- Email: [email protected]
B-Scan Ultrasonography to Screen for Retinal Tears in Acute Symptomatic Age-Related Posterior Vitreous Detachment
B-Scan Ultrasonography to Screen for Retinal Tears in Acute Symptomatic Age-Related Posterior Vitreous Detachment Jose Lorenzo-Carrero, MD, PhD, Ines Perez-Flores, MD, Monica Cid-Galano, MD, Marta Fernandez-Fernandez, MD, Fernando Heras-Raposo, MD, Ramon Vazquez-Nuñez, MD, Marta Lopez-Fuentes, MD Objective: To evaluate the performance characteristics of B-scan ultrasonography (US) as a diagnostic test for the detection of retinal tears in acute symptomatic age-related posterior vitreous detachment (PVD). Design: Evaluation of a diagnostic test through a cross-sectional study with prospective data collection. The study intended to meet the 14 items proposed by the Quality Assessment of Diagnostic Accuracy Studies panel. Participants: Two hundred thirty-nine patients with acute-onset age-related PVD were consecutively enrolled in a nonreferral hospital. Testing: Comprehensive eye examination including vitreous and retinal biomicroscopy was performed on an emergency basis followed by blind B-scan kinetic US. Sensitivity, specificity and predictive values of the index test (B-scan US) were analyzed and compared with the standard reference (baseline examination). In cases of disagreement between both diagnostic methods, a new gold standard was established based on the findings of subsequent directed indirect ophthalmoscopy based on the echographic findings. Positive and negative likelihood ratios and a likelihood nomogram with pretest and posttest odds of retinal tears were calculated for B-scan US. Main Outcome Measures: Index test performance for the detection of retinal tears secondary to agerelated PVD. Results: Both diagnostic methods performed comparably. The sensitivity of B-scan US for detection of retinal tears was 96% and that of baseline examination was 89%. Both methods had similar negative predictive values of 99%. B-scan US specificity was 98%. The estimated pretest and posttest probability for a positive B-scan US were 10.8% and 89%, respectively. Conclusions: Proper B-scan kinetic US is a noninvasive and accurate diagnostic method for the detection of retinal tears that can be reliably used in no view or small pupil cases with symptomatic PVD. Financial Disclosure(s): The authors have no proprietary or commercial interest in any of the materials discussed in this article. Ophthalmology 2009;116:94 –99 © 2009 by the American Academy of Ophthalmology.
Posterior vitreous detachment (PVD) is the most common age-related event that occurs in the vitreous and the principal predisposing risk factor for the development of a rhegmatogenous retinal detachment. Approximately 10%–14% of nonreferred patients with acute symptomatic PVD have a retinal tear on initial examination.1,2 Approximately 30%– 50% of symptomatic retinal breaks with persistent vitreoretinal traction will cause a clinical retinal detachment if left untreated.3 The early detection of retinal tears is crucial to providing early treatment that might prevent retinal detachment. No symptoms can reliably distinguish a PVD with an associated retinal break from those without. Therefore, accurate vitreous and retinal examinations are mandatory in eyes with symptomatic PVD.1– 4 The preferred method of evaluating peripheral vitreoretinal pathology is indirect ophthalmoscopy combined with scleral depression (Preferred Practice Patterns. Posterior vitreous detachment, retinal breaks, and Lattice degeneration. American Academy of Ophthalmology 2003). This method, mostly used by vitreoretinal specialists, requires adequate skill, pupillary dilatation, media
94
© 2009 by the American Academy of Ophthalmology Published by Elsevier Inc.
transparency, and patient cooperation. Developing accurate imaging procedures for detecting retinal tears in the setting of an acute PVD would allow for adequate evaluation of patients who are, otherwise, poor candidates for standard visual examination or get unreliable results. Recent advances in ultrasound (US) resolution have improved the acoustic detection of subtle structures in the vitreous cavity such as the posterior vitreous cortex and retinal tears, but a test validation in the setting of acute age-related PVD remains to be done.5–9 The purpose of this study was to determine the performance of B-scan US to screen for retinal tears in symptomatic age-related PVD.
Methods We conducted a cross-sectional blind comparative study with prospective data collection. The study was designed to meet the 14 items proposed by an expert panel for evidence based quality assessment of accuracy of diagnostic tests.10 The study group consisted of a consecutive series of symptomatic patients presentISSN 0161-6420/09/$–see front matter doi:10.1016/j.ophtha.2008.08.040
Lorenzo-Carrero et al 䡠 Echographic Testing for Retinal Tears in Symptomatic PVD
Figure 1. A, Complete PVD shown in a longitudinal echogram. Anterior limit of PVD at the vitreous base (arrow). Weiss’ ring (arrow head). B, PVD as shown in a transverse echogram. C–G, Traction retinal tears attached to the posterior vitreous cortex are seen as highly reflective echoes and clearly individualized from the adjacent retina. H–J, These echograms show discrete echogenic retinal ridges adherent to the posterior vitreous cortex that were interpreted as suspicious retinal tears, but no retinal tear could be confirmed on ophthalmoscopy. K, Intense vitreous synchysis with a posterior lacuna and PVD. L, Retinoschisis and PVD. PVD ⫽ Posterior vitreous detachment.
