- Email: [email protected]
Clinical Predictors of Scleral Rupture After Blunt Ocular Trauma
Clinical Predictors of Scleral Rupture After Blunt Ocular Trauma Jan A. Kylstra, M.D., Jeffrey
c. Lamkin, M.D., and Desmond K. Runyan, M.D.
We conducted a two-part study to define better the clinical predictors of scleral rupture after blunt trauma. In part 1 we ascertained the prevalence of scleral rupture among a population of patients examined in an ophthalmic emergency room with severe blunt ocular trauma over a six-month period. Scleral rupture was diagnosed in ten of 283 patients (3.5%). In part 2 we compared the clinical findings in 29 patients with scleral rupture to those of 273 patients with no scleral rupture after blunt trauma. We noted that eyes with visual acuity of light perception or less, an intraocular pressure of 5 mm Hg or less, an abnormally deep or shallow anterior chamber, or a media opacity preventing a view of fundus details by indirect ophthalmoscopy, should be considered ruptured when severe intra- or periocular hemorrhage is present. This diagnostic algorithm had a sensitivity of 100.0% (98.7% to 100.0%), specificity of 98.5% (97.1% to 99.9%), and a positive predictive value of 71.4% (66.3% to 76.5%).
T HE DIAGNOSIS of scleral rupture after blunt ocular trauma can be dlfficult.v" Intraocular hemorrhage can prevent visualization of a posterior rupture site and extensive subconjunctival hemorrhage can hide an anterior scleral defect. Several studies have examined the clinical character istics of eyes with scleral ruptures caused by blunt trauma.i" Other studies have compared clinical findings in ruptured eyes Accepted for publication Jan. 5, 1993. From the Departments of Ophthalmology (Dr. Kylstra) and Social Medicine and Pediatrics (Dr. Runyan), University of North Carolina School of Medicine, Chapel Hill, North Carolina; and the Department of Ophthalmology, Massachusetts Eye & Ear Infirmary, Boston, Massachusetts (Dr. Lamkin). Reprint requests to Jan A. Kylstra, M.D., UNC School of Medicine, 617 Clinical Science Bldg., CB No. 7040, Chapel Hill, NC 27599-7040.
530
with findings in eyes that were presumed to be ruptured but that were discovered, at operation, to have intact sclera.P Though these studies provide useful general information, they do not give the clinician a specific set of guidelines for deciding whether to explore surgically or simply observe an eye which has sustained severe blunt trauma. We formulated a clinical algorithm, with known sensitivity, specificity, and positive predictive value, to diagnose scleral ruptures in patients who have sustained severe blunt ocular trauma.
Material and Methods Our study consisted of two distinct parts. In part 1, we estimated the prevalence of scleral rupture in the population of patients examined in the emergency room of the Massachusetts Eye and Ear Infirmary after severe blunt ocular trauma. The records of all patients seen in the emergency room during the six-month period between Sept. 1, 1988, and Feb. 28, 1989, were reviewed by one of us (J.eL.) to select eyes that had been exposed to severe blunt ocular trauma. To meet this qualification, patients had to have a history of ocular or periorbital contact with a blunt object within seven days before initial ophthalmic examination. Additionally, each patient had to be given one or more diagnoses consistent with the history of blunt trauma; these included and were limited to the following: subconjunctival hemorrhage, corneal abrasion, traumatic iritis, iris sphincter tears or other focal defects, iridodialysis, hyphema, lens dislocation, cataract, vitreous base avulsion, vitreous hemorrhage, commotio retinae, retinal tear, retinal dialysis, retinal detachment, choroidal rupture, traumatic optic neuropathy, orbital fracture, and orbital contusion. All patients in whom there was documented or potential sharp or lacerating trauma, as well as those with any obvious lacerating injuries in or
©AMERICAN JOURNAL OF OPHTHALMOLOGY
115:530-535,
APRIL,
1993
Vol. 115, No.4
Scleral Rupture After Blunt Ocular Trauma
around the eye were excluded. Any patient initially examined more than seven days after injury was likewise excluded from the study. An eye was classified as ruptured only if a scleral defect was directly visualized (either at initial examination or intraoperatively). An eye on which an exploratory operation was performed was classified as not ruptured if no scleral defect was noted. An eye on which an exploratory operation was not performed was classified as not ruptured only if the entire anterior scleral surface could be visualized and demonstrated no evidence of scleral defect and only if the entire fundus could be visualized to the ora serrata and no scleral defect was seen. Any patient whose documented findings at initial examination or on follow-up examination did not exclude the possibility of undetected scleral rupture was not included in the study. In part 2 of the study, we compared clinical findings, on initial examination, in patients with scleral ruptures after blunt trauma (cases) to findings in patients in whom scleral rupture had been adequately excluded (controls). The control population consisted of patients identified in part 1 of the study as having no scleral rupture. The cases consisted, in part, of those patients in part 1 of the study in whom scleral rupture was surgically documented. To increase the number of cases, operating-room records were reviewed to identify blunt scleral ruptures managed over the 18 months before the study period. These 19 cases (which had been included in a previously published study-) were added to the first group of cases to form the entire case sample. We reviewed the clinical records of all cases and control subjects and recorded the mode of injury, the visual acuity in each eye, the intraocular pressure in each eye, and the results of ophthalmoscopy. We also noted the presence or absence of an afferent pupillary defect, asymmetry in anterior chamber depth between the two eyes, extraocular motility disturbance, hemorrhagic chemosis, hyphema, vitreous hemorrhage, or orbital fracture. Not all records were complete and, therefore, data for each clinical finding were not recorded for each patient. Absent data were recorded as such. The following assumptions, however, were made. If a hyphema was not explicitly noted on the chart, it was recorded as absent. If an orbital fracture was not documented radiographically, it was recorded as absent. The data were entered into a microcomputer spreadsheet program and then analyzed using
