1143)=2.26,pb .05). Individuals who refused to participate in the currentstudy did not differ from participators in gender (c2=1.50,df=1, ns), length of hospital admission (t(1571)=.92, ns), ISS(t(1571)=1.46, ns), or mechanism of injury (c2=1.01, df=1,ns). Refusers were younger than participants (mean=35.18±14.17 vs 37.89±13.67, t(1561)=3.44, p=.001).

1.3. Measures

1.3.1. Psychiatric historyMini International Neuropsychiatric Interview (MINI)

Version 5.5 (Sheehan et al., 1998) was used to assess pastpsychopathology. The MINI is a short, structured diagnosticclinical interview based on the DSM-IV and the ICD-10classification of mental illness, especially designed to beused in clinical trials and epidemiological studies. The MINIconsists of a set of screening questions andmodules; modulesare administered if a patient responds positively to thescreening question. In the current study we used the MINI toidentify a past history of major depressive episode (MDE),dysthymia, panic disorder, agoraphobia without panic, socialphobia, obsessive compulsive disorder, PTSD, generalisedanxiety disorder, alcohol abuse and dependence, and mar-ijuana abuse and dependence. When MINI diagnosis werecompared to the CIDI all but simple phobia and GAD hadkappa values above .50. Sensitivity is above .70 for alldiagnoses except panic and agoraphobia and specificity ishigher than .70 for all diagnoses (Sheehan et al., 1998).

1.3.2. Acute anxiety and depressionThe Hospital Anxiety and Depression Scale (HADS)

(Zigmond and Snaith, 1983) was used to assess acute levelsof anxiety and depression symptoms in the acute hospitalenvironment. The HADS is a particularly useful tool withinjured populations for it does not measure somatic symp-toms. Both the depression and anxiety subscales were foundto be internally consistent with values of Cronbach's coeffi-cient (a) being .76 and .80 respectively (Mykletun et al.,2001). It has excellent discriminant validity and internalconsistency and a stable factor structure (Bjelland et al.,2002).

1.3.3. Injury characteristicsInjury characteristics were obtained frommedical records.

These included ISS, presence of an mTBI, length of hospitalstay and number of days in ICU.

1.3.4. Posttraumatic stress symptomsThe Clinician Administered PTSD Scale (CAPS) (Blake et

al., 1998) was used to assess PTSD symptom severity postinjury. The CAPS was specifically anchored to the event inwhich the individual was injured so the PTSD diagnosed wasspecific to the injury event. The CAPS is one of the mostwidely used tools for diagnosing PTSD and measuring PTSDseverity, and has excellent reliability and validity (Blake et al.,1998; Weathers et al., 2001). When compared to the SCID theinternal consistency for the severity scores ranged from .85 to.87 (Blake et al., 1995). In line with common practice(O'Donnell et al., 2003; Schnyder et al., 2001), item 8 of theCAPS (“Inability to recall an important aspect of the trauma”)was excluded from the scoring process of PTSD due to highlevels of mTBI in the sample, and the difficulty differentiatingorganic from psychogenic amnesia (O'Donnell et al., 2003;Huskisson, 1974). Severity was computed by summing thefrequency and intensity scores across the remaining 16symptoms. Participants were deemed to have subsyndromalPTSD if they met criteria for the re-experiencing symptomcluster and either the avoidance or arousal symptom clusters.

