BACKGROUND: Retinal hemorrhages (RHs) occur in inflicted traumatic brain injury (ITBI), accidental traumatic brain injury (ATBI), and some medical conditions, although the reported number, distribution, type, and frequency vary greatly between these different etiologies. We hypothesize that these RH characteristics reliably help to distinguish ITBI from ATBI and nontraumatic etiologies.
METHODS: A 6-year prospective observational study using wide-field retinal imaging (RetCam) was conducted within 24 hours of admission to PICU, on serially recruited children with traumatic and nontraumatic encephalopathies. “Definite” and “probable” ITBI cases were confirmed by multiagency child protection case conferences. Image analysis used digital color and grayscale images for retinal “zoning” and “layering” of hemorrhages.
RESULTS: Significant differences were found between the mean numbers of hemorrhages in ATBI/ITBI, and ITBI/nontraumatic etiologies for the 3 retinal zones (range, P = .003–.009) and for the dot-blot hemorrhages (range P = .001–.002). The mean numbers of RHs per ITBI patient in the peripapillary, macula, and peripheral zones were 14, 28, and 31 respectively. RHs in ATBI were near the optic disc and more superficial than in ITBI, where hemorrhages involved deeper layers (range, P = .003–.039) and were more peripheral (P = .03). The positive predictive value for ITBI in children <3 years with >25 dot-blot (intraretinal) hemorrhages was 93%.
CONCLUSIONS: This prospective study, which included all potential causes of RHs, with objective retinal methodology, has confirmed that a young age and a high dot-blot count are strong predictors of ITBI. This high predictive value may support medicolegal deliberations.
- ATBI —
- accidental traumatic brain injury
- IR —
- ITBI —
- inflicted traumatic brain injury
- NFL —
- nerve fiber layer
- “other” NTE —
- “other” nontraumatic etiology
- RH —
- retinal hemorrhage
- TNTC —
- to numerous to count
What’s Known on This Subject:
Retinal hemorrhages occur in accidental and inflicted traumatic brain injury (ITBI) and some medical encephalopathies. Large numbers and peripherally located retinal hemorrhages are frequently cited as distinguishing features of ITBI in infants, but the predictive value has not been established.
What This Study Adds:
This prospective retinal imaging study found that a diagnosis of ITBI in infants and children can be distinguished from other traumatic and nontraumatic causes by the presence of >25 dot-blot (intraretinal layer) hemorrhages (positive predictive value = 93%).
Retinal hemorrhages (RHs) are a recognized accompaniment of both accidental and inflicted childhood trauma and of many other nontraumatic encephalopathies. Most investigators have only recognized differences in the retinal findings in inflicted, compared with accidental traumatic brain injury, by the presence or absence of RHs,1,2 or if there were very large numbers of hemorrhages described as “too numerous to count” (TNTC), but few have defined precise RH characteristics.
Three prospective studies3–5 have assessed RHs from funduscopy with or without retinal drawings sketched by the examiner in traumatic brain injury in children. Pierre-Kahn et al3 studied children presenting only with a subdural hematoma and found that RH characteristics were not statistically different in those patients presumed to have been shaken in comparison with those with signs of head trauma without a relevant history, and additionally found no RHs in 7 accidental traumatic brain injury cases. Bechtel et al4 included only children with accidental and inflicted trauma and found RHs in abuse were more frequent, more bilateral, more pre-retinal, and more extending to the periphery. Vinchon et al5 included all causes of traumatic injury to children from birth to 24 months and found that the sensitivity value for those patients with inflicted head injury who have a RH (presence or absence) was 75%, and specificity was 93.2%. A number of other retrospective clinical and autopsy studies have been reported,6–15 including 2 studies that used retinal photographs.9,15
We aimed, by means of a prospective study of infants and children admitted to a PICU, to determine, by RetCam examination within 24 hours of admission, the frequency with which RHs occur in serious accidental traumatic brain injury (ATBI) and inflicted traumatic brain injury (ITBI) and those children admitted with “other” nontraumatic etiologies (“other” NTEs). We further aimed to document the number of RHs, their retinal location/s and retinal layer involved, and any other retinal injury from the first retinal image obtained after admission. We sought to estimate the relative risk of these RH characteristics for predicting ITBI in comparison with ATBI and other nontraumatic diagnoses.