531
Systat, a microcomputer-based statistical package (Systat, Inc., Evanston, Illinois). We analyzed our data in three ways. First, we performed univariate analysis on each of the following dichotomous variables: visual acuity of light perception or worse, intraocular pressure less than 10 mm Hg, intraocular pressure less than 6 mm Hg, afferent pupillary defect, ocular motility disturbance, hemorrhagic chemosis, hyphema, anterior chamber depth asymmetry, ability to visualize fundus details (retinal vessels or optic disk) with the indirect ophthalmoscope, and orbital fractures. Eyes with missing data were not included in the analysis. (The number of eyes with missing data for each of the clinical findings examined can be determined from the Table by subtracting the denominator of the fraction found in the second column from 29, for the ruptured eyes, and the denominator of the fraction in the third column from 273, for the control eyes.) We calculated the sensitivity, specificity, and a positive likelihood ratio for each clinical feature against a standard of diagnosis at operation. We determined significance by using the chi-square statistic. With the prevalence data, obtained in part 1, we were able to calculate positive predictive values for each clinical sign as a diagnostic test. Ninety-five percent confidence intervals were calculated for all proportions. Our second step was to combine these variables in various ways to produce clinical algorithms for diagnosing scleral rupture after blunt trauma. We then evaluated these algorithms to determine sensitivity, specificity, positive likelihood ratio, and (by using the prevalence data from part 1) positive predictive value. All algorithms were required to have 100% sensitivity. We chose the algorithm with the highest specificity as our clinical test or guideline for evaluating eyes subjected to severe blunt trauma. Our final step was to reanalyze our clinical algorithm by performing logistic regression analysis, using combinations of the previously mentioned variables, to predict the presence of scleral rupture. Because a considerable amount of data was not recorded, especially in the ruptured group, only five variables were included in this analysis (we excluded variables that were recorded for fewer than 20 of the 29 ruptured eyes). Additionally, for purposes of this multivariate analysis alone, it was assumed that indirect ophthalmoscopy was performed in all eyes and that a failure to note fundus details meant that no fundus details were visible.
532
April, 1993
AMERICAN JOURNAL OF OPHTHALMOLOGY
Results In part 1 of the study, we determined the prevalence of scleral rupture in 294 eyes of 294 consecutive patients who were examined in the emergency room over a six-month period for having sustained severe blunt ocular trauma. Eleven eyes were not included in this study because of insufficient follow-up. Of the remaining 283 patients, ten were identified as having sustained a scleral rupture. Thus, the prevalence of scleral rupture in our population of patients who had sustained severe blunt ocular trauma was 3.5% (1.4% to 5.6%). Fifteen of the 283 eyes were suspected of having a scleral rupture and were explored and repaired in the operating room. Five of these eyes were shown not to be ruptured. The falsepositive diagnosis rate was therefore 33.3% (9.0% to 57.0%). None of the eyes initially considered nonruptured turned out on followup to be ruptured. The false-negative diagnosis rate was therefore 0.0% (0.0% to 0.5%). In part 2 of the study, we compared clinical findings among eyes with and without scleral rupture and produced a diagnostic algorithm.
Univariate analysis-We evaluated each of the ten clinical factors separately as a test for diagnosing scleral ruptures (Table). The clinical findings with the greatest positive predictive value were as follows: intraocular pressure of 5 mm Hg or less, visual acuity of light perception or less, anterior chamber depth asymmetry, and inability to visualize fundus details with the indirect ophthalmoscope. The clinical findings found most commonly in eyes with scleral ruptures were as follows: hyphema, inability to see fundus details, visual acuity of light perception or less, and hemorrhagic chemosis. We combined these variables in different ways to produce a clinical test for diagnosing scleral rupture after blunt ocular trauma. The best result we obtained was the following: an eye should be presumed ruptured if the visual acuity is light perception or less, if there is anterior chamber depth asymmetry, if the intraocular pressure is 5 mm Hg or less, or if no fundus details can be seen by indirect ophthalmoscopy provided that there is evidence of severe ocular bleeding (hyphema, no fundus visible, or hemorrhagic chemosis). This test correctly identified 29 of the 29 ruptured globes
TABLE CLINICAL FINDINGS AS PREDICTORS OF SCLERAL RUPTURE NO. WITH FINDING/TOTAL
POSITIVE
POSITIVE
RUPTURED
INTACT
SENSITIVITY
SPECIFICITY
PREDICTIVE
LIKELIHOOD
FINDING
EYES
EYES
(%)
(%)
VALUE (%)'
RATIO (%)
Vision < hand motions Intraocular pressure < 10 mm Hg Intraocular pressure < 6 mm Hg Afferent pupillary defect Motility disturbance!
