If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
To determine if routinely performed computed tomographic (CT) scanning in patients with aneurysmal subarachnoid hemorrhages (aSAHs) is sufficient to identify patients at high risk of vision loss due to Terson syndrome (TS).
Consecutive patients with a diagnosis of aSAH admitted to the neurologic intensive care unit of a regional referral hospital over a 3-year period were prospectively evaluated. Head CT scans performed in the emergency department were assessed for the presence of a “crescent sign” (evidence of significant subinternal limiting membrane hemorrhage). Dilated funduscopic examinations were performed by an ophthalmologist, masked to the results of the CT scan, to identify retinal and vitreous hemorrhages consistent with TS. Retinal hemorrhages were categorized according to size—those smaller than 2 mm in diameter were deemed low risk (lrTS) for vision loss and those larger than 2 mm in diameter were deemed high risk (hrTS) for vision loss.
One hundred seventeen patients with aSAH were enrolled in the study. The overall incidence of TS was 24.9% (29 of 117 patients; 12 were bilateral). Compared to patients without TS, those with TS had a higher Fisher Hemorrhage Grade and a lower mean (±standard deviation) GCS score (8.66 ± 4.97 vs 12.09 ± 1.10; P < 0.001). The CT crescent sign was positive in 7 patients (6.0%), 6 (5.1%; 2 were bilateral) of whom were found to have hrTS. Of the 110 patients without a CT crescent sign, 88 (75.1%) patients did not have TS, 21 had lrTS, and 1 patient had hrTS in one eye. The CT crescent sign was highly sensitive (85.7%) and specific (99.1%) for diagnosing hrTS.
The CT crescent sign is a highly sensitive and specific marker for hrTS. CT scanning may replace routine ophthalmologic examinations to identify patients at risk of vision loss due to aSAH.
The association between subarachnoid hemorrhage (SAH) and retinal hemorrhages was first described by Litten (1881) and was expanded by Terson (1900) to include vitreous hemorrhage, a condition now commonly known as Terson syndrome (TS).
but this low percentage probably reflected several factors, including high mortality and difficulty in adequately examining the fundus with the rudimentary equipment available at the time. Although many physicians still consider TS to be vitreous hemorrhage in patients with SAH, most authors require only the presence of retinal hemorrhages to invoke this diagnosis.
The authors noted correlations between TS and a high Fisher score (3.0 vs 2.32; P = .008) and a low Glasgow Coma Sale (GCS) score (5.55 vs 12.87; P < .001). Bilateral hemorrhages were seen in 6 patients whereas monocular hemorrhages (right eye only) were seen in 5 patients. Seif and coworkers
diagnosed TS in 10 of 46 (21%) patients admitted due to SAH; median GCS and Fisher's scores were, respectively, 6.5 and 4 in patients with TS, and 14 and 3 in patients without TS. For every 1-point increase in the GCS, there was a 0.81-times decrease in the risk of having TS. Other risk factors for the development of SAH (hypertension, smoking, and diabetes) were equal between TS and non-TS groups. Associations between aneurysms of the anterior communicating artery and TS have been reported,
The underlying mechanism responsible for retinal hemorrhages in patients with SAH has been the subject of considerable debate for decades. Some authors have speculated that intraretinal and subhyaloid hemorrhages are composed of blood from the subarachnoid space that is forced into the globe by elevated intracranial pressure.
They suggest that subarachnoid blood may pass into the cerebrospinal fluid surrounding the optic nerves or cause veins within the optic nerve sheath to rupture. Blood subsequently dissects into the sub–internal limiting membrane (ILM) space from the optic nerve margin and tracks through the nerve fiber layer to produce a flame-shaped or domelike hemorrhage between the nerve fiber layer and the ILM.
Subsequent splitting or rupture of the ILM allows blood to accumulate in the subhyaloid space, and an additional episode of intracranial hypertension, such as occurs with an aneurysmal rebleed, may cause subhyaloid blood to extend into the vitreous.
Ballantyne, however, stated that blood within the cerebral subarachnoid space does not pass through the optic foramen into the subarachnoid space of the optic nerve.