1.3.5. AUDITThe AUDIT was administered during the acute hospital

admission to measure alcohol use prior to the injury, andreadministered at 3 months. The AUDITcore questionnaire wasdeveloped by the World Health Organization (WHO) for theidentification of currently active, hazardous or harmful alcoholuse (Babor et al., 1989). The AUDIT has high test–retestreliability (Selin, 2003), and performs similar to or better thanother self-report alcohol screening tests (Rumpf et al., 2002).The 10 items are scored according to their frequency ofoccurrence on a five-point scale (e.g., ‘never’ = 0, ‘daily’ = 4),and refer to the past 12months. This wasmodified to 3monthsin the 3month survey. TheAUDIT total score is derived from thesummation of three subscale scores, measuring hazardousdrinking (quantity and frequency of consumption), alcoholdependence (drinking indicative of addiction) and harmful orproblem drinking (consequences suggesting harm to self orothers). Total AUDIT score was used to split the cohort into 3groups; AUDITgroup1 comprised thosewith a score of 0 or 1, toobtain a score of 1 the maximum amount an individual canconsume is 2 drinks once a month; AUDIT group 2 comprisedthose who scored between 2 and 15 and AUDIT group 3comprised those who score 16 or more. The AUDIT guidelines(Babor et al.,1989) state that a score of greater than15 indicatesa high level of alcohol problems.

In analysing subscale scores, participants were grouped intorisk categories according to the AUDIT Guidelines. For con-sumption, participants who scored zero or one formed theminimal/abstainer group. Participants who scored between 2and 5 (females) or 6 (males) formed the moderate drinkinggroup. Consumption in this range is indicative of a moderaterisk of harm. Those who scored over 5 (females) or 6 (males)which indicates a high risk of harm, formed the highconsumption group. Any score on the ‘Dependence’ and‘Problem’ subscales indicates the ‘presence or incipience ofdependence… (and)…that alcohol related harm is alreadybeing experienced’ (Babor et al., 1989). On this basis Depen-dence and Problem groups were comprised of those with ascore of zero vs those with any score.

Table 1Acute PTSD, anxiety and depression scores.

AUDIT consumption score ANOVA

Outcomevariable

0–1 M (SE)N=208

2–5/6 M (SE)N=533

N 5/6 M (SE)N=281

F

CAPS 19.41 (.996) 16.4 (.66) 17.82 (.92) 3.47 .032HADS-a 6.28 (.27) 5.28 (.18) 5.87 (.25) 5.8 .003HADS-d 5.53 (.27) 5.03 (.18) 5.63 (.25) 2.71 ns

Dependence subscale score ANOVA

Outcomevariable

N0 M (SE)N=242

0 M (SE)N=780

F

CAPS 16.71 (.56) 20.07 (.98) 9.78 .002HADS-a 5.44 (.15) 6.43 (.26) 11.81 .001HADS-d 5.14 (.15) 5.85 (.26) 6.27 .012

Problem subscale score ANOVA

Outcomevariable

N0 M (SE)N=349

0 M (SE)N=673

F

CAPS 15.71 (.59) 19.09 (.83) 6.12 .014HADS-a 5.43 (.16) 6.14 (.22) 7.52 .006HADS-d 5.16 (.16) 5.59 (.22) 2.79 ns

Note. M = Estimated Marginal Means.ns = Non significant.

169A.C. McFarlane et al. / Journal of Affective Disorders 118 (2009) 166–172

In the current study, the ‘standard drink’ referred to in thefirst and second items of the AUDIT, was defined by Australianguidelines (National Health and Medical Research Council,2001) as being one containing 10 g or 12.5 ml (.44 fluidounces) of pure alcohol which contrasts to the USA standarddrink of .6 fluid ounces.

1.3.6. Blood alcohol samplesThe state of South Australia has legislation that requires

medical practitioners to take a sample of blood from anypatient aged 14 years or more who is injured in a vehicleaccident. This sample is taken as soon as practically possiblyafter admission, usually in the emergency department.Wherever possible in the current study, participants BALwas obtained from pathology reports with the appropriateconsents.

1.4. Procedure

This research project was approved by the Royal AdelaideHospital and University of Adelaide Human Research EthicsCommittees and carried out in accordance with the currentversion of the Declaration of Helsinki. Informed consent wasobtained from all participants.