95
Ophthalmology Volume 116, Number 1, January 2009 ing to the emergency department of a nonreferral hospital who were diagnosed of acute age-related PVD and satisfied the criteria for inclusion. Patients who had symptoms suggestive of recent onset of PVD (visual floaters and/or photopsia) were referred to the ophthalmology department by primary emergency care practitioners. Comprehensive ophthalmic examination, undertaken by a general ophthalmologist in an emergency basis, included slit-lamp vitreous biomicroscopy with noncontact and/or 3-mirror lenses. Diagnosis of complete PVD was established when biomicroscopy confirmed the presence of a Weiss’ ring. The peripheral retina was examined with either slit-lamp biomicroscopy and 3-mirror contact lens or indirect binocular ophthalmoscopy with or without scleral indentation at the examiner’s discretion. Patients were excluded if they had onset of symptoms ⬎1 month, if they had ocular diseases thought to modify vitreous condition (e.g., history of recent blunt trauma, eye surgery, diabetes, uveitis, or tumors in any eye) or if they had media opacity that precluded appropriate ophthalmoscopic examination (cataract, vitreous hemorrhage, corneal opacities, or narrow pupil). Investigational review board approval was obtained and all subjects signed an informed consent form. After baseline examination, all participants were referred for blind B-scan US evaluation the next working day. Blind echographic examinations were performed by means of B-scan V Plus/Biovision (Quantel Medical, Clermont-Ferrand, France), which has a focused probe of 10 Mhz with a mechanical sector scanning. High-sensitivity gain settings of 105 dB specifically designed for vitreous examinations were selected. Examinations were performed by the same examiner under topical anesthesia with the patient in the supine position. The patient’s eyelids were kept open and the probe was placed directly on the ocular surface behind the limbus through 2.5% methylcellulose as a coupling gel. The probe was positioned in transverse sections (cross-sections) all around the conjunctiva and then progressively shifted to examine more peripheral aspects of the vitreoretinal interface. For adequate characterization of vitreoretinal relationships, movement of the vitreous was evaluated during voluntary saccadic motion of the patient’s globe while the probe was held stationary (kinetic examination). Patients who were initially diagnosed with PVD on biomicroscopy but showed posterior vitreous attachment to the optic nerve head on B-scan US were considered to be incorrectly diagnosed and excluded from study analysis. If a suspected retinal tear was found, longitudinal probe orientation (radial section) was used to determinate its location. According to dynamic echographic examination, the examiner classified and photographed the presence of retinal tears as (1) “suspicious” when the US image depicted an echogenic retinal traction ridge that was adherent to the posterior vitreous cortex, but a retinal flap could not be clearly individualized (Fig 1H–J); (2) “definite” when the US image showed a echogenic retinal flap or operculum continuous with the posterior vitreous cortex (Fig 1C–G); or (3) “absent” when no vitreoretinal adherences could be detected (Fig 1A, B, K). For analysis purposes, retinal tears classified as “definite” and “suspicious” were considered as a positive diagnosis. Patients who presented conflicting results between the US examination and the baseline examination were subsequently referred for directed indirect ophthalmoscopy based on the echographic findings by a vitreoretinal specialist. The results on final indirect ophthalmoscopy served as the new reference standard with which the results of B-scan echography and the baseline funduscopic examination were compared.
Diagnostic Test Characteristics Sensitivity, specificity, and positive and negative predictive values were calculated with confidence intervals for baseline funduscopic
96
examination and B-scan US. Likelihood ratios and the posttest probability of retinal tears given the pretest probability and test characteristics were calculated with confidence intervals for B-scan US according to a Bayesian analysis model. Graphical calculations of Bayes’ theorem were displayed in a Fagan’s nomogram.11 Demographic characteristics of patients presenting retinal tears (age, gender, and eye’s axial length) were compared with those who had uncomplicated PVD. Tests of significance included the Pearson chi-square and Fisher tests for nominal data and t-test for continuous data. If P⬍5%, the result was considered significant.
Results From previous pilot studies performed in this setting, (Invest Ophthalmol Vis Sci 2004;45:E-Abstract 2009) we concluded that a sample of 250 patients would allow to estimate the prevalence of retinal tears in patients with PVD with a confidence interval of ⫾4% and ␣ error ⫽ 0.05. Of 250 consecutively enrolled patients, 11 were excluded from the study analysis because of incorrect initial diagnosis of PVD. The study population comprised 239 patients of mean age 65 years (range, 40 – 82). There were 96 (40%) men and 143 (60%) women. Of 239 patients, 26 (10.8%) had final evidence of ⱖ1 retinal tears associated with PVD. In 22 eyes, there was a solitary retinal flap tear; 2 other eyes had a tear with a free operculum, and 2 eyes showed 2 retinal tears each. Demographic characteristics of patients presenting retinal tears (age, gender, and eye’s axial length) were similar to those who had uncomplicated PVD (Table 1). Table 2 summarizes the test performance of baseline biomicroscopy/ophthalmoscopy and kinetic B-scan US for the analysis of retinal tears. The prevalence of patients with retinal tears was 10.8% (26/239 patients). Baseline funduscopic examination detected 25 retinal tears, whereas blind B-scan kinetic US detected 30 retinal tears, among which 25 were classified as “definite” and 5 were classified as “suspicious” by the examiner. There were 3 false-positive retinal tears, which had been classified as “suspicious” on B-scan US by the examiner (Fig 1H–J). There were 3 additional retinal tears initially missed on baseline funduscopic evaluation that were detected by blind B-scan US and confirmed by unblinded ophthalmoscopy. There was also 1 false negative on B-scan US, which was confirmed by final indirect ophthalmoscopy. Analysis of sensitivity for detection of retinal tears was done according to the total number of retinal tears. The sensitivity of B-scan US for detection of retinal tears was 96.4% (27/28) and the sensitivity of baseline funduscopic examination was 89.2% (25/28). The specificity of B-scan US was 98% (210/213) and the specificity of baseline funduscopic examination was 100% (213/ 213). The positive and negative predictive values for B-scan US were 89% (25/28) and 99% (210/211), respectively; for baseline funduscopic examination, they were 100% (24/24) and 99% (213/ 215), respectively.
Table 1. Demographic Characteristics of Study Patients with and without Retinal Tears
Gender Mean age (SD) Mean axial length (SD)
Patients with RT
Patients without RT
P
14 M/12 F 65.04 (⫾5.9) 23.9 (⫾2)
84 M/129 F 65.18 (⫾6.9) 23.4 (⫾1.2)
0.53 0.92 0.12
F ⫽ female; M ⫽ male; RT ⫽ retinal tear; SD ⫽ standard deviation.
