Optical coherence tomography assessment for attached posterior vitreous in eyes with macular holes and epiretinal membranes

Abstract

Purpose: The purpose of this study was to assess the reliability of spectral-domainoptical coherence tomography (OCT) as a diagnostic tool in the detection of an attachedposterior vitreous.

Methods: A double-blind comparison of the vitreous imaged by OCT acquired on theday of surgery with the status of the posterior vitreous assessed at the time of surgery.

Results: A total of 103 patients were recruited to the study, 67 of whom had completedata and were included in the analysis. 37 eyes had macular holes (MHs), 23 eyes hadepiretinal membranes (ERM), and the remaining 7 eyes had other macular pathology,including vitreomacular traction syndrome and lamellar holes. The overall sensitivity ofOCT in detecting an attached posterior vitreous was 80% and the specificity was 56%.For MH, this was 88% and 60%, respectively, whereas for ERM, it was only 45% and50%.

Conclusions: The assessment of attachments of the posterior hyaloid membrane onOCT can reliably predict attached posterior vitreous in MH but not in ERM.

Introduction

For a number of macular pathologies such as macular holes(MHs) and epiretinal membranes (ERM), vitreoretinal surgeryis required to restore or preserve central vision. An importantpart of surgery is to detect any attachments of the posteriorvitreous and induce a posterior vitreous detachment (PVD)before peeling any macular membranes. The pre-operativediagnosis of attached posterior vitreous is not always reliable andmany times the true state of posterior vitreous attachments isfound at surgery.

A number of imaging modalities have so far been reportedas diagnostic tools for assessing the posterior vitreous. Theseinclude the scanning laser ophthalmoscope, B-scan ultrasound,time-domain, and spectral-domain OCT.^[1] Kicova et al. foundthat B-scan ultrasonography and examination of the patient withslit lamp biomicroscopy were more reliable than OCT alone indiagnosing a PVD.^[2] The authors showed that only a proportionof patients (30%) can have their posterior hyaloid imaged usingOCT, with a poor correlation between pre-operative assessmentof the posterior hyaloid and intraoperative findings duringvitrectomy.^[2] However, their study was based on a small sampleof only 30 eyes in 30 patients. Other studies have contradictedthe findings by Kicova et al. demonstrating a high correlationbetween pre-operative OCT imaging and intraoperative findingsof the status of the PVD (κ = 0.947).^[3] In this latter study, therewas also 100% agreement between the two observers of the OCTfindings preoperatively. Gallemore et al. reviewed OCT imagesof 139 eyes retrospectively and showed that OCT was better ableto identify vitreoretinal adhesion than B-scan ultrasonography.This suggests that OCT is more accurate in determining thestatus of the posterior hyaloid.^[4] OCT has also been used tosubclassify the progressive stages of a partial PVD developinginto a complete PVD.^[5] In a study by Uchino et al., eyes withfully detached vitreous slit lamp examination had no identifiablevitreous signal on OCT because the posterior hyaloid was at agreater distance from the retinal surface than the OCT was ableto detect within its scan range.

The aim of this study was to further explore the ability ofOCT to detect the status of the posterior hyaloid preoperativelyin patients undergoing vitrectomy and to define OCT diagnosticcriteria for predicting an attached posterior hyaloid at surgery.The sensitivity and specificity of OCT in pre-operative predictionof the posterior hyaloid status was compared to intraoperativefindings during vitrectomy.

Methods

This is a prospective cross-sectional study conducted at theBristol Eye Hospital, Bristol, UK. The study complies with theDeclaration of Helsinki, and ethical approval was obtained fromthe local ethics review board (Ref OP/2008/2994, CentralBristol Research Ethics Committee). All patients gave writteninformed consent before taking part in the study.

