Effect of Intravitreal Triamcinolone Acetonide or Bevacizumab on Choroidal Thickness in Eyes With Diabetic Macular Edema

Abstract

Purpose.: We evaluated the effect of intravitreal triamcinolone acetonide (IVTA) or intravitreal bevacizumab (IVB) on subfoveal choroidal thickness (SFCT) in eyes with diabetic macular edema (DME).

Methods.: In this prospective, randomized, interventional comparative study, 51 DME eyes of 51 patients were randomized to receive either IVTA or IVB. The central macular thickness (CMT) and SFCT were determined by optical coherence tomography at 24 hours, 7 days, and 4, 8, and 12 weeks. The SFCT at 1500 and 3000 μm nasal or temporal to the central fovea also was measured. The values obtained before were compared to those obtained 12 weeks after the injections.

Results.: The eyes were randomly assigned to the IVTA (25 eyes) and IVB (26 eyes) groups. The SFCT was reduced significantly in the IVTA group from 24 hours to 12 weeks. The average ± SD of the SFCT expressed as the ratio to baseline thickness decreased to 94.8% ± 5.6% (P < 0.01) at 24 hours after IVTA and remained unchanged up to 12 weeks (91.8% ± 10.5%, P < 0.01, Wilcoxon signed-rank test). In the IVB group, no significant difference was found in the SFCT after IVB for 12 weeks. The CMT decreased significantly in both groups from 24 hours to 4 weeks; however, the decrease was not significant at 8 weeks or later in the IVB group.

Conclusions.: The decrease in choroidal thickness in eyes with DME after IVTA suggests that the choroidal pathology in diabetic retinopathy might be due to steroid-sensitive factors rather than vascular endothelial growth factor. (www.umin.ac.jp/ctr number, clinical trials number UMIN000009854.)

Introduction

Recent studies have shown that the treatment of diabetic retinopathy (DR) has been greatly improved, but diabetic macular edema (DME) still is a vision-threatening disease in patients with DR. ^1,2 The etiology of DME is multifactorial and is associated with VEGF, inflammation, vitreoretinal traction, and ischemia. Because of its multifactorial nature, it is difficult to treat. ²  

Grid laser treatment has been the gold standard treatment for refractory DME, but it has been replaced almost completely by pharmacologic treatments, such as intravitreal steroids or anti-VEGF antibodies. ^2,3 The anti-VEGF drugs are considered to be most effective treatment for DME. ^2,4 Even so, it still is not accepted completely because of the potential serious side-effects on the vascular system, the need for repeated injections, and the cost. ^2,5  

The main pathologic effect of DR is on the retina, and the majority of the studies have concentrated on the retina. However, several recent studies have shown that the choroid has a significant role in the pathological changes in DR. Thus, McLeod et al. ⁶ reported that different types of inflammatory cells were present in the diabetic choroid, and other studies have demonstrated a dropout of microvessels in the choroid at the early stage of diabetes. ^7,8 Nonetheless, the role of the choroid in diabetes is not well understood because of the lack of an effective investigative method to study the choroid noninvasively. 

After the introduction of enhanced depth imaging optical coherence tomography (EDI-OCT) by Spaide, ⁹ there have been significant advancements in the understanding of the pathophysiology of the choroid in retinochoroidal diseases. The choroidal thickness in eyes with DR has been investigated by EDI-OCT; however, the results still are controversial. ^10–13 If the choroid has an important role in DR, the changes in the choroidal thickness to different treatments could be used as an indicator of the effectiveness of the treatment. However, to the best of our knowledge, no study has evaluated the choroidal responses to different treatments. 

Thus, the purpose of this study was to compare the effect of intravitreal triamcinolone acetonide (IVTA), a steroid, and intravitreal bevacizumab (IVB), an anti-VEGF antibody, on eyes with DME especially on the choroid. 

Methods

This prospective study was designed as a 12-week, single center, randomized clinical study, performed to compare effectiveness of a single injection of IVTA or IVB. The protocol was approved by the Ethics Committee of the Kagoshima University Hospital (Kagoshima, Japan), and the procedures conformed to the tenets of the Declaration of Helsinki. All participants signed a written informed consent. Patients were enrolled between March 2013 and October 2013. This study was registered with the University Hospital Medical Network (UMIN)-clinical trials registry. The registration title was, “Randomized controlled clinical trials on the therapeutic effects of triamcinolone acetonide versus bevacizumab for diabetic macular edema” and the registration number was UMIN UMIN000009854. A detailed protocol is available in the public domain at https://upload.umin.ac.jp/cgi-open-bin/ctr/ctr.cgi?function=history&action=list&type=summary&recptno=R000011544&language=J, January 25, 2013. 