Lorenzo-Carrero et al 䡠 Echographic Testing for Retinal Tears in Symptomatic PVD Table 2. Performance Characteristics of Baseline Ophthalmoscopy/Biomicroscopy and B-Scan Ultrasounds (US) for the Detection of Retinal Tears in Symptomatic Age-Related Posterior Vitreous Detachment (239 Patients)
Identified total tears False-negative retinal tears False-positive retinal tears Missed eyes with tears Identified eyes with tears Sensitivity§ Specificity Positive predictive value Negative predictive value Negative likelihood ratio Positive likelihood ratio
Baseline Ophthalmoscopy/ Biomicroscopy
B-Scan US
25* 3‡ 0 2 24* 89% [72%–98%] 100% [98%–100%] 100% [86%–100%] 99% [97%–100%] 0.08 [0.03–0.30] Inf
27† 1 3 1 25† 96% [82%–100%] 98% [96%–100%] 89% [86%–93%] 99% [98%–100%] 0.04 [0.01–0.27] 68 [22–211]
Inf ⫽ infinite. *One eye had 2 retinal tears. † Two eyes were shown to have 2 retinal tears. ‡ One retinal tear was missed in 1 eye that was found to already have 1 retinal tear. § Analysis according to retinal tears.
Table 3 summarizes the clinical findings of eyes that were misclassified by either baseline ophthalmoscopy or B-scan US and required the application of the new gold standard. Figure 2 shows the likelihood Fagan’s nomogram of B-scan US. For a positive test result, the posttest probability is 89% (posttest odds, 8.3), whereas for a negative test result the posttest probability and odds is zero. Hence, a positive test result increases the probability of having a retinal tear from 10% to 89% and a negative result almost excludes the probability of having a retinal tear. Additional findings on baseline funduscopic examination include peripheral retinal abnormalities. Among the 239 eyes with PVD, 74 (31%) had visible retinal abnormalities; 29 (12.2%) had lesions of lattice degeneration. Of these 29 eyes, 4 (13.7%) had associated retinal tears and 3 (10.3%) had atrophic holes of lattice lesions. One eye had degenerative retinoschisis (Fig 1L), and 3 eyes showed tiny round retinal holes with no associated retinal abnormality. Sixteen eyes (6.7%) showed peripheral retinal abnormalities without clinical significance (peripheral cystoid degeneration and cobblestone degeneration).
Discussion Our results indicate that B-scan US is an accurate diagnostic test for detection of retinal tears in symptomatic patients
with acute PVD. The likelihood that a retinal tear would be missed on B-scan US is very low given the high negative predictive value found in this study. Patients with retinal tears are 68 times more likely (positive likelihood ratio ⫽ 68) to have their retinal tears detected on B-scan US than showing a normal echogram. The prevalence of retinal tears associated with PVD in this series was 10.8%, which is similar to that found in other studies that analyzed series of nonreferred patients.1,2 We have used B-scan US to screen symptomatic patients and not as a standard screening test, which would typically apply to asymptomatic patients with a low pretest probability. As long as the prevalence can vary, it modifies predictive values and posttest odds (Bayes’ theorem). The prevalence of complicated PVD seen in referral hospitals is likely to be higher; therefore, the negative predictive values of B-scan US would be lower than that obtained in this study. However, we do not believe this effect would change significantly the test performance of B-scan US. Based on the calculated likelihood nomogram, we can hypothesize that a prevalence up to 3 times higher than that obtained in this study would result in posttest probability of 98% for a positive test and 1% for a negative test, which are not very different from those obtained in this study.12 Comparative studies carried out in eyes with clear media suggest that US may assess the posterior vitreoretinal interface better than biomicroscopy but we could not find any article in a Medline search reporting on a validation of ultrasounds to detect retinal tears in PVD.6 – 8 We have used 105-dB gain setting to ensure less wave attenuation and high signal to noise ratio. B-scan US allows for better kinetic evaluation of peripheral vitreoretinal relationships than biomicroscopy with contact lens or indirect ophthalmoscopy because the ocular movements are less restricted.6,7 The Quality Assessment of Diagnostic Accuracy Studies panel (QUADAS) developed a tool composed of 14 items to evaluate bias, variability, and quality of reporting.10 Although we followed the recommendations to meet the 14 items, several items deserve further discussion. We excluded patients with obscuring vitreous hemorrhage because waiting for clearing of vitreous hemorrhage would make it likely that the target condition would change before the reference test could be performed (QUADAS item 4) because one third of patients with PVD and vitreous hemorrhage at presentation develop new retinal tears during follow-up.13 When PVD is complicated by vitreous hemorrhage, the incidence of retinal tears increases to 40%– 67%.9,14 In
Table 3. Clinical Findings of Misclassified Eyes Age
Gender
Axial Length
Complaints
Baseline Examination
B-Scan Ultrasound
Gold Standard
64 56 67 50 72 63 69
F F M F M M M
23.06 22.22 23.32 30.46 24.01 26.26 23.09
Photopsia ⫹ floaters Floaters Floaters Floaters Photopsia ⫹ floaters Floaters Floaters
1 flap RT (upper TQ) Normal fundus Preretinal hemorrhage 1 RT (inferior NQ) Normal fundus Retinal hole ⫹ lattice Normal fundus
2 flap RT 1operculum 1 flap RT Vitreous synchysis Suspicious RT Suspicious RT Suspicious RT
2 flap RT 1operculum ⫹ RT 1 flap RT 1 RT VR traction Retinal hole ⫹ lattice VR traction
F ⫽ female; M ⫽ male; NQ ⫽ nasal quadrant; RT ⫽ retinal tear; TQ ⫽ temporal quadrant; VR ⫽ vitreoretinal.
