Surgery for Idiopathic Full-Thickness Macular Hole: Two-Year Results of a Randomized Clinical Trial Comparing NaturalHistory, Vitrectomy, and Vitrectomy Plus Autologous Serum: Moorfields MacularHole Study Group Report No. 1

Objectives  To determine the benefits of idiopathic full-thickness macular hole(FTMH) surgery compared with observation and to evaluate the use of autologousserum as an intraoperative adjunct.

Methods  A randomized clinical trial was performed to evaluate the anatomic andvisual benefits of FTMH surgery for lesions of 9 months or less symptom durationand visual acuity of 20/60 or less. We compared surgery with natural historyand determined whether use of intraoperative adjunctive autologous serum improvesthe surgical outcome. Eyes were randomized to (1) observation, (2) vitrectomy,or (3) vitrectomy plus serum and were followed for 24 months to assess anatomicstatus and visual function.

Results  In total, 185 eyes of 174 patients were enrolled. In the observationgroup, spontaneous closure of the FTMH occurred in 7 (11.5%) of 61 patients,with little or no change in overall acuity levels in 24 months. In contrast,the surgical groups had an overall closure rate of 80.6% (100/124) at 24 months,with 45% of eyes achieving Snellen acuity of 20/40 or greater. Surgical eyeshad better median near acuity than observation eyes by 6 lines (N5 vs N14).Use of autologous serum did not seem to affect anatomic or visual results.At 24 months, 72 (58.1%) of 124 surgical eyes had undergone cataract extraction.

Conclusions  Surgery for FTMH is safe and effective and is associated with significantvisual improvement compared with the natural history. Autologous serum applicationdoes not enhance the results of surgery.

Numerous studies^1-38 havedescribed the benefits of surgical treatment for idiopathic full-thicknessmacular hole (FTMH). The rationale for performing a vitrectomy and gas tamponadeis 2-fold. First, vitrectomy allows the removal of direct vitreous tractionfrom the fovea and, second, postoperative gas tamponade results in flatteningand reapposition of the hole edges, which facilitates glial repair after surgery.In this way, after successful anatomic closure of the hole, normal or nearnormal foveal architecture could be reestablished, with an improvement invision.

Although the natural history of FTMH has been assumed to carry a pooranatomic and visual prognosis,^39-47 avariety of studies have described arrest, spontaneous closure of holes, orboth. The incidence of improvement in acuity in untreated holes has been estimatedto be 3.4% to 10.0%.^39-47

Studies^1-38 describingsurgical treatment for FTMH have been mostly uncontrolled; they have estimatedthat the proportion of patients experiencing improvement in visual acuityof 2 or more Snellen lines after surgery for stage 2, 3, or 4 holes is approximately40% to 83%. Other studies have reported improved anatomic and visual resultswith the use of intraoperative adjuncts such as transforming growth factor β2,^3,4 autologous serum,¹⁶ autologousplatelets,^18-20 andthrombin and fibrin mixtures,^21,22,24 with50% to 100% of eyes achieving an improvement of at least 2 Snellen lines.These adjuncts are thought to improve anatomic and visual results by promotingglial repair of the hole after surgery.

Despite these encouraging results, many patients experience "adverse"effects after surgery, some of which may threaten sight, including persistenceof the hole in 0% to 46% of eyes^1-38;acceleration of nucleosclerosis in up to 80%,⁴⁸ with25% requiring cataract surgery at 2 years^1,8;peripheral field defects^49-58;retinal tears (3%)⁵⁹; retinal detachments (RDs)(1.8%-14.0%)^59-61;retinal pigment epithelium damage or phototoxicity (1%-3%)^59,60,62,63;and glaucoma.⁶⁴ The incidence of adverse effectshighlights the importance of randomized clinical trials (RCTs) that includeobservational controls and incorporate strict criteria for ensuring that particularmedical centers involved use only surgeons who have a high level of surgicalexperience.

The Vitrectomy for Macular Hole Study Group (VMHSG) conducted an RCT,including an observational arm.^{10,11,13,60,65} Thestudy demonstrated only a modest benefit in the group undergoing surgery forstage 2, 3, and 4 lesions compared with the observation group. In addition,an unusually high incidence of vision-threatening complications was reported.A variety of issues regarding methodologic design make these data difficultto interpret. The study protocol precluded cataract surgery in the surgicalcases at any stage of the trial. Thus, as nucleosclerosis is common aftersurgery, much of the visual benefit in the treatment cohort would have beenmasked.

Now, although the treatment of choice for idiopathic FTMH is surgery,it is generally agreed that further RCT data are required to evaluate theexact benefit of surgery compared with the natural history of the condition.^32,65,66 This article describesa masked-observer, RCT evaluating (1) the benefit of surgery compared withobservation and (2) the benefit of autologous serum application as an intraoperativeadjunct compared with surgery alone for recent-onset idiopathic FTMH. Thestudy was carried out at a single center, Moorfields Eye Hospital, with anexperienced consultant surgeon (Z.G.) performing all of the operations. Enrollmentwas conducted between November 1, 1993, and October 31, 1997. This study adheresto the CONSORT (Consolidated Standards of Reporting Trials) statement.^67-69

• Prospectively defined hypotheses tested: (1) surgery for idiopathicFTMH results in a better outcome than observation and (2) vitrectomy plususe of autologous serum results in a better outcome than vitrectomy alone

• Single-center, RCT with masked observers

• Patients with recent-onset (≤9 months) idiopathic FTMH wererandomized to 3 arms for intervention: observation, vitrectomy, and vitrectomyplus autologous serum

• Prospectively defined primary outcome measures: anatomic closureof the hole and best-corrected visual acuity

• Outcomes were assessed at 3, 6, 12, 18, and 24 months, with statisticalanalysis 3, 12, and 24 months after enrollment

• A single, independent, experienced consultant surgeon (Z.G.) performedall treatments and was not involved in randomization or assessment of outcomes

• Assessors of anatomic and visual outcomes (E.E. and A.M.) weremasked to the allocation status of individual patients

• Statistical analysis was performed by an independent statistician(J.G.).