He believed that retinal hemorrhages were caused by bleeding into the subarachnoid space from a ruptured cerebral aneurysm that suddenly increased the intracranial pressure. Manshot, however, demonstrated that the subarachnoid space surrounding the optic nerves communicates with that surrounding the brain and that blood passes into the optic nerve subarachnoid space with significant force.
According to each of these theories, sudden elevation of pressure (either intracranial or surrounding the optic nerve) closes the central retinal vein and its choroidal anastomoses. Sudden closure of the central retinal vein where it crosses the intervaginal spaces around the optic nerve halts venous blood flow from the retina and increases intraluminal pressure within all branches of the vein. When coupled with unimpeded arterial blood flow into the retina, the resultant high intraluminal pressure ruptures the veins and causes bleeding into the retina. This theory was later supported by Walsh and Hedges,
and has been the accepted explanation for the development of retinal hemorrhages for the past 64 years. Pressure measurements obtained during endoscopic third ventriculostomies also support the theory that TS results from high intracranial pressure.
Retinal hemorrhages are most commonly seen in the macula and surrounding the optic disc, and are found with decreasing density in the peripheral retina. Hemorrhages may be subretinal, intraretinal, or subhyaloid, all of which satisfy the updated definition of TS.
Iron from hemoglobin catalyzes the conversion of hydrogen peroxide into a hydroxyl radical, the most destructive reactive oxygen species. Photoreceptor damage occurs via lipid peroxidation, cleavage of DNA chains, and biomolecular degradation. Because the primary function of the retinal pigment epithelium is to phagocytose lipid-rich photoreceptor outer segments, both the retina and the retinal pigment epithelium are highly prone to oxidative damage.
A thin layer of subretinal blood will often resolve without permanent sequelae but thick submacular hemorrhages may permanently damage the retinal pigment epithelium. Because of this, many surgeons recommend early evacuation of thick subretinal hemorrhage because of conditions such as exudative age-related macular degeneration. But because patients with TS are often intubated and neurologically unstable without the ability to assume a prone postoperative position, surgical evacuation of subretinal hemorrhage is frequently problematic and is not commonly performed.
Postmortem examinations have supported some of the clinical findings in patients with TS. A patient with a traumatic SAH was found to have blood in the optic nerve sheath, inner retina (outer plexiform layer and inner nuclear layer), and subretinal space.
Histopathologic specimens from patients with TS show erythrocytes and leukocytes in the vitreous, subhyaloid, and sub-ILM spaces and in the retina. Although not found as commonly, subretinal blood has been reported in some studies.
The incidence of subhyaloid hemorrhage in TS ranges from 11% to 33%,
thereby making the vitreoretinal interface particularly susceptible to secondary complications. The growth of epiretinal membranes (incidence of 78%) is one of the most commonly seen long-term retinal complications in patients with TS. Epiretinal membranes form because of hemorrhage-induced fibroblast and glial cell proliferation in the subhyaloid or sub-ILM spaces.
Breakdown of hemoglobin allows some iron to escape but up to 25% remains trapped within ocular tissues. Iron appears to be responsible for liquefaction of the vitreous and for retinal damage in rabbits.
In patients with TS who survive, the retinal hemorrhages usually resolve without leaving residual deficits, but slow-clearing vitreous hemorrhages compromise visual function, impair rehabilitation, and may lead to legal blindness. Timberlake and Kubik reported that 56 of 280 cases of SAH (20%) had retinal hemorrhages but only 4% had vitreous hemorrhages.
but subsequent improvements in neurosurgical techniques and neurologic care have enabled more patients to survive the devastating effects of aSAH, thereby subjecting more to vision loss due to vitreous hemorrhage. Vision loss due to vitreous hemorrhage can be severe and may persist for years because of slow clearing of the blood,
In the pre-vitrectomy era, a wait-and-see approach to vitreous hemorrhage was all that could be done, but modern pars plana vitrectomy surgery enables much earlier visual rehabilitation. Advocates of early surgery believe that all patients with subarachnoid hemorrhage should undergo dilated fundus examinations in the intensive care unit despite the relatively low yield of vision-threatening hemorrhages. Pursuing this strategy is possible in some hospitals, but it identifies few at-risk patients who would benefit from surgery whereas it consumes limited health care resources.