Randomised participants were assessed prior to dischargefrom each trauma centre, an average of 7.22 days (SD=9.64,median 4, range 98) after injury. In the acute assessment theycompleted the CAPS, MINI, HADS and AUDIT (assessing pre-trauma alcohol use) scales. Injury characteristics wereobtained from medical records. At 3 months the CAPS wasconducted via a telephone assessment which has been shownto be a valid administration strategy (Aziz and Kenford, 2004).Self-report questionnaires containing the HADS and AUDITwere administered in a booklet sent to participants andreturned by post. Study interviewers were honour psychologygraduates and were trained in the study protocol by a clinicalpsychologist (MOD). All assessments were digitally audiorecorded to ensure ongoing adherence to the protocol. Fivepercent of all CAPS interviews were re-scored blind to theoriginal scoring to test inter-rater reliability. Overall, the PTSDdiagnostic consistency for the CAPS was 1.00 at 3 months.

1.5. Analysis

Analysis of covariance was used to examine groupdifferences in the AUDIT subscales, controlling for lifetimehistory of any depressive or anxiety disorder as measured bythe MINI. Pair-wise group comparisons were made betweenthe low, moderate and high consumption groups.

2. Results

2.1. Posttraumatic stress symptoms

Participant's CAPS scores at the first assessment rangedfrom 0 to 92 (m 15.63, SD 15.97) with 39 (3.4%) participantsmeeting DSM IV criteria for PTSD excluding the posttraumaticamnesia item and criterion E (duration more than onemonth) and a further 142 (12.3%) meeting criteria for sub-syndromal PTSD. At the 3 month assessment CAPS scoresranged from 0 to 116 (m 18.12, SD 20.6) with 81 (8.2%)

meeting criteria for PTSD and a further 117 (11.9%) meetingcriteria for subsyndromal PTSD.

2.2. Prior alcohol use and subsequent symptoms

Participants' scores on the AUDIT were used to examinethe effect of alcohol use in the 12 months prior to the injuryon PTSD symptoms following trauma.

There was a significant main effect for group on the acutePTSD severity score (CAPS, F(2, 1040)=11, pb .001, anxietyscore (HADS-a, F(2,1044)=19, pb .001) and depression score(HADS-d, F(2, 1044)=8.5, pb .001)). Post hoc comparisons(alpha=.01) showed moderate drinkers as compared tominimum or high consumers in the year prior to injury scoredsignificantly lower on acute PTSD severity and acute anxietyseverity, and significantly lower than those drinking a highamount on acute depression severity.

To further explore the relationship, the cohort was splitaccording to scores on each of the three AUDIT subscales;‘Consumption’, ‘Dependence’ and ‘Problem’. Analysis of variancewas used to examine group differences in the subscales, control-ling for gender and lifetime history of any depressive or anxietydisorder as measures by the MINI. Estimated Marginal MeanPTSD, anxietyanddepressionseverity scores foreach subscale areshown in Table 1. Significant main effects were found betweengroups on each subscale with the exception of depressionseverity scores in theConsumptionandProblemdrinkinggroups.

Post hoc comparisons using Scheffe's procedure showedthe moderate consumption group had significantly lowerPTSD and anxiety severity scores than the minimal consump-tion/abstainer group. Participants who recorded a positivescore on the Dependence subscale had significantly higherPTSD, anxiety and depression severity scores than those whoscored zero on the Dependence subscale, while participantswho recorded a positive score on the Problem subscale hadsignificantly higher PTSD and anxiety severity scores thanthose who scored zero on the Problem subscale.

Table 3Changes in consumption over time.

Change in consumption ANOVA

Outcomevariable

Stable M (SE)N=265

Increase M(SE) N=191

Decrease M(SE) N=368

F Sig

CAPS 18.45 (1.12) 19.48 (1.36) 20.07 (1) .63 nsHADS-a 4.76 (.24) 4.91 (.29) 5.4 (.21) 2.44 nsHADS-d 5.86 (.26) 6.88 (.31) 6.21 (.23) 3.37 .035

Note. M = Estimated Marginal Means.ns = Non significant.