Participants

The inclusion criteria were age ≥ 18 years and a diagnosis of DME involving the central macula in patients with type 1 or 2 diabetes mellitus. The central macular thickness (CMT) had to be ≥250 μm in the central subfield based on the Heidelberg Spectralis-OCT (Heidelberg Engineering, Heidelberg, Germany) images. Participants were required to have a best-corrected visual acuity (BCVA) of 0.097 to 1.0 logMAR units, hemoglobin A1c (HbA1C) ≤ 12%, and reduced vision attributable to foveal thickening from DME without any other causes. In addition, women of childbearing age were included only if they were willing not to become pregnant and to use a reliable form of birth control during the study period. 

The exclusion criteria were history of vitreoretinal surgery, panretinal or macular laser photocoagulation, use of intraocular or periocular corticosteroids or anti-VEGF drugs within 6 months of the screening, vision reduction due to causes other than DME, regressed and currently inactive proliferative DR, ocular inflammation, cataract or other intraocular surgery within 6 months of the screening, laser capsulotomy within 3 months of the screening, aphakia, refractive error (spherical equivalent) greater than −6 diopters (D), or any disease that would compromise the visual acuity, or require medical or surgical intervention during the study period. In addition, patients were excluded if they had active iris neovascularization, vitreous hemorrhage, tractional retinal detachment, preretinal fibrosis involving the macula, visually significant vitreomacular traction or epiretinal membrane evident biomicroscopically or on OCT, structural damage to the center of the macula that would likely preclude improvement in the BCVA acuity after the resolution of macular edema, uncontrolled glaucoma or prior filtration surgery, or infectious blepharitis, keratitis, scleritis, or conjunctivitis. In addition, subjects were excluded if they had any of the following systemic conditions: uncontrolled diabetes mellitus, uncontrolled hypertension, history of cerebral vascular accident or myocardial infarction within 6 months, renal failure requiring dialysis or renal transplant, pregnancy or lactation, history of allergy to fluorescein or povidone iodine, only one functional eye, or an ocular condition in the fellow eye with a poorer prognosis than the study eye. 

Interventions

Patients were assigned randomly to two groups: 25 eyes were placed in the IVTA group and 26 eyes in the IVB group. The 25 eyes in the IVTA group received a single intravitreal injection of 4 mg of triamcinolone acetonide (IVTA, MaQaid; Wakamoto Pharmaceutical Co., Ltd., Tokyo, Japan), and the eyes in the IVB group received the single injection of 1.25 mg bevacizumab (IVB, Avastin; Genentech, Inc., South San Francisco, CA, USA) as we reported in detail. ^14,15 To avoid vitreous opacities caused by triamcinolone acetonide (TA), the TA was injected into the lower part of vitreous as much as possible. ¹⁵ A topical antimicrobial drug (Santen Pharmaceutical Co., Ltd., Osaka, Japan) was administered four times daily for three days before the injections. Topical anesthesia was induced by 0.4% oxybuprocaine hydrochloride (Benoxil; Santen Pharmaceutical Co., Ltd.) before the injections. Following disinfection and draping, 0.1 mL of 4 mg TA or 0.05 mL containing 1.25 mg of bevacizumab was injected into the vitreous cavity using a 30-gauge needle (bevacizmab) or a 27-gauge needle (triamcinolone) 3.5 mm from the limbus. To avoid an increase of the IOP, aqueous humor was removed before the injections. 

Outcomes

The primary outcome measure was a reduction in the subfoveal choroidal thickness (SFCT) from the baseline assessed by spectral-domain optical coherence tomography (SD-OCT) at each time point from 24 hours to 12 weeks. Secondary measures were the changes in the BCVA and the changes in the CMT from the baseline to 12 weeks. As an additional examination, we also measured the choroidal thickness at 1500 and 3000 μm temporally and nasally from the center of the fovea. 

