Article Text

Download PDFPDF

Predicting intracranial lesions by antiplatelet agents in subjects with mild head injury
  1. Andrea Fabbri1,
  2. Franco Servadei2,
  3. Giulio Marchesini3,
  4. Sherman C Stein4,
  5. Alberto Vandelli1
  1. 1Dipartimento dell'Emergenza, Presidio Ospedaliero Morgagni-Pierantoni, Azienda Unità Sanitaria Locale di Forlì, Italy
  2. 2Unità Operativa di Neurochirurgia, Azienda Ospedaliero-Universitaria di Parma, Italy
  3. 3Alma Mater Studiorum, Università di Bologna, Italy
  4. 4Department of Neurosurgery, University of Pennsylvania School of Medicine, Philadelphia, PA, USA
  1. Correspondence to Dr A Fabbri, Dipartimento dell'Emergenza, Presidio Ospedaliero Morgagni-Pierantoni, Azienda USL Forlì, Via Forlanini 34, I-47100 Forlì, Italy; dr.andrea.fabbri{at}


Background The effect of pre-injury antiplatelet treatment in the risk of intracranial lesions in subjects after mild head injury (Glasgow Coma Scale (GCS) 14–15) is uncertain.

Methods The potential risk was determined, considering its increasing use in guidelines on cardiovascular disease prevention, and ageing of the trauma population in Europe.

Patients The interaction of antiplatelet therapy with the prediction variables of main decision aids was analysed in 14 288 consecutive adolescent and adult subjects with mild head injury.

Measurements Any intracranial lesion at CT scan was selected as an outcome measure in a multivariable logistic regression analysis.

Results Intracranial lesions were demonstrated in 880 cases (6.2%), with an unfavourable outcome at 6 months in 86 (0.6%). Antiplatelet drugs were recorded in 10% of the entire cohort (24.7% in the group over 65 years). They increased the risk of intracranial lesions in the univariate analysis (OR 2.6; 95% CI 2.2 to 3.1), interacting with age in the multivariate analysis (antiplatelet OR 2.7 (1.9 to 3.7); age ≥75 years 1.4 (1.0 to 1.9)). The inclusion of these two variables with those included in previous decision aids for CT scanning (GCS, neurodeficit, post-traumatic seizures, suspected skull fracture, vomiting, loss of consciousness, coagulopathy) predicted intracranial lesions with a sensitivity of 99.7% (95% CI 98.9 to 99.8) and a specificity of 54.0% (95% CI 53.1 to 54.8), with a CT ordering rate of 49.3% (undetermined events 0.2:1000).

Interpretation Antiplatelet drugs need to be considered in future prediction models on mild head injury, considering their increasing use and progressive ageing of the trauma population.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Mild head injury (MHI) with a Glasgow Coma Scale (GCS) score of 14–15 is low risk for intracranial complications but the condition remains a critical issue in terms of disease management because of the potential risk of unfavourable outcome. Nearly 95% of adult subjects with MHI show no abnormalities on CT scan and less than 1% need neurosurgical intervention. Every effort should be made to ensure CT scanning for all subjects with intracranial injuries while limiting the number of negative scans.1

An increasing number of investigations have been carried out to identify subjects at higher risk of intracranial lesions, using groups of predictor variables, in order to increase the specificity of CT scans.2–6 Some of the derived algorithms have been externally validated6–9 with influence on clinical practice.10–13 A recent study showed that the operating characteristics of the main decision instruments have reached sensitivities of more than 98%, with specificities ranging from 30% to 50%, depending on available resources.14

Whereas the use of anticoagulants is a well known risk factor, the chronic use of antiplatelet drugs has never been considered in the panel of prediction variables tested by the currently available decision aids. This issue was raised in the National Institute for Health and Clinical Excellence (NICE) guidelines15 but studies of haemorrhagic risk attributable to antiplatelet drugs in different groups showed uncertain results.16 17 However, in the past few years the use of antiplatelet drugs has grown considerably. This is due, in part, to national and international guidelines which promote their widespread use to prevent cardiovascular events in the elderly and other high risk populations.18 19 This development corresponds to changes in the epidemiology of trauma which has progressively shifted towards older age groups20 in whom antiplatelet drugs use is more prevalent.

Since 1999, a complete database of adolescent and adult subjects with traumatic brain injury has been set up in our hospital, now totalling nearly 17 000 cases. The aim of the study was to test the effect of chronic use of antiplatelet drugs on the risk of intracranial lesions in subjects with MHI.