170 A.C. McFarlane et al. / Journal of Affective Disorders 118 (2009) 166–172

2.3. 3 months assessment

Participant's alcohol use in the last 3 months since thetrauma was assessed using the AUDIT. Mean PTSD, anxietyand depression severity scores for the each subscale areshown in Table 2. After controlling for gender and psychiatrichistory, significant main effects were found between groupson each subscale with the exception of PTSD severity in theconsumption groups.

Post hoc comparisons using Scheffe's procedure showedthe moderate consumption group had lower PTSD, anxietyand depression severity scores than the minimal consump-tion/abstainer group. Participants who recorded a positivescore on the Dependence subscale had significantly higherPTSD, anxiety and depression severity scores than those whoscored zero. Similarly participants who recorded a positivescore on the Problem subscale had higher PTSD, anxiety anddepression severity scores than those who scored zero.

2.3.1. Blood alcohol level and PTSDA sub-sample of 167 individuals had BAL measured at the

time of their presentation to emergency departments. Whilethere was a trend towards lower acute PTSD, anxiety anddepression scores for those with a BAL between zero and .05,the only significant finding was that those with a BAL betweenzero and .05 scored significantly lower on acute PTSD severitythan those with a BAL of greater than .05. The relationshipbetween BAL at admission and three months PTSD, depressionand anxiety scores showed no significant effects.

2.3.2. PTSD developmentThere was no difference between the participants who did

not develop PTSD at 3 months post-injury (n=906) and thosewho did develop PTSD (n=81) on anymeasures of alcohol usebefore their injury (AUDIT total score (6.33±5.68 vs 5.99±6.38)). However, participants who developed PTSD at3 months had significantly higher post injury ‘problem

Table 23-Month PTSD, anxiety and depression scores.

Consumption ANOVA

Outcomevariable

0–1 M (SE)N=237

2–5/6 M (SE)N=427

N5/6 M (SE)N=189

F Sig

CAPS 22.19 (1.18) 19.12 (.95) 19.34 (1.4) 2.3 nsHADS-a 6.92 (.27) 5.91 (.22) 6.46 (.32) 4.79 .009HADS-d 6.01 (.25) 4.67 (.2) 5.18 (.29) 9.46 b .001

Dependence subscale score

Outcomevariable

N0 M (SE)N=122

0 M (SE)N=732

CAPS 19.56 (.74) 23.94 (1.7) 5.99 .015HADS-a 6.1 (.17) 7.9 (.38) 19.98 b .001HADS-d 5.06 (.16) 6.13 (.36) 8.25 .004

Problem subscale score

Outcomevariable

N0 M (SE)N=188

0 M (SE)N=666

CAPS 19.43 (.78) 22.74 (1.39) 4.7 .03HADS-a 6.01 (.18) 7.51 (.31) 19.01 b .001HADS-d 5.06 (.16) 5.7 (.29) 3.99 .046

Note. M = Estimated Marginal Means.ns = Non significant.

drinking’ scores (3-month AUDIT problemscore (.63±1.69vs1.16±2.37), t(205.5)=−2.73, p=.007). There was no differ-ence betweenparticipants who developed 3-month PTSD on 3-month total AUDIT scores, 3-month AUDIT consumption scoresor 3 month AUDIT dependency scores.

Participants were grouped according to changes in reportedalcohol consumption between initial and 3-month assessments(Table 3). There was a significant group effect on depressionseverity scores (HADS-d F(2,824)=3.37, p=.035). The overalltrend indicated that participants whose alcohol consumptionremained stable between acute and 3 month assessment, hadlower PTSD, anxiety and depression severity scores than thosewhose consumption increased or decreased.

3. Discussion

In general, the group whose alcohol consumption prior totrauma was in the moderate range experienced less sympto-matic distress than the groups who abstained or those whowere drinking in the at-risk range. This patternwas observed inthe acute phase after the accident and at 3 months, at whichtime the moderate drinkers had significantly lower levels ofPTSD symptoms than either the minimal consumption orproblematic drinking participants, similar to the findings ofMaes et al. (2001). Importantly, this effect remained aftercontrolling for the influence of gender and prior psychiatricdisorder. A similar pattern existed for anxiety and depression inthe acute phase and at 3 months. However, alcohol consump-tion prior to the accident did not predict the development ofPTSD as a diagnostic entity, in contrast to symptom severity.