97
Ophthalmology Volume 116, Number 1, January 2009
Figure 2. Likelihood ratio nomogram B-scan Ultrasound (US). Prior probability (odds) ⫽ 11% (0.1). Positive test: positive likelihood ratio, 68 (22, 211). Posterior probability (odds), 89% (8.3) [73%, 96%]. Negative test: negative likelihood ratio, 0.04 (0.01, 0.27). Posterior probability (odds), 0% (0.0) [0%, 3%].
those cases, echographic evaluation is the preferred method to support clinical decisions. One retrospective study found a sensitivity of 91% and specificity of 97% in identifying retinal tears in eyes with PVD and vitreous hemorrhage.9 However, evaluating test accuracy with such a high prevalence of the target condition could lead to a biased assessment of the test properties and compromise the external validity for a broader spectrum of patients. To extend its clinical applicability to patients with unreliable standard evaluation apart from vitreous hemorrhage, we carried it out in nonreferred patients regardless of symptoms severity (QUADAS item 1). Although the recommended method of evaluating peripheral vitreoretinal pathology is indirect ophthalmoscopy combined with scleral depression, most ophthalmologists in
98
continental Europe rely on biomicroscopy with a 3-mirror contact lens as the reference standard. The study protocol allowed the examiner to choose the method of fundus peripheral examination. Five participating general ophthalmologists acknowledged to have routinely performed slitlamp biomicroscopy with a 3-mirror contact lens; another preferred indirect ophthalmoscopy with scleral depression. Although this could represent a differential verification bias (QUADAS item 6), both are visual examination techniques that do not differ in their definition of the target condition and can be regarded as the same reference standard. The implementation of a new gold standard consisting on unblinding final examination of patients with conflicting results from B-scan US and baseline examination allowed for secondary assessment of hypothetical false-negative results on baseline visual examination that otherwise could be recorded as false positives if they were to be detected on B-scan echography. In fact, this technique resulted in the recategorization of 3 false-negative results in baseline funduscopic examination. This assumption may be objected to because new retinal tears can develop after the initial PVD.1,13,14 A recent meta-analysis has shown that delayed retinal tears occur at a rate of 1.8%, but it is unlikely for a retinal tear to occur in the first 6 weeks after the initial examination.1,15,16 Three eyes in our study were found to have undiagnosed tears after the final unblinded examination. The time interval from presentation was 1, 2, and 3 days, making unlikely that a retinal tear would have developed in this period. One retinal tear was missed by B-scan US in this study. It was a small, solitary retinal tear located in the inferior retinal periphery of a highly myopic eye. In these eyes, the posterior vitreous cortex is usually disrupted and hardly displayed by B-scan US.17 Moreover, scanning the inferior meridians is more demanding as the presence of the superior orbital ring may not allow for a wide surface of the globe on which to place and adequately shift the probe. Both circumstances may represent a limitation of B-scan in cases of highly myopic eyes. In conclusion, indirect ophthalmoscopy and contact lens biomicroscopy are the gold standard for peripheral retinal examination because they allow for a direct view of the retina, which is mandatory for adequate evaluation and treatment of peripheral retinal diseases. However, B-scan US is a reliable alternative to diagnose PVD and to screen for retinal tears in patients with deficient media transparency or inadequate pupil dilation. The accuracy of B-scan US could validate a strategy for combined or parallel testing that would result in an overall increase in tears detection. Recently, a new probe of 20 Mhz has been introduced for clinical use which presumably may improve the test’s accuracy.
References 1. Byer NE. Natural history of posterior vitreous detachment with early management as the premier line of defense against retinal detachment. Ophthalmology 1994;101:1503–13.
Lorenzo-Carrero et al 䡠 Echographic Testing for Retinal Tears in Symptomatic PVD 2. Sharma S, Walker R, Brown G, Cruess A. The importance of qualitative vitreous examination in patients with acute posterior vitreous detachment. Arch Ophthalmol 1999;117:343– 6. 3. Davis MD. The natural history of retinal breaks without detachment. Trans Am Ophthalmol Soc 1973;71:343–72. 4. Hikichi T, Trempe CL. Relationship between floaters, light flashes, or both, and complications of posterior vitreous detachment. Am J Ophthalmol 1994;117:593– 8. 5. Coleman DJ, Silverman RH, Chabi A, et al. High-resolution ultrasonic imaging of the posterior segment. Ophthalmology 2004;111:1344 –51. 6. Fisher YL, Slakter JS, Friedman RA, Yannuzzi LA. Kinetic ultrasound evaluation of the posterior vitreoretinal interface. Ophthalmology 1991;98:1135– 8. 7. Arzabe CW, Akiba J, Jalkh AE, et al. Comparative study of vitreoretinal relationships using biomicroscopy and ultrasound. Graefes Arch Clin Exp Ophthalmol 1991;229:66 – 8. 8. Van Newkirk MR, Johnson MW, Hughes JR, et al. B-scan ultrasonographic findings in the stages of idiopathic macular hole. Trans Am Ophthalmol Soc 2000;98:163–9. 9. DiBernardo C, Blodi B, Byrne SF. Echographic evaluation of retinal tears in patients with spontaneous vitreous hemorrhage. Arch Ophthalmol 1992;110:511– 4. 10. Whiting P, Rutjes AWS, Reitsma JB, et al. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC
11. 12.
13.
14.
15.
16.
17.
Med Res Methodol [serial online] 2003;3:25. Available at: http://www.biomedcentral.com/1471-2288/3/25. Accessed July 25, 2008. Fagan TJ. Nomogram for Bayes theorem [letter]. N Engl J Med 1975;293:257. Fletcher RW, Fletcher SW. Clinical Epidemiology: The Essentials. 4th ed. Baltimore, MD: Lippincott Williams & Wilkins; 2005:35–58. Van Overdam KA, Bettink-Remeijer MW, Klaver CC, et al. Symptoms and findings predictive for the development of new retinal breaks. Arch Ophthalmol 2005;123:479 – 84. Sarrafizadeh R, Hassan TS, Ruby AJ, et al. Incidence of retinal detachment and visual outcome in eyes presenting with posterior vitreous separation and dense fundus-obscuring vitreous hemorrhage. Ophthalmology 2001;108:2273– 8. Coffee RE, Westfall AC, Davis GH et al. Symptomatic posterior vitreous detachment and the incidence of delayed retinal breaks: case series and meta-analysis. Am J Ophthalmol 2007; 144:409 –13. Richardson PS, Benson MT, Kirkby GR. The posterior vitreous detachment clinic: do new retinal breaks develop in the six weeks following an isolated symptomatic posterior vitreous detachment? Eye 1999;13:237– 40. Pierro L, Brancato R, Calori G. Clinical and ultrasound study of peripheral vitreoretinal adhesions. Retina 1997;17:131– 4.
