Outcome and Predictors for 2-Year Visual Acuity in Eyes with Diabetic Macular Edema Treated with Ranibizumab

Introduction

Diabetic retinopathy (DR) is a common complication of diabetes, and its prevalence ranges from 10% to 50% in type 1 and is close to 40% in type 2 diabetes. DR is one of the causes of acquired vision loss among working-age populations. The reasons of vision-threatening DR are severe nonproliferative DR, proliferative DR, and diabetic macular edema (DME).^1,2 DME results from an increased vascular permeability of the perifoveal capillaries and a breakdown of the blood–retinal barrier. Vascular endothelial growth factor (VEGF) plays a critical role in DME development. ³ The clinical trials^4–6 demonstrated the efficacy of different types of anti-VEGF, including bevacizumab, ranibizumab, and aflibercept in the treatment of DME. Ranibizumab is the first VEGF inhibitor to receive US Food and Drug Administration approval for DME in 2012.

In the treatment of DME, gains in visual acuity (VA) obtained by different anti-VEGF drugs and treatment regimens are variable.^7,8 Determining the risk factors that predict the outcomes can help the ophthalmologist to decide which patients should be treated with intravitreal anti-VEGF drugs and to inform the patient about the prognosis. The current study was conducted to determine the role of baseline demographics and ocular and optic coherence tomography (OCT) characteristics in predicting the visual outcome in patients with DME treated with ranibizumab.

Methods

This retrospective study was conducted at the Ondokuz Mayis University Hospital, Samsun, Turkey. The study was approved by the local research ethics committee and was carried out according to the principles outlined in the Declaration of Helsinki.

A review was carried out of the charts of 62 patients (97 eyes) who received pro re nata (PRN) intravitreal ranibizumab (IR; Lucentis^®; Novartis Pharmaceuticals AG, Basel, Switzerland, and Genentech, Inc., South San Francisco, CA) 0.5 mg treatment for macular edema (ME) due to proliferative DR. The patients did not undergo loading phase of IR treatment. Intravitreal injections were administered as needed at the earliest fourth week with a monthly follow-up. If the macula was dry on OCT, no reinjection was given. Reinjection was carried out if the central macular thickness (CMT) was ≥300 μm by OCT and if DME was judged to be the cause of VA loss.

The criteria for inclusion to the study were as follows: age of ≥40 years; type 2 diabetes mellitus; center-involving DME; CMT of ≥300 μm by OCT; absence of macular ischemia on fluorescein angiography (FA); best-corrected visual acuity (BCVA) of 1.0–0.05 log of the minimum angle of resolution (logMAR); and follow-up period of at least 2 years after the first IR treatment. Patients were excluded if they had any of the following: other ocular diseases causing ME; macular ischemia observed on FA; vitrectomized eye; follow-up period of less than 2 years after the first IR treatment; or missing data.

The following data were recorded: baseline demographics; previous treatments; ocular examination, FA, and spectral-domain OCT (SD-OCT) findings of the patients; the numbers of intravitreal injections; additional laser treatments; need for cataract surgery; and development of vitreous hemorrhage (VH). Baseline and follow-up visits included the following: BCVA measurements (Snellen); slit-lamp biomicroscopy; dilated fundoscopy; intraocular pressure measurements (Goldman applanation tonometer); and CMT measurements by SD-OCT (Zeiss Stratus 3; Carl Zeiss Meditec, Inc., Dublin, CA). The SD-OCT images which were taken at baseline and follow-up visits were analyzed to determine the following: morphological characterization of DME (cystoid or diffuse, serous retinal detachment [SRD], hard exudates [HE]); vitreomacular interface abnormalities (VMIA: epiretinal membrane [ERM], vitreomacular adhesion [VMA], vitreomacular traction [VMT], and posterior vitreous detachment [PVD]); and disruption of ellipsoid zone (EZ). Baseline FA of the patients was evaluated to determine presence of leaking microaneurysms (MA) in the perifoveal capillary network up to a distance of ∼500 μm from the fovea in early phase.

For the statistical analysis, Snellen VA was converted to the logMAR. Changes from baseline in the mean BCVA and CMT at the 1st, 2nd, 3rd, 6th, 12th, and 24th months were analyzed. The eyes were divided into 3 groups according to the level of VA at the second year as decreased (Group 1), stable (Group 2), and increased (Group 3) VA. At least 1 line loss in VA was defined as decreased, no change in VA was defined as stable, and at least 1 line increase in VA was defined as increased. The baseline demographics and ocular and OCT findings were analyzed to determine the association with 2-year VA.

