British Journal of Ophthalmology 1996; 80: 282-287

Ocular and cerebral trauma in non-accidentalinjury in infancy: underlying mechanisms andimplications for paediatric practice

M A Green, G Lieberman, C M Milroy, M A Parsons

Department ofForensic Pathology,University of SheffieldM A GreenG LiebermanC M Milroy

Ophthalmic SciencesUnit, University ofSheffieldM A Parsons

Correspondence to:DrM A Parsons,Ophthalmic Sciences Unit,Royal Hallamshire Hospital(Floor 0), Glossop Road,Sheffield S10 2JF.Accepted for publication11 October 1995

AbstractAims-To determine the sites, mech-anisms, and clinical significance ofinjuries to the eyes and brains of childrenwith non-accidental injuries in relation todiffering levels oftrauma.Methods-A forensic pathological study ofinjuries in the eyes and brains of 23 con-secutive children dying of non-accidentalinjuries over a 4 year period (1988-92)under the jurisdiction of Yorkshire andHumberside coroners.Results-Sixteen children died from cere-bral injuries and seven died from non-cerebral causes. There were highincidences of retinal detachment (63%)and subhyaloid (750/0), intraretinal (75%),and perineural (68%) haemorrhages inCNS deaths. Local subhyaloid haemor-rhages and retinal detachment were morecommon at the periphery and optic discthan at the equator. There was a strongcorrelation between CNS and eye traumascores in all 23 children (r=07551,p<O-OOOl). Ranking of injuries by severitysuggests progressively more traumarequired for (a) subdural haemorrhage,(b) subhyaloid, intraretinal, perineuralhaemorrhages, and (c) retinal detach-ment. At highest trauma levels choroidaland vitreous haemorrhages were associ-ated with additional cerebral lacerations,intracerebral and subarachnoid haemor-rhages.Conclusions-In non-accidental (andprobably accidental) infantile head injurythe earliest eye injuries (coinciding withsubdural haemorrhage) could be missed ifindirect ophthalmoscopy is not per-formed. Retinal detachment and multiple(particularly choroidal/vitreous) haemor-rhages may indicate additional cerebrallacerations and/or intracerebral haemor-rhage. Vitreous traction is the likely causeofintraocular pathology.(Br3' Ophthalmol 1996; 80: 282-287)

Retinal lesions associated with head injurywere described first by Purtscher in 1910,1 andin 1946 Caffey noted in infants the associationof subdural haematomas, long bone injuries,and retinal haemorrhages with minimal orabsent signs of external trauma.2 Physical childabuse was first described by Tardieu as longago as 1860,3 but in 1962 Kempe4 coined theterm 'battered child syndrome', and laterCaffey described injuries associated with

shaking infants in 1972,5 and used the term'whiplash shaken infant syndrome' in 1974.6Since these early reports numerous ocularinjuries have been recognised as common innon-accidental injury to children, includingperiorbital haematoma, eyelid laceration, sub-conjunctival haemorrhage, subluxed or dis-located lens, cataracts, glaucoma, anteriorchamber angle regression, iridodialysis, retinaldialysis or detachment, intraocular haemor-rhages, optic atrophy, and papilloedema(reviewed by Kaplan7 and Morris8). Lesionsmay be multiple, and at different stages fromseparate episodes of trauma.

Retinal haemorrhages are a frequent findingin childhood non-accidental injury, andEisenbrey9 and Guilland et all' regard suchhaemorrhages as diagnostic of child abusewhen accompanied by intracranial injuries inthe absence of a verifiable history of severeaccidental head injury. The exact pathogenesisof these haemorrhages is disputed. Severalmechanisms have been postulated, includingdirect jracking of blood from intracranialbleeding," haemorrhage secondary to raisedintracranial pressure,12 and retinoschisis (split-ting) related to vitreous traction forces.13 Theprecise method by which trauma is inducedhas also been disputed, and Duhaime et al 14considered that impact of the head with solidsurfaces was required in addition to shaking ofthe child. However, it is now increasinglyaccepted that shaking alone can cause thecerebral injuries,'5 and that rotational forcesand whiplashing affect the child's head duringshaking,'6 and on this basis a campaign hasbeen launched to educate parents and othercarers about the dangers of shaking infants.'7We initiated our study of ocular and cerebral

injuries in infants who died as a result of non-accidental injury to determine whether itwould be possible, by analysis of the precisesites and extent of injuries, and by rankingpatients in order of severity of traumatic

