Sharma S, Walker R, Brown GC, Cruess AF. The Importance of Qualitative Vitreous Examination in Patients With Acute Posterior Vitreous Detachment. Arch Ophthalmol. 1999;117(3):343-346. doi:10.1001/archopht.117.3.343
To determine whether patients with acute posterior vitreous detachment with pigmented vitreous granules or hemorrhage have a higher likelihood of retinal tear compared with those with qualitatively normal vitreous examination findings.
A multicenter cross-sectional study was performed in 3 peripheral ophthalmic clinics. Patients with acute posterior vitreous detachment were examined for the presence or absence of vitreous pigment granules, vitreous hemorrhage, and horseshoe retinal tear.
Fifty-nine consecutive patients with acute posterior vitreous detachment met our eligibility criteria. Eight patients had a retinal tear, and thus its prevalence in our study was almost 14%. Thirteen patients (22%) had a high likelihood because they had evidence of either pigmented vitreous granules or hemorrhage. The prevalence of retinal tear in the setting of acute posterior detachment associated with vitreous hemorrhage alone, pigment alone, or vitreous hemorrhage and pigment was 54%. Patients with posterior vitreous detachment with pigmented vitreous granules or hemorrhage were significantly more likely to have a retinal tear (odds ratio, 52.0; 95% confidence interval, 5.4-497.0). Patients with a retinal tear were 7 times more likely to have pigmented vitreous granules or hemorrhage (LR + ve=7.4, in which LR + ve indicates positive likelihood ratio; 95% confidence interval, 3.3-16.4).
Patients with posterior vitreous detachment with vitreous pigment granules or hemorrhage are 52 times more likely to have a retinal tear compared with those who have normal findings on qualitative vitreous examination.
ACUTE POSTERIOR vitreous detachment occurs in 63% of patients older than 69 years.1 It is a known risk factor for retinal tear because up to 18% of these patients have evidence of a retinal tear.2 Results of recent studies3,4 conducted in the tertiary care setting demonstrate that patients with abnormal qualitative results of vitreous examination are at higher risk for retinal tear in the setting of acute posterior vitreous detachment. Given that acute posterior vitreous detachment is a common occurrence, and that with our current managed care environment more general ophthalmologists will be screening patients for retinal detachment, we undertook the present study in a secondary care setting to determine whether patients with vitreous cells are at higher risk for retinal detachment in the setting of acute posterior vitreous detachment.
A cross-sectional study was performed at 3 secondary referral centers (defined as centers where assessment of a patient with symptoms suggestive of an acute posterior vitreous detachment was conducted by a general ophthalmologist after referral from a primary care physician). The objective of our study was to determine whether patients with acute posterior vitreous detachment, who had either pigmented vitreous granules or vitreous hemorrhage, had a retinal tear more often than those with qualitatively normal findings on vitreous examination.
Study patients consisted of a consecutive series who had symptoms suggestive of acute posterior vitreous detachment. Patients were included in our study if they (1) had new floaters or photopsia for less than 1 week, (2) had clinical confirmation of a posterior vitreous detachment (defined as the presence of a complete Weiss ring), and (3) were older than 18 years and were able to give informed consent. Patients were excluded for the following reasons: (1) the presence of another ocular condition that increased the risk for retinal tear, including proliferative vitreoretinopathy, sickle cell retinopathy, retinopathy of prematurity, and proliferative retinopathy; (2) an inability to completely visualize the peripheral retina; and (3) the presence of conditions other than vitreous detachment that have been associated with either vitreous pigment granules or vitreous hemorrhage, including previous ocular surgery and previous vitreous hemorrhage.
All patients underwent complete dynamic vitreous examination using slitlamp biomicroscopy with and without a 60-diopter (D) lens and peripheral retinal examination with scleral depression and a Goldmann 3-mirror lens to detect the presence of a horseshoe retinal tear. Patients were classified as having a high likelihood for retinal tear if there were pigmented vitreous granules (the Shafer sign)5 or red blood cells in the vitreous.