Examination and Follow-up Protocol

After obtaining informed consent, eligible participants had a standard examination including refraction and determination of the BCVA, examination of the anterior and posterior segments, measurements of the IOP, and CMT and SFCT measurements by SD-OCT. These evaluations, except the BCVA, were repeated at 24 hours, 7 days, and 4, 8, and 12 weeks after injection. Fluorescein angiography was performed according to standard clinic procedures, and the axial length was measured using AL-2000 (Tomey, Nagoya, Japan) at the screening visit. The BCVA measurements and fundus photographs were taken at the baseline, and at 4 and 12 weeks after the injections. Blood samples for hematology and chemistry panel, and HbA1C were collected at the screening visit. To minimize the effect of circadian variations, all examinations of OCT were performed from 9:00 AM to 11:00 AM. 

Masking and Outcome Assessments

Measurements of the CMT and choroidal thickness were done in a masked fashion by two examiners (HO, NK) as we reported in detail. ^16,17 The inter-rater agreement was evaluated at each time point using two-way mixed-effects model for measurements of absolute agreement. 

The data were collected periodically by the principal investigator (SS). After reviewing for harmful events, the data were sent to the controllers (TS and MS). The remaining investigators were not permitted access to any information on the outcome before completion of the analyses. The complications were recorded as described. ¹⁵ The BCVAs and IOPs were measured by ophthalmic technicians, and the findings were reviewed by the physicians. The ophthalmic technicians were not informed on the purpose of the study or of the assignment schedule. 

SD-OCT Scanning Protocols

All SD-OCT examinations were performed at every visit with a Heidelberg Spectralis-OCT using a standardized protocol as we reported in detail. ^16,17 For each eye, a 31-line (30° × 25°) raster scan centered on the fovea was performed with 12 frames averaged for the CMT measurements. The changes in the SFCT were recorded with the EDI-OCT technique, with retinal scans performed along horizontal lines (7-line, 30° × 10°) through the center of the fovea. For the EDI-OCT technique, the Heidelberg Spectralis OCT instrument was pushed as close as possible to the eye to obtain an inverted image. Each section was obtained using eye tracking, and 100 scans were averaged to improve the signal-to-noise ratio. The CMT was defined as the distance between the internal limiting membrane (ILM) to the RPE, and the SFCT was defined as the distance between the outer border of the hyperreflective line corresponding to the RPE and the outer border of the choroid beneath the center of the fovea. 

Sample Size

We planned our study knowing that the values of the IVTA and IVB were continuously variable. An earlier study showed that the SFCT was almost well correlated with the CMT in DME eyes. ¹⁰ In addition, a previous study from our group showed that the IVTA reduced the CMT to 65% ± 17% of the baseline (35% reduction), while IVB reduced it to 83% ± 18% of the baseline at 4 weeks (17% reduction). ¹⁸ Thus, we estimated that the difference between the two drugs would be approximately 18% with an SD of approximately 20%. If the true difference in the two drugs is 18%, we will need to study ≥20 IVTA subjects and ≥20 IVB subjects to be able to reject the null hypothesis that the population means of the two groups are equal with a probability (power) 0.8. The Type I error probability associated with this test of the null hypothesis then would be 0.05. 

Randomization

Randomization was done by sealed envelopes after the patient was found to be eligible. The enrollment was completed after both groups had ≥20 eyes. 

Adverse Events

Any adverse events during the observation period were recorded by the examining physician and reported to the principal investigator. The adverse events related to the intervention were retinal tears, retinal fibrous membrane formation, retinal detachment, vitreous hemorrhage (VH), sterile endophthalmitis, rubeosis iridis, optic disc damage, corneal diseases, and cataract. The serious adverse events were bacterial endophthalmitis, retinal degeneration, unexplained optic disc damage, unexplained deterioration of vision, or uncontrollable increase in the IOP. The trial was reviewed every 4 weeks, and reevaluations were made when serious adverse events occurred or in cases when unexplainable decrease of visual acuity occurred. 

Withdrawal Criteria

Patients were discontinued from this study if any of the following findings were detected: a 2-line decrease in the BCVA from the previous visit with any increase in the CMT, any sign of infectious endophthalmitis, presence of new angiogenesis or VH, and development of any systemic adverse events. Antiglaucoma eye drops and/or acetazolamide were used when the IOP was >25 mm Hg and persisted for 3 days. If the IOP was not decreased below 20 mm Hg, patients were removed from the study and treated appropriately. 