Key pointsIn non-accidental (and probably accidental) headinjury in infants:* the earliest intraocular injuries are local

subhyaloid haemorrhage and local retinaldetachment; these may be peripheral andmissed without indirect ophthalmoscopy

* ifANY intraocular injury is present subduralhaemorrhage is very likely (and can developlater)

* retinal detachment and multiple haemorrhages(particularly choroidal or vitreous) may indicateadditional cerebral lacerations or intracerebralhaemorrhage

282

on June 9, 2020 by guest. Protected by copyright.

http://bjo.bmj.com

/B

r J Ophthalm

ol: first published as 10.1136/bjo.80.4.282 on 1 April 1996. D

ownloaded from

Ocular and cerebral trauma in non-accidental injury in infancy

lesions, to determine both the relative levels oftrauma required for each injury and the orderin which injuries are sustained with increasinghead trauma. As pathologists we felt that thisapproach was most likely to produce informa-tion about the pathogenesis of individualcerebral and ocular lesions in non-accidentalinjury and accidental head injury. We alsohoped that by ranking cerebral and ocularinjuries in order of severity we might be ableto demonstrate associations between, forexample, ocular clinical signs and specificcerebral injuries which would be useful inclinical paediatric practice.

Subjects and methodsWe examined the detailed files of all childrenup to 3 years old who had died under the juris-diction of the Yorkshire and Humbersidecoroners during the 4 year period 1988-92.We initially selected 25 children for this studywhose death was attributable to unequivocalnon-accidental injury, determined fromreviewed autopsy findings, witness statements,and court transcripts of evidence to establishthe circumstances surrounding each death. Afull and very detailed 'forensic' necropsy hadbeen performed in all cases by members of theDepartment of Forensic Pathology in Sheffieldand Leeds (departments amalgamated into theDepartmnent of Forensic Pathology of theUniversity of Sheffield in 1991; the depart-ment examines all cases of suspicious deathoccurring in Yorkshire and Humberside).Twenty three of the original 25 children dyingof non-accidental injury had eyes removed forexamination, and these 23 cases are the sub-jects of this study.The eyes from all 23 children with non-

accidental injury were examined initially in situafter removal of the roof of the orbit, to deter-mine the extent of orbital and periocular trau-matic abnormalities, and to establish the extentof perineural (subdural) haemorrhage into theoptic nerve sheath. The eyes were thenremoved, using the minimum of trauma, andfixed in formaldehyde, then alcohol. After fixa-tion we examined the eyes externally beforestandard removal of superior and inferiorcalottes, embedding in paraffin wax, andpreparation of five horizontally orientatedhaematoxylin and eosin sections through thecentres of the optic nerve and pupil, and oneadditional section of each of Perls' Prussianblue for haemosiderin and Martius scarlet bluefor fibrin. We recorded traumatic ocularlesions by determining the presence (+) orabsence (0) in one or both eyes of retinaldetachment (using standard histologicalcriteria, and excluding artefactual retinaldetachments attributable to formalin fixationor postmortem trauma); and vitreous, sub-hyaloid, intraretinal, choroidal, and perineural(optic nerve) haemorrhages. These findingswere added to produce an eye trauma score,with and without retinal detachment ('total eyetrauma' and 'haemorrhage' scores respec-tively). Table 1 was produced by placingchildren into groups with and without retinal

detachment, then ranking them according tototal eye trauma scores. Where eye traumascores are the same, ranking was performed bydeeming choroidal haemorrhage more severe(more trauma required) than vitreous or sub-hyaloid haemorrhage.We recorded the presence and extent of

central nervous system (CNS) traumaticlesions by noting the presence (+) or absence(0) of subdural, subarachnoid and intra-cerebral haemorrhage, primary axonal injury,and cerebral lacerations, with very extensivehaemorrhages scoring '+ +'. We produced an'intracranial trauma score' by addition of theseindividual injury scores.