The study population included all patients with idiopathic FTMH whosought care at Moorfields Eye Hospital; they were considered for inclusionaccording to the criteria given in Table1. Patients were examined to confirm the diagnosis of FTMH and todetermine the staging according to the Gass classification.^43,44 Eligiblepatients were counseled, and those willing to participate in the RCT werecontacted within a few days for an "entry" assessment to record the followingbaseline characteristics: date of study entry; patient age and sex; affectedside; nuclear opacity and intraocular pressure; symptom duration; Log MinimumAngle of Resolution (LogMAR), Snellen, and near visual acuities; hole stage(Gass classification); horizontal and vertical hole diameters; and intraretinalcystoid change.

Intervention consisted of vitrectomy alone or vitrectomy plus autologousserum application. Patients randomized to a surgical arm underwent surgerywithin 2 weeks of allocation by an experienced consultant surgeon (Z.G.) atMoorfields Eye Hospital. After a 3-port pars plana vitrectomy, posterior vitreouscortex separation was effected using direct aspiration over the posteriorpole with the vitreous cutter. The vitrectomy was then completed, and epiretinalmembrane peeling was performed using a disposable pick. No attempt was madeto remove the internal limiting membrane. A careful examination of the retinalperiphery was then made, and any breaks were treated by retinopexy at thatstage. This was followed by a fluid-air exchange and "drying" for 10 minutes,with the sclerotomy sites closed with plugs. Any fluid accumulated at theposterior pole was then removed using a 34-gauge extrusion cannula. The procedurewas then completed with a gas exchange using 16% perfluoropropane. Eyes inthe vitrectomy plus serum group underwent the same procedure, with applicationof autologous serum after the 10 minutes of air drying, before the air-gasexchange. The serum was prepared using 10 mL of venous blood drawn from thepatient's antecubital vein, under sterile conditions, and was spun at 3000rpm for 10 minutes. Approximately 0.25 mL of the supernatant was then appliedover the macula through a blunt-ended cannula after the air-drying phase andleft for an additional 10 minutes. The surplus was then removed via a back-flushcannula.

Patients were instructed to posture face down immediately for 2 weeks.Assessment was carried out 1 day, 2 weeks, 6 weeks, and 3 months after primarysurgery, with additional visits as indicated. In eyes in which the FTMH remainedopen after surgery, additional surgery was offered and performed during thefirst 6 weeks after the initial surgery. Subsequent surgery, when necessary,was performed using the pars plana approach, with membrane peeling but notinternal limiting membrane removal. Use of autologous serum during the secondsurgery was according to the original randomization. This was possible, withoutbreaking the code, because the surgeon was not masked to the treatment group.

Observation eyes were first reviewed 3 months after baseline. All patientswere then assessed 6, 12, 18, and 24 months after baseline, and statisticalanalysis was performed at baseline and 3, 12, and 24 months after enrollment.

The 2 outcome measurements, defined prospectively, were anatomic closureof the hole and visual acuity. The hole was defined as closed only if it haddisappeared completely after surgery on clinical examination, in the presenceof a negative Watzke-Allen test result, with confirmation of closure by fluoresceinangiography. Otherwise, the hole was classified as open. Visual function wasassessed using full refraction under standard lighting conditions in 3 ways:Snellen, LogMAR,^70-72 andnear acuities.

The LogMAR acuity was measured using the logarithmic letter chart usedin the Early Treatment Diabetic Retinopathy Study.^70-72 Thismethod has a variety of advantages over the standard Snellen chart, which,although useful for measuring the general level of visual acuity, is difficultto use in measuring changes in acuity.^70-72

The testing conditions, using full refraction, were the same for LogMARand Snellen acuities. Near vision was assessed using a standard near-acuityreading chart, with full refraction, held 30 cm from the patient's eye underthe same conditions as the LogMAR and Snellen acuities. This is a nonlogarithmicscale in which N5 (these letters project the minimum angle of resolution at30 cm for the normal eye) is the smallest print and N48 is the largest. Readingspeed was not measured.

Also recorded were adverse effects of treatment. Intraoperative eventsrecorded were lens touch, macula or retina touch, retinal pigment epitheliumtouch during drying of the FTMH, iatrogenic horseshoe tear, and additionalretinopexy for preexisting lattice degeneration. Postoperative events recordedwere RD or horseshoe tear, a second surgery after unsuccessful primary FTMHsurgery, and progression of nucleosclerosis.

A standard power calculation was used to determine sample sizes. A prioriassumptions, from previously published data, were as follows:

• Observation would result in a 5% improvement

• A treatment effect of 50% was expected in the combined vitrectomygroups compared with the observation group, based on the data of Kelly andWendel and colleagues^1,2 (thebest available in 1993)

• A treatment effect of 40% was expected in the vitrectomy plusserum group compared with the vitrectomy group.