This prospective observational study was conducted over a 6-year period, at the Royal Hospital for Sick Children Edinburgh, in the PICU and Neurosciences Department, on serially admitted children with traumatic and nontraumatic encephalopathies. Children were excluded if there was a previous history of head injury or preexisting eye disease, a clinical contraindication (eg, a clinical necessity for pupil responses where intracranial pressure was being monitored, imminent neurosurgery, etc), an inability to recruit or undertake RetCam examination within 24 hours of admission (eg, the child’s clinical state dictated early withdrawal of ventilation/sedation), early demise, and refusal or withdrawal of consent. Newborns postdelivery (from the maternity hospital) were not routinely included. Initial and sequential RetCam imaging was undertaken after fully informed parental consent.
All photographic imaging was done by using a hand-held digital retinal camera (RetCam, Clarity Medical Systems, Pleasanton, CA), with a 120° lens, while the patient was being ventilated as part of their management. The pupils were dilated, and a local anesthetic and coupling agent (carbomer gel) used. For each eye, a standard selection of views was chosen to cover as much of the retina as possible. Routine practice was indirect ophthalmoscopy before all RetCam imaging, and all suspected ITBI cases were attended by a pediatric ophthalmologist. All RetCam images were downloaded unchanged to a dedicated computer for morphometry and counting by using Photoshop7.0 and routinely applying a green filter to produce a grayscale version of each image in Scion Image Analysis software.
In this study, each RH was allocated a retinal zone and layer. We used a previously published zonal classification, with high interrater κ values, namely the peripapillary circle (zone “D”), posterior pole/macula (zone “M”), and the remainder zone “P,” that is, all the peripheral area, including the midperiphery outside zone M.16
A retinal layer was ascribed to each hemorrhage as previously described (vitreous; pre-retinal; nerve fiber layer [NFL]; intraretinal [IR] (dot-blot)/subretinal)17 by a single ophthalmologist investigator (A.T.), to eliminate any potential interobserver error that might result from multiple raters. Previous studies have highlighted a consistent lack of agreement on identification of subretinal hemorrhages in particular.17,18 Subretinal hemorrhages have been counted separately from intraretinal hemorrhages in this article. For the purposes of analysis, RHs were counted from the single best quality image that showed all 3 zones and were expressed as either RHs per eye or RHs per patient. Any retinal injuries were recorded.
A “centrality score” was devised for each subject, where a score of 1 indicated “adjacent to the disc,” 2 at the macula, and 3 at the periphery. An average value was calculated for each subject, weighted by the number of RHs in each zone, giving a score ranging from 1 (all RHs adjacent to the disc and none elsewhere) to 3 (where all RHs are peripheral and none elsewhere).
Ethical approval (LREC 2004/6/2) was granted by the Lothian Research Ethics Committee. All analyses were performed with the use of SPSS (version 14.0.1, Inc; Chicago, IL) and PASW Statistics Package (version 17.0.2; Chicago, IL). Mann-Whitney U tests were used for all analyses, except tests for asymmetry of RHs (Wilcoxon signed rank tests); primary brain injury and presence/absence of RH (χ2); correlation of “severity” of brain and retinal injury (Spearman rank); and multiple logistic regression for investigating predictive factors for ITBI.
We recruited 114 children to this study of whom 79 (69%) were boys. The median age of all patients was 47.8 months (range, 0.01–212.3 months).
Twenty-one cases were designated ITBI after a convened multiagency Child Protection Case Conference19 recommended making provision for the immediate safety of the child. They were ascribed a degree of medical surety of diagnosis20–23 by the authors as either “definite” (n = 11) or “probable” (n = 10). It is difficult to be certain that a diagnosis of ITBI would have been confirmed by the Child Protection Case Conference, and subsequent legal proceedings, if RHs were not taken into consideration. We reviewed the 21 ITBI cases and found that 3 cases did not have RHs, and, of the remaining 18, 12 would likely have been diagnosed ITBI, without reference to RHs.
The median age for all patients who had any RHs (n = 34) was 5·9 months (range, 0.01–176.5 months). Two of the 34 were unable to proceed to RetCam examination, and RHs were detected and characterized only by indirect ophthalmoscopy. Details of the various etiological groups (ATBI, ITBI, and other NTEs) with the frequency of RHs, are shown in Table 1. Accidental injuries were of similar severity to other nonincluded traumatic admissions to PICU (P = .08), and occurred in 57 children (7 with RHs). Those with RHs were due to (1) motor vehicle accidents (pedestrian 3/15, passengers 1/11, or bicycle accidents, 2/10) and (2) high falls, 1/9. There were no RHs in 12 cases (6 infants with “low” [<1 m] household falls, 3 sports injuries, 2 struck on the head, and 1 other domestic accident).