In order to determine the precise position ofintraocular lesions, we noted (in five sectionsper eye) the presence or absence of retinaldetachment and subhyaloid haemorrhage atfive points between the ora serrata (the point atwhich the peripheral retina meets the parsplana of the ciliary body) anteriorly and theoptic disc posteriorly, using a microscope-linked video camera and Archimedes 420/1computer with 'Digit' image analysis. Resultswere plotted as the percentage of positivefindings at each of five relative retinal distancepoints.We performed statistical analysis of trauma

scores between groups of patients using thenon-parametric Mann-Whitney test. We usedFisher's exact test to determine the significanceof different incidences of specific traumaticlesions between different groups, and the non-parametric Spearman rank correlation methodto quantitate the degree of linear associationbetween CNS and ocular trauma scores. Allp values are two tailed/two sided, and resultswere deemed 'significant' at p values of 0 05 orless.

ResultsOf the 25 children under the age of 3 yearswho died of non-accidental injury under thejurisdiction of the Yorkshire and Humbersidecoroners in the period 1988-92, two wereexcluded from the study because the eyes hadnot been removed for pathology. Death wasattributable to head injury in 16 (70%) of the23 subjects of this study; the remaining seven(30%) died of non-accidental injuries otherthan head injury (including asphyxia, abdomi-nal injuries, etc).

Table 1 illustrates the range and extent oftraumatic eye injuries and central nervoussystem (CNS) injuries in the 16 subjects whodied as a result of head injury, and the singlepatient with ocular lesions from the group ofseven whose deaths were a result of injuriesother than head injury (the six subjects in thelatter group without CNS or ocular lesions arenot shown in this table). The mean age atdeath was 6 9 months, with a range of 0 5 to 24months.

All but one of the 16 infants whose deathwas attributed to head injury had subduralhaemorrhage (94%); in eight of these (53%)subdural haemorrhage was the only localisedCNS injury, but the remaining seven children

283

on June 9, 2020 by guest. Protected by copyright.

http://bjo.bmj.com

/B

r J Ophthalm

ol: first published as 10.1136/bjo.80.4.282 on 1 April 1996. D

ownloaded from

Green, Lieberman, Milroy, Parsons

Table 1 Ocular and intracranial trauma in non-accidental injury in infants

Group A Group B Group C*

1 24 7 4 2 20 1 2-5 4 3 2 10 24 0-5 2 1 10

_1 ~ ~ ~~~~~~~5 7 13 4 12 10 1 1 14 8 9 2 3 15 16 17 6Retinaldetachment + + + + + + + + + + 0 0 0 0 0 0 0Vitreoushaemorrhage 0 0 0 0 0 0 0 0 0 0 0 0 0 0 +Subhyaloid haemorrhage 0 0 0 +Intraretinal haemorrhage 00 0 0 0 +Choroidal haemorrhage _ O O O O O O i { O O OPerineural haemorrhage 0 0 0 0 0 0Haemorrhage score 4 4 4 4 4 4 3 3 3 0 4 3 2 0 0 0 3Total eye trauma score 5 5 5 5 5 5 4 4 4 1 4 3 2 0 0 0 3

Yl 7 3 3 4 6 1 3 3 0 1 1 2 2 1 1 0Subdural haemorrhage + + + + + +± + + + + + ++ 0 + + + + ++ + + 0Subarachnoid haemorrhage 0 + 0 + + 0 + + 0 + 0 0 0 0 0 0 0 0 0Intracerebral haemorrhage 0 + + 0 0 + + + 0 0 + 0 0 0 0 0 0 0 0Primaryaxonal injury 0 + 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0Cerebral lacerations 0 + + 0 0 + 0 0 0 0 0 0 0 0 0 0 0

Shaking+ + + +Shaking/beating + + + + + + + ± + + + +Asphyxia +

Traumatic eye and intracranial injuries, modes of injury, and ages of infants dying of non-accidental injury in Yorkshire and Humberside in the period 1988-92.The notations '+' and '+ +' indicate the presence and severity of injuries, and constitute eye or CNS trauma scores of 1 and 2 respectively. Patients are in threegroups in order of cause of death and severity of intraocular trauma and aetiology of the lesions. Infants in groups A and B have died as a result of cerebral injury,and are divided according to presence (group A) or absence (group B) of retinal detachment. Group C consists of seven infants who died of injuries other thancerebral trauma.*Six patients in group C (cases 18-23 inclusive) have neither cerebral nor eye trauma, and are not included in this table. The illustrated group C case of non-accidental injury was due to asphyxia, and ocular haemorrhages were present. See text for further explanations.

