Long-Term Effects of Intravitreal Ranibizumab Compared With Panretinal Photocoagulation on Optical Coherence Tomography Measured Choroidal Thickness and Vascularity

Abstract

Purpose: Compare choroidal changes in ranibizumab versus panretinal photocoagulation (PRP)–treated eyes with proliferative diabetic retinopathy (PDR).

Methods: DRCR Retina Network Protocol S post hoc analysis evaluated optical coherence tomography change in choroidal thickness (subfoveal and 3mm superior and inferior to the fovea) through five years; choroidal vascularity index (CVI) was assessed at baseline and one year. Mixed linear models for choroidal change included adjustments for the baseline choroidal value and age.

Results: This study included 328 eyes (158 ranibizumab and 170 PRP) from 256 participants (88 ranibizumab and 95 PRP eyes at five years). Mean change in choroidal thickness from baseline to five years at the fovea was −12 µm in ranibizumab versus −8 µm in PRP (difference [95% confidence interval]: −4 [−18 to 10], P = 0.57), superior was −14 µm versus −19 µm (difference: 5 [−8 to 17], P = 0.45) and inferior was -26 µm versus −32 µm [difference: 5 (−9 to 20), P = 0.45]; change at all three points within the ranibizumab group, and the superior and inferior points for PRP, were statistically significant (P < .05). Mean change in CVI at one year was −0.02% in ranibizumab versus −0.95% in PRP (difference: 0.93 [−0.35 to 2.21], P = 0.14).

Conclusions: In patients with PDR, treatment with ranibizumab versus PRP did not result in statistically significant differences in five-year choroidal thickness or one-year CVI change. Both groups had significant decreases in choroidal thickness at five years.

Translational Relevance: Ranibizumab treatment for PDR did not statistically significantly affect choroidal thickness or vascularity differently than PRP.

Introduction

Diabetes mellitus results in microvascular changes to the retina that may lead to proliferative diabetic retinopathy (PDR), which is a common cause of vision loss.^1,2 Panretinal photocoagulation (PRP) and intravitreal antivascular endothelial growth factor (anti-VEGF) agents are commonly used to treat PDR.¹ However, diabetes mellitus may also have microangiopathic effects on the choroid, such as choriocapillaris loss and intrachoroidal neovascularization.^3–5 Histopathological examinations of the choroid in diabetic eyes and animal models have suggested the choriocapillaris dropout may have an inflammatory etiology.⁶ After choriocapillaris dropout, the RPE may become hypoxic and stimulate angiogenesis.⁷ Despite histopathological evidence of these changes, there have been conflicting clinical reports on the extent and spectrum of choroidal changes in eyes with PDR as assessed by optical coherence tomography (OCT).^8–17 Additionally, the effect of PRP versus anti-VEGF (two commonly used treatment approaches for PDR) on the choroid is unclear. 

The choroid plays an important role in the perfusion of the retina, providing blood flow to the outer retina. Abnormal choroidal blood flow may compromise the retina, but the relationship between choroidal thickness and the progression from non-proliferative diabetic retinopathy (NPDR) to PDR remains debated.^14,16,18 Some studies have shown that compared with healthy eyes, subfoveal choroidal thickness may be lower in eyes with NPDR, PDR, or diabetic macular edema (DME).^14,16 Others suggested that choroidal thickness is higher in patients with PDR compared with NPDR, that thicker baseline choroidal thickness may be predictive of worse diabetic retinopathy outcomes, and that PRP may be associated with decreased choroidal thickness.^10,15,19 A limitation of these studies was that many were retrospective, and only a few examined longitudinal changes, so it is unclear if the choroidal thickness change was a cause or effect of the diabetic retinopathy. 