Statistical Analyses

The baseline characteristics of each group were compared with t-tests (age and HbA1c), χ² tests (sex, status of DR, hypertension, hyperlipidemia, status of lens, and history of panretinal photocoagulation), or the Mann-Whitney U test (duration of DR and BCVA). Pre- and postoperative SFCT and CMT were compared with the Wilcoxon signed rank test at each time point. To study the SFCT, the measured values were expressed as a ratio to the baseline values. The ratios of the SFCT to the baseline were compared at each time point. The visual acuity of each group at each time point was compared to that at the baseline using the Wilcoxon signed rank test. To compare the two treatment groups, differences of the SFCT, CME, and visual acuity were examined by the Mann-Whitney U test. The IOP of each group at each time point was analyzed using the repeated measure ANOVA with Tukey postoperative SF. Then, the IOP after IVTA and after IVB were compared using t-tests with Bonferroni correction. The SPSS statistics 19 for Windows (SPSS, Inc., IBM, Somers, NY, USA) was used to perform the statistical analyses. A P value less than 0.05 was regarded as statistically significant. 

Results

Demographics of Subjects

The enrollment started in March 2013 and stopped in October 2013 when both groups had ≥20 participants. A total of 82 patients agreed to participate in the study, but 31 patients were excluded because of the exclusion criteria. The reasons for exclusions were 17 eyes with poor resolution of the OCT images, 4 eyes with prior vitreoretinal surgery, 4 eyes with use of intraocular or periocular corticosteroids or anti-VEGF drugs within 6 months of the screening, 4 eyes with poor control of diabetes mellitus, 3 eyes with cataract surgery within 6 months, 3 eyes with vitreous hemorrhage, 1 eye with active iris neovascularization, 1 eye with high myopia greater than −6 D, and 1 eye with a history of cerebral vascular accident or myocardial infarction within 6 months. Some of the eyes had more than one reason for exclusion. Then, the remaining 51 patients with DME were randomized with 25 eyes in the IVTA group and 26 eyes in the IVB group. Of these, 46 completed the study; four were lost to follow-up, and one dropped out because of treatment failure (Fig. 1). 

Figure 1

 

Schematic diagram of our study on choroidal thickness.

The differences in the demographic characteristics of the two groups were not significant (Table 1). The severity of diabetic retinopathy, the ratio of nonproliferative diabetic retinopathy and proliferative diabetic retinopathy, was not different between two groups. Eyes with ischemic maculopathy were excluded from both groups. 

Table 1

 

Patient Demographic Data and Baseline Characteristics

Variables	Group		P Value
	IVTA, n = 25	IVB, n = 26
Age, y	59.2 ± 12.5	62.9 ± 11.4	0.309
Sex, male	17	17	0.913
Duration of diabetes, y	10.4 ± 4.8	12.7 ± 5.3	0.063
HbA1C, %	7.0 ± 1.1	7.1 ± 1.1	0.696
Status of diabetic retinopathy
NPDR/PDR	10/15	10/16	0.595
Hypertension (%)	14 (56.0)	17 (61.5)	0.595
Hyperlipidemia (%)	15 (60.0)	15 (57.7)	0.958
Status of lens
Phakia/pseudophakia	19/6	17/9	0.277
History of PRP	17	17	0.655
BCVA, logMAR	0.39 ± 0.24	0.47 ± 0.31	0.388

 Values are means with SD. NPDR, nonproliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy; PRP, panretinal photocoagulation.

Subfoveal Choroidal Thickness (SFCT)

The inter-rater agreement at each point was good and nearly perfect (Supplementary Table S1). The baseline values of the mean SFCT group were 307.3 ± 92.1 μm in the IVTA group and 277.8 ± 86.4 μm in the IVB group (P = 0.25). The changes in the choroidal thickness at each time were expressed as the ratio to baseline because the baseline thickness of each individual varied. At 24 hours after IVTA, the ratio of the SFCT was significantly decreased to 94.8% ± 5.6% to the baseline value (P < 0.01) and remained the same up to 12 weeks. At 12 weeks, the SFCT was 91.8% ± 10.5% of the baseline value, which still was a significant decrease compared to the baseline (P < 0.01). 

In the IVB group, the percentage SFCT change was 100.0% ± 3.4% at 24 hours after the injection and remained at the same level throughout the observation period. There was no significant change (P = 0.93). 

There was a significant difference of the SFCT between the two groups at each measurement point (P < 0.05, Fig. 2). The nasal choroidal thickness and temporal choroidal thicknesses had a similar trend as the SFCT (Table 2). 