There were a total of 1867 RHs in 49 eyes where RHs could be numerically counted from RetCam images. There were 16 RHs in 3 eyes counted by indirect ophthalmoscopy. In 6 eyes, the hemorrhages were initially deemed TNTC. However, by factoring up the number of hemorrhages in a radial segment of these eyes (where they were evenly distributed), the total number of RHs could be estimated. The mean number of RHs in these TNTC cases was 303 (range, 146–565).
The maximum number of RHs in each zone for each patient (from either the right or left eye, which ever was the greater for that zone) was summated and compared with etiological groups (Table 2). The analysis excluded data from the TNTC cases, because RHs in these cases could not be confidently ascribed a retinal zone, because retinal zoning required clear retinal landmarks (optic disc and vascular arcades). The clinically more relevant arithmetic mean values for any individual patient, in the different etiological groups, are seen in Table 3.
The centrality score gave a mean value of 1·72 for ATBI and 2·11 for ITBI (P = .033), confirming that RHs in ITBI are more peripheral than in ATBI. The mean for the other NTE group was 2·09, and therefore, in this respect, ATBI RH distribution was different from ITBI and other NTEs.
Regardless of cause, in this series, NFL hemorrhages were predominately peripapillary in location, whereas most intraretinal hemorrhages were in the macula and periphery (96%). Within the limits of our imaging technique, very few preretinal hemorrhages were seen to be exclusively in the peripheral zone “P” (0.5%). This is exemplified in the 7 ATBI cases with RHs, in which there were no vitreous hemorrhages, and almost all (ie, 98.9%) the RHs were in the NFL (65·3%) and IR (33·7%) layer. Their density was not only greatest near the optic disc (46%), they were also more superficial than in ITBI. In the same way, in the 15 ITBI patients (excluding the TNTC cases), the deeper layer IR RHs occurred in the periphery (in 93.1% of all peripheral hemorrhages) and in the macula (in 74.4% of all macula hemorrhages), that is, in ITBI, the numbers of RHs were not only greatest in the peripheral zone (43%) but also involved deeper layers (Fig 1A and 1B)
Retinal Layer of Hemorrhage
Total numbers of RHs in the 4 retinal layers in each etiological group were compared by including those seen on funduscopy (3 eyes), and including the 3 patients (6 eyes) that were from “TNTC” cases, that is, a total of 58 eyes. Based on the summated values of all patients entered into this study, Table 2 shows the significant associations between ITBI and ATBI, and also ITBI and other NTEs for the different RH layers involved. Again, the clinically more relevant arithmetic mean values for each individual patient are shown in Table 4.
ITBI Cases Compared With “Non-ITBI” Cases
There were 10 times as many RHs in ITBI as in all other cases in this study combined. In patients with a RH in at least 1 eye, significant differences exist between the maximum (and mean) RH counts between ITBI and the other 2 combined etiological groups (listed in Table 2).
RHs in Nontraumatic Encephalopathy
For the 9 other NTE cases there were no vitreous hemorrhages, and only 1 pre-retinal hemorrhage. 75% of other NTE RHs were deeper-layer hemorrhages (IR), and 97% of these were in the macular or peripheral region. Table 5 shows details of the other NTE cases with RHs.
Asymmetry of RH Distribution
In this study, unilateral RHs occurred more frequently in the left eye (7 cases) compared with the right eye (3 cases), and there was a nonsignificant greater number of hemorrhages in the left eye than in the right in a ratio of 1:5.2. For all cases in this study with RHs, an “asymmetry ratio” of the number of RHs in left to right eyes, when combining unilateral cases with asymmetry in bilateral cases, was not significant (1:1.16).
Primary Brain and Retinal Injury
The more severe the ATBI (lower Glasgow Coma Score), the greater the likelihood of having a RH (P = .012), and the more severe the ITBI, the greater the severity of RH injury, here defined as total numbers of RHs in both eyes (P = .030).
Other retinal injuries can be difficult to determine in ITBI patients, because the severe hemorrhagic appearance often masks additional injury (15 eyes in this study). Retinal injuries, in addition to the RHs, were seen in 7 cases of ITBI, some with up to 3 retinal injuries. For the purposes of our study, retinoschisis was recognized as traumatic splitting of the NFL, or separation by blood of the inner limiting membrane from the NFL, and occurred in ITBI in 7 eyes. Other ITBI injuries included retinal folds (2 eyes) and choroidal tears (3 eyes); 1 eye had an avulsed optic nerve. One 14-year-old child, who died of a very severe injury from a motor vehicle/bicycle collision, had a large macular schisis (left eye) in zone D, with a well-defined cystic appearance and fluid level. A small pre-retinal hemorrhage (right eye, zone M) was also noted.