To detect these effects at a significance of P =.05 and a power of 0.9, it was calculated that a sample size of 52 eyes perarm (N = 156) was required. Assuming a loss of up to 20% of patients owingto default and death during the study, an additional 29 eyes were required,resulting in a total of 185 eyes.

• The following stopping rules were prospectively defined at thestudy onset:

• All participants were to be followed for a minimum of 24 monthsafter baseline (enrollment) and were discharged from the study at that pointif stable and not requiring further intervention

• When all participants had reached the minimum 24 months of follow-up,the code would be broken and statistical analysis would be performed

• An interim analysis would be performed only in the event of astrong clinical suggestion that either treatment or observation resulted insubstantial morbidity compared with the other groups; in such an event, thestudy would be terminated on ethical grounds if the interim analysis confirmedthese suggestions.

The unit of randomization was 1 eye. If a patient had bilateral FTMHswhen first seen or developed a hole in the fellow eye during follow-up thatmet the eligibility criteria, the second eye was randomized. The allocationschedule involved computer-generated randomization cards using the "block"method. Serially numbered, sealed opaque envelopes were used to conceal individualallocations.

Assignment occurred at the end of the entry visit, when the envelopewas opened. All envelopes were held and opened by the study coordinator (M.D.),who acted as executor and held the allocation sequence code. The executor(M.D.), allocation schedule generators, assessors (E.E. and A.M.), and surgeon(Z.J.G.) were separate individuals.

One observer (E.E.), using the funduscopic, photographic, and fluoresceinangiography, assessed the hole status. As far as funduscopic assessment wasconcerned, it was not possible to postoperatively mask between surgery andobservation, as examination alone would reveal whether a vitrectomy had beenperformed. However, the assessor was masked as far as vitrectomy vs vitrectomyplus serum, as it was not possible to distinguish between them on clinicalgrounds.

Photographic and fundus fluorescein angiographic assessments were performedby examining digitalized images, which were coded separately so that the assessorwas masked to the identity of the patient. Only if the funduscopic, photographic,and fundus fluorescein angiographic criteria were all met could a hole beclassified as closed. In the event of a discrepancy, a second masked assessorwas used. The assessor of visual acuity (A.M.) was masked to the allocationstatus.

The surgeon (Z.J.G.) was not masked to the allocation status. The dataanalyst was an independent statistician (J.G.) who was not masked to the allocationstatus. Patients were not masked in any way. The allocation sequence codewas held solely by an independent study coordinator (M.D.) throughout thestudy. Only when all patients had reached 24 months' follow-up or in the eventthat an interim analysis was required would the code be broken for analysisby the study statistician.

Data comparisons for all baseline variables were made statisticallyamong the 3 groups and between the combined surgical group and the observationgroup to confirm that randomization had been successful. The χ² testwas used to test for associations between categorical variables and randomizationgroups. A 1-way analysis of variance was used to test the hypothesis thatall means were equal among randomization groups for continuous variables.In addition, baseline variables in completers were compared among the randomizationgroups at each time (3, 12, and 24 months) to identify any biases. To excludebiases attributable to missing data, comparisons of baseline variables amongcompleters, defaulters, and those with missing data entries were made at 3,12, and 24 months within and among randomization groups.

The analysis focused on 2 strategies: (1) comparing the combined surgicalgroup (vitrectomy and vitrectomy plus serum) with the observation group todetermine the effect of surgery per se on anatomic hole closure and visualoutcome and (2) comparing all 3 randomization groups separately to determinewhether use of serum improves the results of surgery.

A repeated-measures model was used for each eye at 3, 12, and 24 months.The effects of treatment per se on FTMH closure and vision were analyzed usingthe observation group as the control and the combined surgical group as treatment.A repeated-measures model was used for each eye at 3, 12, and 24 months usingcorrelated binary response data. The model tested 3 possible treatment effects:(1) treatment as a fixed effect (ie, treatment vs nontreatment is primarilyresponsible for the observed outcome), (2) time in months as a fixed lineareffect (ie, the postbaseline period is responsible for the outcome effectrather than treatment itself), and (3) interaction between treatment and time(treatment × month) (ie, whether the effect of treatment on the outcomechanges with the postbaseline period elapsed).

An eye was treated as a random effect, and the correlation between outcomemeasurements for each eye was modeled using a simple power law.

This approach was used to identify any baseline, intraoperative, orpostoperative factors that were prognostic for a favorable outcome after surgery.All eyes were analyzed using an intention-to-treat approach. For categoricaloutcomes, analysis was performed using the χ² test, and forquantitative outcomes, by comparison of means. All significance testing was2-sided. Continuous variables that were strongly skewed were transformed toa logarithmic scale for purposes of analysis.

The flow of eyes through the study is summarized in Table 2. Of 368 eligible eyes, 185 (174 patients) were randomizedbetween November 1, 1993, and October 31, 1997: 61 to observation, 59 to vitrectomy,and 65 to vitrectomy plus serum. Similar numbers of defaulters occurred inall 3 groups at all assessment times. Three patients died during the study,1 in the observation group (between the 12- and 24-month visits) and 2 inthe vitrectomy plus serum group (1 between 3 and 12 months and 1 between 12and 24 months). The other defaults were accounted for by missed visits ratherthan by missing data. At baseline, 3 eyes had missing data for visual acuity(Snellen, LogMAR, and near). These data represent a high capture rate overall.One study entry violation occurred in an eye that had symptoms of 10 months'duration.