In one series, only 40% of eyes with vitreous hemorrhage required vitrectomy and only half of these also needed peeling of the internal limiting membrane because of persistent pre-macular hemorrhage.
Performing a vitrectomy leads to good results in most of these cases, with more than 80% of patients achieving a final VA of 0.8 or more.
A nonrandomized study found that patients undergoing vitrectomy within 90 days of their subarachnoid hemorrhages had better final visual acuity (79.5% were ≥20/40; 59.1% were ≥20/25) and fewer complications than those who underwent surgery beyond 90 days. The best results were achieved in patients under the age of 45 years.
Survivors of SAH are often left with neurologic damage and must endure a long recovery process. Because of the life-threatening severity of the SAH, it is not surprising that TS is often overlooked. Fundus examinations to look for intraocular hemorrhages to establish the diagnosis of TS are not routinely performed in some neurosurgical intensive care units (NICUs)
In an attempt to optimize the efficiency of ophthalmologic screening, Seif and associates suggested the following guidelines based on the patient's general medical status. By using a GCS score of ≤8 or a Hunt and Hess score of ≥3 to trigger an ophthalmologic examination, 27 of their 46 patients (58.7%) would have undergone screening and 9 of 10 patients with TS would have been diagnosed. This corresponds to a sensitivity of 90%, a specificity of 50%, and a negative predictive value of 94.5%.
Unfortunately, this protocol decreases the number of eye examinations by only 41.3%.
Most patients with SAH either do not have TS or they have TS with small hemorrhages that resolve spontaneously and do not require treatment because they are low risk for vision loss. Only patients with large sub-ILM or vitreous hemorrhages stand to benefit from vitrectomy. But the patients with TS that are at high risk of vision loss are also more likely to have had severe SAHs and may have a higher mortality rate. Performing ophthalmologic examinations on every patient with SAH may not be necessary, and accurate screening guidelines or alternative examination techniques that increase the yield for TS would be welcomed.
The initial emergency department evaluation of patients suspected of having SAH includes a noncontrast computed tomography (CT) scan of the brain performed according to stroke protocols. This enables physicians to both diagnose SAH and grade its severity. Scanning is usually performed within hours of the SAH by which time retinal and sub-ILM hemorrhages will have developed but breakthrough into the vitreous has not yet occurred. A fresh, large sub-ILM hemorrhage can be diagnosed by the appearance of a nodular or crescent-shaped “crescent sign” on CT scans. Up to 66.7% of TS may be diagnosed on the screening CT scan,
In a prospective, comparative study of 48 patients with aSAH, ocular ultrasonography had a sensitivity and specificity of 100% for diagnosing TS, whereas CT scanning had a sensitivity and specificity of, respectively, 60% and 96% for vitreous hemorrhage and 32% and 95% for any ocular bleeding. The agreement between observers for the detection of ocular hemorrhage was good (K: 0.78).
In a prospective study of 36 patients with SAH, 6 of whom (16.7%) had TS, ventriculostomy was required in 80.5% of patients, and those with TS had higher intracranial pressures (40 vs 15 cm H2O; P = .003). Terson syndrome was detected on CT scans with a sensitivity of 50%, a specificity of 98.4%, a positive predictive value of 83.3%, and a negative predictive value of 92.4%.
The methodologies used in these studies to determine risk of vision loss have not been sufficiently refined, and the cohorts have been too small to qualify CT scanning as a standalone method of diagnosing patients at risk of vision loss due to TS. In this manuscript, we describe a prospective series of consecutive patients with SAH due to ruptured cerebral aneurysms (aSAH) to determine the incidence of TS. We created a clinical classification for eyes we believe to be at “high risk” and “low risk” of vision loss due to intraocular hemorrhages, and we correlated the findings on screening CT scans with the clinical risk classification to determine the sensitivity and specificity of CT imaging to identify patients at risk of vision loss due to TS. We hypothesized that the CT crescent sign, as noted on the admission screening CT examination, would be able to detect TS, which puts patients at high risk of vision loss.