Anti-VEGF is a well-established, effective treatment for reducing retinal thickening from DME.^1,9,18 Although some studies suggest that anti-VEGF for DME may also reduce choroidal thickness, the effect of anti-VEGF on choroidal thickness has not been clearly defined.^12,18 Some studies have proposed that greater baseline choroidal thickness is related to an improved response of DME to anti-VEGF.²⁰ Few studies have evaluated choroidal changes after anti-VEGF therapy in eyes with PDR and no DME. Short-term and long-term changes in choroidal thickness after PRP treatment for PDR are also not well defined. For example, one prospective study showed increased choroidal thickness one week after receiving PRP.⁸ In comparison, decreased choroidal thickness has been shown one month, three months, six months, and one year after treatment with intravitreal bevacizumab, PRP, or both.^11–13,17 

Because of the variability of the relative contributions of vascular versus stromal tissue to choroidal thickness, the choroidal vascular index (CVI) has been proposed as another quantitative measure of the choroid. The CVI is the vascular area of the choroid divided by the total choroidal area, and it has been suggested to decrease with progression of diabetic retinopathy.²¹ The CVI has also been shown to decrease after anti-VEGF treatment for DME in patients without prior PRP, but not in patients previously treated with PRP.²² Another study suggested CVI was unaffected by PRP.¹⁵ Overall, changes in CVI following treatment for PDR are not yet well understood. OCT images from completed clinical trials provide an opportunity to assess long-term effects on both choroid thickness and CVI in patients with PDR requiring treatment.  

The DRCR Retina Network Protocol S randomized clinical trial compared anti-VEGF intravitreal ranibizumab (with deferral of PRP) with PRP treatment for eyes with PDR and concluded that visual acuity was comparable between the treatment groups at both two and five years.¹ The objective of the current analysis was to evaluate and compare long-term changes in choroidal thickness and CVI (as assessed by OCT) in response to ranibizumab versus PRP for patients with PDR in Protocol S. 

Methods

This study was a post hoc analysis of OCT images. The Protocol S methods were previously reported.¹ The study adhered to Declaration of Helsinki and was approved by institutional review boards. Eyes were randomized 1:1 to receive 0.5 mg intravitreal ranibizumab or PRP treatment for PDR. Study eyes had no prior PRP and best-corrected visual acuity letter score of 24 or better (approximate Snellen equivalent of 20/320 or better). Eyes with and without DME were enrolled, with ranibizumab provided to both groups to treat DME. Data collection took place between February 2012 and February 2018. The post hoc choroidal assessments of the OCT images were conducted from 2018 to 2023. 

Spectral domain volume OCT images (Zeiss Cirrus [Zeiss, Oberkochen, Germany] 512 × 128 scan pattern and Heidelberg Spectralis [Heidelberg Engineering, Heidelberg, Germany] 49 B-scans and Automatic Real-Time setting of 16) were evaluated by graders masked to treatment group assignment and treatments received. Choroidal thickness (between Bruch's membrane and the choroid-sclera junction) was manually measured at a single point using calipers in the devices’ software at the fovea and the superior and inferior outer ring center points (3 mm from the center point, same rings as the ETDRS grid) (Fig. 1). Choroidal thickness was graded at baseline, four weeks, 16 weeks, one year, two years, and five years. CVI was computed using a semi-automated MATLAB-based CVI quantification tool based on previously described image-processing methods on raw OCT B-scans.^21,23,24 In brief, the choroidal area was segmented automatically and the area was binarized using the Otsu Technique.^23,24 The dark pixels corresponded to the luminal region and the bright pixels corresponded to the stromal region. The CVI was calculated as the ratio of luminal area to total area.²¹ Because of the labor-intensive process, the CVI was only evaluated at baseline and one year. Missing choroidal data for participants that completed the study visit could be attributed to missing OCT images, damaged files that could not be opened, or indeterminate choroid-scleral junction. The reasons for ungradable images were not routinely tracked; however, the choroid-scleral junction is difficult to define and degraded image quality due to anterior segment pathology, macular edema, and vitreous hemorrhage all contributed to the inability to quantify choroid characteristics. All study eyes with at least one gradable choroidal assessment were included. 

