Methods

A retrospective case note review of all phakic patients undergoing surgery for IFTMH under one consultant-led surgical team from March 2004 until October 2007 was performed. Only the first operation on the first operated eye of each patient was included in the analysis. Exclusion criteria included macular holes in association with trauma, any previous intraocular surgery and a history of other vision-limiting eye conditions such as age-related macular degeneration, glaucoma and amblyopia.

Data collected included: age, sex, laterality, duration of visual symptoms, preoperative VA, preoperative maximum linear dimension (MLD), grade of operating surgeon, intra- and postoperative complications, postoperative anatomical closure and VA at 3 months (table 1).

Preoperative MLD of the IFTMH was measured using optical coherence tomography (OCT) (Stratus OCT, Zeiss (Carl Zeiss Meditec, Inc., Dublin, California, USA) and Topcon 3D OCT-1000 (Topcon Medical Systems, Paramus, New Jersey, USA)) and was defined as the greatest linear distance along the smallest hole aperture (fig 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Maximum linear dimension measured as the greatest linear distance along the smallest hole aperture (shown as a–b).

All cases underwent phacoemulsification and intraocular lens implantation followed by 20-gauge three-port pars plana vitrectomy with induction of a posterior vitreous detachment when not already present. Indocyanine green (ICG)-assisted internal limiting membrane peeling was performed in all cases, using 0.5 ml of 0.5 mg/ml (0.05%) ICG and minimal retinal exposure time followed by fluid to air to 16% C₃F₈ gas exchange. All patients were instructed to maintain an upright postoperative posture during the day and to strictly avoid sleeping in a supine position. Follow-up was performed at day 1, 2 weeks and 3 months. Anatomical closure was determined at 3 months after surgery using indirect slit lamp biomicroscopy.

Statistical analysis

Snellen acuities were converted to logMAR units for analysis. LogMAR values of 2.0 and 3.0 were used for vision of counting finger and hand movement, respectively. Distribution analysis revealed postoperative vision and MLD to be non-normally distributed. In order to make prediction probabilities easier for application in a clinical setting a binary outcome variable “visual success” was defined, consistent with previous literature,10 as postoperative VA of <0.3 logMAR (better than 6/12) at 3 months. MLD was categorised into quartiles (<350, ⩾350 to <400, ⩾400 to <500, ⩾500 μm), age into decades, and preoperative VA as ⩾0.3 to <0.6, ⩾0.6 to <0.9 and ⩾0.9 logMAR. Logistic regression analysis was used to determine which explanatory variables were to be included in the predictor for visual success. The multivariable model was built using a forward manual stepwise method by adding the most significant variable first. The significance level for removal from the model was p = 0.1. Odds ratios (ORs) and 95% confidence intervals (CIs) were derived from this model. All p values were two-sided. Prediction models using logistic regression were evaluated by establishing a cut-off point; predicted probabilities below the cut-off point were treated as predictors of no event and predictions at or above the cut-off point were considered to be predictors of the event.11 A cut-off point of 0.50 was usually chosen. Using the 0.5 prediction cut-off points, we calculated the proportion of patients correctly classified by the model. The proportion of correct predictions is the proportion of patients in the total sample whose observed status (visual success) agreed with their predicted vital status. The predictive ability of the final model was assessed in three ways. First, receiver-operating characteristic (ROC) curves were calculated; second, the McNemar test was used to compare the percentages of correct predictions (see below); and third, goodness of fit was assessed using the Hosmer–Lemeshow test.12

Results

Of the 132 eyes (of 132 subjects), 96 were those of women (72.7%). Demographic distribution is shown on table 1. The overall anatomical closure rate was 86% (113/132). Median VA improvement was 0.39 logMAR (interquartile range (IQR) 0.16–0.6), with 33% (44/132) achieving visual success of better than 20/40.

Univariable analysis suggested that anatomical closure (p = 0.022), decreasing hole size (p<0.001), better preoperative VA (p<0.001) and younger age (p = 0.002) were significantly associated with visual success (table 2). Univariable and multivariable ORs for these explanatory variables are shown in table 2. A highly significant association of MLD and anatomical outcome was found (<350 μm, 96.6% anatomical closure; ⩾350 to <400 μm, 89.7%; ⩾400 to <500 μm, 87.5%; ⩾500 μm, 67.6% anatomical closure, p<0.001). This confounding factor in the adjusted analysis was no longer statistically significant (p = 0.1), and anatomical closure was therefore not included in the predictor equation.