As the total AUDIT score is a combined measure of levels ofconsumption, dependence and problem drinking, it wasdecided to test whether the apparent negative effects of thoserecording either high or low scores was evident across allsubscales. Minimal consumption and abstinence were asso-ciated with a higher level of acute symptoms compared withmoderated consumption. While at 3 months minimal con-sumption and abstinencewere associatedwith a higher level ofPTSD symptoms compared with bothmoderate and high levelsof consumption. In comparison, dependence and problemdrinking had a general association with more symptoms,suggesting that the pattern of alcohol consumption is a riskfactor for greater levels of PTSD symptoms when it wasassociated with dependence. However, when the pre-traumaand post injury levels of consumption are examined, these datasuggest that theremaybe adirect pharmacological benefit fromalcohol consumption, even at unsafe levels of consumption, aslong as it is not associated with problems or dependence asdefined by the AUDIT.

171A.C. McFarlane et al. / Journal of Affective Disorders 118 (2009) 166–172

The finding that higher levels of PTSD symptoms at threemonths were associated with both increased and decreasedconsumption in the aftermath of traumaexposure suggests thatthere is no unidirectional relationship with PTSD symptomsand alcohol consumption. The presence of a sub-groupsuffering from PTSD symptomatology who increased theiralcohol consumption is consistent with the self-medicationhypothesis. As has been previously reported, we found moreindividuals decreased than increased their consumption fol-lowing their injury. A longer period of follow up is required tosee if there is a differential outcome, between those groupswhoincrease and decrease their alcohol consumption.

No effectwas foundbetweenblood alcohol at the timeof theinitial hospital admission and symptomsat 3months. Thereforethese data did not replicate previous studies which suggested aprotective effect of acute alcohol consumption (Wolfsohn,1918;Maes et al., 200; Mellman et al., 1998) although there was atrend for less acute distress in this group. This finding does notsupport the proposition that alcohol in the immediate wake oftrauma exposure limits fear conditioning by its potential toattenuate noradrenergic activation (Shirao et al., 1988).Additionally we found no support for the adverse effects, assuggested by Zatzick et al. (2002) and Roy-Byrne et al. (2004),whichmay be due to their linking of the use of illicit drugs andalcohol, an association which we did not examine.

The finding that moderate alcohol use was associated withbetter outcome may be interpreted in two ways. First, it maybe argued the moderate alcohol use is protective againstdeveloping PTSD. This interpretation is consistent withevidence that alcohol attenuates noradrenergic activation inthe amygdala and locus coeruleus (Shirao et al., 1988), whichcontribute to fear conditioning of traumamemories. Althoughwe found no protective function of BAL, reflecting alcohollevels at the time of the traumatic injury, it is possible thatmoderate levels of alcohol consumption in the period afterthe traumatic injury served a function in limiting the extent towhich fear conditioning occurred in the aftermath of theinitial injury. A second interpretation is that instead ofmoderate drinking protecting individuals from PTSD, minimaland excessive alcohol consumption patterns represent riskfactors for PTSD development. That is, excessive drinking maycontribute to further problems, impede adequate processingof trauma memories, or reflect an underlying avoidancepattern that impedes optimal recovery from a traumaticexperience. Minimal drinking may also be a risk factor forPTSD because it deprives an individual from engaging in abehaviour that affords some form of self-medication thatassists coping in the aftermath of trauma. Additionally, it ispossible that both minimal and excessive drinking may beassociated with other unmeasured variables that pose risk forPTSD. One such possibility is that moderate drinking may beindicative of a coping style that is indicative of a capacity toself-regulate distress and high levels of impulse control. Theseassociated coping processes require due consideration andattention in the further elucidation of the relationshipsbetween patterns of symptoms and alcohol use.