Possible confounders, such as the presence or absence of floaters and photopsia and the degree of myopia (greater or less than 3 D), were treated in a dichotomous fashion. The main study variables (likelihood stratification and retinal tear status) are summarized in a contingency table. The effect of likelihood status on the presence or absence of a retinal tear was estimated by an odds ratio. A 95% confidence interval (CI) was calculated around our odds ratio using the logit method.6 Likelihood ratios were also calculated, as were 95% CIs.7,8
Fifty-nine consecutive patients who met our inclusion criteria were studied. The mean age of our study population was 61.8 years (Table 1); 47% were men, and 30% had greater than 3 D of myopia.
Of our 59 patients, 8 (14%) had evidence of a retinal tear associated with posterior vitreous detachment (Table 2). Thirteen patients were classified as having a high likelihood for retinal tear because they had either pigment granules or hemorrhage in their vitreous (8 patients had vitreous hemorrhage alone, 2 patients had pigmented vitreous granules alone, and 3 patients had evidence of both) (Table 3). The remaining 46 patients had no evidence of either pigment granules or hemorrhage in their vitreous and, thus, were classified as having a low likelihood for retinal tear.
Seven of 8 patients with posterior vitreous detachment who had a retinal tear had qualitatively abnormal examination findings. If we consider the presence or absence of vitreous pigment granules or hemorrhage to be a diagnostic "test," our test has a sensitivity of 88% (7/8). Because 45 of 51 patients with posterior vitreous detachment without retinal tear were at low risk, our diagnostic test has a specificity of 88% (45/51). The positive and negative predictive values of likelihood stratification are 54% (7/13) and 98% (45/46), respectively. All patients with pigmented vitreous granules had a retinal tear, as did 45% (5/11) of patients with vitreous hemorrhage.
Patients with either pigment granules or vitreous hemorrhage were 52 times more likely to have a retinal tear (odds ratio, 52.0; 95% CI, 5.4-497.0). There was no statistically significant difference between likelihood groups with respect to age, sex, presence of floaters or photopsia, or prevalence of myopia greater than 3 D. Our calculation of the odds ratio is robust because its magnitude was minimally altered on sensitivity analyses (Table 4).
The presence of pigment granules or vitreous hemorrhage generated a likelihood ratio of 7.40 (LR + ve=7.40, in which LR + ve indicates positive likelihood ratio; 95% CI, 3.32-16.40). The absence of these characteristics generated a likelihood ratio of 0.14 (LR − ve=0.14, in which LR − ve indicates negative likelihood ratio; 95% CI, 0.07-0.30).
The incidence of retinal tear associated with posterior vitreous detachment in our series was almost 14%. Although our prevalence of retinal tear was similar to that noted in other studies,1,2,9 ours is the first study, to our knowledge, to be performed in a secondary care setting as opposed to a tertiary care setting. This is important because prevalence figures derived from the tertiary care setting should tend to overestimate the prevalence of retinal tear in patients with posterior vitreous detachment because these patients represent a selected group of patients.
Patients with high-likelihood characteristics were 52 times more likely to have a retinal tear. This result is statistically significant (P<.001) because our 95% CI did not include 1. In fact, because the 95% CI ranged from 5.4 to 497.0, patients with high-likelihood characteristics are at least 5 times more likely to have a retinal tear and may be as much as 500 times more likely. Sensitivity analyses revealed little change in the magnitude of our odds ratio when the prevalence of retinal tear in the population varied (Table 4).