Using the results of the multivariable analysis, the probabilities of visual success are shown on table 3.

The multivariable model correctly classified outcome in 79.7% of cases. The ROC curve in fig 2 shows an area under the curve of 82% (ie good accuracy).13 Finally, we evaluated the goodness of fit of the models by using the Hosmer–Lemeshow test, which yielded a p value of 0.623, indicating that the model fits well.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Receiver-operating characteristic (ROC) curve analysis of multivariable model.

Anatomical closure was not achieved in 19 eyes. In 5/19 there was a ⩾2 lines of logMAR worsening of the visual acuity. Major complications included one eye with unsuccessful closure developing peri-papillary choroidal neovascularisation and associated pigment epithelial detachment, and another developing retinal detachment requiring repeat surgery with silicone oil.

Discussion

One-third of patients in this series of standardised surgeries for IFTMH attained visual success of better than 20/40 Snellen. There are many different thresholds for visual success defined for macular hole surgery, including mean number of Snellen lines gained1415 and proportions of patients with a VA of either 20/50 or better16 or 6/12 or better.1718 The uneven scaling of Snellen charts renders the former measure inappropriate. The absence of a 20/50 (6/15) line on most UK Snellen charts precludes the latter. We selected an acuity of better than 6/12 to avoid the rounding up errors of 6/12 “part” being accorded a clinic acuity of 20/40. The overall anatomical closure rate in this study was 86% with one-third of patients achieving an acuity of better than 20/40 Snellen. These results from a training institution are broadly comparable to those reported by others for combined cataract and macular hole surgery: in such series rates of anatomical success range from 92% to 100%,192021 with 20% achieving >6/12,22 improvement of 2–5.7 lines from baseline.2324

We are not aware of any other studies that have similarly modelled visual success in order to produce a simple probability chart. Visual outcome has been previously shown to be correlated with macular hole diameter,25 but that study did not allow prediction of visual outcome.

Our mathematical model of visual success (better than 20/40; 6/12 Snellen) is easy to apply clinically as it is based on three standard clinical measurements: age, VA and OCT hole maximum minimum linear dimension (table 3). The predicted range of visual success varies from 2% in patients >80 years of age with large holes and a presenting VA of 6/60 or worse to 93% in subjects <60 years of age with a presenting VA of 20/80 or better and a hole of <350 μm MLD. The model appears to be an accurate predictor of outcome, with an area under the ROC curve of 82% and 80%, respectively, of observed cases being correctly classified.

There are a number of caveats that should be considered before widely generalising our results. The surgeries conducted here were of a standardised format (ie combined phaco-vitrectomy, ICG-assisted internal limiting membrane peeling and no prone posturing); thus the model may not be applicable in other circumstances. It has been suggested that ICG toxicity may adversely affect visual outcome; we are not, however, aware of any evidence of such toxicity when a concentration of 0.5 mg/ml is used. It has recently been shown that 5 days of 8 h/day prone-posturing results in an improved rate of anatomical closure in patients with holes of ⩾400 μm MLD.26 None of the patients in this study underwent postoperative prone posturing. In holes of ⩽400 μm our anatomical closure rate was 93% with visual success rates of 42% and for holes >400 μm our results were 77% and 20%, respectively. Accordingly the proportion of patients achieving visual success might have been higher had such prone posturing been advocated for patients with holes >400 μm. Continued visual improvement has been observed in patients with successful macular hole repair beyond 1 year postoperatively;27 our relatively short follow-up time (3 months) must be noted. We have also retrospectively analysed clinic Snellen acuities measured by several technicians rather than best corrected logMAR results. The statistical validity of the model suggests robustness; however, a prospective study to formal validate and test the accuracy of the predicted probabilities is required. Last, biometric factors28 have been associated with FTMH and inclusion of these data may improve the model.

Conclusion

The data presented in this paper may facilitate preoperative patient counselling on the likely visual outcome of macular hole surgery. This novel predictor is easy to use and can provide the patient with guiding information when debating whether to undertake FTMH surgery. However, the limitations of the predictor must also be taken into account when providing guidance.