Eleven patients had both eyes randomized. Seven of these patients initiallyhad unilateral holes and subsequently developed an FTMH in the fellow eye,which was then randomized. The other 4 patients had bilateral FTMHs when firstseen and had both eyes randomized on the same date. Of these 11 patients,5 had an observation/vitrectomy plus serum pairing, 2 an observation/vitrectomypairing, 2 a vitrectomy/vitrectomy pairing, 1 a vitrectomy plus serum/vitrectomyplus serum pairing, and 1 a vitrectomy/vitrectomy plus serum pairing.

The baseline characteristics recorded for patients and eyes at entryare summarized in Table 3 and Table 4. No significant differences inbaseline characteristics were apparent among the randomization groups. Therewas a greater proportion of men in the observation group (26.2%) comparedwith the vitrectomy (20.3%) and vitrectomy plus serum (13.8%) groups, butthis difference did not reach statistical significance (P = .10). The vitrectomy group had a greater proportion of stage 3holes (57.6%) than stage 2 holes (30.5%) vs the vitrectomy plus serum group(41.5% and 49.2%, respectively). Observation eyes had similar proportionsof stage 3 (45.9%) and stage 2 (39.3%) lesions. These differences did notreach statistical significance (P = .07). Similaranalyses were conducted for completers at 3, 12, and 24 months to explorewhether missing data affected baseline comparability, and no differences wereevident.

The anatomic results in the randomization groups at 3, 12, and 24 monthsare given in Table 5In the observation group, most FTMHs remained open, although 7 (11.5%)of 61 closed spontaneously during the study: 6 within 3 months and 1 between3 and 12 months. None of the 7 FTMHs subsequently reopened on follow-up.

In the vitrectomy group, the FTMH closure rate was 78.0% (46/59) at3, 12, and 24 months. In the vitrectomy plus serum group, the closure ratewas 87.7% (57/65) at 3 months, 86.2% (56/65) at 12 months, and 83.1% (54/65)at 24 months. Overall, in the combined surgical groups, the closure rate was83.1% (103/124) at 3 months, 82.3% (102/124) at 12 months, and 80.6% (100/124)at months. In 2 eyes, 1 in the vitrectomy group and 1 in the vitrectomy plusserum group, the hole reopened after initially successful surgery, at 20 monthsand 14 months, respectively. Both holes were closed after a second operation.Thus, the overall reopening rate after initial closure with surgery was 2.0%(2/100) at 24 months.

In the repeated-measures model analysis, the only significant factorfor hole closure was treatment effect (χ²₂ = 279.1; P<.001), confirming that the outcomeeffect, that is, anatomic closure, depends only on whether surgery was performed.Time in months did not seem to affect hole closure (χ²₂ = 0.38; P = .83), confirming that the postbaselineperiod per se did not affect anatomic outcome. A strong correlationbetween hole status for each eye between different times (correlation between3 and 12 months was 0.958, between 3 and 24 months was 0.938, and between12 and 24 months was 0.962) showed that individual eyes rarely have a changein anatomic hole status between different postbaseline times. No significantinteraction was apparent between treatment effect and time (treatment ×month) (χ²₄ = 0.26; P =.99), confirming that the effect of surgery did not change with time afterbaseline.

The analyses were repeated to examine the difference between vitrectomyand vitrectomy plus serum application, using vitrectomy as control and vitrectomyplus serum as treatment. No significant differences were found.

Similar repeated-measures model analyses were carried out to evaluatethe effects of treatment per se on FTMH diameter. The analysis included alleyes in each group so that in the 2 surgical groups, both successfully andunsuccessfully treated eyes were included. Table 6 summarizes horizontal and vertical hole diameters in the3 randomization groups at baseline and 3, 12, and 24 months. In the observationgroup, the mean (SD) horizontal diameter increased predominantly during thefirst 12 months, from 386.05 (154.27) µm at baseline to 433.19 (195.42)µm at 3 months and 482.76 (213.37) µm at 12 months, with minimalenlargement thereafter to 494.49 (227.69) µm at 24 months. This representsan overall increase of 25% during the first 12 months and 29% in 24 months.A similar increase was evident in vertical diameter.

There was strong evidence that horizontal and vertical diameters werepositively skewed, so a logarithmic scale was used for the analysis. Treatmentwas associated with a significant reduction in horizontal (F_2,67 =2.55; P = .008) and vertical (F_2,67 =4.50; P = .02) diameters compared with observation.In addition, there was also a significant increase in horizontal (F_1,130 = 7.23; P = .008) and vertical (F_1,130 = 10.02; P = .002) diameters in all 3 groupsbetween 3 and 24 months. This reflected the overall increase in hole sizein the observation group with time and the increase in diameter in failedsurgical eyes in the vitrectomy and vitrectomy plus serum groups. There wasno apparent association between treatment and time (treatment × month)for horizontal or vertical diameter in any group. The analysis was also performedto compare the vitrectomy plus serum group with the vitrectomy group as acontrol, and no significant differences were found.