Materials and methods

Selection of participants and interventions

Consecutive subjects aged 10 years or more who attended the emergency department (ED) of the local health district of Forlì (level I: 48 000 children and adults per year from a population of 171 000 inhabitants in an area of 1380 km2), within 24 h of traumatic brain injury, were investigated. Subjects were enrolled for outcome purposes in a prospective registry and treated according to the recommendations of the Neurotraumatology Committee of the World Federation of Neurosurgical Societies.14


The Head Injury Registry consists of 16 799 consecutive medical records of all subjects attending within 24 h from the event (from January 1999 to December 2007). MHI was defined as an acute injury of the head, other than any superficial injury to the face, GCS 14 or 15 in adolescent and/or adult inhabitants, aged ≥10 years. A total of 2511 records (14.9%) were excluded for the following reasons: (a) unclear history of injury as the primary event (1100 cases); (b) unstable vital signs (378 cases); (c) GCS <14 (302 cases); (d) penetrating injuries at presentation (33 cases); (e) discharge against medical advice (388 cases) and (f) duplicate records of subjects re-attending for post-concussive syndrome within 7 days of admission for MHI (310 cases).

The use of aspirin as an antiplatelet drug (any dosage) was systematically recorded, independent of time of exposure. Indobufen (a second antiplatelet drug used in Italy) was also considered. In contrast, we did not consider the potential antiplatelet activity of other anti-inflammatory agents. Finally, clopidogrel was not included in the analysis because it was not on the market at the time the protocol was set up and was later very rarely used in this old population because of regulatory limitations in Italy.

Treatment protocol

Subjects were treated according to a predefined diagnostic and observational protocol as previously published.21 Briefly, low risk patients were sent home after examination without any radiological imaging. Medium risk patients underwent CT scan according to the physician's judgement and were discharged after at least 3–24 h from trauma in the event of negative results. High risk patients underwent CT imaging and were managed within the first 24–48 h strict observation period, regardless of CT findings. Age over 60 years was not considered a risk factor per se.

All subjects re-attending the ED had a CT scan.

CT scans showing intracranial lesions were reviewed urgently by a neurosurgeon at the nearest neurosurgical unit (Cesena, 17 km away), consulted via a teleradiology system. The protocol was carried out according to the Helsinki Declaration and approved by the ethics committee of the local health district of Forlì.

Variable definition and outcome measures

The analysis was based on the pool of variables considered in main decision instruments in the literature14 with the chronic use of antiplatelet therapy as an additional variable (table 1).

Table 1

Main clinical characteristics associated with a high risk of clinically significant outcomes in subjects with mild head injury

The main outcome measure was the diagnosis of any post-traumatic lesion at CT within 7 days of the trauma. Lesions included: (a) intracerebral haematomas/brain contusion(s) (ICH); (b) traumatic subarachnoid haemorrhage; (c) subdural haematoma (SDH); (d) epidural haematoma (EDH) and (e) depressed skull fracture.3

In subjects who revisited for complaints, a time limit of 7 days was set to consider any lesion as directly related to the initial head injury. Similarly, haematoma evacuation and/or skull fracture elevation were only considered within 7 days after diagnosis. Late events were not considered in the present analysis.

The 6 month outcome was carried out by a systematic search of all patients by checking the death certificates and medical databases of our local health district. Patients without post-traumatic lesions and not readmitted for complaints were considered fully recovered. In addition, a member of the ED staff contacted all patients with a defined post-traumatic lesion and rated the GCS by means of a structured telephone interview.22 Outcome was considered unfavourable if the patient died of head injury related causes, remained in a vegetative state or had a permanent severe disability.


Mean values (SD, median, IQR and frequencies) were used to describe data distribution. We used multivariable logistic regression with forward stepwise selection with a p value greater than 0.05 for removal of variables. The OR and 95% 95% CI were also calculated. Sensitivity and specificity were evaluated, together with the negative predictive value. We tested associations of each risk factor with the primary outcome measure using χ2 tests for nominal variables, the Mann–Whitney U test for ordinal variables and the unpaired two tailed t test for continuous variables using SPSS software, V.17.0 (SPSS Inc). We also assessed the operating characteristics by calculating the area under the receiver operating characteristic curve for variables selected by the multivariable logistic regression analysis. Proportions were compared between protocols by means of Fisher's exact test.


Patient population

The main predicting variables associated with intracranial injuries are reported in table 2.

Table 2

Risk of intracranial lesions after mild head injury in relation to the presence of the main predicting variables identified by previous decision instruments

Median age of the 14 228 subjects was 49 years (IQR 29–75), with 5180 (36.4%) patients ≥65 years and as many as 3701 (26.0%) ≥80 years. Most subjects were injured following a road accident (5990 (42.1%)) or falls or accidental injury (5379 cases, 37.8%). In the remaining subjects the head injury was work related in 1022 (7.2%), assault in 503 (3.5%) and related to sport and other causes in 732 (5.1%). Adherence to the local protocol was nearly complete, with a CT scan ordering rate of 9056 (63.6% of total cases).