Figure 1.

 

Choroidal thickness was manually measured at a single point using digital calipers at (A) 3 mm superior to the fovea, (B) the fovea, and (C) 3 mm inferior to the fovea.

Statistical Analysis

The primary outcomes were change in subfoveal, superior, and inferior choroidal thickness from baseline through five years and CVI change from baseline to one year. Adjusted least squares means and mean differences were estimated with mixed linear models that included fixed effects for the treatment group, baseline age, the baseline value of the choroidal outcome, and the randomization stratification factors (study eye laterality and baseline central subfield thickness). A random participant-level intercept modeled the potential correlation among participants with two study eyes (one in each treatment group). Adjusted means included two-sided 95% confidence intervals (CI) and P values reflecting a two-sided test of superiority. Although statistical significance was arbitrarily set at the 5% alpha level, there were no formal adjustments for multiplicity and the results from this study should be interpreted as exploratory. This post hoc analysis was complete-case (no imputation of missing data) and followed the intent-to-treat principle (analyzing all eyes according to randomized treatment group assignment, irrespective of treatments received). 

Supplemental analyses included an evaluation of the baseline differences in choroidal thickness measured at the fovea compared with the superior and inferior points, as well as the baseline choroidal thicknesses by age and presence of center-involved DME (CI-DME). Five-year choroidal thickness change by baseline age and a longitudinal analysis of choroidal change that excluded eyes after receiving treatment for DME were also examined. Analyses were conducted with SAS SAS/STAT 15.1 (SAS Institute, Inc, Cary, NC, USA). 

Results

This study included 328 eyes of 256 participants (72 bilateral) across 51 U.S. sites, comprising 83% (328 of 394) of the study eyes from the original Protocol S study. Participants’ mean (±SD) age at baseline was 52 ± 12 years and 55% (140 of 256) were male. Baseline visual acuity (VA) was 75.7 ± 12.5 letters (Snellen Equivalent approximately 20/32), and 20% (64 of 328) had CI-DME with visual acuity 20/32 or worse. Baseline characteristics appeared mostly comparable between overall inclusion/exclusion in this study, and equally distributed between treatment groups among those included (Table). Among the eyes included in this study with OCT images sent for choroidal assessment, at least one choroidal outcome was gradable for 93% (302 of 328) of the eyes at baseline, and among eyes with OCT images at follow-up visits, gradability was 92% (264 of 287) at one year, 92% (254 of 275) at two years, and 87% (183 of 211) at five years (Fig. 2). Across all time points, 95% (643 of 679) of the OCT images on Heidelberg Spectralis and 88% (722 of 823) on Zeiss Cirrus had a graded choroidal value. 

Table.

 

Baseline Participant and Ocular Characteristics

 

Figure 2.

 

Flowchart of eyes in the Protocol S clinical trial that were included in the substudy on choroidal anatomic changes.

The mean choroidal thickness at baseline was 254 ± 55 µm at the fovea, 273 ± 62 µm in the superior, and 271 ± 61 µm in the inferior points. Subfoveal measurements were significantly thinner than both the superior (mean difference [95% CI]: 14 µm [7 to 21], P < 0.001) and the inferior (12 µm [5 to 18], P < 0.001) assessments at baseline (Supplementary Table S1). Higher baseline age was associated with thinner choroidal thickness at all three points (mean decrease for every one-year increase in age: −0.8 µm [−1.5 to −0.1], P = 0.02 (subfoveal); −1.3 µm [−1.9 to −0.6], P < 0.001 (superior); and −1.4 µm [−2.0 to −0.7], P < 0.001 (inferior)). However, baseline age did not have a statistically significant effect on the change from baseline to five years for any of the choroidal thickness points (Supplementary Table S2). 