An initial pilot study was performed to define OCT diagnosticcriteria for attached posterior vitreous and PVD. 19 patientsunderwent full biomicroscopic slit lamp examinations and OCTscanning of the macula. Patients were scanned by a maskedinvestigator using the Topcon three-dimensional (3D) OCT1000 (Topcon Corporation, Japan). The cube protocol at fourpositions at the posterior fundus was implemented: The macula,the disk, the superior-temporal, and the inferior-temporalvascular arcades. Each cube scan consisted of single parallelOCT scans, with no averaging, forming a 6/6 mm area. TheOCT scan was assessed for the presence of a vitreous signal,using the 3D mode; where the vitreous signal is more visiblesince OCT signals from adjacent line scans are used to build athree-dimensional structure. Following the OCT scanning, thepatient was assessed at the slit lamp by a masked physician andthe agreement between the slit lamp examination and the OCTwas thereafter assessed.

For the second part of the study, patients with macularpathology requiring vitrectomy were recruited. The inclusioncriteria for the study were as follows: (1) Listing for macularsurgery, (2) clear optical media, (3) refractive error < 6 diopters,(4) no previous retinal surgery, (5) no proliferative diabeticretinopathy, and (6) no retinal detachment. The exclusioncriteria were as follows: (1) No OCT scan performed on the dayof surgery, (2) missing data on the operative form, (3) < 21 yearsof age, and (4) unable to give informed consent. On the day ofsurgery, patients had a pre-operative OCT scan with the sameprotocol as the pilot study detailed above. During the vitrectomy,the surgeon, who was masked to the study OCT scan, assessedthe following surgical signs: The presence of a visibly detachedposterior hyaloid membrane attaching to the vitreous base andthe presence of a fish-strike sign using a Flynn flute flexiblesilicone-tipped cannula and the induction of a PVD duringsurgery. OCT scans were assessed by a separate clinician whowas masked to the surgical findings. OCT scans were assessedfor the signs defined by the pilot study. The 3D mode was chosento visualize the presence of any vitreous signal within the scanvolume. The macular diagnosis for each case was based on theOCT findings for this study.

Data processing and statistical analysis

Data were treated statistically with excel spreadsheets (Excel2010, Microsoft Corporation). Two-tailed Fisher's exact testwas performed using an online calculator (http://research.microsoft.com/en-s/um/redmond/projects/mscompbio/fisherexacttest/).

Results

In the pilot study, 11 of 19 patients were found not to have aPVD and for 10 of these patients, the OCT found evidence ofan attached vitreous. The remaining eight patients were foundto have a PVD, and for seven of these, there was no evidence ofattached vitreous on OCT. This gave a 91% (10/11) sensitivityof the OCT detecting an attached posterior vitreous and 88%(8/9) specificity.

In the pilot study, a number of OCT signs were found whichsupported the finding of an attached posterior vitreous includingthe visualization of an attached posterior hyaloid face or vitreousstrands to the disk with or without attachments to the retinaor blood vessels [Figure 1]. The absence of any vitreous signalwithin the scan volume was an indication of a PVD [Figure 2].

In total, 103 patients were recruited. 53 patients weresubsequently excluded: 14 patients were found not to meet thestudy inclusion criteria, while 10 patients had no OCT scanperformed on the day of surgery, and 26 patients had missingdata on the operative form.

Of the 67 patients who completed the study, 37 patientshad a full thickness MH (FTMH), while 23 had ERM, 4 hadvitreomacular traction syndrome, and 2 had lamellar MHassociated with an ERM, and one case had macular edemawithout an apparent ERM. Of the 23 ERM seen, 13 extendedbeyond the temporal vascular arcades as seen on OCT, while 10appeared to remain confined within the arcade vessels.

Of the 67 recruited patients, 49 eyes were found to have anattached posterior vitreous at surgery and the remaining 18 hada PVD present. The OCT analysis showed 42 eyes had evidenceof an attached posterior vitreous and 14 eyes had no evidenceof attached posterior vitreous. In 11 eyes, there were no OCTdiagnostic criteria met as the OCT could not detect the posteriorhyaloid membrane, but there was possible vitreous signal withinthe scan area. These results are summarized in Chart 1.