Physical findings can themselves be thought of as diagnostic tests because their presence or absence increases or decreases the estimate of disease probability.10,11 Our high-likelihood vitreous characteristics—the presence of vitreous cells in the setting of acute posterior vitreous detachment—can be thought of as a diagnostic test with sensitivity and specificity of 88% (7/8 and 45/51, respectively). Given these accuracy factors, the likelihood ratio generated given qualitatively abnormal findings on examination was 7.4 (95% CI, 3.3-16.4). Conversely, the likelihood ratio generated given the absence of pigment or vitreous hemorrhage was 0.14 (95% CI, 0.07-0.30).
To better understand the clinical significance of our likelihood ratios, we can calculate posttest probabilities on hypothetical patients. Assuming a pretest probability of 13% (meaning a patient has little more than a 1-in-10 chance of having a retinal tear), if we were to detect either pigment or hemorrhage in the vitreous, the patient would have a 52.5% (+39.5 percentage of probable increase) probability of having a retinal tear. The same patient without vitreous pigment or hemorrhage would have a 2.1% (−10.9 percentage of probable decrease) chance of having a retinal tear. Our diagnostic test, when applied to a patient with a pretest probability of 50% (Table 5), has the ability to modify the probability of disease by 72 percentage points. This, in our opinion, is an excellent diagnostic test. Table 5 includes posttest probabilities of 2 other hypothetical patients.
As with any cross-sectional study, there are some potential study flaws. All of our patients underwent a peripheral retinal examination, which included contact lens biomicroscopy. This may not be standard practice in all centers; however, after surveying many general ophthalmologists during the design phase of our study, this method was included in our study because many perform it routinely to rule out peripheral tears in the setting of acute posterior vitreous detachment. The examiners for this study were all general ophthalmologists, as opposed to retinal specialists, because our study was designed to be generalized to the general ophthalmologist. Although this decision may have resulted in underdetection or overdetection of retinal tears, results of our sensitivity analyses reveal that our estimate of association is relatively independent of retinal tear prevalence. Our decision to treat the presence of vitreous cells as a homogeneous group, as opposed to differentiating cells based on qualitative assessment, was also based on our intent to generalize our results to the general ophthalmologist.
Our study was performed on 59 consecutive patients with posterior vitreous detachment. Given that certain cells of our contingency table contained small numbers, it is possible that small alterations in these values could have altered our overall estimate of association. Although the width of our CI was wide, it did definitively exclude the null value.
There was no statistical difference between our 2 groups with respect to the following possible clinical confounders: age, clinical history, and presence of myopia. There was, however, a trend toward a higher proportion of patients with qualitatively abnormal examination findings being men. Novak and Welch12 previously noted a higher prevalence of retinal tear in men with posterior vitreous detachment. Although the proportional difference in sex noted between our 2 study groups was not statistically significant (P>.05), it may have achieved statistical significance if our sample size had been larger.
Retinal tear status was evaluated by investigators (S.S. and R.W.) who were not masked with respect to the qualitative vitreous assessment. Although it is possible that this method could have induced a biased evaluation of retinal tear status, we believe that this bias was minimal because all retinal evaluations were performed under protocol, with all patients undergoing scleral depression and dynamic examination with a 3-mirror examination.
The prevalence of retinal tear in the setting of acute posterior vitreous detachment was almost 14%. The prevalence of retinal tear in the setting of acute posterior vitreous detachment associated with vitreous hemorrhage alone, pigment alone, or vitreous hemorrhage and pigment was 54%. Patients with acute posterior detachment, with the presence of either vitreous pigment granules or vitreous hemorrhage, are 52 times more likely to have a retinal tear compared with those with qualitatively normal findings on vitreous examination.
Accepted for publication September 22, 1998.
This study was supported in part by the Retina Research and Development Foundation, Philadelphia, Pa; the E. A. Baker Foundation, Toronto, and the Patient Research Foundation (Dr Sharma), Kingston, Ontario; and the Ron Michels Foundation (Dr Sharma), Baltimore, Md.
Reprints: Sanjay Sharma, MD, MSc(Epid), Room 3011, Etherington Hall, Queen's University, Kingston, Ontario, Canada K7L 3N6.