The LogMAR, Snellen, and near visual acuities showed similar trendsover time between 3 and 24 months (Table7). For each response, eyes in the treatment groups had consistentlybetter visual outcomes than those in the observation group. Snellen and nearvisual acuity were transformed to a logarithmic scale for the purpose of statisticalanalysis. In the analysis, repeated measures for each response at 3, 12, and24 months were analyzed using statistical software (PROC MIXED in SAS; SASInstitute Inc, Cary, NC). The model included treatment as a fixed effect,time (months) as a linear fixed effect, and an interaction between time andtreatment (treatment × month). Each eye was treated as a random effect,and the correlation between measurements for each eye was modeled using asimple power law. In the initial analysis, treatment per se was compared withobservation as the control. In the observation group, there was a mild decreasein LogMAR and Snellen acuities between 3 and 24 months, whereas the vitrectomyand vitrectomy plus serum groups showed an increase in LogMAR and Snellenacuities in the same period. Near vision decreased in observation eyes butremained stable in surgical eyes between 3 and 24 months.

Compared with baseline, observation eyes retained meanLogMAR around 0.7 (Snellen equivalent, 20/100), with median Snellen acuitydropping 1 line, from 20/120 at baseline to 20/200 at 24 months, and mediannear acuity dropping from N10 to N14 during the 24 months. In contrast, surgicaleyes (vitrectomy and vitrectomy plus serum) improved in mean LogMAR acuityby 0.26 (equivalent to 2 Snellen lines), in median Snellen acuity from 20/120to 20/60, and in median near acuity even more dramatically from N10 to N5between baseline and 24 months.

In the combined surgical group (vitrectomy and vitrectomy plus serum),the number of eyes with Snellen acuity of 20/40 or better (the legal requirementfor driving in the United Kingdom) increased from 0 of 124 at baseline to41 (33%) of 124 at 3 months, 42 (34%) of 124 at 12 months, and 55 (44%) of124 at 24 months.

This is in marked contrast to the observation group, in which the numberof eyes with Snellen acuity of 20/40 or better increased from 0 of 61 at baselineto 3 (5%) of 61 at 3 months, 6 (10%) of 61 at 12 months, and 4 (7%) of 61at 24 months. In surgical eyes, LogMAR and Snellen visual acuities improvedbimodally between baseline and 3 months and between 12 and 24 months, withlittle change between 3 and 12 months, whereas practically all near visualacuity improvement occurred between baseline and 3 months.

At 24 months, surgical eyes bettered observation eyes by a mean of 0.26LogMAR (0.42 vs 0.68), by a median of 3 Snellen lines (20/60 vs 20/200), andby a median of 5 near lines (Table 7).The beneficial effects of surgery were more dramatic for near visual acuity.There were no significant differences in visual responses between the vitrectomyand vitrectomy plus serum groups.

In the repeated-measures model analysis, all visual responses showeda significant surgical effect, with surgical eyes showing a better visualoutcome than observation eyes. There was a significant correlation betweenvisual improvement in the surgical eyes compared with the observation eyeswith time and a significant interaction between treatment and time (Table 8). The analysis was repeated tocompare vitrectomy plus serum eyes with vitrectomy eyes, and no differenceswere evident for any measure of visual acuity.

The incidence of intraoperative and postoperative adverse events observedduring the trial is summarized in Table9 and Table 10. Therewere no differences between the vitrectomy and vitrectomy plus serum groups.Intraoperative u-tears, caused iatrogenically during posterior vitreous cortexpeeling or at the entry sites, occurred in 4 (3.2%) of 124 surgical eyes andrequired intraoperative retinopexy. In another 6 eyes, preoperative latticedegeneration was treated with retinopexy at the time of surgery, as a prophylacticmeasure, although no retinal tears had occurred in these areas. Postoperatively,RD occurred in 7 (5.6%) of 124 surgical eyes (2 in the vitrectomy group and5 in the vitrectomy plus serum group). In 5 eyes, the detachment occurredduring the first 6 weeks after initial FTMH surgery. The other 2 eyes developedlate RD: 2 months after uncomplicated phakoemulsification cataract surgery(12 months after initial FTMH surgery) in 1 eye and 16 months after initialsurgery in the other eye. Of the 5 early RDs, 3 were due to inferior u-tears,1 to a superonasal u-tear, and 1 to the FTMH itself. The 2 late RDs were associatedwith new superior u-tears around the entry sites. Of the 7 RDs, 3 were successfullytreated with an RD operation involving scleral buckling, 1 with an operationinvolving internal search and gas tamponade, and 2 with a procedure involvingsilicone oil (owing to moderate or marked proliferative vitreoretinopathy),which was later removed without complication. The other eye developed advancedproliferative vitreoretinopathy and required 3 operations, including an inferior270° retinectomy and silicone oil, for reattachment.

In 4 (3.2%) of 124 surgical eyes, a symptomatic inferotemporal visualfield defect was reported in the early postoperative period. This complicationwas an unexpected finding initially (it had not been reported in the literatureat the time, to our knowledge), and after the first patient to report thiscomplication had been evaluated, all surgical eyes underwent preoperativeand postoperative (at 3 months) field analysis with Goldmann perimetry. Perimetricanalysis revealed that another 15 (12.1%) of 124 eyes had minor, asymptomatic,peripheral, wedge-shaped defects involving less than 10° of field.

A second operation, after initially unsuccessful FTMH surgery, was performedin 18 (14.5%) of 124 surgical eyes, with equal numbers in the 2 surgical groups,and resulted in closure in 14 eyes (78%). Of the other 4 eyes in which a secondoperation was unsuccessful, 3 patients declined further surgery and 1 hada third operation that was unsuccessful.

