Introduction

Subjects admitted to the emergency department (ED) with intracranial lesions followinghead injury are a special challenge for emergency physicians. They represent aheterogeneous group of patients with large variability as to injury severity, clinicalcourse, neurological recovery and overall outcome [1].

Worsening detected by imaging and clinical deterioration are associated with anunfavorable outcome, and a group of predictor variables has been related toworsening-type lesions and future events [2–4]. In a few cases, progression is extremely rapid and the ultimate outcomemight be unfavorable because of delayed transfer to neurosurgical units; in other casesthe lesions do not progress and the final outcome is usually favorable.

In the last decade the use of antithrombotic therapy with antiplatelet drugs has grownconsiderably, as an effect of national and international guidelines promoting theirwidespread use to prevent cardiovascular events in high-risk populations andparticularly in older people [5, 6]. In the same period, the epidemiology of the trauma population has alsochanged, with a larger and larger prevalence of older age-groups [7], where antiplatelet drug use is more prevalent, in the presence ofcomorbidities [8, 9].

The aim of this study was to test the effect of pre-injury antiplatelet therapy onshort- and long-term outcomes in subjects with head injury and a positive computedtomography (CT) scan at first evaluation.

Methods

Study design and settings

This multicenter observational study included all adult subjects, who attended 32Italian EDs of community and regional hospitals for mild, moderate or severe headinjury and intracranial lesions within 24 hours from the event (from January toDecember 2009). The participating centers represented a wide variety of facilities,distributed across the country, to increase external validity and to make the resultsgeneralizable to the majority of subjects observed for head injury. The centersincluded hospitals with neurosurgical units, hospitals with teleradiology-consultingsystems connected with a neurosurgical center and hospitals without neurosurgical andteleradiology facilities.

Adult subjects ≥ 18 years old with mild (Glasgow coma scale (GCS) = 15 to 14)or moderate to severe (GCS ≤ 13) head injury within 24 hours of trauma and apositive head CT scan at their first evaluation in the ED were included in the study.The subjects were all consecutive patients with a positive head CT scan withoutindication of urgent (within 7 days) neurosurgical hematoma/hemorrhage evacuation(Marshall category 2 to 4 at entry).

Excluded were subjects in the presence of: a) an initial head CT scan requiringurgent neurosurgical intervention (Marshall category 5) or not-operated mass lesion(Marshall 6 category); b) GCS = 3 and bilateral, fixed and dilated pupils; c) anunclear history of the mechanism of injury as the primary event; d) hypotension, thatis, systolic blood pressure persistently < 90 mmHg during the observation period;e) the need for cardiopulmonary resuscitation; f) penetrating injuries atpresentation; and g) discharge against medical advice.

The use of antiplatelet drugs was systematically recorded, independent of time ofexposure. Aspirin (usual dose, 100 mg), ticlopidine, indobufen (a popularantithrombotic drug used in Italy) and clopidogrel were considered, as well as thepotential antiplatelet activity of other anti-inflammatory agents. During theobservation period there was no specific indication for rescue therapy with humanprothrombin complex or platelet transfusions in subjects treated withanticoagulant/antiplatelet agents, and no patients received this supporttreatment.

Treatment protocol

From the ED, subjects were transferred for observation and treatment to a highdependency unit, ordinary admitting unit, neurosurgical unit or ICUs. Afteradmission, all patients were submitted to additional head CT scan within 6 to 24hours from injury according to local protocols. Furthermore, CT was always repeatedin the case of clinical or neurological deterioration. The time interval betweentrauma and the initial head CT scan was dictated by emergency procedures of theindividual centers. For the purpose of the present study, all head CT scans wereretrospectively reviewed in a temporal sequence by an independent expertneuroradiologist in a blinded fashion to confirm the initial diagnosis and toevaluate possible worsening in the head CT scan at 6 to 24 hours. CT scans wereclassified according to the criteria of Marshall [10], modified according to the revision of the European Brain InjuryConsortium (EBIC) [11].

The protocol was carried out according to the Helsinki Declaration and approved bythe ethical committee of the Local Health District of Forlì. All data weretransferred from the peripheral centers to the coordinating unit in a completelyanonymous form. According to the Italian law on privacy protection and use ofpersonal data (Dls n. 85, March 1, 2012), informed consent is not needed wheneverhandling is carried out in an anonymous form on retrospective data on file and itwould be technically impossible to trace people for signing consent forms.