[IMAGE OMITTED: SEE PDF]

[IMAGE OMITTED: SEE PDF]

[IMAGE OMITTED: SEE PDF]

For all pathologies, 49 of 67 eyes had attached posteriorvitreous at surgery. Including equivocal results, the OCThad 80% sensitivity and 56% specificity in detecting attachedposterior vitreous, which gave a 93% positive predictive valueand 71% negative predictive value. Excluding ungradable results,sensitivity rose to 91% and specificity was 71% [Table 1].

For FTMH, 32 of 37 eyes had attached posterior vitreousat surgery. 11 eyes (26%) had an equivocal OCT result. OCTincluding equivocal OCT results had 88% sensitivity and 60%specificity in detecting attached posterior vitreous, which gavea 93% positive predictive value and 75% negative predictivevalue. Excluding ungradable OCTs sensitivity rose to 97% andspecificity was 60% [Table 2].

For patients with ERM, 11 of 23 eyes had attached posteriorvitreous at surgery. Eight eyes (35%) had an equivocal OCTresult. OCT including equivocal OCT results had 45% sensitivityand 50% specificity in detecting attached posterior vitreous,which gave an 83% positive predictive value and 33% negativepredictive value. Excluding equivocal results, sensitivity rose to63% and specificity was 86% [Table 3]. Figure 3 shows a falsepositivecase and Figure 4 a false-negative case.

Cases of FTMH were more likely to have an OCT meetinggrading criteria, compared to ERM (Table 4, Fisher's exact test:P = 0.0153 [two-tailed], statistically significant association).

Optical coherence tomography (OCT) data were obtainedfrom the four locations on the optic disk, macula, superiortemporal, and inferior temporal vascular arcade. For patientswith confirmed attachments of the posterior vitreous at surgery,only the MH group had enough number of eyes for analysis. Allpatients with no PVD had on OCT the vitreous visibly attachedto the optic disk (which was the OCT definition of attachedposterior vitreous). Half of cases had detached posterior hyaloidfrom the macular surface and slightly more than that haddetached posterior hyaloid from the retina around the temporalvascular arcades. Table 5 summarizes these results.

Discussion

For MH, there was a good agreement between surgical findingsand the OCT interpretation of the posterior hyaloid. Thediagnostic ability of spectral-domain OCT mainly relies on thedetection of a partially detached posterior hyaloid membrane asthis serves as the primary interface where signal is detected. Thepathology of an FTMH depends at least on a partial separation ofthe posterior hyaloid membrane from the surface of the macula.Detecting the attachment of the posterior hyaloid membraneto the disk with OCT is a sensitive sign of an attached posteriorvitreous. The attachments of the posterior vitreous to theposterior retina were variable in cases of FTMH. In around halfof the cases, the only attachment of the vitreous to the posteriorpole was at the optic disc, and not at the macula or temporalvascular arcades. This finding is in keeping with previous studieson the evolution of PVD both in normal eyes^[5] and fellow eyes ofpatients with a preexisting unilateral MH.^[6]

The difference in sensitivity and specificity of OCT as adiagnostic tool for attached posterior vitreous differs greatlybetween cases of FTMH and cases of ERM. For the former,OCT carries a good diagnostic sensitivity of predicting anattached vitreous at surgery, if the vitreous signal is detected onOCT. OCT diagnostic criteria were met in the vast majority ofFTMH cases (34 of 37 eyes). On the other hand, OCT could notpredict the state of posterior vitreous attachment for ERM, withmany examples of apparent vitreous signal and a PVD observedat surgery and vice versa, no OCT signal seen with attachedposterior vitreous. There were cases where the posterior vitreouswas completely attached with no separation of the posteriorhyaloid membrane from the retinal surface. The OCT did nothave enough definition to distinguish between cortical vitreousand a posterior vitreous cavity free from vitreous. Moreover,undulations of the ERM were mistaken for the posterior hyaloidmembrane and attached vitreous was predicted in cases wherea full PVD was found at surgery. Therefore, the detection of anapparent posterior hyaloid membrane is not a reliable sign inpredicting attached vitreous in ERM.