(47%) had one or more additional CNSinjuries which included subarachnoid haemor-rhage, intracranial haemorrhage, and cerebrallacerations. When we compared the groups ofchildren dying from head injury with deathsfrom other causes there was a very strongassociation between head injury death and thepresence of subdural haemorrhage (p<00001,Fisher's exact test), but not with the presenceof the other recorded traumatic CNS lesions.The presence or absence of cerebral oedemawas difficult to assess with any accuracy fromthe pathological findings which we were able toreview, and for this reason was not recorded.In one case (case 8), however, cerebral oedemawas recorded as the only demonstrable cere-

bral abnormality secondary to shaking of theinfant. We found no evidence of scalphaematomas or soft tissue contusions on thehead in any of the children who died fromCNS injury.There is a strong association between CNS

injury and traumatic ocular lesions. Eyeinjuries were present in 81% of the 16 childrenwho died ofhead injuries but only 17% (one ofseven) of those whose death was due to othercauses (p<0005, Fisher's exact test). Childrenwho died of CNS injuries had a significantlyhigher 'total' eye trauma score (3-25 (SD 0 50)versus 0O43 (043), p<0Q004) and 'haemor-rhage' score (2-63 (042) versus 043 (043),p=001) than those without head injury

(Mann-Whitney test). We also founda very strong positive correlation in all 23children between CNS trauma scores and both'total' eye trauma scores (Spearman rank cor-

relation, r=0-7551, p<00001), and 'haemor-rhage' scores (Spearman rank correlation,r=0-7254, p<00001).The incidences of subhyaloid, intraretinal,

and perineural (optic nerve) haemorrhageswere significantly increased in children whodied as a result of CNS trauma, and in thosewith subdural haemorrhage (Table 2).Although all of the children with perineural(subdural) optic nerve (sheath) haemorrhageshad co-existing subdural haemorrhages, theperineural haemorrhages involved only thedistal segment of the nerve, and did not involvethe proximal part of the optic nerve within theapex of the orbit, confirming that there was nodirect connection between these two separatesites of haemorrhage. Although the incidenceof retinal detachment was significantlyincreased in children dying with CNS injury(Table 2), the increased incidence of retinaldetachment in those with subdural haemor-rhage alone was not quite significant(p=0074).We were interested to determine whether

there was a relation between retinal detach-ment and CNS injury and other eye injuries inthe children who had died of CNS trauma.There were proportionally more individual

Table 2 Specific traumatic eye injuries in infants dying of central nervous system injuries in non-accidental injury

CNS deaths Subdural haemorrhage

No Yes Statistical No Yes StatisticalEye injuries (n= 7) (n= 16) significance (n=8) (n= 15) significance

Retinal detachment 0 10 (6255%) p<005 1 (12-5%) 9 (60%) NSVitreous haemorrhage 0 2 (12-50/) NS 1 (12-5%) 2 (13%) NSSubhyaloid haemorrhage 1 (140/o) 12 (75%) p<0 02 1 (12-5%) 12 (80%) p<0 006Intraretinal haemorrhage 1 (14%) 12 (75%) p<0 02 1 (12-5%) 12 (80%) p<0O006Choroidal haemorrhage 0 5 (31-30/%) NS 0 5 (33%) NSPerineural haemorrhage 0 11 (68-8%) p<0005 0 11 (73%) p<0 002

There is a significant association between subhyaloid, intraretinal, and perineural optic nerve haemorrhage and death due to bothcentral nervous system (CNS) injury and subdural haemorrhage. Retinal detachment is significantly increased in association withdeath due to CNS injury, but its increased incidence in subdural haemorrhage is not quite significant statistically (p=0 074)(Fisher's exact test, two sided p values). NS=not statistically significant (p>0 05).