This study was performed with the approval of the Mayo Clinic Investigational Review Board. Written consent allowing for the collection, evaluation, and publication of data in this cohort of patients was given by patients (whenever possible) or by their families or individuals responsible for directing their care.
Recruitment of Patients
All patients with SAH admitted to the NICU at the Mayo Clinic Hospital in Jacksonville, FL, between March 2, 2011, and May 30, 2014, were evaluated, and those diagnosed with cerebral aneurysms were enrolled in this study. At the time of enrollment, most patients with SAH were intubated, sedated, or debilitated because of altered mental status, so consent to participate in this study was often obtained from family members.
Neurologic and Neurosurgical Care
Patients were initially evaluated and stabilized in the emergency department according to the usual stroke protocol. The Neurology service was emergently consulted and the patients underwent noncontrast CT scanning of the brain on a 128-slice Siemens Edge machine. Radiologic diagnosis of intraventricular hemorrhage prompted admission to the NICU. The Neurology service evaluated the CT scan for evidence of intraocular hemorrhage and documented the presence or absence of a “crescent sign” in the retina of each eye (Figure 1). This assessment was performed prior to an ophthalmoscopic examination. Evaluation of patients included determination of the Fisher grade of hemorrhage (Table 1) and the GCS score (Table 2).
Table 1Modified Fisher Scale for Grading Subarachnoid Hemorrhages
Table 2Glasgow Coma Scale, Which Is Used to Assess Patients’ General Neurologic Status
Score for eye opening ± score for best verbal response ± score for best motor response = Glasgow Coma Scale score What the score means: Maximum score is 15—best prognosis Minimum score is 3—worst prognosis Scores of 3-8 are usually said to be in a coma Scores of 8 or above have good chance for recovery Scores of 3-5 are potentially fatal, especially if accompanied by fixed pupils or absent oculovestibular responses
Eye Opening Response Spontaneous—open with blinking at baseline. Opens to verbal command, speech, or shouts. Opens to pain, not applied to face
4 points 3 points 2 points 1 point
Verbal Response Oriented Confused conversation, but answers to questions. Inappropriate responses, words discernible Incomprehensible speech 5 points 4 points 3 points 2 points 1 point
Motor Response Obeys commands for movement Purposeful movement to painful stimulus Withdraws from pain Abnormal (spastic) flexion, decorticate posture Extensor (rigid) response, decerebrate posture 6 points 5 points 4 points 3 points 2 points 1 point
The initial score correlates with the severity of the brain injury and the patient's prognosis.
The Neurosurgery service was consulted and further investigations including magnetic resonance angiography and intra-arterial cerebral angiography were performed as needed to identify a bleeding site. Aneurysms were usually closed with intra-arterial “coiling” or external clamping, depending on the anatomic location of the lesion and experience of the neurosurgeon. Throughout the treatment period, the patient's systemic and neurologic status were carefully monitored, and treatment was administered as needed.
Ophthalmologic Assessment and Determination of Risk of Vision Loss
One of the authors (M.W.S.) was consulted between 1 and 7 days after admission of each patient to perform a dilated fundus examination. Patients’ eyes were dilated with topical tropicamide 1% and neosynephrine 2.5%, indirect ophthalmoscopy was performed at the bedside, and ocular abnormalities were noted and recorded in the medical record. The fundus assessment for hemorrhages was performed without knowledge of the CT findings. Most patients received only one funduscopic evaluation, with follow-up examinations performed at the discretion of the ophthalmologist and the NICU team.
The diagnosis of TS was based on the presence of retinal hemorrhages. The risk of vision loss—low risk or high risk—was based on the size of the intraretinal, subretinal, or subhyaloid hemorrhages, and/or the presence of vitreous hemorrhage. Eyes with retinal hemorrhages <2 mm in diameter (approximately the vertical diameter of the optic disc or smaller) were classified as low-risk Terson syndrome (lrTS), whereas eyes with retinal hemorrhages >2 mm in diameter or vitreous hemorrhages were classified as high-risk Terson syndrome (hrTS). The rationale for this classification scheme stemmed from the author's previous experience with TS patients in which hemorrhages <2 mm in diameter resolved spontaneously, never broke through the ILM into the vitreous, and did not leave central visual scotomas, whereas subhyaloid hemorrhages >2 mm in diameter usually included the fovea and frequently evolved into vitreous hemorrhages. Follow-up examinations were performed on surviving hrTS patients who were able to return to our Clinic, and eyes that developed nonresolving vitreous hemorrhages underwent elective pars plana vitrectomies.