3.1. Strengths and limitations

We note several limitations to this study. First, while theparticipation rate of 75% and subsequent follow up was high,

the possibility cannot be excluded that some non-participa-tion was related to patterns of alcohol intake. Second alcoholconsumption could impact upon injury severity and thenature of injury, due to the known effects of alcohol in causingMVA's, accidents and interpersonal violence (Milne et al.,2007). However therewere no significant differences in termsof injury severity, between those with different patterns ofdrinking prior to the injury. Third, BAL could only be obtainedin two hospitals from motor accident due to differences instate legislation. Fourth, we relied on retrospective reports ofalcohol use obtained shortly after the traumatic injury toindex prior alcohol use; these reports may have beeninfluenced by mood states at the time of recall. Fifth, ourfollow-up was limited to 3 months and there is a need forlonger-term assessment of the relationship between prioralcohol use and subsequent posttraumatic adjustment. Finallywe do not know about the effect of alcohol related interven-tions in this sample and what role clinical interventionsplayed in modifying consumption patterns following theinjury. The data from the AUDIT need to be interpreted withreference to the fact that a standard drink varies betweennations, being larger in the USA than in Australia.

4. Implications

In summary, this study raises the possibility that moderateconsumption may have a protective role against the develop-mentofposttraumatic anxiety, depression andPTSD symptoms.The apparent J shaped relationship between alcohol use andsymptom development suggests that the inconsistent findingsin the previous literature arise in part from not anticipating thisnon-linear relationship. Perhaps strength is added to theargument about apparent beneficial impact of moderateconsumption because these were highly stressed individualsat risk of psychiatric disorder, in contrast to communityepidemiological studies in which no such population widerisk exists. Furthermore, these findings emphasise the impor-tance of differentiating the effects of typical patterns ofconsumption prior to trauma exposure, the BAL at the time ofthe trauma and drinking patterns in the recovery phase.

From a public health perspective, these data suggest thatpopulations who have been admitted to hospital due to injuryshould be screened with an instrument such as the AUDIT todetect the individuals with problem and dependent drinkingas being at risk of worse outcomes. Given the importance ofalcohol misuse, as a cause of traumatic injury, preventiveinterventions should be targeted at this population followingscreening (Bertholet et al., 2005). The emergence of increaseduse of harmful alcohol consumption in those with PTSDindicates the importance of screening and early preventiveinterventions with this group, given the prevalence of the co-morbidity of PTSD and alcohol abuse (Stewart et al., 1998;McFarlane, 1998; Jacobsen et al., 2001). Furthermore, ratherthan suggesting abstinence following exposure to traumaticevents the importance of moderate drinking should beemphasised, as this behaviour may have some benefit inminimising distress. Replication of these findings is impor-tant. The neurobiological mechanism of action of alcohol inminimising symptom development also warrants furtherinvestigation.

172 A.C. McFarlane et al. / Journal of Affective Disorders 118 (2009) 166–172

Role of funding sourceThis projectwas funded throughanAustralianGovernmentNationalHealth

andMedical ResearchCouncil programgrant; theNHMRChad no further role inthe study design; in the collection, analysis and interpretation of data; in thewriting of the report; and in the decision to submit the paper for publication.

Conflict of interest

All authors declare that they have no actual or potential conflict ofinterest in this research project.

References

Alati, R., Lawlor, D.A., Najman, J.M., Williams, G.M., Bor, W., O'Callaghan, M.,2005. Is there really a ‘J-shaped’ curve in the association between alcoholconsumption and symptoms of depression and anxiety? Findings from theMater-University Study of Pregnancy and its outcomes. Addiction 100 (5),643–651.

American Congress of Rehabilitation Medicine, 1993. Definition of mildtraumatic brain injury. J. Head Trauma Rehabil. 8 (3), 86–87.

Aziz, M., Kenford, S., 2004. Comparability of telephone and face-to-faceinterviews in assessing patients with posttraumatic stress disorder.J. Psychiatr. Pract. 10 (5), 307–313.