23{29
1{268
79.3
99.6
88.0
198
11{17
12{247
64.7
95.2
32.9
13.5
9{17
0{247
52.9
100.0
100.0
>999
6{11
2{254
54.5
99.2
71.3
68.1
2{10 19{28
25{240 24{273
20.0 67.8
90.6 91.2
7.4 21.9
2.1 7.7
24{27 17{23
27{273 1{273
88.9 73.9
90.1 99.6
24.8 87.2
9.0 185
18{21 2{29
3{273 38{273
85.7 6.9
98.9 86.0
74.3 3.8
77.9 0.5
Hemorrhagic chemosis Hyphema Anterior chamber asymmetry Fundus invisible Facial fracture
'Using an estimated prevalence of 3.5%. !O '> .05 by chi-square (all others P < .05).
Vol. 115, No.4
Scleral Rupture After Blunt Ocular Trauma
as ruptured and incorrectly identified only four of 273 nonruptured eyes as ruptured. Thus, in our patient population, this clinical test had a sensitivity of 100% (98.7% to 100.0%), a specificity of 98.5% (97.1 % to 99.9%), a positive likelihood ratio of 68 (36 to 999) and a positive predictive value of 71.4% (66.3% to 76.5%). Regression analysis-To evaluate the appropriateness of our clinical test (which was derived by trial and error) we performed linear and logistic regression analyses on our data by using the software program. We first constructed a linear regression model to predict the presence of a scleral rupture on the basis of the following variables: hemorrhagic chemosis, inability to see fundus details, visual acuity less than hand motions, anterior chamber asymmetry, and hyphema. We confirmed this model with a logistic model. Analysis confirmed that all of these variables, except hemorrhagic chemosis, independently contributed to the accuracy of the model (P < .05). The r value was .932 for a model that used these variables. The most useful variables were inability to visualize fundus details, visual acuity of light perception or less, and anterior chamber depth asymmetry.
Discussion Our study provides a useful test for diagnosing scleral rupture in eyes that have sustained severe blunt trauma: all eyes with severe ocular hemorrhage (intraocular or periocular) that have either visual acuity of light perception or worse, media opacification precluding a view of fundus details by indirect ophthalmoscopy, abnormally deep or shallow anterior chamber, or intraocular pressure of 6 mm Hg or less should be surgically explored. All other eyes should be observed. This test was designed to be sensitive (100.0% in this study population) because we believe that the potential benefit of rapid diagnosis and repair of these severely injured eyes outweighs the risk of an exploratory operation in patients with nonruptured eyes. Despite this high sensitivity, the test is also highly specific. It should be noted, however, that the use of this algorithm should not replace good clinical judgment in examining individual patients. Though our clinical guideline was derived by trial and error, we began the process with several concepts in mind. Scleral ruptures al-
533
ways involve rupture of the underlying highly vascular choroid or ciliary body. Severe hemorrhage, either intraocular or subconjunctival, or both, must therefore always accompany scleral rupture. For this reason our clinical guideline states that the presence of either hyphema, hemorrhagic chemosis, or severe vitreous hemorrhage (as evidenced by inability to visualize fundus details) is necessary, but not sufficient to diagnose scleral rupture. We chose the other factors for our guideline by including those clinical findings that were most specific for scleral rupture as determined by the univariate analysis. In order not to compromise sensitivity, we structured the guideline so that a positive diagnosis is made if anyone of the four most specific factors is present in an injured eye (provided that severe hemorrhage is present). We determined the prevalence of scleral rupture in a population of patients treated by ophthalmologists for ocular complications of blunt trauma. Our estimate of 3.5%, however, is similar to the results obtained in a recent study in which one of 94 consecutive patients (1.1 %) who were referred for ophthalmic examination after hospitalization for multiple blunt injuries was found to have a ruptured globe." It appears, therefore, that scleral rupture after severe blunt trauma is infrequent. We also calculated the sensitivity, specificity, and positive predictive values for individual clinical predictors of globe rupture. Two previous studies have attempted to define clinical guidelines for diagnosing blunt ocular rupture by comparing clinical findings in ruptured eyes with findings in nonruptured eyes. 1,2 These studies, however, only examined patients who were operated on, most of whom had ruptured globes. These studies are limited for three reasons. Possibly, some patients whose eyes were not suspected of being ruptured and who were not operated on, did have ruptured globes. Clinical guidelines derived from these studies, therefore, might not be sensitive enough to detect the less obvious cases. It is also possible that, because of the limited size of the control group, clinical characteristics found only in ruptured eyes in these studies are found in many nonruptured but traumatized eyes that were not included in the studies. Guidelines derived from these studies might, therefore, also not be specific enough to avoid a considerable number of false-positive diagnoses. Finally, these studies did not determine the prevalence of scleral rupture in the population at risk. Without this information, the positive pre-
534
AMERICAN JOURNAL OF OPHTHALMOLOGY