A 3-year study was anticipated with the goal of enrolling at least 100 subjects. Characteristics of the study population (means, standard deviations) were determined with descriptive statistics (Excel 2010; Microsoft, Redmond, WA).
Sensitivity and specificity analyses were performed according to the following equations:
Sensitivity = (no. of true positives) (no. of true positives ± no. of false negatives)–1
Specificity = (no. of true negatives) (no. of true negatives ± no. of false positives)–1
One hundred thirty-three patients were admitted due to SAH, of which 117 (234 eyes) were found to have aSAH. Informed consent was obtained from all patients and/or health care directors, and all were admitted to the study. The mean age of these patients was 55.3 years (SD 14.4), and 37 (31.6%) were male (P < .0001; power = 0.98). Only 4 patients (3.42%) were receiving systemic anticoagulants at the time of hemorrhage, and 1 patient tested positive for lupus anticoagulant. Important demographic and clinical characteristics of the cohort are listed in Table 3.
Table 3Demographic and Clinical Characteristics of the Cohort
Sex, n (%)
P < .0001
Age, y, mean ± SD
55.3 ± 14.4
Patients taking systemic anticoagulants, n
Ophthalmoscopic examination, n (%)
Glasgow Coma Scale score, mean ± SD
9.00 ± 1.4
8.66 ± 4.97
12.09 ± 1.10
P < .0001 vs Non-TS
Fisher hemorrhage grade
All patients, n (%)
P = .0467 vs Non-TS
Grade 3, n
Grade 4, n
hrTS, n (%)
lrTS, n (%)
Non-TS, n (%)
Location of aneurysms, n (%)
Cerebrospinal fluid diversion, % of all patients
CT crescent sign, no. of patients (no. of eyes)
Mortality, n (%)
P = .105 vs Non-TS P = .091 vs Non-TS
The results of important probability analyses are listed in the right-hand column.
Severity of Subarachnoid Hemorrhage and Clinical Status
Overall, 2 patients (1.7%) had Fisher grade 1 hemorrhages, 4 (3.4%) had grade 2, 27 (23.1%) had grade 3, and 85 (71.8%) had grade 4 (Figure 2). The locations of aneurysms were varied: 58 (49.6%) were in the anterior circulation, 56 (47.9%) were in the posterior circulation (Figure 3), and 3 patients (2.6%) had aneurysms in both the anterior and posterior circulations.
The average GCS score for all patients was 9.0 (SD 1.4). GCS scores for the 7 patients with hrTS were as follows: 3 (3 patients); 4 (1 patient); 6 (1 patient); 8 (1 patient); and 9 (1 patient). Patients with TS had significantly lower GCS scores 8.66 ± 4.97 (mean ± SD) compared to those without TS (12.09 ± 1.10; P < .0001; power = 0.945). Central spinal fluid diversion was performed in 99 (84.6%) patients because of elevated intracranial pressure.
Ophthalmoscopic Findings and “Crescent Sign”
CT images of the brain were examined by the NICU service to determine the presence or absence of the crescent sign. Of the 117 patients, positive crescent signs were seen in 7 patients (6.0%). Two of these patients had positive crescent signs in both eyes and 5 were read as positive in one eye only. CT scans were read as negative for CT crescent signs in both eyes in the remaining 110 patients (94.0%).
On indirect ophthalmoscopic examination, 29 patients (24.79%) had retinal hemorrhages consistent with TS. Of these, 7 had hrTS (6% of all aSAH patients; 24.1% of all TS patients). Two of these patients had hrTS in both eyes, 3 had hrTS in one eye and lrTS in the fellow eye, and 2 had hrTS in one eye but no hemorrhages in the fellow eye. The remaining 22 patients (18.79% of all aSAH patients; 75.9% of all TS patients) had lrTS. Seven of these had lrTS in both eyes and 15 had lrTS in one eye but no hemorrhages in the fellow eye.