Babor, T., Fuente, J., Saunders, J., Grant, M., 1989. The Alcohol Use DisordersIdentification Test: Guidelines for Use in Primary Health Care. WorldHealth Organization, Geneva, Division of Mental Health.

Baker, S.P., O'Neil, B., Haddon,W., Long,W.B., 1974. The injury severity score: amethod for describing patients with multiple injuries and evaluatingemergency care. J. Trauma 14 (2), 187–196.

Bertholet, N., Daeppen, J.B., Wietlisbach, V., Fleming, M., Burnand, B., 2005.Reduction of alcohol consumption by brief alcohol intervention inprimary care: systematic review and meta-analysis. Arch. Intern. Med.165 (9), 986–995.

Bjelland, I., Dahl, A.A., Tandgen Haug, T., Neckelmann, D., 2002. The validity ofthe hospital and depression scale: an updated literature review.J. Psychosom. Res. 52, 69–77.

Blake, D.D., Weathers, F.W., Nagy, L.M., Kaloupek, D.G., Gusman, F.D., Charney,D.S., Keane, T.M., 1995. The development of a Clinician-AdministeredPTSD Scale. J. Trauma. Stress 8 (1), 75–90.

Blake, D.D.,Weathers, F.W.,Nagy, L.M., Kaloupek,D.G., Charney, D.S., Keane, T.M.,1998. Clinician-Administered PTSD Scale for DSM-IV. Boston, NationalCenter For Posttraumatic. Stress Disorder 1–19.

Brewin, C.R., Andrews, B., Valentine, J.D., 2000. Meta-analysis of risk factorsfor posttraumatic stress disorder in trauma-exposed adults. J. Consult.Clin. Psychol. 68 (5), 748–766.

Fear, N.T., Iversen, A., Meltzer, H., Workman, L., Hull, L., Greenberg, N., Barker, C.,Browne, T., Earnshaw, M., Horn, O., Jones, M., Murphy, D., Rona, R.J., Hotopf,M.,Wessely, S., 2007. Patternsof drinking in theUKArmedForces. Addiction102 (11), 1749–1759.

Grillon, C., Southwick, S.M., Charney, D.S., 1996. The psychobiological basis ofposttraumatic stress disorder. Mol. Psychiatry 1 (4), 278–297.

Horsley, K., Wilson, E., Hawthorne, G., 2006. The first known case–controlstudies: rum rations protects against war neurosis and positive familyhistory predisposes. Australian Society for Traumatic Stress Studies/Australian Centre for Posttraumatic Mental Health Conference. Adelaide,South Australia.

Huskisson, E.C., 1974. Measurement of pain. Lancet 4, 1127–1131.Jacobsen, L.K., Southwick, S.M., Kosten, T.R., 2001. Substance use disorders in

patients with posttraumatic stress disorder: a review of the literature.Am. J. Psychiatry 158 (8), 1184–1190.

Maes, M., Delmeire, L., Mylle, J., Altamura, C., 2001. Risk and preventivefactors of post-traumatic stress disorder (PTSD): alcohol consumptionand intoxication prior to a traumatic event diminishes the relative risk to

develop PTSD in response to that trauma. J. Affect. Disord. 63 (1–3),113–121.

Mason, S., 2006. Risk factors for psychological distress following injury. B. J. Clin.Psychol. 45 (2), 217–230.

McFarlane, A.C., 1998. Epidemiological evidence about the relationshipbetween PTSD and alcohol abuse: the nature of the association. Addict.Behav. 23 (6), 813–825.

Mellman, T.A., Ramos, J., David, D., Williams, L., Augenstein, J.S., 1998. Possibleinhibition of family PTSD symptoms by alcohol intoxication. Depress.Anxiety 7 (3), 145.

Milne, B., Bell, J., Lampropoulos, B., Towns, S., 2007. Alcohol, drugs andAustralian young people. Int. J. Adolesc. Med. Health 19 (3), 245–253.