Results

Thirty-five (56.5%) males and 27 (43.5%) females with a mean age of 62.6 ± 8.3 (47–86) years were included in the study. The data for 97 eyes with DME of 62 patients were analyzed. The mean durations of diabetes and DME were 14.6 ± 3.6 (7–21) years and 5.1 ± 3.5 (1–12) months, respectively. Thirty-one (32%) eyes had received previous macular focal/grid laser therapy. Fifteen (15.5%) eyes were pseudophakic, and 55 (56.7%) eyes had lens opacities at baseline. Fifty-seven (58.8%) eyes had received previous panretinal photocoagulation (PRP), and remaining 40 (41.9%) eyes received PRP during the IR treatment.

BCVA increased from 0.54 ± 0.2 (0.05–1.0) to 0.41 ± 0.3 (0.0–1.0) logMAR within 2 years (P < 0.001), and CMT decreased from 448.6 ± 113.6 (306–829) to 340.2 ± 96.8 (185–775) μm within 2 years (P < 0.001). The improvement in mean BCVA and CMT from baseline at the 1st, 2nd, 3rd, 6th, 12th, and 24th months was statistically significant (P < 0.001) (Table 1). There was no significant further improvement in the mean BCVA and CMT after the first month (P > 0.003). The proportion of eyes gaining at least 3 lines of BCVA was 19.6%. The proportion of eyes gaining less than 3 lines of BCVA was 38.1%.

ME was cystoid in 58 (59.8%) and diffuse in 39 (40.2%) eyes. At baseline, the proportions of leaking MA, SRD, HE, VMA, ERM, and disrupted EZ were 69.1%, 32%, 45.4%, 30.9%, 13.4%, and 17.5%, respectively. During the follow-up, ERM, VMT, PVD, and VH developed in 5 (5.2%), 7 (7.2%), 9 (9.3%), and 3 (3.1%) eyes, respectively. A decline in the HE was seen in 7.2% of the eyes.

The average number of injections in the 2-year period was 6.6 ± 2.2 (2–11) (3.7 ± 1.3 for the first year and 2.9 ± 1.2 for the second year, P = 0.000). Nineteen (19.6%) eyes received additional macular laser therapy. Cataract surgery was required in 4 (4.9%) of 82 phakic eyes. No other ocular or systemic side effects were detected in any of the patients. In eyes with or without leaking MA, the mean numbers of IR injections (6.8 ± 2.4 vs. 6.3 ± 1.8, P = 0.275) and the rates of macular laser administrations (34.4% vs. 26.7%, P = 0.608) were not statistically different.

At the end of the 2-year period, VA decreased in 13 (13.4%) eyes (Group 1), was stable in 29 (29.9%) eyes (Group 2), and increased in 55 (56.7%) eyes (Group 3). In Group 1, cystoid ME (P = 0.023) and development of VMT (P = 0.018) were statistically higher. In Group 2, female gender (P = 0.036) and HE (P = 0.017) were statistically higher. In Group 3, absence of HE (P = 0.026) and absence of disrupted EZ (P = 0.030) were statistically higher. There was no association between the visual outcome and the other characteristics, including age, durations of diabetes, and ME (P > 0.005). There was no association between the visual outcome and the other findings, including baseline BCVA, baseline CMT, lens status, presence of leaking MA, presence of SRD, presence of the other VMIA, development of VH, mean number of injections, PRP administration time, and need for additional macular laser therapy and cataract surgery (P > 0.005). In the groupwise analysis, the relationships between the characteristics and findings of the patients and visual outcome are presented in Supplementary Tables S1 and S2.

Multivariate linear regression analysis demonstrated that the predictors of final VA were age (P = 0.012), gender (P = 0.018), baseline BCVA (P < 0.001), presence of leaking MA (P = 0.018), development of VMT (P = 0.001), development of PVD (P = 0.040), and disruption of EZ (P = 0.007). For the poor final VA, in addition to older age and female sex, the eyes with lower baseline VA, VMT development, and EZ disruption were more at risk than eyes without these findings. Development of PVD and leaking MA are predictors for the good final VA. The importance of HE and type of ME could not be confirmed in the multivariate model (Table 2).