dictive value of any set of clinical guidelines or of any single clinical finding cannot be determined. Thus, though previous studies have demonstrated an association between the presence of a hyphema, for example, and the presence of scleral rupture, they do not allow the clinician to determine the risk of globe rupture in a patient with a hyphema. Our study suggests that at least one specific recommendation found in the published reports is incorrect. One study! states that all eyes with intraocular pressures of less than 10 mm Hg after blunt trauma should be surgically explored. Our data suggest, however, that almost 5.0% of unruptured eyes that have sustained severe blunt trauma will have intraocular pressures below 10 mm Hg and that two thirds of eyes operated on because of this clinical finding will not be ruptured. Our results do demonstrate that low intraocular pressure is a specific clinical sign of scleral rupture. Analysis of our intraocular pressure data, however, suggests that the intraocular pressure must be 5 mm Hg or less before it becomes a specific clinical sign of scleral rupture. Previous studies have suggested that visual acuity of light perception or worse, intraocular pressure less than 10 mm Hg, the presence of a hyphema, the presence of hemorrhagic chemosis, and an asymmetry in anterior chamber depth are all more common in ruptured than in nonruptured eyes.!" Our study confirms these results. Our study also suggests that 25.0% of eyes with hyphema and 22.0% of eyes with hemorrhagic chemosis will have a scleral rupture. This information is useful for nonophthalmologists working in the emergency room setting who can easily document these clinical findings, even in uncooperative patients, and treat eyes with these findings as if they were ruptured, that is, place a protective shield over the eye and consult an ophthalmologist. Our study is retrospective in design and is therefore flawed because of missing data. The problem of missing data is of particular concern in two areas. First, 11 of the patients in the prevalence part of this study were excluded from analysis because not enough follow-up information was available to place them in either the ruptured or the nonruptured group. It is possible that all of these patients had ruptured globes and that the true prevalence of ruptures in our population was 21 of 294 or 9.9%. However, there is no reason to suspect that the patients with missing data differ markedly from the study group either in prevalence
April, 1993
of ruptures or in distribution of the clinical findings. If any difference exists it could be argued that the eyes with missing data were less likely to be ruptured than the eyes included in the study because the patients did not return for follow-up and because patients with serious injuries would be more likely to return for examination. The effect of missing data in the form of individual clinical observations is a second weakness of this study and is a problem inherent in retrospective studies in general. We decided to handle these missing data, in most cases, by simply recording them as absent. Of the clinical factors included in our algorithm, only intraocular pressure and afferent pupillary defect had a considerable number of missing entries in both the ruptured and nonruptured groups. Additionally, many missing entries were also found for ocular motility in the ruptured group. However, the only one of these three factors we included in our clinical algorithm for predicting blunt rupture was intraocular pressure. Intraocular pressure was included in the algorithm because of its apparently great specificity as a predictor; that is, an intraocular pressure of less than 6 mm Hg was not found in any of the nonruptured globes in which an intraocular pressure value was recorded. Reasons for the absence of an intraocular pressure value on a patient's chart could include lack of patient cooperation caused by age or mental status, presence of a corneal abrasion, or simply oversight on the part of the clinician. Because none of these factors is associated with low intraocular pressures it is unlikely that any of the patients in the nonruptured group whose intraocular pressure data were missing had extremely low intraocular pressures (one would expect that the distribution of intraocular pressures in the nonruptured eyes with recorded intraocular pressure values is the same as the distribution of intraocular pressures in nonruptured eyes without recorded values). Thus, our finding that an intraocular pressure of less than 6 mm Hg is specific for scleral rupture is most likely true, despite the missing data. For these reasons, missing data do not markedly affect the validity of our algorithm for diagnosing blunt ocular ruptures. Missing data did prevent a more comprehensive application of multivariate analysis techniques which, in this study, were only useful as a confirmation that the clinical factors we chose to include in our algorithm were the appropri-
Vol. 115, No.4
Scleral Rupture After Blunt Ocular Trauma
ate ones. It also may have prevented the design of a simpler and more accurate clinical test. These shortcomings can only be solved by performing a prospective study. The use of diagnostic imaging modalities, such as computed tomography and ultrasound, might improve diagnostic accuracy beyond that obtainable with routine clinical examination techniques.l" We did not include these tests in our analysis because few of our patients were examined by using these modalities. Our results suggest, however, that routine clinical examination can be both a sensitive and specific method for diagnosing scleral ruptures caused by blunt trauma.
References 1. Russell, S. R., Olsen, K. R., and Folk, J. c.. Predictors of scleral rupture and the role of vitrectomy in severe blunt ocular trauma. Am. J. Ophthalmol. 105:253, 1988.
535
2. Kylstra, J. A.: Management of suspected ocular laceration or rupture. Can. J. Ophthalmol. 26:224, 1991. 3. Cherry, P. M. H.: Rupture of the globe. Arch. Ophthalmol. 88:498, 1972. 4. - - : Indirect traumatic rupture of the globe. Arch. Ophthalmol. 96:252, 1978. 5. Riffenburgh, R. S.: Contusion rupture of the sclera. Arch. Ophthalmol. 69:722, 1963. 6. Joseph, E., Zak, R., Smith, S., Best, W. R., Barnelli. R. L., and Dries, D. J.: Predictors of blinding or serious eye injury in blunt trauma. J. Trauma 33:19,1992. 7. Benjamin, L., and Wormald, R.: CT diagnosis of scleral rupture. Eye 1:757,1987. 8. Eide, N., and Syrdalen, P.: Contusion rupture of the globe. Acta Ophthalmol. 65(suppl. 182):169, 1987. 9. Fisher, Y. L.: Advances in contact ophthalmic ultrasonography. Ocular trauma and intraocular foreign body patients. Dev. Ophthalmol. 18:69, 1989. 10. Sevel, D., Krausz, H., Ponder, T, and Centeno, R.: Value of computed tomography for the diagnosis of a ruptured eye. J. Comput. Assist. Tomogr. 7:870, 1983.
c. Lamkin, M.D., and Desmond K. Runyan, M.D.