Footnotes and Financial Disclosures Originally received: January 29, 2008. Final revision: August 18, 2008. Accepted: August 18, 2008. Available online: November 13, 2008.
Financial Disclosure(s): The authors have no proprietary or commercial interest in any materials discussed in this article. Manuscript no. 2008-145.
From the Ophthalmology Department, POVISA Hospital, Vigo, Spain. Presented at: American Academy of Ophthalmology Annual Meeting, November, 2006, New Orleans, Louisiana.
Correspondence: Dr Jose Lorenzo Carrero, Departamento de Oftalmologia, Hospital Povisa, C/Salamanca 5, Vigo 36211, Pontevedra, Spain. E-mail: josel.carrero@ yahoo.es.
99
Posterior vitreous detachment (PVD) is the most common age-related event that occurs in the vitreous and the principal predisposing risk factor for the development of a rhegmatogenous retinal detachment. Approximately 10%–14% of nonreferred patients with acute symptomatic PVD have a retinal tear on initial examination.1,2 Approximately 30%– 50% of symptomatic retinal breaks with persistent vitreoretinal traction will cause a clinical retinal detachment if left untreated.3 The early detection of retinal tears is crucial to providing early treatment that might prevent retinal detachment. No symptoms can reliably distinguish a PVD with an associated retinal break from those without. Therefore, accurate vitreous and retinal examinations are mandatory in eyes with symptomatic PVD.1– 4 The preferred method of evaluating peripheral vitreoretinal pathology is indirect ophthalmoscopy combined with scleral depression (Preferred Practice Patterns. Posterior vitreous detachment, retinal breaks, and Lattice degeneration. American Academy of Ophthalmology 2003). This method, mostly used by vitreoretinal specialists, requires adequate skill, pupillary dilatation, media
94
© 2009 by the American Academy of Ophthalmology Published by Elsevier Inc.
transparency, and patient cooperation. Developing accurate imaging procedures for detecting retinal tears in the setting of an acute PVD would allow for adequate evaluation of patients who are, otherwise, poor candidates for standard visual examination or get unreliable results. Recent advances in ultrasound (US) resolution have improved the acoustic detection of subtle structures in the vitreous cavity such as the posterior vitreous cortex and retinal tears, but a test validation in the setting of acute age-related PVD remains to be done.5–9 The purpose of this study was to determine the performance of B-scan US to screen for retinal tears in symptomatic age-related PVD.
Methods We conducted a cross-sectional blind comparative study with prospective data collection. The study was designed to meet the 14 items proposed by an expert panel for evidence based quality assessment of accuracy of diagnostic tests.10 The study group consisted of a consecutive series of symptomatic patients presentISSN 0161-6420/09/$–see front matter doi:10.1016/j.ophtha.2008.08.040
Lorenzo-Carrero et al 䡠 Echographic Testing for Retinal Tears in Symptomatic PVD
Figure 1. A, Complete PVD shown in a longitudinal echogram. Anterior limit of PVD at the vitreous base (arrow). Weiss’ ring (arrow head). B, PVD as shown in a transverse echogram. C–G, Traction retinal tears attached to the posterior vitreous cortex are seen as highly reflective echoes and clearly individualized from the adjacent retina. H–J, These echograms show discrete echogenic retinal ridges adherent to the posterior vitreous cortex that were interpreted as suspicious retinal tears, but no retinal tear could be confirmed on ophthalmoscopy. K, Intense vitreous synchysis with a posterior lacuna and PVD. L, Retinoschisis and PVD. PVD ⫽ Posterior vitreous detachment.
95
Ophthalmology Volume 116, Number 1, January 2009 ing to the emergency department of a nonreferral hospital who were diagnosed of acute age-related PVD and satisfied the criteria for inclusion. Patients who had symptoms suggestive of recent onset of PVD (visual floaters and/or photopsia) were referred to the ophthalmology department by primary emergency care practitioners. Comprehensive ophthalmic examination, undertaken by a general ophthalmologist in an emergency basis, included slit-lamp vitreous biomicroscopy with noncontact and/or 3-mirror lenses. Diagnosis of complete PVD was established when biomicroscopy confirmed the presence of a Weiss’ ring. The peripheral retina was examined with either slit-lamp biomicroscopy and 3-mirror contact lens or indirect binocular ophthalmoscopy with or without scleral indentation at the examiner’s discretion. Patients were excluded if they had onset of symptoms ⬎1 month, if they had ocular diseases thought to modify vitreous condition (e.g., history of recent blunt trauma, eye surgery, diabetes, uveitis, or tumors in any eye) or if they had media opacity that precluded appropriate ophthalmoscopic examination (cataract, vitreous hemorrhage, corneal opacities, or narrow pupil). Investigational review board approval was obtained and all subjects signed an informed consent form. After baseline examination, all participants were referred for blind B-scan US evaluation the next working day. Blind echographic examinations were performed by means of B-scan V Plus/Biovision (Quantel Medical, Clermont-Ferrand, France), which has a focused probe of 10 Mhz with a mechanical sector scanning. High-sensitivity gain settings of 105 dB specifically designed for vitreous examinations were selected. Examinations were performed by the same examiner under topical anesthesia with the patient in the supine position. The patient’s eyelids were kept open and the probe was placed directly on the ocular surface behind the limbus through 2.5% methylcellulose as a coupling gel. The probe was positioned in transverse sections (cross-sections) all around the conjunctiva and then progressively shifted to examine more peripheral aspects of the vitreoretinal interface. For adequate characterization of vitreoretinal relationships, movement of the vitreous was evaluated during voluntary saccadic motion of the patient’s globe while the probe was held stationary (kinetic examination). Patients who were initially diagnosed with PVD on biomicroscopy but showed posterior vitreous attachment to the optic nerve head on B-scan US were considered to be incorrectly diagnosed and excluded from study analysis. If a suspected retinal tear was found, longitudinal probe orientation (radial section) was used to determinate its location. According to dynamic echographic examination, the examiner classified and photographed the presence of retinal tears as (1) “suspicious” when the US image depicted an echogenic retinal traction ridge that was adherent to the posterior vitreous cortex, but a retinal flap could not be clearly individualized (Fig 1H–J); (2) “definite” when the US image showed a echogenic retinal flap or operculum continuous with the posterior vitreous cortex (Fig 1C–G); or (3) “absent” when no vitreoretinal adherences could be detected (Fig 1A, B, K). For analysis purposes, retinal tears classified as “definite” and “suspicious” were considered as a positive diagnosis. Patients who presented conflicting results between the US examination and the baseline examination were subsequently referred for directed indirect ophthalmoscopy based on the echographic findings by a vitreoretinal specialist. The results on final indirect ophthalmoscopy served as the new reference standard with which the results of B-scan echography and the baseline funduscopic examination were compared.