Mott, F.W., 1919. War Neurosis and Shell Shock. Oxford Medical Publications.Mykletun, A., Stordal, E., Dahl, A.A., 2001. Hospital Anxiety and Depression

(HAD) scale: factor structure, item analyses and internal consistency in alarge population. Br. J. Psychiatry 179, 540–544.

National Health and Medical Research Council, 2001. Australian AlcoholGuidelines: Health Risks and Benefits. Commonwealth of Australia,Canberra.

O'Donnell, M.L., Creamer, M., Bryant, R.A., Schnyder, U., Shalev, A., 2003.Posttraumatic disorders following injury: an empirical and methodolo-gical review. Clin. Psychol. Rev. 23 (4), 587–603.

Peele, S., Brodsky, A., 2000. Exploring psychological benefits associated withmoderate alcohol use: a necessary corrective to assessments of drinkingoutcomes? Drug Alcohol Depend. 60 (3), 221–247.

Richmond, T.S., Kauder, D., 2000. Predictors of psychological distressfollowing serious injury. J. Trauma. Stress 13 (4), 681–692.

Rodgers, B., Korten, A.E., Jorm, A.F., Jacomb, P.A., Christensen, H., Henderson,A.S., 2000. Non-linear relationships in associations of depression andanxiety with alcohol use. Psychol. Med. 30 (2), 421–432.

Roy-Byrne, P.P., Russo, J., Michelson, E., Zatzick, D., Pitman, R.K., Berliner, L.,2004. Risk factors and outcome in ambulatory assault victims presentingto the acute emergency department setting: implications for secondaryprevention studies in PTSD. Depress. Anxiety 19 (2), 77–84.

Rumpf, H., Hapke, U., Meyer, C., John, U., 2002. Screening for alcohol usedisorders and at-risk drinking in the general population: psychometricperformance of three questionnaires. Alcohol Alcohol. 37 (3), 261–268.

Schnyder, U., Moergeli, H., Klaghofer, R., Buddeberg, C., 2001. Incidence andprediction of posttraumatic stress disorder symptoms in severely injuredaccident victims. Am. J. Psychiatry 158 (4), 594–599.

Selin, K., 2003. Test–retest reliability of the Alcohol Use Disorder IdentificationTest in a general population sample. Alcohol. Clin. Exp. Res. 27, 1428–1435.

Sheehan, D.V., Lecrubier, Y., Harnett-Sheehan, K., Amorim, P., Janavs, J., Weiller,E., Hergueta, T., Baker, R., Dunbar, G., 1998. The Mini InternationalNeuropsychiatric Interview (M.I.N.I.): the development and validation ofa structureddiagnosticpsychiatric interview. J. Cli. Pscyhiatry59 (suppl20),22–33.

Shirao, I., Tsuda, A., Ida, Y., Tsujimaru, S., Satoh, H., Oguchi, M., Tanaka, M.,Inanaga, K., 1988. Effect of acute ethanol administration on noradrenalinemetabolism in brain regions of stressed and nonstressed rats. Pharmacol.Biochem. Behav. 30 (3), 769–773.

Stewart, S.H., Pihl, R.O., Conrod, P.J., Dongier, M., 1998. Functional associationsamong trauma, PTSD,andsubstance-relateddisorders. Addict. Behav. 23(6),797–812.

Weathers, F.W., Keane, T.M., Davidson, J., 2001. Clinician-Administered PTSDScale: a review of the first ten years of research. Depress. Anxiety 13 (3),132–156.

Wolfsohn, J.M., 1918. The predisposing factors of war psycho-neuroses.Lancet 191 (4927), 177–181.

Zatzick, D.F., Kang, S.M., Muller, H.G., Russo, J.E., Rivara, F.P., Katon,W., Jurkovich,G.J., Roy-Byrne, P., 2002. Predicting posttraumatic distress in hospitalizedtrauma survivors with acute injuries. Am. J. Psychiatry 159 (6), 941–946.

Zigmond, A., Snaith, R., 1983. The Hospital Anxiety and Depression Scale. ActaPsychiatr. Scand. 67, 361–370.