Correlation Among Indirect Ophthalmoscopy, CT Scans, and Clinical Course
Of the 7 patients with positive CT crescent signs, 6 had hrTS on indirect ophthalmoscopy and 1 had lrTS. Of the 110 patients with negative CT crescent signs, 109 had lrTS or negative fundus findings and only 1 patient had one eye with hrTS. The sensitivity of the CT crescent sign for diagnosing hrTS was calculated to be 85.7%. The specificity of the CT crescent sign for ruling out hrTS was calculated to be 99.1%. Of the 109 patients with negative CT crescent signs and without hrTS, 21 patients had lrTS and 88 had no ophthalmoscopic evidence of TS (Figure 4).
Twenty-three patients (19.67%) in the total cohort expired during the initial hospitalization. The mortality rate among all patients with TS (8 of 29 [31.03%]) was not significantly higher than among those without TS (15 of 88 [17.05%]; P = .105, power = 0.378) and neither was it significantly higher for patients with hrTS (3 of 7 [42.86% vs 17.05%]; P = .091).
The prevalence of TS was also assessed with respect to the severity of aSAH, using the Fisher grading system. All patients with TS had a Fisher grade of 3 or 4, and the incidence of TS among aSAH patients with a Fisher grade of 4 was 34.5%. Only 4 of 29 patients (13.79%) with TS patients had a Fisher grade of 3 or less, whereas 29 of 88 patients (32.95%) without TS had a Fisher grade of 3 or less (P = .0467, power = 0.514). Patients with TS had a mean Fisher score of 3.86, whereas those without TS had a mean Fisher score of 3.60 (Figure 2).
Taking anticoagulants at baseline (TS: 1 of 29; non-TS: 3 of 88) was not associated with the risk of developing TS, but the numbers were too small to be relevant. The location of the causative aneurysm (anterior vs other) did not appear to correlate with the risk of developing TS (TS: 18 of 29; non-TS: 40 of 88; P = .12; power = 0.338).
One of the hrTS patients underwent bilateral vitrectomy, with visual acuities of 20/30 OD and 20/80 OS at the 6-month follow-up examination. None of the other patients with either hrTS or lrTS were known to have undergone vitrectomy, but some patients in each group lived remotely from the Clinic and were lost to follow-up after discharge from the hospital. Complete follow-up data regarding final visual acuities and need for vitrectomy were not available for these patients.
Terson syndrome may be responsible for 5.5% of all nondiabetic and nontraumatic vitreous hemorrhages,
and though it may be caused by several conditions, most cases are due to ruptured cerebral aneurysms. During the 3-year enrollment period of our study, 85% of the patients who were admitted to the NICU of our adult referral hospital with SAH had cerebral aneurysms as the underlying cause of their bleeding. Most patients in our series were female (68.4%), which is similar to the 70% rate reported in a series of 213 patients,
and our patients tended to be relatively young (mean age of 55.3 years) with well-formed, highly viscous vitreous that often prevents spontaneous clearing of vitreous hemorrhage. Similar to the findings from most previously published series, we failed to find an association between TS and the location of the aneurysm.
The incidence of TS in our series—according to the broad definition that includes any retinal hemorrhage—was 24.79%, though only 6% of patients had what we termed hrTS. Because retinal hemorrhages result from a rapid increase in intracranial pressure, one would expect the retinal circulation of each eye to be equally affected. But we noted considerable asymmetry in both the distribution and severity of retinal hemorrhages. Most of our patients with TS (55.17%) had hemorrhages in only one eye, and only 2 of 7 patients with hrTS had high-risk hemorrhages in both eyes.