The most common complication in the postoperative periodwas the development of nucleosclerosis, which occurred with a similar incidencein the vitrectomy and vitrectomy plus serum groups (Table 10). Nucleosclerosis was detected clinically as early as 3months after FTMH surgery, with the mean overall grading increasing from 0.919at baseline to 1.396 at 3 months, 2.180 at 12 months, and 2.400 at 18 months.By 24 months there was a small decline to 2.348 compared with 18 months. Thecumulative cataract surgery rate increased from 1.6% (2/124) of surgical eyesat 6 months to 16.1% (20/124) at 12 months, and 58.1% (72/124) at 24 months.Thus, 72.2% (52/72) of cataract extractions were performed between 12 and24 months. This accounted for the second bimodal rise in visual function notedduring this period (Table 7).

Cataract surgery was performed in each eye according to clinical indications,with all cataract extractions being carried out by an experienced surgeonusing phakoemulsification, with minimal complications. Posterior capsularrupture occurred in 1 (1.4%) of 72 eyes. In 1 eye, RD occurred 2 months afteruncomplicated phakoemulsification, as described earlier. In no eyes did theFTMH reopen after cataract surgery.

Multiple regression analysis showed that general baselinefactors (age, sex, and side) had no effect on anatomic and visual outcomesat 3, 12, and 24 months for the surgery and observation groups. However, FTMHstage and diameter (horizontal and vertical) were inversely correlated toanatomic and visual outcomes (an advanced stage and larger diameter had aworse outcome) but were dependent. Cystic change had no effect onoutcome.

Better Snellen and LogMAR acuities at baseline were associated withbetter anatomic and visual outcomes but depended on hole stage and size. Baselinenear acuity did not affect either outcome. Symptom duration also did not seemto affect either outcome. Thus, it seems that the major prognostic baselinefactor for anatomic and visual outcomes after surgery and observation is holestage. Factors such as FTMH diameter and baseline visual acuity are surrogatefactors for FTMH stage in this analysis.

A similar analysis showed that intraoperative complications (Table 9) did not affect either outcome.Of the 7 eyes with RD, 4 retained good visual acuity (LogMAR, 0.1-0.5; Snellen,20/30-20/60; and near, N5-N12). These were eyes in which reattachment wasachieved with 1 operation. The other 3 had macular involvement and significantproliferative vitreoretinopathy requiring extensive surgery with siliconeoil, and they retained only poor LogMAR (0.7-1.0) and Snellen (20/200 to countingfingers) acuities. Retinal detachment did not affect closure, with all 7 eyesretaining closed holes. The lack of statistical correlation between the occurrenceof an RD and a poor visual outcome is related to the small number of eyeswith this complication.

Eyes in which FTMH closure was achieved after a second operation hadslightly worse LogMAR and Snellen acuities, but similar near acuities, thaneyes in which closure had been achieved after the first operation. The presenceof a postoperative field defect, whether symptomatic or asymptomatic, didnot seem to affect either outcome. Eyes that had undergone cataract surgeryhad similar anatomic outcomes as those that did not, indicating that cataractextraction does not cause substantial reopening of the hole. Similarly, theseeyes had LogMAR, Snellen, and near acuities comparable to those that had notundergone cataract surgery. Thus, the most important complication of treatmentthat resulted in a poor visual outcome was RD with macular involvement andproliferative vitreoretinopathy.

Further analysis, using the strategy described in the "Data Analysis"subsection, was performed on the study cohort with eyes divided into 2 subgroups:Gass stage 2 lesions and Gass stage 3 and 4 lesions.

For stage 2 lesions, 5 (21%) of 24 eyes had spontaneous closure in theobservation group, whereas 47 (96%) of 49 treated lesions were closed at 24months' follow-up, representing a significant treatment effect (χ²₂ = 146.6; P<.001) (Table 11). There was no treatment effectrelated to time or an effect for treatment × month. In treated eyes,similar closure rates were evident for the vitrectomy group (18 [100%] of18 eyes) and the vitrectomy plus serum group (29 [94%] of 31 eyes).

For stage 3 and 4 lesions, 2 (6%) of 36 eyes had spontaneous closurein the observation group (both eyes were stage 3 at baseline) at 24 monthscompared with 53 (77%) of 69 treated eyes, again representing a significanttreatment effect (χ²₂ = 153.3; P<.001), with no effect related to time or an effect for treatment× month (Table 12). Closurerates were again similar for vitrectomy eyes (28 [74%] of 38 eyes) and vitrectomyplus serum eyes (25 [81%] of 31 eyes).

For stage 2 lesions, treated eyes had a mean LogMAR acuity 2.5 lines(P = .03) and a mean Snellen acuity 2.0 lines (P = .03) greater than observation eyes at 24 months, withno significant difference between vitrectomy and vitrectomy plus serum eyes.For stage 3 and 4 lesions, treated eyes had a mean LogMAR acuity 3.0 lines(P = .02) and a mean Snellen acuity 2.0 lines (P = .004) greater than observation eyes at 24-month follow-up,with no significant difference between vitrectomy and vitrectomy plus serumeyes.

The RCT described herein evaluated the benefit of surgery for FTMH comparedwith the natural history and determined whether use of intraoperative autologousserum improves the outcome of surgery. The results show conclusively thatsurgery is associated with a better anatomic closure rate (81% vs 11%) anda better visual outcome compared with observation, and use of autologous serumdoes not seem to improve the outcome of surgery. The data also corroboratedfindings from previous studies^1-47 showingthat the outcome for stage 2 lesions is more favorable than that for stage3 and 4 lesions.