Discussion

In patients with DME, BCVA and CMT improved from the baseline values 1 month after the treatment with IR, and these improvements were maintained with low number of injections throughout the 24 months. In the RIDE and RISE Studies, ⁵ the proportions of eyes gaining ≥15 Early Treatment Diabetic Retinopathy Study (ETDRS) letters in BCVA score from baseline were 39.2% and 45.7% by monthly IR treatment at 24 months in DME. In the RESTORE extension study (month 12–36), ⁹ the patients with DME received individualized IR treatment and concomitant laser treatment according to the ETDRS guidelines. In the 3-year results of the study, the proportions of patients gaining ≥15 BCVA letters were 27.7% and 30.1% with mean of 14.2 and 13.5 IR applications in the prior IR group and prior IR + laser group. In the Diabetic Retinopathy Clinical Research (DRCR) Network Protocol-I Study, ¹⁰ the rates of eyes gaining ≥15 letters in BCVA score from baseline at 24 months were 26% and 29% with median of 11 and 13 IR applications in the ranibizumab + prompt laser and ranibizumab + deferred laser groups. In the 5-year results of the Protocol-I Study, ¹¹ it was seen that the mean VA gain at 1 year was maintained through 5 years accompanying a progressively diminishing number of treatments. The median number of injections from years 3 to 5 was 0 and 1 in the ranibizumab + prompt laser and ranibizumab + deferred laser groups. In the recent study, the proportion of eyes gaining ≥3 lines in BCVA was 19.6% with a mean of 6.6 ± 2.2 IR injections over 2 years. This ratio was lower than in the aforementioned studies; however, the numbers of injections especially in the first year were also lower. These results suggest that at the first year of intravitreal anti-VEGF therapy for DME, more intensive intravitreal treatment or more strict reinjection criteria can be better for more gain in VA. Many reports in the literature have considered cohorts, including patients with ME due to nonproliferative DR. In the prospective randomized studies, 1 out of 3 or 4 patients had proliferative DR,^5,10,11 whereas in this study, the consideration of only patients with proliferative DR may have affected the results. The therapeutic effect was poor compared with other studies. Although the clinical significance of retinopathy severity is uncertain in this issue, deterioration of retinopathy is associated with poor visual outcomes. In addition, in the eyes with proliferative DR, the PRP treatment destroys the retina and may result in vision and visual field reduction. ¹² In this study, no association was found between visual outcome and PRP application before or during IR treatment. However, the effect of PRP on the visual outcome could not be evaluated because there were no patients who did not receive PRP. In addition, in the recent study, the significant reduction in the mean CMT obtained by PRN IR treatment at month 1 sustained through 24 months.

When comparing the different treatment agents and treatment regimens, the results are reliable if the patients have similar baseline characteristics that might affect the outcomes. It is therefore important to identify the baseline clinical features that may affect the outcome of treatment with an agent. In this study, the baseline clinical features affecting visual outcome of the PRN IR treatment in patients with DME were investigated. Age, gender, baseline BCVA, and 3 OCT findings were identified as associated factors with 2-year visual outcome: VMT, PVD, and disruption of EZ.

When the literature is reviewed, it is generally seen that younger age and male sex are associated with better visual outcome in DME patients treated with anti-VEGF therapy. ¹³ In the post hoc analysis of READ-2 study, ¹⁴ it was found that there was a higher percentage of females in the poor outcome group of DME patients treated with IR. However, it has also been reported that age and gender are not predictors of long-term VA.^15,16 In the current study, we found that older age and female sex are predictors for poor 2-year VA. However, it is unknown why female patients had poor visual outcomes.

DRCR Network Protocol-T Study ⁷ demonstrated that baseline VA had an interaction on treatment effect in the DME patients treated with intravitreal anti-VEGF. A post hoc analysis of Protocol-I Study ¹⁷ participants who were treated with IR for DME revealed that the eyes with lower baseline VA scores were more likely to have greater gain in VA at 1-year period. After adjusting for baseline VA, they reported that the younger age, less severe DR, and absence of surface wrinkling retinopathy were associated with a larger VA treatment benefit. We found a relationship between baseline VA and visual outcome similar to literature. However, in the current study, we found that presence of ERM and VMA had no effect on visual prognosis. Wong et al. ¹⁸ demonstrated that ERM and more severe retinopathy at baseline were predictive of less visual improvement in patients with DME treated with IR. Patients with VMA had better final VA than those without VMA. Nevertheless, in patients with VMA, baseline VA was also better than in those without VMA. Chang et al. ¹⁹ reported that the incidence of VMIA formation in DME eyes treated with anti-VEGF (Bevacizumab and Ranibizumab) was 6.43% and that there was no difference between eyes with and without VMIA formation in improvement of BCVA and CMT. Sadiq et al. ²⁰ assessed the role of VMA in 6-month visual and anatomic outcomes in DME patients treated with IR in the READ-3 study participants. They showed that patients with VMA had a greater potential for improvement in only visual outcomes. Ciulla et al. ²¹ proposed that VMA-associated traction can cause localized inflammation and/or that VMA can act as a diffusion barrier for VEGF and oxygen. However, they reported that VMA or VMT was not associated with VA at baseline and follow-up in neovascular age-related macular degeneration patients treated with anti-VEGF. The relationship between the VMIA and the response to anti-VEGF treatment is still unclear and is variable in the studies.