284

on June 9, 2020 by guest. Protected by copyright.

http://bjo.bmj.com

/B

r J Ophthalm

ol: first published as 10.1136/bjo.80.4.282 on 1 April 1996. D

ownloaded from

Ocular and cerebral trauma in non-accidental injury in infancy

CNS injuries in the whole group of subjectswith retinal detachment (42%) than in thegroup of those with CNS injury but no retinaldetachment (25%) (Table 1), but this is notquite statistically significant (p=0 533,Fisher's exact test. NB, '+ +' counted as '+' forthis calculation). Similarly, although the 'CNStrauma' score and 'haemorrhage' score arehigher in subjects with retinal detachment thanin those without (3- 1 (0 69) versus 1 -33 (021),p=0 093 and 3-3 (0-40) versus 1-50 (0 72),p=0-056 respectively), this was not quitestatistically significant. There were no statisti-cally significant differences between incidencesof the individual intracranial injuries listed inTable 1 and presence or absence of retinaldetachment. There were proportionally moreindividual 'haemorrhagic' ocular injuries(shaded areas in Table 1) in the whole groupof subjects with retinal detachment (66%)than in the group of those without retinaldetachment (66% versus 30% respectively,p<0003, Fisher's exact test). Although thereare much higher proportions of subjects withindividual 'non-detachment' eye injuries withretinal detachment than without detachment(Table 1, shaded areas) the numbers are small,and perineural (optic nerve) haemorrhage isthe only individual lesion with a statisticallysignificantly increased incidence in childrenwith retinal detachment than without detach-ment (90% versus 33% respectively, p<004,Fisher's exact test).

SITES OF LESIONSFigure 1 shows the incidence and relativeretinal sites of true (non-artefactual) retinaldetachment and subhyaloid haemorrhage inthe 10 and 12 subjects, respectively, with theselesions. The incidence of both of these lesionsis lowest at the equator of the eye, but is almostdoubled at the periphery of the retina/oraserrata, and is increased (but less so) pos-teriorly around the optic disc and in relation tothe macula.

DiscussionWe believe that this study of infants with non-accidental injury, the first to combine the iden-tification of the nature and precise site ofocular injuries with quantitation of the extentof injuries in the CNS and eye, provides newinformation which indicates the mechanism ofthe injury, and has important implications forclinical paediatric practice.

It is important to place the extent andlimitations of this study in context. Our find-ings are based on the study of 23 consecutiveinfants with non-accidental injury in a 3 yearperiod, 16 of whom died of central nervoussystem trauma and seven of whom died ofinjuries to other areas of the body. We felt thatit was vital to the accuracy of the study toexclude cases where the diagnosis of non-acci-dental injury was in any doubt, and also toexclude cases where there was no opportunityfor either histological examination of the eyes,or pathological examination of the CNS. The

50 -m Subhyaloid haem

40 _ Retinal detachment

30-

20-

10-

01 2 3 4 5

Relative retinal positionFigure 1 Relative positions of retinal lesions (subhyaloidhaemorrhage and local retinal detachment) in infants dyingof non-accidental head injury. The incidence of these lesions(% positivefindings in five sections per eye) is highestanteriorly in the peripheral retina/ora serrata (position 1),next highest posteriorly at the disc/macula (position 5), andlowest at the equator (position 3). This suggests thattraction by the vitreous at the sites of its attachment, the oraserrata and the optic disc, causes the injuries at these sites.

study is also biased towards more seriousocular and CNS injuries, as 'survivors' of non-accidental injury (with, presumably, lessserious injury) are excluded. We believe, how-ever, that we have obtained valid results fromthis study by statistical analysis of non-selectedconsecutive cases of unequivocal non-acciden-tal injury deaths in infancy.

It is clear that all except one of the infantswho died as a result of head injury had sub-dural haemorrhages which ranged from smallto extensive, and this finding is in keeping withmany previous studies.6 10 11 14 15 The singlecase (case 8) without subdural haemorrhage,however, died of documented cerebral oedemaalone, and it is likely that cerebral oedema waspresent as an important factor in many 'headinjury' deaths in our study, by analogy withhigh incidences reported by Munger et al'8and Bundenz et al.19 It is interesting to notethat subarachnoid haemorrhages, intracerebralhaemorrhages, and cerebral lacerations werenever present without coexisting subduralhaemorrhages in the infants with head injury inthis series, although 63% of the infants hadsubdural haemorrhage as the only localisedintracranial injury. This suggests that lesstrauma is required for the development of sub-dural haemorrhage than for subarachnoid andintracerebral haemorrhage or cerebral lacera-tion, a theory in keeping with the knownpathogenesis of these injuries in adults.Subarachnoid haemorrhage, intracerebralhaemorrhage, and cerebral lacerations werefound only in the group of patients with retinaldetachment. Infants with retinal detachmentalso had a significantly higher incidence ofboth perineural haemorrhage and total 'non-detachment' ocular injuries, and this supportsthe supposition that more trauma is requiredfor intracranial traumatic lesions other thansubdural haemorrhage. All patients with peri-neural (subdural) optic nerve haemorrhagehad coincident intracranial subdural haemor-rhages, although these were also present infour of the five patients without perineuraloptic nerve haemorrhage. Examination ofonly the anterior half of the intraorbital opticnerve might lead to the conclusion thatthese intracranial and perineural subdural