To date, the only randomized trial comparing surgery with observationfor FTMH is that described by the VMHSG, which did not show any significantbenefit of surgery.^{10,11,13,60,65} Furthermore,this study had a relatively high rate of sight-threatening complications aftersurgery. As far as the role of autologous serum is concerned, although anuncontrolled retrospective analysis conducted by the VMHSG showed that whereautologous serum was used it had little or no significant effect on outcome,^{10,11,13,60,65} itsrole has never been subjected to RCT evaluation.

Our study had a masked-observer design rather than a double-masked designbecause sham surgery to the observation group would not have been possibleor desirable ethically. A variety of design characteristics, which strengthenedthe study, deserve emphasis. First, the study was conducted in a single tertiaryreferral center, with all initial FTMH operations and subsequent operationsbeing performed by an independent, experienced surgeon. Second, FTMHs of allstages (including stage 4) were included to evaluate the benefit of surgeryfor the full spectrum of lesions and to avoid selection biases, which havebeen a feature of many retrospective or nonrandomized series, where only earlyor smaller lesions are included. Third, strict clinical and angiographic criteriawere used to determine whether an FTMH had closed, and full refraction wasused to determine visual acuity, during follow-up. This is again in contrastto some uncontrolled or retrospective series, in which only subjective methodswere used to assess anatomic closure (including retrospective medical chartreview) and uncorrected acuity measurements were used to assess visual function.Finally, the study design allowed patients to receive treatment and postoperativemanagement, including management of complications such as cataracts, as clinicallyindicated. In particular, no restrictions were placed in terms of carryingout cataract surgery in surgical eyes. This is in contrast to the VMHSG trial,which precluded all surgical eyes from having cataract surgery in the postoperativeperiod, thus leading to cataracts "masking" the visual benefit achieved byFTMH surgery.^{10,11,13,60,65}

The overall closure rate of 81% is similar to the 76% reported by theVMHSG^{10,11,13,60,65} butlower than the rates reported in other studies.^3-38 Thisis probably accounted for by 2 factors. First, the latter studies probablyhad a selection bias, as most included only stage 2 and 3 lesions and notstage 4 lesions. Second, these studies used more subjective and less stringentcriteria to define hole closure. Most studies defined closure as "flattening"of the hole rim rather than as complete resolution of the full-thickness retinaldefect. Thus, lesions in which the subretinal fluid cuff resolved with flatteningof the hole rims and in which the retinal defect was still present would beclassified as closed. In our study, such a lesion would be classified as open.This is supported by the fact that some studies^73-77 havereported significantly higher incidences of the hole reopening after successfulsurgery. In our study, only 2 (1.6%) of 124 lesions reopened after successfulsurgery, both occurring late (at 14 and 24 months). There were no cases ofreopened holes during the first 12 months. This is somewhat lower than therate of 4% to 5% in other series that included eyes with longer follow-up.In addition, it may have been that in other series, some holes classifiedafter surgery as closed were in fact open, thus increasing the initial successrates.

In the observation group, spontaneous closure was associated with thedevelopment of a posterior vitreous detachment in all 7 eyes. Most occurredin stage 2 lesions (5 in stage 2 and 2 in stage 3), suggesting that spontaneousclosure occurs in lesions in which a posterior vitreous detachment (with reliefof vitreous traction) occurs early in the natural history, while the defectis small and amenable to glial repair. Hole diameter analysis confirmed thatalthough the FTMH remains open in 88.5% of observation eyes, only a modestincrease in size occurs after baseline, in accordance with previous studies,^{13,39-46,78-80} ofapproximately 25% during the first 12 months and 29% in 24 months.

In the surgical groups, mean horizontal and vertical diameters decreasedby 70% to 75% overall after surgery. The mean residual diameter in the combinedsurgical cohort is attributable to eyes with persistent FTMH after unsuccessfulsurgery. The lack of significant differences between the vitrectomy and vitrectomyplus serum groups confirms that serum has little effect on the anatomic outcomeafter surgery.

The clear difference in anatomic outcome between observation and surgicaleyes was also translated into a significant difference in visual function.Whereas the observation group had relatively stable mean LogMAR (decreaseof 0.02) and median Snellen (decrease of 1 line) acuities and a 2-line declinein median near acuity in 24 months, surgical eyes demonstrated a substantialimprovement. The latter showed a biphasic improvement in LogMAR (0.26, equivalentto 2 Snellen lines) and Snellen (2 lines) acuities. The first phase occurredimmediately after FTMH surgery, during the first 3 months, owing to the effectof FTMH closure after surgery resulting in immediate improvement because ofrealignment and functional recovery of foveal photoreceptors. The period between3 and 12 months was characterized by relatively stable vision in the surgicalgroup owing to further visual acuity improvement, associated with gradualand slow recovery of photoreceptor function at the macula being masked bynuclear cataract progression. The second phase of visual recovery, between12 and 24 months, owing to "unmasking" by phacoemulsifications, 72% of whichwere carried out during this period, accounted for 50% of the overall improvementin surgical eyes, confirming the importance of cataract surgery in the postoperativeperiod. This also accounts for the poor visual improvement in surgical eyescompared with observation eyes reported by the VMHSG,^{10,11,13,60,65} wheresurgical eyes were precluded from cataract extraction after macular hole surgery.

The marked functional benefit of surgery was also confirmed by the numberof eyes achieving Snellen acuity of 20/40 (the legal minimum requirement fordriving in the United Kingdom) or better, with 45% of surgical eyes but only7% of observation eyes achieving this level.