Study Treatments

All eyes included in this study in the ranibizumab group received ranibizumab, and all in the PRP group received PRP. At five years, the mean number of ranibizumab injections in the ranibizumab group was 18.4 ± 10.8, with 48% (42 of 88) of eyes indicating that at least one injection was for the treatment of DME, and 16% (14 of 88) having received PRP. In the PRP group, the mean number of injections at 5 years was 8.3 ± 7.2 among the 53% (50 of 95) of eyes that received ranibizumab for DME, with the remaining eyes receiving no injections through five years. Additional treatment information is in Supplementary Table S3. Although there was treatment crossover in both directions between the groups, there did not appear to be substantial differences in the mean changes in choroidal thickness over time when comparing strict treatment group assignments, censoring eyes out after receiving the alternate groups treatment, with the overall (intent-to-treat) randomized treatment group assignments (Supplementary Fig. S1). 

Subfoveal Choroidal Thickness

From baseline to one year, the mean (95% CI) decrease in choroidal thickness at the fovea was statistically significant in the ranibizumab group (−11 µm [−18 to −3], P = 0.008) but not the PRP group (−7 µm [−14 to 0], P = 0.07) and the mean difference between the ranibizumab and PRP groups was not statistically significant (−4 µm [−13 to 6], P = 0.44). At two years, neither within nor between group mean differences were statistically significant. At five years, the mean subfoveal choroidal thickness was again significantly reduced in ranibizumab (−12 µm [−23 to −1], P = 0.03) but not PRP (−8 µm [−20 to 3], P = 0.14) treated eyes, and the difference between groups remained not statistically significant (−4 µm [−18 to 10], P = 0.57) (Fig. 3A, Supplementary Table S1). 

Figure 3.

 

Change in (A) subfoveal, (B) superior, and (C) inferior choroidal thickness over 5 years for eyes randomized to the ranibizumab (RBZ) and panretinal photocoagulation (PRP) treatment groups. The least squares (LS) means, mean differences between treatment groups (RBZ-PRP), and 95% confidence intervals (CI) were estimated from mixed linear models with fixed effects for the treatment group, baseline age, the baseline choroidal thickness measure, the randomization stratification factors (study eye laterality and baseline central subfield thickness) and participant-level random intercepts for correlation among participants with two study eyes.

Superior Choroidal Thickness

The mean change in superior thickness was not statistically significant within or between treatment groups at one or two years. Both groups decreased at five years (−14 µm [−23 to −6], P = 0.002 [ranibizumab] vs. −19 µm [−30 to −8], P = 0.001 [PRP]), but there was not a statistically significant difference in five-year change between the groups (5 µm [−8 to 17], P = 0.45) (Fig. 3B, Supplementary Table S1). 

Inferior Choroidal Thickness

There were no statistically significant differences between treatment groups in mean inferior thickness change at one year. At two years, both groups had statistically significant mean reductions (−15 µm [−24 to −5], P = 0.003 [ranibizumab] vs. −13 µm [−21 to −5], P = 0.002 [PRP]) without a statistically significant between-group difference (−2 µm [−14 to 10], P = 0.78). Similarly, there were substantial reductions in inferior thickness for both groups at five years (−26 µm [−38 to −15], P < 0.001 [ranibizumab] vs. −32 µm [−44 to −20], P < 0.001 [PRP]), but the difference between groups was again not statistically significant (5 µm [−9 to 20], P = 0.45) (Fig. 3C, Supplementary Table S1). There were also no statistically significant differences between the treatment groups in mean choroidal thickness change from baseline to four or 16 weeks for any of the subfoveal, superior, or inferior points (Supplementary Table S1). 