We conducted a two-part study to define better the clinical predictors of scleral rupture after blunt trauma. In part 1 we ascertained the prevalence of scleral rupture among a population of patients examined in an ophthalmic emergency room with severe blunt ocular trauma over a six-month period. Scleral rupture was diagnosed in ten of 283 patients (3.5%). In part 2 we compared the clinical findings in 29 patients with scleral rupture to those of 273 patients with no scleral rupture after blunt trauma. We noted that eyes with visual acuity of light perception or less, an intraocular pressure of 5 mm Hg or less, an abnormally deep or shallow anterior chamber, or a media opacity preventing a view of fundus details by indirect ophthalmoscopy, should be considered ruptured when severe intra- or periocular hemorrhage is present. This diagnostic algorithm had a sensitivity of 100.0% (98.7% to 100.0%), specificity of 98.5% (97.1% to 99.9%), and a positive predictive value of 71.4% (66.3% to 76.5%).
T HE DIAGNOSIS of scleral rupture after blunt ocular trauma can be dlfficult.v" Intraocular hemorrhage can prevent visualization of a posterior rupture site and extensive subconjunctival hemorrhage can hide an anterior scleral defect. Several studies have examined the clinical character istics of eyes with scleral ruptures caused by blunt trauma.i" Other studies have compared clinical findings in ruptured eyes Accepted for publication Jan. 5, 1993. From the Departments of Ophthalmology (Dr. Kylstra) and Social Medicine and Pediatrics (Dr. Runyan), University of North Carolina School of Medicine, Chapel Hill, North Carolina; and the Department of Ophthalmology, Massachusetts Eye & Ear Infirmary, Boston, Massachusetts (Dr. Lamkin). Reprint requests to Jan A. Kylstra, M.D., UNC School of Medicine, 617 Clinical Science Bldg., CB No. 7040, Chapel Hill, NC 27599-7040.
530
with findings in eyes that were presumed to be ruptured but that were discovered, at operation, to have intact sclera.P Though these studies provide useful general information, they do not give the clinician a specific set of guidelines for deciding whether to explore surgically or simply observe an eye which has sustained severe blunt trauma. We formulated a clinical algorithm, with known sensitivity, specificity, and positive predictive value, to diagnose scleral ruptures in patients who have sustained severe blunt ocular trauma.
Material and Methods Our study consisted of two distinct parts. In part 1, we estimated the prevalence of scleral rupture in the population of patients examined in the emergency room of the Massachusetts Eye and Ear Infirmary after severe blunt ocular trauma. The records of all patients seen in the emergency room during the six-month period between Sept. 1, 1988, and Feb. 28, 1989, were reviewed by one of us (J.eL.) to select eyes that had been exposed to severe blunt ocular trauma. To meet this qualification, patients had to have a history of ocular or periorbital contact with a blunt object within seven days before initial ophthalmic examination. Additionally, each patient had to be given one or more diagnoses consistent with the history of blunt trauma; these included and were limited to the following: subconjunctival hemorrhage, corneal abrasion, traumatic iritis, iris sphincter tears or other focal defects, iridodialysis, hyphema, lens dislocation, cataract, vitreous base avulsion, vitreous hemorrhage, commotio retinae, retinal tear, retinal dialysis, retinal detachment, choroidal rupture, traumatic optic neuropathy, orbital fracture, and orbital contusion. All patients in whom there was documented or potential sharp or lacerating trauma, as well as those with any obvious lacerating injuries in or
©AMERICAN JOURNAL OF OPHTHALMOLOGY
115:530-535,
APRIL,
1993
Vol. 115, No.4
Scleral Rupture After Blunt Ocular Trauma
around the eye were excluded. Any patient initially examined more than seven days after injury was likewise excluded from the study. An eye was classified as ruptured only if a scleral defect was directly visualized (either at initial examination or intraoperatively). An eye on which an exploratory operation was performed was classified as not ruptured if no scleral defect was noted. An eye on which an exploratory operation was not performed was classified as not ruptured only if the entire anterior scleral surface could be visualized and demonstrated no evidence of scleral defect and only if the entire fundus could be visualized to the ora serrata and no scleral defect was seen. Any patient whose documented findings at initial examination or on follow-up examination did not exclude the possibility of undetected scleral rupture was not included in the study. In part 2 of the study, we compared clinical findings, on initial examination, in patients with scleral ruptures after blunt trauma (cases) to findings in patients in whom scleral rupture had been adequately excluded (controls). The control population consisted of patients identified in part 1 of the study as having no scleral rupture. The cases consisted, in part, of those patients in part 1 of the study in whom scleral rupture was surgically documented. To increase the number of cases, operating-room records were reviewed to identify blunt scleral ruptures managed over the 18 months before the study period. These 19 cases (which had been included in a previously published study-) were added to the first group of cases to form the entire case sample. We reviewed the clinical records of all cases and control subjects and recorded the mode of injury, the visual acuity in each eye, the intraocular pressure in each eye, and the results of ophthalmoscopy. We also noted the presence or absence of an afferent pupillary defect, asymmetry in anterior chamber depth between the two eyes, extraocular motility disturbance, hemorrhagic chemosis, hyphema, vitreous hemorrhage, or orbital fracture. Not all records were complete and, therefore, data for each clinical finding were not recorded for each patient. Absent data were recorded as such. The following assumptions, however, were made. If a hyphema was not explicitly noted on the chart, it was recorded as absent. If an orbital fracture was not documented radiographically, it was recorded as absent. The data were entered into a microcomputer spreadsheet program and then analyzed using