Patients with TS had significantly higher Fisher scores and lower GCS scores compared to those without TS. Because the Fisher score is a radiologic assessment of the subarachnoid hemorrhage density, it probably serves as an indicator of the rapidity and severity of the cerebral hemorrhage and the resultant increase in intracranial pressure. GCS scores for patients with hrTS ranged from 3 to 9, which suggests that patients with higher scores may not require ophthalmologic screening. But because several patients with lrTS had GCS scores greater than 9, we are uncomfortable making this recommendation, and if ophthalmoscopic examinations were limited to patients with GCS scores of ≤9, the number of examinations would be reduced by 68.4%. Alternatively, one might consider not performing eye examinations on patients with GCS scores of 3 because of the high associated mortality. Three hrTS patients had a GCS score of 3; 2 of these patients died during hospitalization but 1 patient experienced a remarkable neurologic recovery, underwent bilateral vitrectomies, and had excellent postoperative visual acuity. Therefore, we believe that all patients with a positive CT crescent sign should undergo ophthalmologic evaluation, regardless of the GCS score.
Retinal hemorrhages were found within the nerve fiber layer and inner retina of our patients, but none were subretinal. Because most hemorrhages were not contiguous with the optic nerve, they probably originated from the superficial and deep retinal capillary plexuses. This is consistent with the theory that retinal hemorrhages occur because elevated cerebral pressure increases central retinal venous pressure, but is inconsistent with the theory that blood from the subarachnoid space surrounding the optic nerve dissects directly into the eye.
Some patients may develop vitreous hemorrhage within the first hour after the intracranial hemorrhage because of the sudden rise in ICP,
These descriptions of delayed hemorrhages are similar to the findings in our series. Our earliest eye examination was approximately 24 hours after the SAH, which prevented us from commenting on the early appearance of retinal hemorrhage. In our 7 cases with hrTS, all patients developed large (>2 mm) subhyaloid hemorrhages within the first week, but only 1 patient had a minor vitreous hemorrhage during this period.
Our series was not designed to determine the long-term visual results in patients with hrTS. Only 7 patients in our series developed vitreous hemorrhage or large subhyaloid hemorrhage, and most of these patients died during the original hospitalization or were not seen after discharge because of their distance from our clinic. But one of our patients, a 24-year-old man with nonclearing bilateral vitreous hemorrhages, underwent vitrectomy surgeries from 5 to 6 months after his aSAH and improved his visual acuities from counting fingers OU to 20/30 OD and 20/80 OS.
From our series we are unable to determine which patients should undergo vitrectomy surgery or what constitutes the optimal timing for surgery, and randomized controlled trials have not been reported. Persistent vitreous hemorrhage limits rehabilitation efforts, which can be problematic because these patients frequently have neurologic deficits resulting from high Fischer grade hemorrhages and low GCS scores. When one also considers that these patients are at risk of vitreomacular interface abnormalities and that modern vitrectomy techniques are remarkably safe and effective, early vitrectomy is appropriate. If the vitreoretinal surgeon determines that the vitreous hemorrhage has improved little over 3 months and the patient's neurologic status suggests that survival is likely and surgery poses little systemic risk, vitrectomy surgery should be considered.
Though aSAH are rare, they are frequently fatal, with an estimated mortality rate of 40% within the first 48 hours. Some authors have reported increased mortality in aSAH patients who also have TS. A prospective study found a 90% mortality rate in patients with SAH and TS, compared to only 10% in patients with SAH but without TS.
We found that patients with TS and those with hrTS tended to have higher mortality rates compared to those without TS, but the differences were not statistically significant.
Diagnosing TS may be important for 2 reasons. First and probably more importantly, it identifies patients at risk of vision loss, particularly those with bilateral hemorrhages that may hinder rehabilitation.
Second, establishing the diagnosis of TS helps physicians predict neurologic outcomes. Indirect ophthalmoscopy through dilated pupils is the gold standard for diagnosing TS, but some authors believe that ocular ultrasonography can be used to make the diagnosis of TS (Figure 5).
Therefore, ultrasonography may be useful for diagnosing hrTS but will not likely detect the small, thin hemorrhages that characterize lrTS. Although ultrasonography may have a role in the diagnosis of hrTS, it has important limitations. Most ophthalmic ultrasonographs are performed by ophthalmologists or ophthalmic technicians in the outpatient setting, and general ultrasonographic technicians and radiologists may be unfamiliar with performing and interpreting ophthalmic examinations. Therefore, high-quality ocular ultrasonographs are not available in most hospitals.