Some studies investigated the association between PVD and DME. Sivaprasad et al. ²² reported that intravitreal triamcinolone acetonide (IVTA) perhaps plays a role in visual improvement by promoting vitreous detachment in DME. In their study, patients with perifoveal vitreous detachment with traction had the worst visual prognosis. They suggested that IVTA may induce changes in vitreomacular relationship in addition to the pharmacological effects on ME and that injection procedure may trigger mechanical effects or induce a local inflammatory reaction. Gaucher et al. ²³ suggested that vitreous might play a role in the pathogenesis of the DME by applying traction on the macula during perifoveolar PVD. Yamaguchi et al. ²⁴ reported the cases of DME resolution after spontaneous PVD and that this resolution, which is caused by vitreofoveal separation, supports the contribution of vitreous traction to CME. In the current study, development of PVD was found as a predictor for good visual outcome, and development of VMT was found as a predictor for poor visual outcome. Vitreous traction during incomplete PVD, which can be triggered or accelerated by intravitreal injections, may be an aggravating factor for DME and may cause poor response to anti-VEGF treatments.

A review ¹³ about predicting outcome to anti-VEGF treatments in DME revealed that patients with subretinal fluid seemed to achieve higher visual and anatomic gains and that intraretinal cystoid fluid did not affect the outcomes. Guthoff et al. ¹⁵ found that the presence of cystoid ME was an unfavorable factor for the long-term visual outcome in DME eyes treated with intravitreal bevacizumab (IB) and/or triamcinolone acetonide. Itoh et al. ²⁵ investigated the relationship between SD-OCT features and functional outcomes for DME undergoing treatment with IB. Severity of intraretinal fluid and EZ integrity were associated with functional outcomes. Zur et al. ²⁶ investigated OCT biomarkers for DME treated with intravitreal dexamethasone implant. The presence of subretinal fluid, absence of hyperreflective foci, and integrity of the EZ layer were predictive of better visual outcome. The study revealed that size and location of intraretinal cysts did not change the functional outcome. They proposed that the remaining tissue between cysts in the central macula seems to be more important than cyst size for VA. Nevertheless, Murakami et al. ²⁷ showed that the areas beneath the cystoid spaces in the outer plexiform layer (OPL) had a longer length of disrupted EZ and disrupted external limiting membrane lines in DME patients. The significant association between cystoid spaces and photoreceptor degeneration beneath them suggests that the cystoid spaces in the OPL might be an important mechanism for photoreceptor damage and visual impairment. They demonstrated that the length of disrupted EZ at the fovea was correlated with the logMAR more than the association between foveal thickness and logMAR. Similarly, in our study, disrupted EZ was associated with poor visual prognosis. Cystic ME was greater in patients with decreased VA, but the association between cystic ME and visual outcome could not be confirmed in the multivariate analysis. However, in this study, the relationship between cystoid edema and disruption of EZ was not investigated, and SRD did not affect the visual outcome.

In the study reported by Bressler et al., ¹⁷ at 1 year after IR treatment of DME, HE were associated with favorable OCT improvement but was not associated with vision outcome. DRCR Network Protocol-I Study ¹⁰ determined that HE in the macula were correlated with both higher VA gain and anatomical improvement on OCT in DME treated with IR. In the current study, although its significance could not be confirmed in a multivariate model, absence of HE was associated with increased vision. These results suggest that HE accompanied by ME are a limiting factor to the increase of VA in DME treatment. Sigurdsson and Begg ²⁸ stated that the accumulation of HE in the OPL—near the pigment epithelium—may be a condition which causes focal alteration in the pigment epithelium, which leads to organization with scar formation. Studies on the functional consequences of the changes caused by HE are necessary.

In the combination therapy using IR and short pulse focal/grid laser photocoagulation for the treatment of DME, Hirano et al. found that eyes without leaking MA required fewer IR injections than eyes with leaking MA. ²⁹ In contrast, in the recent study, the mean numbers of IR injections were similar in eyes with or without leaking MA. The rates of macular laser administrations were also similar. In contrast to Hirano et al.'s study, an association was found between presence of leaking MA and visual outcome in this study. There is a need for further studies on the effect of DME accompanying leaking MAs on treatment response.

The current study needs to be viewed in light of the following limitations: it was not a prospective study; medications and systemic factors were related to diabetes, and some recent OCT findings—such as disruption of retinal inner layers and retinal hyperreflective foci—have not been considered as prognostic factors.

In conclusion, PRN IR treatment provides significant benefits in VA gain and anatomic improvement in eyes with DME at first month of therapy, and these gains sustain through 2 years. Age, gender, baseline BCVA, development of VMT and PVD, and disruption of EZ were found as associated factors with 2-year visual outcome.