Diagnostic Test Characteristics Sensitivity, specificity, and positive and negative predictive values were calculated with confidence intervals for baseline funduscopic
96
examination and B-scan US. Likelihood ratios and the posttest probability of retinal tears given the pretest probability and test characteristics were calculated with confidence intervals for B-scan US according to a Bayesian analysis model. Graphical calculations of Bayes’ theorem were displayed in a Fagan’s nomogram.11 Demographic characteristics of patients presenting retinal tears (age, gender, and eye’s axial length) were compared with those who had uncomplicated PVD. Tests of significance included the Pearson chi-square and Fisher tests for nominal data and t-test for continuous data. If P⬍5%, the result was considered significant.
Results From previous pilot studies performed in this setting, (Invest Ophthalmol Vis Sci 2004;45:E-Abstract 2009) we concluded that a sample of 250 patients would allow to estimate the prevalence of retinal tears in patients with PVD with a confidence interval of ⫾4% and ␣ error ⫽ 0.05. Of 250 consecutively enrolled patients, 11 were excluded from the study analysis because of incorrect initial diagnosis of PVD. The study population comprised 239 patients of mean age 65 years (range, 40 – 82). There were 96 (40%) men and 143 (60%) women. Of 239 patients, 26 (10.8%) had final evidence of ⱖ1 retinal tears associated with PVD. In 22 eyes, there was a solitary retinal flap tear; 2 other eyes had a tear with a free operculum, and 2 eyes showed 2 retinal tears each. Demographic characteristics of patients presenting retinal tears (age, gender, and eye’s axial length) were similar to those who had uncomplicated PVD (Table 1). Table 2 summarizes the test performance of baseline biomicroscopy/ophthalmoscopy and kinetic B-scan US for the analysis of retinal tears. The prevalence of patients with retinal tears was 10.8% (26/239 patients). Baseline funduscopic examination detected 25 retinal tears, whereas blind B-scan kinetic US detected 30 retinal tears, among which 25 were classified as “definite” and 5 were classified as “suspicious” by the examiner. There were 3 false-positive retinal tears, which had been classified as “suspicious” on B-scan US by the examiner (Fig 1H–J). There were 3 additional retinal tears initially missed on baseline funduscopic evaluation that were detected by blind B-scan US and confirmed by unblinded ophthalmoscopy. There was also 1 false negative on B-scan US, which was confirmed by final indirect ophthalmoscopy. Analysis of sensitivity for detection of retinal tears was done according to the total number of retinal tears. The sensitivity of B-scan US for detection of retinal tears was 96.4% (27/28) and the sensitivity of baseline funduscopic examination was 89.2% (25/28). The specificity of B-scan US was 98% (210/213) and the specificity of baseline funduscopic examination was 100% (213/ 213). The positive and negative predictive values for B-scan US were 89% (25/28) and 99% (210/211), respectively; for baseline funduscopic examination, they were 100% (24/24) and 99% (213/ 215), respectively.
Table 1. Demographic Characteristics of Study Patients with and without Retinal Tears
Gender Mean age (SD) Mean axial length (SD)
Patients with RT
Patients without RT
P
14 M/12 F 65.04 (⫾5.9) 23.9 (⫾2)
84 M/129 F 65.18 (⫾6.9) 23.4 (⫾1.2)
0.53 0.92 0.12
F ⫽ female; M ⫽ male; RT ⫽ retinal tear; SD ⫽ standard deviation.
Lorenzo-Carrero et al 䡠 Echographic Testing for Retinal Tears in Symptomatic PVD Table 2. Performance Characteristics of Baseline Ophthalmoscopy/Biomicroscopy and B-Scan Ultrasounds (US) for the Detection of Retinal Tears in Symptomatic Age-Related Posterior Vitreous Detachment (239 Patients)
Identified total tears False-negative retinal tears False-positive retinal tears Missed eyes with tears Identified eyes with tears Sensitivity§ Specificity Positive predictive value Negative predictive value Negative likelihood ratio Positive likelihood ratio
Baseline Ophthalmoscopy/ Biomicroscopy
B-Scan US
25* 3‡ 0 2 24* 89% [72%–98%] 100% [98%–100%] 100% [86%–100%] 99% [97%–100%] 0.08 [0.03–0.30] Inf
27† 1 3 1 25† 96% [82%–100%] 98% [96%–100%] 89% [86%–93%] 99% [98%–100%] 0.04 [0.01–0.27] 68 [22–211]
Inf ⫽ infinite. *One eye had 2 retinal tears. † Two eyes were shown to have 2 retinal tears. ‡ One retinal tear was missed in 1 eye that was found to already have 1 retinal tear. § Analysis according to retinal tears.