Prediction of ITBI
The presence of RHs was a very strong diagnostic factor for ITBI in the whole study sample (Table 6). Among those with RH, both young age and high dot-blot count were strongly diagnostic, with young age having greater sensitivity and high dot-blot count greater specificity. A high dot-blot count remained independently significant (P = .017) among those with both RH and young age.
This study has confirmed a frequency of RHs in children with severe ITBI (43% in PICU) of 85.71%. Published frequencies from other prospective studies report 37.5% to 77.5%.1–3,5 We have also found a highly significant greater number of RHs in all retinal zones and retinal layers in ITBI in comparison with cases of ATBI. Average number of RHs in each zone and layer for each individual patient has been tabulated in Tables 3 and 4, and may be useful bedside indicators for further investigation of possible abuse. TNTC is the description used to indicate the extreme end of the hemorrhagic retinopathy appearance in ITBI, and an estimate of the numbers of RHs in the 3 cases in this study show a mean RH count of 303 (range, 146–565).
RH Characteristics in ITBI
In ITBI cases, 93% of all the peripheral hemorrhages were in the IR layer, as were 74% of all macula hemorrhages. Only 2 subretinal hemorrhages were recognized in this study, both located in the same eye (zones M and P) of an 18-month-old child with ITBI. A centrality score was devised, and this confirmed that the RH distribution was significantly more peripheral (ie, a centrifugal pattern) than the more central location in ATBI cases. Bechtel et al4 and Gilliland10 reported a similar peripheral distribution in ITBI. RHs in ITBI cases are therefore more peripheral and involve the deeper layers. The distribution of RH is also influenced to some extent by the retinal layer involved. It is known that the NFL diminishes in thickness toward the ora serrata, and we have found that, regardless of cause, NFL hemorrhages are predominately peripapillary, and 96% of the intraretinal hemorrhages are found in the macula/peripheral zone, all suggesting that the distribution and layer are dictated by factors other than solely the mechanism of injury.
Bilateral RHs are known to be a feature of ITBI,4,24,25 and this study confirms bilateral RHs in 78% of cases, similar to frequencies reported by other investigators.4,26 Unilateral RHs are also described in ITBI,4,9,12,24,25,27–29 and they occurred in 4 cases in this study, but were additionally seen in 3 ATBI cases and in 3 medical patients. The unilateral RHs more often involved the left eye (7/10 cases), a figure similar to the 65% reported by Arlotti et al.27 We have found a nonsignificant greater number of RHs in the left eye in unilateral cases, in bilateral RH cases, and in ITBI cases. An asymmetry ratio combining RH counts from unilateral cases with bilateral RH cases showed little asymmetry (left to right ratio of 1.16:1).
RHs have been reported in ATBI with a frequency range of 0% to 15%.2,4,5,7,30–33 These have been described as few in number, unilateral, at the posterior pole, and involving pre-retinal and intraretinal layers. We report a RH rate of 12.2% in ATBI (90% in PICU); 98% of these were either in NFL (68%) or in the IR layer (33%). Again, the average numbers of RHs seen in children after ATBI in the different retinal zones and layers (Tables 3 and 4) could be clinically helpful. RHs in ATBI therefore were near the optic disc (ie, centripetal pattern) and predominantly superficial.
RHs occurred in children with nontraumatic medical encephalopathies in 9/36 (25%) in this series, including 1 child with a blocked ventriculoperitoneal shunt and raised intracranial pressure, 2 children after cardiorespiratory arrest and cardiopulmonary resuscitation, and 4 with meningococcal meningitis. Dinakaran et al34 described RHs in 5/12 children with meningococcal septicemia. These were flame shaped and dot-blot, and they numbered <20 in each fundus, compared with a mean of 7 intraretinal hemorrhages per patient counted in the macula zone in our study. The large number of RHs in multiple layers (much greater than 25), in a neonate with a large congenital brain tumor (one of the other NTE cases), was the reason the prediction rate for ITBI in our study was not higher than that which we report. RH numbers in these other NTE cases combined with RH numbers in ATBI cases were still significantly fewer than the RHs occurring in ITBI.