[IMAGE OMITTED: SEE PDF]

[IMAGE OMITTED: SEE PDF]

[IMAGE OMITTED: SEE PDF]

[IMAGE OMITTED: SEE PDF]

[IMAGE OMITTED: SEE PDF]

[IMAGE OMITTED: SEE PDF]

[IMAGE OMITTED: SEE PDF]

The anatomic detail visible with spectral-domain OCT appearsto be a limiting factor with respect to its ability to visualize thevitreous. The posterior hyaloid membrane was clearly seen inmany eyes, mainly with FTMH, but in eyes with ERM, the OCTwould not reliably detect vitreous cortex if the hyaloid membranewas not partly detached from the retinal surface. This was morecommon in cases of ERM. The OCT machine used in this studyuses single OCT line scans without averaging. OCT platforms fromother manufacturers are able to average a number of line scans toimprove anatomic definition and reduce noise. This may improvethe diagnostic sensitivity of the OCT. Swept source OCT (ss-OCT)is now providing even better imaging capabilities of the posteriorsegment. The visualization of the vitreoretinal interface appears tobe very good although ss machines use longer wavelengths to imagestructures deeper to the retina, which limits the amount of anatomicdetail visible at the level of the posterior vitreous and retina.^[7]Indeed, as many as 25.6% of patients can still not have the status ofthe PVD graded on ss-OCT.^[8] Until validation studies are carriedout using these new OCT techniques on cases with ERM, we advisecaution in interpreting posterior vitreous attachments. Indeed,spectral-domain OCT is currently the most widely used OCTtechnology, and this study confirms its use in predicting the state ofposterior vitreous attachment in MH. However, it also highlights itslimitations in determining the vitreous state in cases of ERM.

References

Mirza RG, Johnson MW, Jampol LM. Optical coherencetomography use in evaluation of the vitreoretinal interface: A review.Surv Ophthalmol 2007;52:397-421.

Kicova N, Bertelmann T, Irle S, Sekundo W, Mennel S. Evaluationof a posterior vitreous detachment: A comparison of biomicroscopy,B-scan ultrasonography and optical coherence tomographyto surgical findings with chromodissection. Acta Ophthalmol2012;90:e264-8.

Rahman R, Chaudhary R, Anand N. Verification of posterior hyaloidstatus during pars plana vitrectomy, after preoperative evaluation onoptical coherence tomography. Retina 2012;32:706-10.

Gallemore RP, Jumper JM, McCuen BW 2nd, Jaffe GJ, Postel EA,Toth CA, et al. Diagnosis of vitreoretinal adhesions in macular diseasewith optical coherence tomography. Retina 2000;20:115-20.

Uchino E, Uemura A, Ohba N. Initial stages of posterior vitreousdetachment in healthy eyes of older persons evaluated by opticalcoherence tomography. Arch Ophthalmol 2001;119:1475-9.

Ma F, Arcinue CA, Barteselli G, Cheng L, Ezon I, Lee SN, et al.Optical coherence tomography findings of the vitreoretinal interfacein asymptomatic fellow eyes of patients with acute posterior vitreousdetachment. Retina 2014;34:447-54.

Mrejen S, Spaide RF. Optical coherence tomography: Imaging ofthe choroid and beyond. Surv Ophthalmol 2013;58:387-429.

Stanga PE, Sala-Puigdollers A, Caputo S, Jaberansari H, Cien M,Gray J, et al. In vivo imaging of cortical vitreous using 1050-nmswept-source deep range imaging optical coherence tomography.Am J Ophthalmol 2014;157:397-40400.