Table 3 summarizes the clinical findings of eyes that were misclassified by either baseline ophthalmoscopy or B-scan US and required the application of the new gold standard. Figure 2 shows the likelihood Fagan’s nomogram of B-scan US. For a positive test result, the posttest probability is 89% (posttest odds, 8.3), whereas for a negative test result the posttest probability and odds is zero. Hence, a positive test result increases the probability of having a retinal tear from 10% to 89% and a negative result almost excludes the probability of having a retinal tear. Additional findings on baseline funduscopic examination include peripheral retinal abnormalities. Among the 239 eyes with PVD, 74 (31%) had visible retinal abnormalities; 29 (12.2%) had lesions of lattice degeneration. Of these 29 eyes, 4 (13.7%) had associated retinal tears and 3 (10.3%) had atrophic holes of lattice lesions. One eye had degenerative retinoschisis (Fig 1L), and 3 eyes showed tiny round retinal holes with no associated retinal abnormality. Sixteen eyes (6.7%) showed peripheral retinal abnormalities without clinical significance (peripheral cystoid degeneration and cobblestone degeneration).
Discussion Our results indicate that B-scan US is an accurate diagnostic test for detection of retinal tears in symptomatic patients
with acute PVD. The likelihood that a retinal tear would be missed on B-scan US is very low given the high negative predictive value found in this study. Patients with retinal tears are 68 times more likely (positive likelihood ratio ⫽ 68) to have their retinal tears detected on B-scan US than showing a normal echogram. The prevalence of retinal tears associated with PVD in this series was 10.8%, which is similar to that found in other studies that analyzed series of nonreferred patients.1,2 We have used B-scan US to screen symptomatic patients and not as a standard screening test, which would typically apply to asymptomatic patients with a low pretest probability. As long as the prevalence can vary, it modifies predictive values and posttest odds (Bayes’ theorem). The prevalence of complicated PVD seen in referral hospitals is likely to be higher; therefore, the negative predictive values of B-scan US would be lower than that obtained in this study. However, we do not believe this effect would change significantly the test performance of B-scan US. Based on the calculated likelihood nomogram, we can hypothesize that a prevalence up to 3 times higher than that obtained in this study would result in posttest probability of 98% for a positive test and 1% for a negative test, which are not very different from those obtained in this study.12 Comparative studies carried out in eyes with clear media suggest that US may assess the posterior vitreoretinal interface better than biomicroscopy but we could not find any article in a Medline search reporting on a validation of ultrasounds to detect retinal tears in PVD.6 – 8 We have used 105-dB gain setting to ensure less wave attenuation and high signal to noise ratio. B-scan US allows for better kinetic evaluation of peripheral vitreoretinal relationships than biomicroscopy with contact lens or indirect ophthalmoscopy because the ocular movements are less restricted.6,7 The Quality Assessment of Diagnostic Accuracy Studies panel (QUADAS) developed a tool composed of 14 items to evaluate bias, variability, and quality of reporting.10 Although we followed the recommendations to meet the 14 items, several items deserve further discussion. We excluded patients with obscuring vitreous hemorrhage because waiting for clearing of vitreous hemorrhage would make it likely that the target condition would change before the reference test could be performed (QUADAS item 4) because one third of patients with PVD and vitreous hemorrhage at presentation develop new retinal tears during follow-up.13 When PVD is complicated by vitreous hemorrhage, the incidence of retinal tears increases to 40%– 67%.9,14 In
Table 3. Clinical Findings of Misclassified Eyes Age
Gender
Axial Length
Complaints
Baseline Examination
B-Scan Ultrasound
Gold Standard
64 56 67 50 72 63 69
F F M F M M M
23.06 22.22 23.32 30.46 24.01 26.26 23.09
Photopsia ⫹ floaters Floaters Floaters Floaters Photopsia ⫹ floaters Floaters Floaters
1 flap RT (upper TQ) Normal fundus Preretinal hemorrhage 1 RT (inferior NQ) Normal fundus Retinal hole ⫹ lattice Normal fundus
2 flap RT 1operculum 1 flap RT Vitreous synchysis Suspicious RT Suspicious RT Suspicious RT
2 flap RT 1operculum ⫹ RT 1 flap RT 1 RT VR traction Retinal hole ⫹ lattice VR traction
F ⫽ female; M ⫽ male; NQ ⫽ nasal quadrant; RT ⫽ retinal tear; TQ ⫽ temporal quadrant; VR ⫽ vitreoretinal.
97
Ophthalmology Volume 116, Number 1, January 2009
Figure 2. Likelihood ratio nomogram B-scan Ultrasound (US). Prior probability (odds) ⫽ 11% (0.1). Positive test: positive likelihood ratio, 68 (22, 211). Posterior probability (odds), 89% (8.3) [73%, 96%]. Negative test: negative likelihood ratio, 0.04 (0.01, 0.27). Posterior probability (odds), 0% (0.0) [0%, 3%].