Antiplatelet drug use (excluding clopidogrel) was recorded in 10% of the entire cohort (24.7% in the group ≥65 years) and was associated with the risk of intracranial lesions (OR 2.6; 95% CI 2.2 to 3.1; p<0.001) (table 2).

A total of 1201 injuries were detected in 880 subjects (6.1%), with a single lesion in 600 (4.0%, respectively). Lesions included 591 intraparechymal haemorrhagic contusions, 329 acute subdural, 63 acute epidural haematomas, 193 subarachnoid haemorrhages and 25 depressed skull fractures.

Neurosurgical intervention was performed in 153/880 (17.3%) subjects within 7 days for acute subdural haematomas (66 cases), depressed skull fractures (17 cases), epidural hematomas (26 cases) and intracerebral haematomas/brain contusion (96 cases) in different combinations.

Follow-up data were obtained in 715 cases with intracranial lesions (87.0%). Of the remaining 107 cases, 16 subjects died because of head injury unrelated complications, 18 were lost to follow-up and in 73 GCS was unreliable, due to either a previous disability (68 cases) or a disability unrelated to the head injury (five cases). The outcome was unfavourable in 86 cases (0.6%), with 51 aged ≥75 years. Thirty-six subjects died, two remained in a permanent vegetative state (0.1%) and 62 cases (0.4%) were severely disabled. Only 76 cases had a moderate disability and 600 were fully recovered.

Outcome prediction

Eight of the 14 items proposed were independently associated with intracranial lesions (table 3) but only four (GCS, suspected skull fracture, neurodeficit and loss of consciousness) showed a good discriminant capacity (receiver operating characteristic areas >0.65) in predicting an intracranial lesion, the remaining having little or no discriminant ability.

Table 3

Multivariable model for prediction of intracranial lesions in subjects after mild head injury

The discriminating operating characteristics area of the items selected was 0.697 (95% CI 0.690 to 0.705), sensitivity 98.4 (95% CI 97.3 to 99.0), specificity 41.1 (95% CI 40.2 to 41.9) with a theoretical CT ordering rate of 61.4% with only 14 subjects with intracranial lesions remaining undetermined (EDH two cases; ICH four cases; SDH seven cases; SDH+ICH one case) (table 4).

Table 4

Sensitivity, specificity, operating characteristics curve and negative predictive value of the items selected by different variable combinations in the prediction of intracranial lesions in subjects with mild head injury

When the covariate pre-injury use of antiplatelet drugs was added into the model, it showed a significant additional effect (OR 2.8 (95% CI 2.0 to 3.9); p<0.001). Antiplatelet therapy entered the model as a predictor variable excluding age ≥65 years (OR 1.2, 95% CI 0.9 to 1.7; p=0.202) (table 4). The discriminating operating characteristics area of the new item combination was higher in comparison with that of the classical items (0.807, 95% CI 0.803 to 0.814; Fisher's exact test; p<0.001) (table 4). This increased specificity by 18.6% lowered the CT ordering rate but six more cases remained undetermined compared with the classical model (ICH three cases; SDH three cases) (table 4). All of these cases were ≥80 years old, had non-neurosurgical lesions and had a good recovery with the exception of one patient who died at 6 months.

The interrelationship of age with antiplatelet use was further tested at different age cut-offs (online only table AI). Finally, a model was built, including both age ≥75 years and antiplatelet therapy that showed a significantly lower discriminating operating characteristics area 0.783 (95% CI 0.761 to 0.775; p<0.001), a 9.8% decrease in specificity, without significant differences in sensitivity (table 4). The CT ordering rate was now around 50%, and only three cases remained undiagnosed (EDH two cases, 20 years and 59 years; ICH one case, 63 yrs). Of these, only the 59-year-old man with EDH was operated on with an unfavourable outcome. The remaining two subjects had a good recovery at 6 months. The three cases had no symptoms or risk factors, and their lesions would remain undetected by any clinical decision instrument.


This observational analysis shows that pre-injury antiplatelet therapy significantly increases the risk of intracranial lesions in subjects after MHI and should be considered in future reviews of the predictive algorithms on this topic. The inclusion of antiplatelet drugs in our prediction model shows good accuracy and reduces to a minimum the number of undiagnosed lesions (<0.2/1000) with a CT ordering rate of 49%. Also, the type of items (age≥75 years and chronic antiplatelet therapy) may be more reliable than the association of age ≥65 years in the presence of amnesia or loss of consciousness.2

Antiplatelet therapy, mainly aspirin, is commonly prescribed to reduce cardiovascular morbidity and mortality in subjects with cardiovascular disease over 45 years of age.18 Its use has increased considerably in the general Italian population in the past few years, namely in high risk subjects. A recent analysis of drug prescribing in over 300 000 Italian subjects with diabetes showed an increased use of antiplatelet drugs from 15% to 52% in the period 1997–2006,23 and in 2008 over 4% of the Italian population was treated with aspirin.24 The outcome was determined 6 months after injury by checking the death certificates and hospital database and by a structured telephone interview to determine the presence of disability.