285

on June 9, 2020 by guest. Protected by copyright.

http://bjo.bmj.com

/B

r J Ophthalm

ol: first published as 10.1136/bjo.80.4.282 on 1 April 1996. D

ownloaded from

Green, Lieberman, Milroy, Parsons

haemorrhages were related. However, wefound, as did Munger et al18 that there was noperineural optic nerve haemorrhage at the apexof the orbit, and conclude that perineural opticnerve haemorrhage is a separate entity in thesechildren, and not the result of direct tracking ofintracranial subdural haemorrhage into theoptic nerve sheath.

INTRAOCULAR TRAUMATIC LESIONSOur findings indicate that there is a significantlyhigher proportion of 'whole group' intraocularinjuries in infants with retinal detachment thanin those without retinal detachment, althoughperineural optic nerve haemorrhage was theonly individual ocular lesion with a statisticallysignificantly increased incidence in this study. Alarger study is required to determine whether, aswe suspect, there is a relation between thepresence of retinal detachment and other indi-vidual intraocular injuries, and with CNStrauma scores. The incidences and patient dis-tribution of subhyaloid and intraretinal haemor-rhages are very similar to the incidence ofperineural optic nerve haemorrhage but,because they are both present in one morepatient without retinal detachment, this is notquite significant statistically. Choroidal haemor-rhage is about half as frequent as subhyaloid,intraretinal, and perineural haemorrhages, indi-cating either that greater trauma is required forchoroidal haemorrhage (as the case rankingsuggests), or that the choroid is more distantfrom the source of trauma than the other typesof haemorrhage. Vitreous haemorrhage is theleast common form ofintraocular haemorrhage,and was present in the two (of five) infants withhighest eye trauma scores (but withoutchoroidal haemorrhage) and probably requiresthe somewhat idiosyncratic rupture of the pos-terior hyaloid membrane.

RELATION BETWEEN CNS AND OCULAR INJURIESOur results indicate conclusively that there is adirect relation between the presence andseverity of both intracerebral trauma andintraocular injuries. These various injuries canbe scored and ranked in order of severity,which allows inferences to be made about theorder in which the injuries are sustained withincreasing severity of head injury or levels ofhead trauma. Thus, with increasing trauma tothe head (Table 3), an infant is likely to

Table 3 Head trauma and developing central nervous system (CNS) and eye injuries innon-accidental injury in children

Traumalevel CNS injury Eye injury

0"' None INone

develop subdural haemorrhage first, followedby (at slightly higher levels of trauma) sub-hyaloid and intraretinal haemorrhages andperineural (subdural) optic nerve sheathhaemorrhage. Slightly more trauma willproduce retinal detachment, and only the mostsevere trauma will cause choroidal and vitreoushaemorrhages, which coincide with develop-ment of subarachnoid and intracerebralhaemorrhages and/or cerebral lacerations. Thissequence of events with increasing traumagives some insight into the mechanism ofthe injuries. It is well recognised thatshaking/whiplash injury to the head producessubdural haemorrhage, as the inertia of thebrain (accelerating from a stationary position)or its momentum (when the moving head issubject to rapid deceleration) results in move-ment ofthe brain relative to the meninges, withresulting trauma to thin veins, and subduralhaemorrhage.With slightly more head trauma than that

required for subdural haemorrhage, ocularinjury occurs in the form of subhyaloid andintraretinal haemorrhage, and haemorrhage inthe most anterior position of the optic nervesheath. With further trauma, retinal detach-ment occurs. We believe that these four lesionscan all be explained by invoking a mechanismsimilar to that causing subdural haemorrhage- namely, the effect of inertia/momentum ofthe vitreous body within the eye during cyclesof acceleration/deceleration during shaking.Our findings on plotting the exact site of theretinal lesions, which show that subhyaloidhaemorrhage and retinal detachment are mostfrequent at the periphery of the retina, andposteriorly in relation to the optic nerve, sup-port this hypothesis, as these are the precisesites at which the vitreous is most stronglyattached to the retina.20 Thus, the mass of thevitreous (which forms a large proportion of themass of the globe) acts upon the points atwhich it is attached during both acceleration ordeceleration, with up/down, left/right, axial orrotational movement, producing first smallhaemorrhages in the retina (intraretinal andsubhyaloid haemorrhage) then retinal detach-ment (directly, with haemorrhage having anadditive effect).