CVI

The mean CVI was 61.1% ± 3.0% in the ranibizumab group and 60.9% ± 3.2% in the PRP group at baseline, and 60.6% ± 3.6% in the ranibizumab group and 60.4% ± 3.2% in the PRP group at one year (Supplementary Table S4). The adjusted mean change from baseline to one year was not statistically significant in the ranibizumab group (−0.02% [−0.94 to 0.90], P = 0.96), and although there was a slight decline in the PRP group (−0.95% [−1.79 to −0.11], P = 0.03) the difference between the ranibizumab and PRP groups was not statistically significant (0.93% [−0.35 to 2.21], P = 0.14). Other measures of choroidal vascularity, including choroidal, stromal, and luminal area are in Supplementary Table S4. 

DME

There were no statistically significant differences in baseline choroidal thickness or CVI by presence of CI-DME with visual acuity 20/32 or worse (Supplementary Table S5). Among eyes without CI-DME and vision loss at baseline, fewer eyes in the ranibizumab group (18% [23 of 129]) compared with the PRP group (32% [43 of 135]) developed vision-impairing CI-DME (on OCT) within five years. However, among all eyes included in this study, within five years 51% (80 of 158) of eyes in ranibizumab and 56% (95 of 170) in PRP received treatment for DME (focal/grid laser or at least one ranibizumab injection indicated for DME). Excluding eyes after receiving DME treatment did not appear to alter the overall mean trajectories for change in choroidal thickness within either treatment group (Supplementary Fig. S2). Additionally, CVI change from baseline to one year among eyes that received DME treatment was comparable to eyes that did not, and the interaction between treatment group and receiving DME treatment on one-year CVI change was not statistically significant (P = 0.60) (Supplementary Table S6). 

Discussion

In this post hoc analysis of a randomized clinical trial comparing ranibizumab with PRP treatment for PDR, there were no statistically significant differences between the treatment groups in mean choroidal thickness at any of the three measured points in the macula at four or 16 weeks or one, two, or five years. Both groups had statistically significant decreases from baseline in the superior point at five years and the inferior point at two years and five years. For the subfoveal point, only the ranibizumab group had statistically significant reductions at 1 and 5 years. However, the reductions were not statistically significantly different from the PRP group change at one or five years (meaning although the mean changes in the ranibizumab group were statistically significantly less than zero they were not statistically significantly less than the PRP group). There were no differences in CVI change at one year between the treatment groups. Although over half of the eyes in both groups were treated for DME, the results did not appear to differ whether eyes treated for DME were included or excluded after treatment. 

In this study, the choroid tended to be thinner subfoveally than inferiorly or superiorly at baseline. In contrast, previous studies have shown that in eyes with PDR, the choroid is thicker subfoveally than at the temporal or nasal macula.¹⁶ One reason for higher baseline choroidal thicknesses noted in the superior and inferior points at the macula in this study may be due to closer proximity to areas of peripheral retinal ischemia or nonperfusion. However, normal eyes also have been reported to have choroidal thickness greatest superiorly.²⁵ Other reasons for inconsistent findings may be due to the difficulty in segmenting the choroid, which is only well defined in about 60% of healthy eyes when using spectral domain OCT.²⁶ Enhanced depth imaging, additional image averaging, and swept-source OCT may improve visualization of the choroid-scleral junction and may account for some variation in published results between OCT instruments.^27,28 

Although the reductions in superior and inferior thickness from baseline to 5 years were statistically significant within both treatment groups, it is difficult to determine whether these reductions are wholly due to treatment or also impacted by the natural progression of diabetic eye disease and/or age-related choroid changes. Participants were a mean of 52 years old at baseline, and it has been shown that a decline in choroidal thickness occurs most dramatically in the fifth and sixth decades.²⁶ Although higher age was related to thinner choroidal tissue at baseline, it was not associated with any more or less change from baseline to five years, suggesting choroidal thickness change during the study did not vary greatly depending on age at the beginning of the study. Still, baseline age and baseline choroidal thickness were adjusted for in all analyses. Therefore, although some contributions from participant aging during the study may have affected the choroidal tissue, an age-related change is unlikely to have confounded any treatment effects, especially with respect to the PRP versus ranibizumab comparisons. It is impossible to know whether choroidal thickness in these eyes would have decreased by a different amount had they not received any treatment (PRP or ranibizumab). However, these results suggest that, on average, one treatment compared with the other did not lead to different choroidal thickness changes over five years. 