Near visual acuity was improved even more dramatically than distanceacuity after surgery, with surgical eyes achieving a median of N5 at 24 monthsand observation eyes achieving a median of N14, a median difference of 6 lines.Near acuity showed only a uniphasic pattern of improvement in the surgicalgroups, with essentially all improvement occurring immediately after FTMHsurgery. Nuclear cataract and eventual extraction had little impact on nearacuity, which is not unexpected, as nucleosclerosis is known to have a morepronounced effect on distance acuity.

Advanced FTMH stage was the most significant predictor of adverse anatomicand visual outcomes for observation and surgery at 24 months. Although a largeFTMH diameter was also predictive of an adverse outcome, regression analysissuggested it to be not independent of hole stage, or a surrogate marker. Betterbaseline LogMAR and Snellen acuities were associated with a better outcomefor observation and surgery, probably reflecting more preserved foveal photoreceptorpopulations at baseline. These findings broadly confirm previous data fromcontrolled and uncontrolled natural history^{13,39-46,78-80} andsurgery^13-38 studies.

The lack of effect of symptom duration on outcome can be accounted forin 2 ways. First, the study only included recent-onset lesions (≤9 months'duration), and in such lesions, duration of 0 to 9 months may not be an importantfactor compared with duration in lesions of more than 1 year, in which wehave shown in a previous study⁸¹ that thisvariable is associated with a worse outcome. Second, symptom duration, inrecent-onset lesions, is a relatively inaccurate and subjective preoperativevariable, particularly as most are unilateral, with patients perhaps remainingasymptomatic until the lesion is discovered incidentally.

In general, the incidence of intraoperative and postoperative sight-threateningcomplications was considerably lower than that reported in the VMHSG.^{10,11,13,60,65} TheRD rate of 5.6% was considerably lower than the 14% reported by the VMHSG^{10,11,13,60,65} buthigher than the rates in other uncontrolled data. Most (5 of 7) RDs occurredduring the first 6 weeks after FTMH surgery. In 3 eyes, detachment was relatedto inferior breaks, which could have been missed during surgery. Alternatively,they may have arisen in the early postoperative period as a result of superiorvectoral tractional forces, at the inferior vitreous bases, generated by surfacetension at the inferior surface of the bubble as it contracts over time. Althoughinferior breaks and RDs have been observed with a higher frequency after surgerywith long-acting gas tamponade, such as macular hole surgery, the underlyingpathogenesis remains unclear. The surgical technique used in our study includeda thorough examination of the retinal periphery using scleral indentationbefore the fluid-gas exchange. This is an important step in ensuring thatiatrogenic tears are treated.

The other adverse effect observed in our study was postoperative visualfield defects. At the onset of the study, this had not been observed by ourgroup, or indeed by any other group, and had not been reported in the literature,to our knowledge. After the first patients reported field defects, which wereconfirmed by field analysis, Goldmann perimetry was performed before surgery(baseline) and at 3 months in all surgical eyes. In due course, it becameevident that visual field defects were relatively rare and did not constitutesufficient reason to terminate the study. Overall, only 4 (3.2%) of 124 surgicaleyes developed symptomatic field defects, although another 15 (12.1%) of 124developed very minor, peripheral wedge defects that were asymptomatic. Thesefindings led to a separate study⁴⁹ to evaluatethe mechanisms responsible for the defects. These data and a variety of otherstudies^49-58 suggestthat the most likely cause is mechanical damage to inner retinal nerve fiberscaused by either vitreopapillary traction during posterior vitreous cortexseparation or the effects of air drying during surgery.

In summary, this article described a masked-observer RCT to evaluatethe benefit of FTMH surgery compared with the natural history and determinedwhether application of intraoperative adjunctive autologous serum improvesthe results of surgery. The data showed successful randomization, with a highcapture rate during follow-up. There was significant anatomic and visual improvementin surgical eyes compared with observation eyes. Treatment was associatedwith a low incidence of sight-threatening complications, confirming that FTMHsurgery is a safe and effective procedure that should be considered and offeredwhere appropriate. Use of autologous serum did not seem to have a significantimpact on the results, and its role in FTMH surgery remains unproved.

Clinical Center: Vitreoretinal Service, MoorfieldsEye Hospital. Lead Clinicians: Eric Ezra, MD, FRCS,FRCOphth, and Zdenek J. Gregor, FRCS, FRCOpth. Study Clinicians: William G. Aylward, MD, FRCS, FRCOphth; David G. Charteris, MD, FRCS,FRCOphth; Steven D. Schwartz, MD; and John A. Wells III, MD. Study Coordinator: Michael Davis. Optometrist: AndrewMilliken, MSc, FCOptom. Statistician: John Grego,PhD, Department of Statistics, University of South Carolina, Columbia. Medical photography: Medical Illustration Department, MoorfieldsEye Hospital.

Corresponding author and reprints: Eric Ezra, MD, FRCS, FRCOphth,Vitreoretinal Service, Moorfields Eye Hospital, City Road, London EC1V 2PD,England (e-mail: ericezra@hotmail.com).

Submitted for publication December 28, 2001; final revision receivedApril 23, 2003; accepted May 30, 2003.

This study was supported by grant GREZ0195 from the Guide Dogs for theBlind Association, London, England, and grant EZRE0311 from the Stringer Bequestto the Special Trustees of Moorfields Eye Hospital.