531
Systat, a microcomputer-based statistical package (Systat, Inc., Evanston, Illinois). We analyzed our data in three ways. First, we performed univariate analysis on each of the following dichotomous variables: visual acuity of light perception or worse, intraocular pressure less than 10 mm Hg, intraocular pressure less than 6 mm Hg, afferent pupillary defect, ocular motility disturbance, hemorrhagic chemosis, hyphema, anterior chamber depth asymmetry, ability to visualize fundus details (retinal vessels or optic disk) with the indirect ophthalmoscope, and orbital fractures. Eyes with missing data were not included in the analysis. (The number of eyes with missing data for each of the clinical findings examined can be determined from the Table by subtracting the denominator of the fraction found in the second column from 29, for the ruptured eyes, and the denominator of the fraction in the third column from 273, for the control eyes.) We calculated the sensitivity, specificity, and a positive likelihood ratio for each clinical feature against a standard of diagnosis at operation. We determined significance by using the chi-square statistic. With the prevalence data, obtained in part 1, we were able to calculate positive predictive values for each clinical sign as a diagnostic test. Ninety-five percent confidence intervals were calculated for all proportions. Our second step was to combine these variables in various ways to produce clinical algorithms for diagnosing scleral rupture after blunt trauma. We then evaluated these algorithms to determine sensitivity, specificity, positive likelihood ratio, and (by using the prevalence data from part 1) positive predictive value. All algorithms were required to have 100% sensitivity. We chose the algorithm with the highest specificity as our clinical test or guideline for evaluating eyes subjected to severe blunt trauma. Our final step was to reanalyze our clinical algorithm by performing logistic regression analysis, using combinations of the previously mentioned variables, to predict the presence of scleral rupture. Because a considerable amount of data was not recorded, especially in the ruptured group, only five variables were included in this analysis (we excluded variables that were recorded for fewer than 20 of the 29 ruptured eyes). Additionally, for purposes of this multivariate analysis alone, it was assumed that indirect ophthalmoscopy was performed in all eyes and that a failure to note fundus details meant that no fundus details were visible.
532
April, 1993
AMERICAN JOURNAL OF OPHTHALMOLOGY
Results In part 1 of the study, we determined the prevalence of scleral rupture in 294 eyes of 294 consecutive patients who were examined in the emergency room over a six-month period for having sustained severe blunt ocular trauma. Eleven eyes were not included in this study because of insufficient follow-up. Of the remaining 283 patients, ten were identified as having sustained a scleral rupture. Thus, the prevalence of scleral rupture in our population of patients who had sustained severe blunt ocular trauma was 3.5% (1.4% to 5.6%). Fifteen of the 283 eyes were suspected of having a scleral rupture and were explored and repaired in the operating room. Five of these eyes were shown not to be ruptured. The falsepositive diagnosis rate was therefore 33.3% (9.0% to 57.0%). None of the eyes initially considered nonruptured turned out on followup to be ruptured. The false-negative diagnosis rate was therefore 0.0% (0.0% to 0.5%). In part 2 of the study, we compared clinical findings among eyes with and without scleral rupture and produced a diagnostic algorithm.
Univariate analysis-We evaluated each of the ten clinical factors separately as a test for diagnosing scleral ruptures (Table). The clinical findings with the greatest positive predictive value were as follows: intraocular pressure of 5 mm Hg or less, visual acuity of light perception or less, anterior chamber depth asymmetry, and inability to visualize fundus details with the indirect ophthalmoscope. The clinical findings found most commonly in eyes with scleral ruptures were as follows: hyphema, inability to see fundus details, visual acuity of light perception or less, and hemorrhagic chemosis. We combined these variables in different ways to produce a clinical test for diagnosing scleral rupture after blunt ocular trauma. The best result we obtained was the following: an eye should be presumed ruptured if the visual acuity is light perception or less, if there is anterior chamber depth asymmetry, if the intraocular pressure is 5 mm Hg or less, or if no fundus details can be seen by indirect ophthalmoscopy provided that there is evidence of severe ocular bleeding (hyphema, no fundus visible, or hemorrhagic chemosis). This test correctly identified 29 of the 29 ruptured globes
TABLE CLINICAL FINDINGS AS PREDICTORS OF SCLERAL RUPTURE NO. WITH FINDING/TOTAL
POSITIVE
POSITIVE
RUPTURED
INTACT
SENSITIVITY
SPECIFICITY
PREDICTIVE
LIKELIHOOD
FINDING
EYES
EYES
(%)
(%)
VALUE (%)'
RATIO (%)
Vision < hand motions Intraocular pressure < 10 mm Hg Intraocular pressure < 6 mm Hg Afferent pupillary defect Motility disturbance!