Optic nerve sheath haemorrhage is, webelieve, the result of angular, rotational, oraxial movement of the eye about a point in themost anterior part of the optic nerve, posteriorto the sclera (the posterior half of the opticnerve is relatively fixed by the bony marginsof the posterior orbit and by the extraocularmuscles). Again, the mass of the vitreous formsthe majority of the mass of the eye, and distor-tion and stretching of the small veins com-municating between the optic nerve and itsdura (sheath) are likely to be the major sourceof this haemorrhage.More trauma is likely to be required to inflict

the ocular injuries than the trauma needed fordevelopment of subdural haemorrhagebecause, although the angular momen-tum/angular acceleration and deceleration areidentical at all sites, the size and mass of thebrain (rotating about the top of the cervical

The table indicates the types of CNS and eye injuries with increasing levels of head trauma.Note that the trauma scale is relative but not necessarily linear (point 1 represents major traumawhich increases further to point 4), and is intended to indicate a trend only.

286

on June 9, 2020 by guest. Protected by copyright.

http://bjo.bmj.com

/B

r J Ophthalm

ol: first published as 10.1136/bjo.80.4.282 on 1 April 1996. D

ownloaded from

Ocular and cerebral trauma in non-accidental injury in infancy

spine) together lead to greater relative speed,forces, and distance travelled by the brainrelative to meninges than vitreous relative toretina, or optic nerve relative to nerve sheath.The pattern of sequential development and

coexistence of the cerebral and ocular injurieshave important clinical and prognostic impli-cations for paediatric practice. Both ocular andcentral nervous system injuries such as we havedescribed have been demonstrated in acciden-tal and non-accidental childhood injury byGilliland et al,°0 and we agree that, in theabsence of a verifiable history of accidentaltrauma, the combination of life threateningCNS injuries and ocular haemorrhage arelikely to be the result of child abuse, althoughretinal haemorrhages can be associated with avariety of other causes of raised central venouspressure including epileptic fits, asphyxia (as inour case 6), and subarachnoid haemorrhage,and can be present in normal neonates.21

In addition, however, our results giveclinically important new insights into the typesof lesions which may coexist in children withhead injury of accidental or non-accidentalaetiology. For example, an unconscious childwith a combination of intraocular (vitreous,subhyaloid, intraretinal, and choroidal)haemorrhages and retinal detachment has avery high chance of intracerebral haemorrhageor lacerations in addition to subdural haemor-rhage, the most common complication, andthis has important implications for surgicalmanagement and prognosis. With lesserdegrees of, albeit serious, head injury theearliest and most clinically apparent injuries inthe eyes, subhyaloid, and/or intraretinalhaemorrhage followed by small retinal detach-ments, may precede the onset of signs ofcerebral injury.22 Our findings indicate that themost likely site for these intraocular lesions willbe in the peripheral retina (which is only visibleby indirect ophthalmoscopy), and that inci-dence of these lesions will be lower in the equa-torial retina and in relation to the optic nerve(where angular momentum and vitreousattachment strengths are relatively smaller). Asboth subhyaloid and intraretinal haemorrhagesrequire more trauma for their production thansubdural haemorrhage, this intracranial com-plication should be suspected in all childrenwhere these haemorrhages or retinal detach-ment are present.

CONCLUSIONSWe believe that our study provides, for the firsttime, firm evidence that intraocular injuries innon-accidental head injury are due to the effectof vitreous traction forces, and that a similarmechanism is likely in accidental head injury

accompanied by ocular injuries. The earliestintraocular lesion to accompany subduralhaemorrhage in accidental or non-accidentalhead injury is likely to be peripheral subhyaloidhaemorrhage, with or without local retinaldetachment, and these peripheral retinallesions could be missed on clinical examina-tion if indirect ophthalmoscopic examination isomitted. At higher trauma levels intraocularhaemorrhage at multiple sites combined withretinal detachment may indicate the presenceof cerebral lacerations or intracranial haemor-rhage in addition to subdural haemorrhage,with important clinical and prognostic implica-tions.