The finding of no statistically significant differences between the treatment groups in one-year CVI change is consistent with cross-sectional studies that have shown no differences in CVI between PDR patients with or without PRP treatment.^22,29 However, unlike previous reports suggesting ranibizumab treatment for DME decreases CVI only in patients without prior PRP, the one-year CVI change among eyes treated with ranibizumab for DME in this study were not statistically significant between PRP and ranibizumab groups.²² 

This study had several limitations. First, it was a post hoc analysis that was not designed to detect differences in choroidal anatomic changes. This study may have been underpowered to detect some baseline changes statistically significant at the 5% level, and at the same time, some statistically significant findings may be spurious because type 1 error wasn't strictly controlled at 5%. However, the mean differences between the two treatment groups tended to be small and probably not clinically meaningful. Second, there was a treatment crossover between the groups, but only a small proportion (16%) of the ranibizumab group received PRP, and ranibizumab treatment for DME in both groups was part of the study design. Third, imaging was performed with spectral domain OCT for retinal imaging rather than with enhanced depth imaging or swept-source imaging OCT of the choroid. Fourth, there was a significant loss to follow-up, especially at five years, and missing data were not imputed. Fifth, choroidal thickness was measured manually. Sixth, the thickness of the choroid can vary depending on the measurement location, participant age, disease severity, and treatment, and these factors may have contributed to differences found in other studies.^16,25 Seventh, CVI was not evaluated beyond one year because CVI assessment is a time-consuming and laborious process, and there were only minimal changes in CVI at one year. 

Despite these limitations, this analysis was in a large, multicenter, prospective cohort of patients with PDR, with and without DME at baseline, that had randomized treatment groups, standardized OCT acquisition by certified personnel and standardized OCT choroidal evaluations by trained and masked graders. 

Conclusions

In patients with PDR, treatment with ranibizumab versus PRP resulted in no statistically significant differences between the two treatments in changes in choroidal thickness at 3 distinct locations within the macula (subfoveal, superior, and inferior) through five years or in CVI at one year. At baseline, mean choroidal thickness was greater in the superior and inferior macular regions than in the subfoveal region. At five years, both treatment groups had a statistically significant reduction in superior and inferior choroidal thickness above and below the fovea. Future studies may elucidate whether these decreases in peripheral macular choroidal thickness are the result of aging, diabetic eye disease progression, treatment, or a combination of these factors. 

Acknowledgments

Supported through a cooperative agreement from the National Eye Institute and the National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, U.S. Department of Health and Human Services EY14231. Genentech provided ranibizumab for the study and funds to the DRCR Retina Network to defray the study's clinical site costs. 

The National Institutes of Health participated in oversight of the conduct of the study and review of the manuscript but not directly in the design or conduct of the study nor in the collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication. Per the DRCR Retina Network Industry Collaboration Guidelines (available at http://www.drcr.net), the DRCR Retina Network had complete control over the design of the protocol, ownership of the data, and all editorial content of presentations and publications related to the protocol. 

Disclosure: B. Lee, None; K. Josic, NEI (R), Genentech (R), JDRF (R), Regeneron (R); M.G. Nittala, None; S.B. Velaga, None; A. Karamat, None; S. Srinivas, None; F. Corvi, None; G. Singh, None; S. Sadda, None; J.K. Sun, None; M. Ip, Alimera (C), Allergan (C), Amgen (C), Apellis (C), Clearside (C), Genentech (C), Novartis (C), OCCURX (C), Regeneron (C), Astellas (F), Biogen (F), Genentech (F), Lineage Cell Therapeutics (F), Regenxbio (F), Splice Bio (F), 4DMT (F)