We hypothesized that the presence or absence of a crescent sign on the admission CT scan would enable us to identify with high sensitivity and specificity patients at high risk of vision loss due to TS. Small retinal hemorrhages, though sufficient to establish the diagnosis of TS, are of low concern because they do not put patients at risk for vision loss (Figure 6). Based on our previous clinical experience, we categorized eyes as low risk (lrTS) and high risk (hrTS) for vision loss based on the size of the retinal hemorrhages (<2 or >2 mm). Using these definitions, we found that 6% of our patients with aSAH had hrTS and 18.79% had lrTS.
We determined that the CT scan was highly sensitive (85.7%) and specific (99.1%) for identifying patients at high risk of vision loss. Had we performed ophthalmologic examinations only on patients with a positive CT crescent sign, only 1 patient (out of 117) would have had an unnecessary eye examination (a false positive crescent sign in an eye with lrTS) and only 1 patient with hrTS would have been missed (a false negative crescent sign). The very high specificity (99.1%) of the CT crescent sign in this study nearly completely eliminates the need for eye examinations in patients without hrTS. The sensitivity (85.7%) is also high, though it is based on a sample size of only 7 patients. Nevertheless, this sensitivity exceeds that of all previously published studies that have used CT imaging as a screening tool for TS. Despite this impressive sensitivity score, basing the need for an ophthalmoscopic examination solely on the CT crescent sign would miss 1 in 7 patients with hrTS. This may be an acceptable risk when one considers that 94% of eye examinations would be eliminated if they were performed solely on the basis of the CT crescent sign. Based on these data, only 1 patient in 7 with hrTS would not have been diagnosed within the first week of hospitalization, but surgeries for hrTS patients are usually deferred until patients are healthier (generally 3 months), and opportunities for later diagnosis and timely intervention are still possible.
As a result of these findings, Mayo Clinic Florida no longer performs routine eye examinations on patients with aSAH. Examinations are requested by the Neurology team only when the crescent sign is seen on the screening CT scan. We believe that the results of this study are compelling enough to limit eye examinations to patients with crescent signs on screening CT scans. We look forward to other prospective, masked studies that may validate our conclusions.
We acknowledge several strengths and weaknesses of this study. The 3-year, prospective, consecutive patient design of this study constitutes one of its strengths. Masking of the physicians was never broken—the participating neurologists determined the presence or absence of a CT crescent sign before the ophthalmology examination was performed, and the examining ophthalmologist was masked as to the interpretation of the CT scan. Although the cohort of 117 patients with aSAH was large, the small hrTS cohort (7 patients) limits the strength of the sensitivity calculation. Long-term follow-up of our patients was poor because many returned to distant homes after discharge. Because most patients received only 1 ophthalmologic examination (during the first week), we cannot be sure that delayed, repeat hemorrhages did not result in new cases of lrTS or hrTS, or worsening of previous hemorrhages.
Perhaps the designation of hrTS and lrTS based on the size of the largest sub-ILM hemorrhage constitutes the greatest weakness. This classification was based on decades of experience by the senior author; simply stated, small hemorrhages (<2 mm) had never been observed to cause decreased vision or breakthrough into the vitreous, whereas large hemorrhage (>2 mm—and most hrTS hemorrhages are much larger) usually involve the macula and frequently break through the ILM into the vitreous. We feel comfortable with the merits of this classification scheme but we admit that it needs to be validated by other investigators.
In conclusion, we found that the presence of a crescent sign on screening CT examinations is highly sensitive and specific for identifying patients at high risk of vision loss due to TS from aSAH.
Funding/Support: This study received no funding. Financial Disclosures: Dr Stewart is a consultant for Alkahest; receives institutional research support from Allergan , Kanghong , and Regeneron ; and is on the advisory board of Bayer. All authors attest that they meet the current ICMJE criteria for authorship.
Ueber Einige von Allgemein-Klinischen Standpunkt aus Interessante Augenveränderungen [in German].