1 Purtscher 0. Noch unbekannte Befunde nach Schadeltrauma. Ber Zusammenkunft Dtsch Ophthalmol Ges 1910;36: 294-301.

2 Caffey J. Multiple fractures in the long bones of infantssuffering from chronic subdural haematomas. Am JRoentgenol 1946; 56: 163-73.

3 Tardieu A. Etude medico-legale sur les services et mauvaistraitments exerces sur des enfants. Ann Hyg Pub Med Leg1860; 13: 361-98.

4 Kempe CH, Silverman FN, Steele BF, Droegemueller W,Silver HK. The battered-child syndrome. JAMA 1962;181: 14.

5 Caffey J. On the theory and practice of shaking infants: itspotential residual effects of permanent brain damage andmental retardation. AmJDis Child 1972; 124: 161-9.

6 Caffey J. The whiplash infant syndrome: manual shakingby the extremities with whiplash-induced intracranialand intraocular bleeding linked with permanent braindamage and mental retardation. Paediatrics 1974; 54:396-403.

7 Kaplan B. Ophthalmic issues in family violence and neglect.In: Albert DM, Jakobiec FA, eds. Principles and practice ofophthalmology. Philadelphia: WB Saunders, 1994; 5:3757-62.

8 Morris DA. Ocular trauma. In: Gamer A, Klintworth GK,eds. Pathobiology ofocular disease. A dynamic approach. 2nded. New York: Marcel Dekker, 1994: 387-432.

9 Eisenbrey AB. Retinal haemorrhage in the battered child.Child's Brain 1988; 5: 40-4.

10 Gilliland MGF, Luckenbach MW, Chenier TC. Systemicand ocular findings in 169 prospectively studied childdeaths: retinal haemorrhages usually mean child abuse.Forensic Science International 1994; 68: 117-32.

11 Brain RB. Diseases of the nervous system. 5th ed. Oxford:Oxford University Press, 1955.

12 Gilkes MJ, Mann TP. Fundi of battered babies. Lancet1967; 2: 468-9.

13 Greenwald MJ, Weiss A, Oesterle CS, Friendly DS.Traumatic retinoschisis in battered babies. Ophthalmology1986; 93: 618-25.

14 Duhaime A-C, Gennarelli TA, Thibault LE, Bruce DA,Margulies SS, Wiser R. The shaken baby syndrome, aclinical pathological and biochemical study. Jf Neurosurg1987; 66: 409-15.

15 Hadley MN, Sonntag VKH, Rekate HL, Murphy A. Theinfant whiplash-shake injury syndrome: a clinical andpathological study. Neurosurgery 1989; 24: 536-40.

16 Kleinman PK. Diagnostic imaging in infant abuse. Am J7Roentgenol 1990; 155: 703-12.

17 Carty H, Ratchiffe J. The shaken infant syndrome. Parentsand other carers need to know of its dangers. BMJ 1995;310: 344-5.

18 Munger CE, Peiffer RL, Bouldin TW, Kylstra JA,Thompson RL. Ocular and associated neuropathologicobservations in suspected whiplash shaken infant syn-drome. A retrospective study of 12 cases. Am J ForensicMed Pathol 1993; 14: 193-200.

19 Budenz DL, Farber MG, Mirchandani HG, Park H, RorkeLB. Ocular and optic nerve hemorrhages in abusedinfants with intracranial injuries. Ophthalmology 1994;101: 559-65.

20 Sebag J. Anatomy and pathobiology of the vitreo-retinalinterface. Eye 1992; 6: 541-52.

21 Cavanagh N. Non-accidental injury. In: Taylor D, ed.Pediatric ophthalmology. Boston: Blackwell, 1990: 545-50.

22 Spaide RF, Swengel RM, Scharre DW, Mein CE. Shakenbaby syndrome. Am Fam Physician 1990; 41: 1145-52.

287

on June 9, 2020 by guest. Protected by copyright.

http://bjo.bmj.com

/B

r J Ophthalm

ol: first published as 10.1136/bjo.80.4.282 on 1 April 1996. D

ownloaded from