those cases, echographic evaluation is the preferred method to support clinical decisions. One retrospective study found a sensitivity of 91% and specificity of 97% in identifying retinal tears in eyes with PVD and vitreous hemorrhage.9 However, evaluating test accuracy with such a high prevalence of the target condition could lead to a biased assessment of the test properties and compromise the external validity for a broader spectrum of patients. To extend its clinical applicability to patients with unreliable standard evaluation apart from vitreous hemorrhage, we carried it out in nonreferred patients regardless of symptoms severity (QUADAS item 1). Although the recommended method of evaluating peripheral vitreoretinal pathology is indirect ophthalmoscopy combined with scleral depression, most ophthalmologists in
98
continental Europe rely on biomicroscopy with a 3-mirror contact lens as the reference standard. The study protocol allowed the examiner to choose the method of fundus peripheral examination. Five participating general ophthalmologists acknowledged to have routinely performed slitlamp biomicroscopy with a 3-mirror contact lens; another preferred indirect ophthalmoscopy with scleral depression. Although this could represent a differential verification bias (QUADAS item 6), both are visual examination techniques that do not differ in their definition of the target condition and can be regarded as the same reference standard. The implementation of a new gold standard consisting on unblinding final examination of patients with conflicting results from B-scan US and baseline examination allowed for secondary assessment of hypothetical false-negative results on baseline visual examination that otherwise could be recorded as false positives if they were to be detected on B-scan echography. In fact, this technique resulted in the recategorization of 3 false-negative results in baseline funduscopic examination. This assumption may be objected to because new retinal tears can develop after the initial PVD.1,13,14 A recent meta-analysis has shown that delayed retinal tears occur at a rate of 1.8%, but it is unlikely for a retinal tear to occur in the first 6 weeks after the initial examination.1,15,16 Three eyes in our study were found to have undiagnosed tears after the final unblinded examination. The time interval from presentation was 1, 2, and 3 days, making unlikely that a retinal tear would have developed in this period. One retinal tear was missed by B-scan US in this study. It was a small, solitary retinal tear located in the inferior retinal periphery of a highly myopic eye. In these eyes, the posterior vitreous cortex is usually disrupted and hardly displayed by B-scan US.17 Moreover, scanning the inferior meridians is more demanding as the presence of the superior orbital ring may not allow for a wide surface of the globe on which to place and adequately shift the probe. Both circumstances may represent a limitation of B-scan in cases of highly myopic eyes. In conclusion, indirect ophthalmoscopy and contact lens biomicroscopy are the gold standard for peripheral retinal examination because they allow for a direct view of the retina, which is mandatory for adequate evaluation and treatment of peripheral retinal diseases. However, B-scan US is a reliable alternative to diagnose PVD and to screen for retinal tears in patients with deficient media transparency or inadequate pupil dilation. The accuracy of B-scan US could validate a strategy for combined or parallel testing that would result in an overall increase in tears detection. Recently, a new probe of 20 Mhz has been introduced for clinical use which presumably may improve the test’s accuracy.
References 1. Byer NE. Natural history of posterior vitreous detachment with early management as the premier line of defense against retinal detachment. Ophthalmology 1994;101:1503–13.
Lorenzo-Carrero et al 䡠 Echographic Testing for Retinal Tears in Symptomatic PVD 2. Sharma S, Walker R, Brown G, Cruess A. The importance of qualitative vitreous examination in patients with acute posterior vitreous detachment. Arch Ophthalmol 1999;117:343– 6. 3. Davis MD. The natural history of retinal breaks without detachment. Trans Am Ophthalmol Soc 1973;71:343–72. 4. Hikichi T, Trempe CL. Relationship between floaters, light flashes, or both, and complications of posterior vitreous detachment. Am J Ophthalmol 1994;117:593– 8. 5. Coleman DJ, Silverman RH, Chabi A, et al. High-resolution ultrasonic imaging of the posterior segment. Ophthalmology 2004;111:1344 –51. 6. Fisher YL, Slakter JS, Friedman RA, Yannuzzi LA. Kinetic ultrasound evaluation of the posterior vitreoretinal interface. Ophthalmology 1991;98:1135– 8. 7. Arzabe CW, Akiba J, Jalkh AE, et al. Comparative study of vitreoretinal relationships using biomicroscopy and ultrasound. Graefes Arch Clin Exp Ophthalmol 1991;229:66 – 8. 8. Van Newkirk MR, Johnson MW, Hughes JR, et al. B-scan ultrasonographic findings in the stages of idiopathic macular hole. Trans Am Ophthalmol Soc 2000;98:163–9. 9. DiBernardo C, Blodi B, Byrne SF. Echographic evaluation of retinal tears in patients with spontaneous vitreous hemorrhage. Arch Ophthalmol 1992;110:511– 4. 10. Whiting P, Rutjes AWS, Reitsma JB, et al. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC
11. 12.
13.
14.
15.
16.
17.
Med Res Methodol [serial online] 2003;3:25. Available at: http://www.biomedcentral.com/1471-2288/3/25. Accessed July 25, 2008. Fagan TJ. Nomogram for Bayes theorem [letter]. N Engl J Med 1975;293:257. Fletcher RW, Fletcher SW. Clinical Epidemiology: The Essentials. 4th ed. Baltimore, MD: Lippincott Williams & Wilkins; 2005:35–58. Van Overdam KA, Bettink-Remeijer MW, Klaver CC, et al. Symptoms and findings predictive for the development of new retinal breaks. Arch Ophthalmol 2005;123:479 – 84. Sarrafizadeh R, Hassan TS, Ruby AJ, et al. Incidence of retinal detachment and visual outcome in eyes presenting with posterior vitreous separation and dense fundus-obscuring vitreous hemorrhage. Ophthalmology 2001;108:2273– 8. Coffee RE, Westfall AC, Davis GH et al. Symptomatic posterior vitreous detachment and the incidence of delayed retinal breaks: case series and meta-analysis. Am J Ophthalmol 2007; 144:409 –13. Richardson PS, Benson MT, Kirkby GR. The posterior vitreous detachment clinic: do new retinal breaks develop in the six weeks following an isolated symptomatic posterior vitreous detachment? Eye 1999;13:237– 40. Pierro L, Brancato R, Calori G. Clinical and ultrasound study of peripheral vitreoretinal adhesions. Retina 1997;17:131– 4.
Footnotes and Financial Disclosures Originally received: January 29, 2008. Final revision: August 18, 2008. Accepted: August 18, 2008. Available online: November 13, 2008.
Financial Disclosure(s): The authors have no proprietary or commercial interest in any materials discussed in this article. Manuscript no. 2008-145.
From the Ophthalmology Department, POVISA Hospital, Vigo, Spain. Presented at: American Academy of Ophthalmology Annual Meeting, November, 2006, New Orleans, Louisiana.
Correspondence: Dr Jose Lorenzo Carrero, Departamento de Oftalmologia, Hospital Povisa, C/Salamanca 5, Vigo 36211, Pontevedra, Spain. E-mail: josel.carrero@ yahoo.es.
99