In our database, nearly 10% of the whole population was treated with antiplatelet agents which increased to 24.7% in the group of subjects over 65 years. This problem is expected to become more and more relevant, following guidelines indicating the need for antiplatelet drug therapy in a large proportion of older subjects.18 Although the effects on cardiovascular thrombosis is definitely proved, antiplatelet drugs were reported to increase the occurrence of chronic subdural haemorrhage in the elderly,25 26 as well as morbidity and mortality in older subjects with intracranial post-traumatic haemorrhage.27 28

There is a small body of literature on intracranial haemorrhage and aspirin use outside spontaneous events. In a small sample with mild to moderate head injury, the mortality rate of subjects receiving aspirin was higher than normal16 but in a prospective study, low dose aspirin prophylaxis showed no effect in the frequency or types of intracerebral or meningeal haemorrhage in subjects over 60 years following mild and moderate head injury.17

Based on these considerations and the availability of a large database, we looked at chronic antiplatelet drug treatment as a risk factor, potentially adding to the 14 variables considered by the main clinical decision strategies.5 14 In the multivariable analysis, the use of antiplatelet therapy improved the prediction of the covariate panel in the logistic model, and increased the age cut-off to 75 years in a sensitivity analysis (table 4). The age limit of 75 years was empirically selected on the basis of repeated analyses testing of age at 5 year intervals. Data were not corrected for multiple testing and should be confirmed in different settings. The decision about which age threshold should be used (ie, the minimum desired sensitivity) should ideally be based on a cost/benefit analysis of CT scanning.14 In the clinical setting, there is evidence that prediction rules are more and more frequently used by busy physicians as decision aids—that is, they use predicted probabilities of an outcome in the decision making process.14 The inclusion of antiplatelet therapy and age over 75 years as new cut-offs, irrespective of clinical symptoms, to main accepted variables, provides the best sensitivity (99.7%), a good level of specificity (54%) and an acceptable CT ordering rate (49.3%), thus reducing to a minimum the number of undetermined lesions (table 4).

In a few models, age is considered a prediction variable only in the presence of amnesia or loss of consciousness.2 3 29 However, the accepted age cut-offs of the main guidelines (60–65 years) are associated with a very low specificity and an unacceptably high CT ordering rate. In our protocol, old age, independent of signs and symptoms, was not considered an a priori risk factor since it would increase CT ordering rate over 70% without improving the number of undetermined lesions.7 Also, limiting the use of age to subjects with loss of consciousness or amnesia gives a CT ordering rate of over 60%. Old subjects are more likely to develop a head injury by a fall mechanism,29 have mortality rates twofold higher than younger subjects,30 a higher number of complications and need of extended care after discharge.31 However, not age per se, but a series of conditions associated with age probably accounts for the increased risk and the concurrent use of antiplatelet therapy might be pivotal.

Limitations of the study include firstly, the fact that the clinical characteristics of subjects were obtained from variables registered on a prospective database, but all analyses were retrospectively performed, potentially creating a bias in the results. Secondly, the results are based on a single centre large database, with only one district reference hospital, with possible selection bias. An external validation in different settings is needed. Thirdly, elderly subjects may be unable to recollect their medications, and a few cases might have been missed. Given that not all subjects had a CT scan and the assumption that no lesions were present in subjects who did not have a scan and were not readmitted later, the above sensitivity refers to clinically significant lesions. This is likely to underestimate the true sensitivity of predictor variables for intracranial lesions, but missing clinically silent lesions is acceptable to reduce costs, provided that a rescue mechanism is operative. Our ED is the core of a community hospital, serving a well defined population, and we previously reported that possible missing cases did not have an unfavourable outcome, as assessed by death certificates.32

In conclusion, antiplatelet drug therapy is a significant risk factor for the development of intracranial lesions after MHI. The use of these drugs has increased in the ageing population which also explain the lack of significance reported in previous analyses and should be considered in future guideline revisions. In our hands, antiplatelet drug use, coupled with age ≥75 years, independent of amnesia and loss of consciousness, when included as risk factors together with the classical main prediction variables, reduces to a minimum (0.2:1000) the number of cases with undetermined intracranial lesions, with a CT ordering rate of 49%. These results need external validation in different settings to be proposed in a new decision aid.


We are grateful to the Emergency Medical System personnel of the Azienda USL of Forlì for helpful support.



  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the local ethics committee of Azienda USL, Forlì, Italy.

  • Provenance and peer review Not commissioned; externally peer reviewed.