Variables definition

A few items were selected as the variables potentially associated with outcomes. Weconsidered age, sex, type of injury (motor vehicle accidents, falls or accidental,work-related, assault, sport injuries and other causes), coagulation (by prothrombintime) and neurological status (by GCS), as well as the use of antithrombotic agentsas described above. In the antiplatelet group, we also considered the few cases inwhich other non-steroidal anti-inflammatory drugs (NSAIDs) with a definiteantiplatelet activity had been administered in the three days before trauma for otherreasons.

Comorbidities, although common and associated with outcome in spontaneousintracerebral hemorrhage [8, 9], were not considered in the present analysis. In previous studies ontraumatic brain injury comorbidities did not predict short- or long-term outcome [1].

The intracranial injuries considered for analyses were: traumatic subarachnoidhemorrhage (t-SAH), subdural hematoma (SDH), epidural hematoma (EDH), intracerebralhemorrhage/contusion (ICH) or depressed skull fracture (DSF) and intraventricularhemorrhage (IVH) [12, 13]. IVH was considered a distinctive intracranial injury, but no subjectswere considered with positive head CT scans for this type of injury as a uniquelesion. In all cases IVH resulted in different combinations with other types ofintracranial injury.

Patients' coagulation status (prothrombin time) was determined by protocol in allcases. Values of the International Normalized Ratio (INR) > 1.5 were considered atrisk of hemorrhage.

Outcome measures

Short-term outcome measures were: a) intracerebral injuries with worseningcharacteristics, indicated by a change of at least one point in Marshall categorybetween initial and follow-up CT scan performed during serial controls within 24hours; and b) the need for neurosurgical intervention because of clinical and/orradiological deterioration during the observation period. This period was limited tothe first seven days after diagnosis in order to exclude delayed complication ofinjury (chronic subdural hematomas, hygromas or hydrocephalus) [12].

As a long-term outcome measure we considered the Glasgow outcome scale (GOS) at sixmonths. For ease of analysis and reporting, the five-point GOS score was categorizedas either favorable (moderate disability or good recovery - GOS 4 to 5) orunfavorable (dead, vegetative, or severely disabled - GOS 1 to 3). The follow-up GOSwas rated by an expert physician unaware of the study protocol, on the basis of theresponse to a structured telephone call [12]. Main outcome measures were then related to the different hospitalfacilities, for example, hospital with neurosurgical unit, hospital with telemedicineconsultation only (no neurosurgical unit) and hospital without both neurosurgicalunit and telemedicine consultation.

Statistical analysis

A data mining method was chosen to select relevant patterns between predictorvariables and main outcomes by Weka software (University of Waikato, Hamilton, NZ).We used a decision tree technique, in which nodes indicate decision points, chanceevents, or branch terminals. Branches correspond to each decision alternative orevent outcome emerging from a node. The root nodes are the first set of decisionalternativeness. The construction of a decision tree was obtained by a 'recursivepartitioning' analysis [14].

Mean value, SD and frequencies were used to describe data distribution. We usedmultivariable logistic regression analysis with a P value greater than 0.05for removal of variables. A score for the risk of unfavorable outcome was calculatedfor each patient on the basis of the coefficients computed by the logistic regressionderived from variables entering the stepwise procedure. The accuracy of the riskscore was then evaluated by the area under the receiver operating characteristic(ROC) curves. The odds ratio (OR) and 95% CI were also calculated. We tested theassociation of each variable with the primary outcome measure using Chi-square testsfor nominal variables, the Mann-Whitney U test for ordinal variables, and theunpaired two-tailed t-test for continuous variables (SPSS software, version17.0 - SPSS Inc., Chicago, IL, USA). The relative risk (RR) of different outcomes wasalso calculated.

Results

Patients

The mean age of the 1,558 subjects with intracranial lesions was 65 years (SD 21),with 288 (18.5%) patients under 40 and 664 subjects (42.6%) over 75. The vastmajority of subjects (1,123 cases, 72.1%) had a mild head injury with GCS 14 to 15,420 cases (24.9%) had a moderate injury (360 cases with GCS 13 to 11 and 60 with GCS10 to 9). The last group of 15 subjects (1.0%) had a Marshall category 2 to 4 andsevere head injury (GCS < 9) (Table 1).