23{29
1{268
79.3
99.6
88.0
198
11{17
12{247
64.7
95.2
32.9
13.5
9{17
0{247
52.9
100.0
100.0
>999
6{11
2{254
54.5
99.2
71.3
68.1
2{10 19{28
25{240 24{273
20.0 67.8
90.6 91.2
7.4 21.9
2.1 7.7
24{27 17{23
27{273 1{273
88.9 73.9
90.1 99.6
24.8 87.2
9.0 185
18{21 2{29
3{273 38{273
85.7 6.9
98.9 86.0
74.3 3.8
77.9 0.5
Hemorrhagic chemosis Hyphema Anterior chamber asymmetry Fundus invisible Facial fracture
'Using an estimated prevalence of 3.5%. !O '> .05 by chi-square (all others P < .05).
Vol. 115, No.4
Scleral Rupture After Blunt Ocular Trauma
as ruptured and incorrectly identified only four of 273 nonruptured eyes as ruptured. Thus, in our patient population, this clinical test had a sensitivity of 100% (98.7% to 100.0%), a specificity of 98.5% (97.1 % to 99.9%), a positive likelihood ratio of 68 (36 to 999) and a positive predictive value of 71.4% (66.3% to 76.5%). Regression analysis-To evaluate the appropriateness of our clinical test (which was derived by trial and error) we performed linear and logistic regression analyses on our data by using the software program. We first constructed a linear regression model to predict the presence of a scleral rupture on the basis of the following variables: hemorrhagic chemosis, inability to see fundus details, visual acuity less than hand motions, anterior chamber asymmetry, and hyphema. We confirmed this model with a logistic model. Analysis confirmed that all of these variables, except hemorrhagic chemosis, independently contributed to the accuracy of the model (P < .05). The r value was .932 for a model that used these variables. The most useful variables were inability to visualize fundus details, visual acuity of light perception or less, and anterior chamber depth asymmetry.
Discussion Our study provides a useful test for diagnosing scleral rupture in eyes that have sustained severe blunt trauma: all eyes with severe ocular hemorrhage (intraocular or periocular) that have either visual acuity of light perception or worse, media opacification precluding a view of fundus details by indirect ophthalmoscopy, abnormally deep or shallow anterior chamber, or intraocular pressure of 6 mm Hg or less should be surgically explored. All other eyes should be observed. This test was designed to be sensitive (100.0% in this study population) because we believe that the potential benefit of rapid diagnosis and repair of these severely injured eyes outweighs the risk of an exploratory operation in patients with nonruptured eyes. Despite this high sensitivity, the test is also highly specific. It should be noted, however, that the use of this algorithm should not replace good clinical judgment in examining individual patients. Though our clinical guideline was derived by trial and error, we began the process with several concepts in mind. Scleral ruptures al-
533
ways involve rupture of the underlying highly vascular choroid or ciliary body. Severe hemorrhage, either intraocular or subconjunctival, or both, must therefore always accompany scleral rupture. For this reason our clinical guideline states that the presence of either hyphema, hemorrhagic chemosis, or severe vitreous hemorrhage (as evidenced by inability to visualize fundus details) is necessary, but not sufficient to diagnose scleral rupture. We chose the other factors for our guideline by including those clinical findings that were most specific for scleral rupture as determined by the univariate analysis. In order not to compromise sensitivity, we structured the guideline so that a positive diagnosis is made if anyone of the four most specific factors is present in an injured eye (provided that severe hemorrhage is present). We determined the prevalence of scleral rupture in a population of patients treated by ophthalmologists for ocular complications of blunt trauma. Our estimate of 3.5%, however, is similar to the results obtained in a recent study in which one of 94 consecutive patients (1.1 %) who were referred for ophthalmic examination after hospitalization for multiple blunt injuries was found to have a ruptured globe." It appears, therefore, that scleral rupture after severe blunt trauma is infrequent. We also calculated the sensitivity, specificity, and positive predictive values for individual clinical predictors of globe rupture. Two previous studies have attempted to define clinical guidelines for diagnosing blunt ocular rupture by comparing clinical findings in ruptured eyes with findings in nonruptured eyes. 1,2 These studies, however, only examined patients who were operated on, most of whom had ruptured globes. These studies are limited for three reasons. Possibly, some patients whose eyes were not suspected of being ruptured and who were not operated on, did have ruptured globes. Clinical guidelines derived from these studies, therefore, might not be sensitive enough to detect the less obvious cases. It is also possible that, because of the limited size of the control group, clinical characteristics found only in ruptured eyes in these studies are found in many nonruptured but traumatized eyes that were not included in the studies. Guidelines derived from these studies might, therefore, also not be specific enough to avoid a considerable number of false-positive diagnoses. Finally, these studies did not determine the prevalence of scleral rupture in the population at risk. Without this information, the positive pre-
534
AMERICAN JOURNAL OF OPHTHALMOLOGY