A predictive algorithm based on RH characteristics indicated that the chance of a particular case being due to ITBI, if RHs are observed and if the maximum number of IR layer hemorrhages (predominantly dot-blot) was >25, is 93%; and chance of a particular case being due to ITBI is 92% if the patient is <18 months of age. A young age and a high dot/blot (IR) count are therefore strong indicators of ITBI. The relatively low sensitivity value for these characteristics is a reflection of the fact that ITBI may be diagnosed from other clinical features and a different pattern of retinal findings.
Although children with previous head injuries were excluded from entry into this study, no child with ITBI and only 1 with ATBI were actually excluded on this basis. Underreporting of abnormalities may occur in all zones,18 but it was a consistent technical limitation for counting from zone P (and ora serrata) when using the 120° RetCam lens, because of the inherent fading and reduced contrast of the image toward the periphery owing to the oblique viewing angle, and from using a single “centered” photographic image. Future developments in imaging software may allow a combination of multiple images from 1 eye to be used. In future Optical coherence tomography should clarify the depth of the hemorrhages within the retina more accurately than two-dimensional retinal photographs. This article presents the first prospective study with objectively measured RHs from all possible etiologies in children and predicts inflicted brain injury from certain RH characteristics alone.
We are grateful for the support of the Trustees of the R. S. Macdonald Charitable Trust, and thank the medical and nursing staff of PICU and Neurosciences ward at Royal Hospital for Sick Children Edinburgh, Ophthalmology registrars and fellows, parents/guardians of children recruited to the study, and child protection specialists over the 6-year study period.
- Accepted July 2, 2012.
- Address correspondence to Robert A. Minns, FRCP(Edin), FRCPCH, PhD, Emeritus Professor of Pediatric Neurology, College of Medicine and Veterinary Medicine, Child Life and Health, University of Edinburgh, 20 Sylvan Place, Edinburgh, EH9 1UW Scotland. E-mail:
Dr Minns obtained the research grant, designed the project, obtained ethics approval and patients’ consent, obtained clinical data and substantially contributed to the analysis and interpretation, literature search, manuscript writing and submission; Dr Jones obtained patients’ consent, obtained clinical data and substantially contributed to the analysis and interpretation, literature search, manuscript writing, submission, and prepared figures and tables; Dr Tandon undertook clinical (photographic) data collection, categorization of imaging data, and manuscript review; Dr Fleck obtained ancillary funding, undertook clinical (photographic) data collection, and manuscript editing and review; Dr Mulvihill undertook clinical (photographic) data collection, manuscript editing and review; and Dr Elton provided statistical advice and analysis, and critical review of manuscript.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Funding was obtained from the R.S. Macdonald Charitable Trust Edinburgh, from the Royal College of Surgeons Edinburgh (Ophthalmology Grant), and from the Friends of The Sick Kids Foundation, Edinburgh, Scotland. The funders had no role in study design; in the collection, analysis, and interpretation of data; or in the writing of the report or the decision to submit the article for publication.
- Duhaime AC,
- Alario AJ,
- Lewander WJ,
- et al
- Ewing-Cobbs L,
- Kramer L,
- Prasad M,
- et al
- Bechtel K,
- Stoessel K,
- Leventhal JM,
- et al
- Budenz DL, Farber MG, Mirchandani HG, Park H, Rorke LB. Ocular and optic nerve hemorrhages in abused infants with intracranial injuries. Ophthalmology. 1994;101(3):559–565
- ↵Gilliland MGF. Extent and distribution of retinal hemorrhages in abusive and non-abusive head injury. In: Proceedings of the 55th. Annual Meeting of the American Academy of Forensic Sciences; Chicago, IL, February 17–22, 2003; Abstract G86:232–233
- Fleck BW,
- Tandon A,
- Jones PA,
- Mulvihill AO,
- Minns RA
- Mulvihill AO,
- Jones PA,
- Tandon A,
- Fleck BW,
- Minns RA
- Tandon A,
- McIntyre S,
- Yu A,
- et al
- Keenan H
- Thomas SA,
- Rosenfield NS,
- Leventhal JM,
- Markowitz RI
- Maguire S,
- Pickerd N,
- Farewell D,
- Mann M,
- Tempest V,
- Kemp AM
- Shaw AD,
- Watts P,
- Maguire S,
- Holden S,
- Mann M,
- Kemp AM
- Barry GP,
- Romero KE,
- Forbes BJ,
- Christian CW,
- Binenbaum G
- Copyright © 2012 by the American Academy of Pediatrics