Table 1 Clinical characteristics of subjects according to worsening characteristicsbetween initial and follow-up CT scan
Full size table

A total of 708 subjects (45.4%) were injured by falls or accidents with 474 (30.4%)following a road accident. In the remaining subjects the head injury was work-related(83 cases, 5.3%) or following an assault (46, 3.0%), or related to sports and othercauses (247, 15.8%) (Table 1).

At the first evaluation, 1,328 subjects (85.2%) had an intracranial injury withMarshall category 2, 168 subjects (10.8%) had category 3, and only 62 cases (4.0%)had category 4 (Table 1). A single lesion was recorded in 886subjects (56.9%), 2 lesions in 430 cases (27.6%) and 3 or more lesions in theremaining 237 cases (15.2%). The frequency distribution of type of lesion was: ICH(766 cases; 49.2%), SDH (604; 38.8%), t-SAH (776; 49.8%), EDH (157; 10.1%) and IVH(94; 6.0%) (Table 1).

Pre-injury antiplatelet therapy was recorded in 537 subjects (34.5%) of the entirecohort (454, 49.1% in the group ≥ 65 years old). Aspirin was the mostfrequently used antiplatelet medication (439 subjects, 28.2%), followed byticlopidine (69, 4.4%), clopidogrel (28, 1.8%), NSAIDs (20, 1.3%) and low molecularweight heparin (10, 0.6%). A group of 129 cases (8.3%) had INR > 1.5 because ofsimultaneous treatment with warfarin.

Outcome prediction

Short-term outcome

In 201/1,558 subjects (12.9%) head CT scan comparison documented a worseninglesion in the short-term. Antiplatelet therapy increased the risk of worseningtwo-fold (n = 106, 19.7% of treated versus 95, 9.4% of untreated cases),corresponding to a relative risk (RR) of 2.09, 95% CI 1.63 to 2.71 (Figure 1). Compared with untreated subjects, the risk was particularlyhigh in subjects on clopidogrel (RR 5.76, 95% CI 3.88 to 8.54), independent of theassociation with aspirin (15 cases, 8 with worsening lesions; RR 5.73, 95% CI 3.44to 9.55; P < 0.001).

Figure 1
figure 1

Distribution of worsening events in relation to antiplatelet therapy insubjects with intracranial lesions following head injury. Significantoutcomes in the decision tree analysis are reported as white text on a greybackground.

Full size image

On multivariable logistic regression analysis a group of 5/14 items wasindependently associated with worsening lesion (Table 2).The discriminating operating characteristics area of the selected items was 0.777(95% CI 0.755 to 0.797; P < 0.001).

Table 2 Logistic model of variables considered in predicting subjects with worseninglesions after head injury
Full size table

Data mining analysis selected the following relevant patterns between predictorvariables and main outcomes: a) in subjects with mild head injury (GCS 15 to 14),antiplatelet therapy increased the risk of worsening two-fold when the number oflesions at the first CT scan was ≤ 2, (6.90% treated versus 3.70% nottreated; RR 1.86, 95% CI 1.06 to 3.30, P = 0.032) and further increasedthe risk of worsening when the number of lesions was ≥ 3 (34.8% treatedversus 10.4% not treated; RR 3.34, 95% CI 1.74 to 6.40, P = 0.003)(Figure 2); b) in subjects with moderate-severe head injuryantiplatelet therapy increased the risk of worsening when the number of lesions atthe first CT scan was ≤ 2 (37.6% treated versus 21.8% not treated, RR 1.72,95% CI 1.21 to 2.45; P = 0.002 (Figure 2).

Figure 2
figure 2

Data mining analysis: relevant patterns of variables predicting caseswith worsening lesions in relation to severity of head injury (mildversus moderate-severe head injury), number of intracranial lesions andantiplatelet therapy. Significant variables are reported as whitetext on a grey background.

Full size image

Worsening seen on serial CT scans resulted in neurosurgical intervention in 46subjects (2.9%). The intervention was required for EDH (8 cases), SDH (30 cases)and ICH (8 cases). Neurosurgical intervention was needed more frequently insubjects treated with antiplatelet drugs (21.2% treated versus 11.2% untreated, RR1.90, 95% CI 1.35 to 2.66; P < 0.001). On multivariable logisticregression analysis 8/15 items (male sex, younger age, mechanism of injury, INR >1.5, antiplatelet therapy, GCS, Marshall category, and type of lesions) wereindependently associated with worsening and the need for neurosurgicalintervention.