dictive value of any set of clinical guidelines or of any single clinical finding cannot be determined. Thus, though previous studies have demonstrated an association between the presence of a hyphema, for example, and the presence of scleral rupture, they do not allow the clinician to determine the risk of globe rupture in a patient with a hyphema. Our study suggests that at least one specific recommendation found in the published reports is incorrect. One study! states that all eyes with intraocular pressures of less than 10 mm Hg after blunt trauma should be surgically explored. Our data suggest, however, that almost 5.0% of unruptured eyes that have sustained severe blunt trauma will have intraocular pressures below 10 mm Hg and that two thirds of eyes operated on because of this clinical finding will not be ruptured. Our results do demonstrate that low intraocular pressure is a specific clinical sign of scleral rupture. Analysis of our intraocular pressure data, however, suggests that the intraocular pressure must be 5 mm Hg or less before it becomes a specific clinical sign of scleral rupture. Previous studies have suggested that visual acuity of light perception or worse, intraocular pressure less than 10 mm Hg, the presence of a hyphema, the presence of hemorrhagic chemosis, and an asymmetry in anterior chamber depth are all more common in ruptured than in nonruptured eyes.!" Our study confirms these results. Our study also suggests that 25.0% of eyes with hyphema and 22.0% of eyes with hemorrhagic chemosis will have a scleral rupture. This information is useful for nonophthalmologists working in the emergency room setting who can easily document these clinical findings, even in uncooperative patients, and treat eyes with these findings as if they were ruptured, that is, place a protective shield over the eye and consult an ophthalmologist. Our study is retrospective in design and is therefore flawed because of missing data. The problem of missing data is of particular concern in two areas. First, 11 of the patients in the prevalence part of this study were excluded from analysis because not enough follow-up information was available to place them in either the ruptured or the nonruptured group. It is possible that all of these patients had ruptured globes and that the true prevalence of ruptures in our population was 21 of 294 or 9.9%. However, there is no reason to suspect that the patients with missing data differ markedly from the study group either in prevalence
April, 1993
of ruptures or in distribution of the clinical findings. If any difference exists it could be argued that the eyes with missing data were less likely to be ruptured than the eyes included in the study because the patients did not return for follow-up and because patients with serious injuries would be more likely to return for examination. The effect of missing data in the form of individual clinical observations is a second weakness of this study and is a problem inherent in retrospective studies in general. We decided to handle these missing data, in most cases, by simply recording them as absent. Of the clinical factors included in our algorithm, only intraocular pressure and afferent pupillary defect had a considerable number of missing entries in both the ruptured and nonruptured groups. Additionally, many missing entries were also found for ocular motility in the ruptured group. However, the only one of these three factors we included in our clinical algorithm for predicting blunt rupture was intraocular pressure. Intraocular pressure was included in the algorithm because of its apparently great specificity as a predictor; that is, an intraocular pressure of less than 6 mm Hg was not found in any of the nonruptured globes in which an intraocular pressure value was recorded. Reasons for the absence of an intraocular pressure value on a patient's chart could include lack of patient cooperation caused by age or mental status, presence of a corneal abrasion, or simply oversight on the part of the clinician. Because none of these factors is associated with low intraocular pressures it is unlikely that any of the patients in the nonruptured group whose intraocular pressure data were missing had extremely low intraocular pressures (one would expect that the distribution of intraocular pressures in the nonruptured eyes with recorded intraocular pressure values is the same as the distribution of intraocular pressures in nonruptured eyes without recorded values). Thus, our finding that an intraocular pressure of less than 6 mm Hg is specific for scleral rupture is most likely true, despite the missing data. For these reasons, missing data do not markedly affect the validity of our algorithm for diagnosing blunt ocular ruptures. Missing data did prevent a more comprehensive application of multivariate analysis techniques which, in this study, were only useful as a confirmation that the clinical factors we chose to include in our algorithm were the appropri-
Vol. 115, No.4
Scleral Rupture After Blunt Ocular Trauma
ate ones. It also may have prevented the design of a simpler and more accurate clinical test. These shortcomings can only be solved by performing a prospective study. The use of diagnostic imaging modalities, such as computed tomography and ultrasound, might improve diagnostic accuracy beyond that obtainable with routine clinical examination techniques.l" We did not include these tests in our analysis because few of our patients were examined by using these modalities. Our results suggest, however, that routine clinical examination can be both a sensitive and specific method for diagnosing scleral ruptures caused by blunt trauma.
References 1. Russell, S. R., Olsen, K. R., and Folk, J. c.. Predictors of scleral rupture and the role of vitrectomy in severe blunt ocular trauma. Am. J. Ophthalmol. 105:253, 1988.
535
2. Kylstra, J. A.: Management of suspected ocular laceration or rupture. Can. J. Ophthalmol. 26:224, 1991. 3. Cherry, P. M. H.: Rupture of the globe. Arch. Ophthalmol. 88:498, 1972. 4. - - : Indirect traumatic rupture of the globe. Arch. Ophthalmol. 96:252, 1978. 5. Riffenburgh, R. S.: Contusion rupture of the sclera. Arch. Ophthalmol. 69:722, 1963. 6. Joseph, E., Zak, R., Smith, S., Best, W. R., Barnelli. R. L., and Dries, D. J.: Predictors of blinding or serious eye injury in blunt trauma. J. Trauma 33:19,1992. 7. Benjamin, L., and Wormald, R.: CT diagnosis of scleral rupture. Eye 1:757,1987. 8. Eide, N., and Syrdalen, P.: Contusion rupture of the globe. Acta Ophthalmol. 65(suppl. 182):169, 1987. 9. Fisher, Y. L.: Advances in contact ophthalmic ultrasonography. Ocular trauma and intraocular foreign body patients. Dev. Ophthalmol. 18:69, 1989. 10. Sevel, D., Krausz, H., Ponder, T, and Centeno, R.: Value of computed tomography for the diagnosis of a ruptured eye. J. Comput. Assist. Tomogr. 7:870, 1983.