Long-term outcome

A complete six-month follow-up was obtained in 1,222/1,558 subjects (78.4%). Atotal of 336 cases (21.6%) were lost at follow up and in 115 (7.4%) cases GOS wasunreliable due to previous disability or trauma-related disability not dependenton head injury.

Outcome was unfavorable in 78 cases (5.0%): 26 patients (1.7%) died during thesix-month follow up, 9 patients (0.6%) were judged in a permanent vegetative stateand 43 (2.8%) were severely disabled. The majority of subjects (n =1,144, 73.4%) had a favorable outcome, with moderate disability being present inonly 168 cases (10.8%). At follow up, the risk of unfavorable outcome at sixmonths increased by 50% in the group treated with antiplatelet therapy (9.7%treated versus 4.4% untreated; RR 1.58, 95% CI 1.28 to 1.95; P <0.001) (Figure 3).

Figure 3
figure 3

Unfavorable outcomes in subjects with head injury and intracraniallesions in relation to antiplatelet therapy. Significant outcomes arereported as white text on a grey background.

Full size image

On multivariable logistic regression analysis only 4/14 items (GCS, Marshallseverity, age in decades, intracerebral hemorrhage/contusion) were selected aspredictors of unfavorable outcome (Table 3). Thediscriminating operating characteristics area of the selected items was 0.891 (95%CI 0.860 to 0.921); P < 0.001.

Table 3 Logistic model of variables considered in predicting unfavorable outcome insubjects with head injury
Full size table

These results were confirmed by ordinal regression analysis: five items (Marshallseverity, GCS, age in decades, antiplatelet therapy and type of injury) wereselected for the prediction of GOS with a discriminating operating characteristicsarea of 0.716 (95% CI 0.645 to 0.786; P < 0.001).

These results were not significantly different after exclusion of subjects treatedwith warfarin or low molecular weight heparin (LMWH) (discriminating operatingcharacteristics area of 0.783 (95% CI 0.746 to 0.820; P < 0.001)) andin the subgroup of subjects fully recovered or with moderate disability at sixmonths, that is, GOS 4 to 5 (0.716, 95% CI 0.645 to 0.786; P <0.001).

Data mining analysis did not select any relevant pattern in relation to differenthospital facilities, that is, neurosurgery versus telemedicine systems versusnone; P test for trend = 0.144).

Discussion

This observational study derived from Italian EDs shows that pre-injury antithrombotictherapy is associated with negative outcomes in subjects with head injury andintracranial lesions with an indication of observation and conservative treatment; inthe short-term progression of lesions was seen on the CT scan, in the long-term the riskof unfavorable outcome increased. The risk of lesion worsening was particularly highwhen subjects were treated with clopidogrel, independent of the concomitant use of otherantiplatelet agents.

The prognosis of subjects with head injury and intracranial lesions with an indicationfor conservative treatment is extremely variable, depending on the progression ofinjury, the size of the lesion and secondary injury responses that may worsen theprimary lesion [15]. The earlier the initial CT scan, the greater the likelihood that the lesionswill progress at follow-up. Progression generally occurs within the first 12 hours, butmay occur as late as three to four days after trauma. Small contusions that progress areusually clinically silent and are less likely to require neurosurgical intervention [16], whereas large contusions in subjects with low GCS scores are more likely toevolve [15].

Injury progression was defined by worsening of the Marshall category, a validated toolto assess the outcome of subjects with head injury [17–19]. According to EBIC [11], an increase in the Marshall CT category at the follow-up CT scan may beconsidered a sign of disease progression. Whenever the initial CT scan shows a diffuseinjury without swelling or shift worsening to a mass lesion with need of neurosurgicalintervention, the outcome becomes definitely unfavorable (62% versus 38%) [20].

Our data confirmed that the risk of imaging progression is associated with the severityof the initial Marshall category with 10.2% of cases worsening in the group of subjectsclassified as Marshall category 2, 25.6% of subjects in category 3 and 37.1% of subjectscategorized as Marshall 4. Our worsening rate is, however, much lower than that reportedin different series from neurosurgical facilities, where approximately 50% of patientswith lesions who were admitted for conservative treatment showed progression [21–24]. This difference is probably due to the selection of more severe and youngerpatients in neurosurgical units, including Marshall 5 cases, compared to those observedin a general ED. This hypothesis is also confirmed by a larger use of neurosurgicalevacuation reported in those settings, whereas neurosurgery may be contraindicated inolder persons, although this issue is not settled.

A group of variables (injury severity, anticoagulant therapy, need for cardiopulmonaryresuscitation in the field, older age, short duration between injury and the first CTscan, multiple lesions, midline shift and injuries with need of neurosurgicalprocedures) had been indicated as predictors of radiological progression [22, 23, 25, 26]. Our study confirms the importance of clinical and radiological itemsselected by previous studies in predicting lesions likely to evolve after head injuryand indicates antiplatelet therapy as a relevant, additional predictor.

The negative effect of antiplatelet therapy might depend on a number of factors, such asminimum continued bleeding or a microvascular dysfuntion, exaggerated by reducedplatelet function, favoring edema and brain swelling, producing a midline shift [15].

We defined progression on the basis of the Marshall classification. Differences inradiological progression may depend on the criteria used: 100% [26], 30% [21] or 25% increase [22] in hematoma dimension, but different cuts and angulation may introduce animportant bias with the use of strict criteria, such as a 25% to 30% enlargement [11]. The Marshall category, although crude, provides a very easy-to-defineclinical index of progression. It was selected as an outcome measure in the short termand, combined with GCS, it was the only variable associated with unfavorable outcome atsix months, confirming the clinical importance of this item.

Among antiplatelet agents both aspirin and, particularly, clopidogrel increased the riskof evolving lesions, but their combined use did not further increase the risk. Bycontrast, ticlopidine, largely used in Italy in the past, did not increase the risk. Therisk associated with clopidogrel is of particular concern, considering its increasinguse. The advantages of clopidogrel on cardiovascular outcomes have made it a lifesavingdrug in subjects over 45 years old who have cardiovascular disease [5, 27]. Its use was later extended from coronary artery disease to cerebrovascularand peripheral artery disease, thus being largely diffused in the elderly population [28]. An analysis of drug prescriptions in more than 300,000 Italian subjects withdiabetes showed an increased prevalence of antiplatelet drug use from 15% to 52% in theperiod 1997 to 2006 [29], and in 2008 more than 4% of the general Italian population was treated withaspirin [30]. In our series, approximately 35% of the subjects were being treated withantiplatelet agents, and this figure increased to 53% in the group of subjects more than75 years old. The use of these drugs is likely to increase further in the future,following guidelines indicating antiplatelet drug therapy in a large proportion of oldersubjects [5, 27].

Two recent reviews have summarized the available evidence on the risk of unfavorableoutcomes of antiplatelet medications, especially in subjects with severe head injury andolder age. Beynon et al. [31], on the basis of the scarce available evidence, concluded that these agentsincrease the risk of an unfavorable outcome, particularly in cases of severe traumaticbrain injury. In a meta-analysis of five studies, Batchelor and Grayson compared themortality rates of patients with blunt head trauma who were on aspirin or clopidogrelversus cases not on antiplatelet agents [32]. They found a significant heterogeneity and a moderately increased overallrisk of death for both drugs, which did not reach statistical significance. However, thelow number of events precludes any firm conclusion and further work is required.

Event rates constitute an even more significant drawback in studies on mild-to-moderatebrain injury. Nonetheless, the mortality rate of subjects receiving aspirin was alsoreported to be higher than normal [33]. In cases observed in EDs, pre-injury antiplatelet therapy was recently shownto increase significantly the risk of intracranial lesions in subjects after mild headinjury [34], whereas in a prospective study of mild and moderate head injury in subjectsmore than 60 years old, low-dose aspirin prophylaxis had no effect on the frequency ortypes of intracerebral or meningeal hemorrhage [35]. The initial size of the contusion and the presence of SDH were selected aspredictors of radiological progression, and the initial GCS and younger age aspredictors of good disposition at discharge [34], but much more evidence is required before a firm conclusion can be drawn.Anticoagulation and antiplatelet therapy were not included in any study model [21].

Age represents an important issue in head injured patients. In Italy, age is notformally considered a criterion for admitting patients to hospitals with differentlevels of care, but in clinical practice older patients frequently have limited accessto conservative observation in neurosurgical units and to interventions. Our analysisselected older age as a significant, independent predictor of long-term outcome. In theItalian database, 924 subjects (59.3%) were ≥ 65 years old and 42% were older than75, a figure completely different from previously published studies. In a widely citedstudy [36], subjects older than 65 years were excluded from the analyses, the median ageof subjects was 33 when treated in neurosurgical units and 31 in those admitted tonon-neurosurgical centers. Our database reflects the 'real world' of head injurysubjects with Marshall 2 to 4, observed in the Italian EDs, with a median age of 72years and with 30% of the cases older than 80, as previously reported [37, 38].

The growing elderly population and the expanding indications for anticoagulant therapymight produce more complications associated with anticoagulant treatment, challengingthe emergency physicians more and more. A very recent study showed that oralanticoagulants may also be safely used in older patients at risk of fall [39], but in a previous report we showed that anticoagulation increased the riskof intracranial lesions by more than four times, independent of other variables [12]. In the present study, anticoagulant treatment did not significantly predictworsening in the 126 cases (8.1%) on anticoagulants with an INR above 1.5, but aselection bias may be operative. In subjects on oral anticoagulants, the initial lesionmight be so severe (that is, Marshall 5 or 6) that is excludes them from the analysis.The progressive use of rapid anticoagulation reversal will clarify this problem.

A few limits should be considered. Firstly, selection biases might be present because ofthe retrospective nature of the analysis of clinical records and different extractionprocedures according to software available in the various EDs. These biases might beamplified by incomplete recording of drug use and/or incomplete reporting by patients.Underreporting of drug use might also increase in relation to incomplete anamnesis byphysicians, unaware of the possible risk associated with antiplatelet drugs.

Secondly, the time lag between head trauma and CT scanning was variable between a fewhours to 24 hours. Both trauma-to-admission and admission-to-CT times were variable,according to clinical judgment, with an influence on the natural history of lesions. Asdiscussed above, these biases might be reduced by the use of the Marshallclassification, in which category changes imply evident changes in the imagingappearance of lesions.

Thirdly, the history of antiplatelet drug use might be completed by the analysis ofantiplatelet activity. In a series of 84 subjects treated with aspirin, 2.4% of caseshad normal platelet function, and 42% of subjects without a documented history ofaspirin use had platelet inhibition. Aspirin resistance is a multifactorial phenomenon,associated with comorbidities, leading to reduced platelet activation and aggregation [40, 41]. However, aspirin history and the measured activity of platelet inhibitionwere associated with only a marginal risk of CT scan progression, craniotomy, mortalityor poor outcome at multivariable analysis [42].

Fourthly, we did not consider comorbidities in our analysis. Comorbidities have adefinite importance in hemorrhagic stroke [8] and spontaneous intracerebral hemorrhage [9], whereas their importance in traumatic lesions is doubtful. The use ofantiplatelet drugs might identify subjects with more prevalent cardiovascular disease,at higher risk of spontaneous cerebro-vascular events, independent of antiplatelet use.Apparently, this does not apply to traumatic brain lesions and the Charlson index ofcomorbidities was not associated with outcome in a previous study in Italian EDs [1, 2].

Finally, an increase in the Marshall CT classification score would not always representlesion extension or disease progression. The hierarchy of Marshall class is based on theabsence/presence of signs of raised intracranial pressure, such as brain swelling,midline shift and mass lesions which need neurosurgiucal evacuation. This is very likelyto determine an unfavorable outcome in the long term, but this is not always the case.Lack of significance of variables other than Marshall category and GCS versus outcomedetermined by GOS categories does not exclude possible clinical relevance.

Conclusions

Our data, which are derived from a representative number of Italian EDs, show thatpre-injury antithrombotic therapy is associated with an increased risk of short-termradiological worsening and six-month unfavorable outcome in subjects with a positivehead CT scan, particularly in subjects treated by clopidogrel. The results should beconsidered in predictive algorithms of future guidelines of diagnosis and treatment ofhead injury.

Key messages

• In subjects with mild or moderate-severe head injury and a positive head CT scanwith indications for conservative treatment, 12.9% of subjects worsened by CT comparison(change of at least one point in the Marshall category) at 6 to 24 hours.

• A group of 5/14 items (GOS, Marshall category, antiplatelet therapy, IVH, numberof lesions) were independently associated with short term (6 to 24 hours) worsening.

• Pre-injury antiplatelet therapy increased the risk of short term worseningtwo-fold. The risk was particularly high in subjects on clopidogrel, independen of theassociation with other antiplatelet drugs.

• At long-term follow up (six months), only 4/14 items (GCS, Marshall severity,age in decades, intracerebral hemorrhage/contusion) were selected as predictors ofunfavorable outcome. The risk increased by 50% in the group treated with antiplatelettherapy.