Figure 2

 

Chronological change of SFCT. Choroidal thickness at each measurement point is expressed as the ratio to that at baseline. At 24 hours after IVTA, the SFCT significantly decreased (P < 0.01, Wilcoxon signed rank test) of the baseline value and remained the same up to 12 weeks. After IVB, SFCT ratio at 24-hour injection was the same as the baseline value and remained at the same level throughout the observation period. There was no statistically significant difference (P = 0.88, Wilcoxon signed rank test). There was a significant difference of SFCT at each measurement point between the two groups (P < 0.05, Mann–Whitney U test). Significant value of †Wilcoxon signed rank test or *Mann–Whitney U test.

Table 2

 

Chronological Change of Choroidal Thickness at Different Points

Treatment	3-mm Nasal	1.5-mm Nasal	SFCT	1.5-mm Lateral	3-mm Lateral
IVTA
Baseline	161.1 ± 54.2	246.7 ± 85.6	307.3 ± 92.1	257.0 ± 75.2	221.3 ± 64.3
4 wk	143.4 ± 57.7	234.5 ± 79.5	279.6 ± 95.4	246.0 ± 74.7	210.4 ± 61.3
12 wk	138.6 ± 54.0	229.9 ± 81.6	282.9 ± 97.2	243.6 ± 72.5	208.4 ± 60.1
IVB
Baseline	156.4 ± 69.1	233.2 ± 72.5	277.8 ± 86.5	248.6 ± 78.1	208.2 ± 68.8
4 wk	151.8 ± 68.0	230.1 ± 71.5	275.2 ± 78.4	244.7 ± 77.7	205.2 ± 69.7
12 wk	152.2 ± 67.7	230.1 ± 68.9	276.6 ± 79.8	246.0 ± 77.3	205.7 ± 70.7

 Thickness is expressed as the average ± SD (μm).

Best-Corrected Visual Acuity (BCVA)

The baseline mean BCVA was 0.40 ± 0.25 logMAR units in the IVTA group and 0.48 ± 0.32 logMAR units in the IVB group. The BCVA in the IVTA group was significantly improved to 0.31 ± 0.23 logMAR units at 4 weeks (P < 0.01) and remained at the same value for up to 12 weeks. The BCVA in the IVB group also was significantly improved to 0.40 ± 0.25 logMAR units (P < 0.01) at 4 weeks and remained at the same value up to 12 weeks. The difference in the BCVA in the two groups was not statistically significant at any time (Fig. 3). 

Figure 3

 

Chronological changes of BCVA. The BCVA in the IVTA group significantly improved at 4 weeks (P < 0.01) and remained at the same level up to 12 weeks. The BCVA in the IVB group also significantly improved (P < 0.01) at 4 weeks and was maintained at the level up to 12 weeks. There was no statistically significant difference at any point between the two groups. †Significant value of Wilcoxon signed rank test.

Central Macular Thickness (CMT)

The mean baseline CMT was 503.9 ± 171.4 μm in the IVTA group and 495.7 ± 195.3 μm in the IVB group (P > 0.05). At 24 hours after the IVTA, the CMT decreased significantly to 452.1 ± 151.5 μm (P < 0.01). At 12 weeks, the CMT was 389.4 ± 209.4 μm, which also was significantly thinner than the baseline thickness (P < 0.01). In the IVB group, the mean CMT at 24 hours after injection was significantly decreased to 459.0 ± 173.9 μm (P < 0.01) and remained at the same thickness up to 4 weeks. However, at 8 and 12 weeks, the CMT increased, and at 12 weeks the CMT was 449.7 ± 212.2 μm, which was not significantly different than the baseline CMT (P = 0.12). The differences in the CMT in the two groups were not statistically significant at any time (Fig. 4). 

Figure 4

 

Chronological changes of CMT. There was no significant difference in baseline values of the mean CMT between the two groups. In the IVTA group, the CMT significantly decreased after 24 hours (P < 0.01, Wilcoxon signed rank test), and remained the same up to 12 weeks. In the IVB group, the CMT at 24 hours after injection significantly decreased (P < 0.01, Wilcoxon signed rank test) and remained same level up to 4 weeks. The differences in the CMT in the two groups were not statistically significant at any time. †Significant value of Wilcoxon signed rank test.

Serious Adverse Events and Alterations of IOP

No serious adverse events related to vitreous injection were detected in both groups during the 12-week follow-up period. The average baseline IOP was 13.7 ± 2.4 mm Hg in the IVTA group and 13.1 ± 2.9 mm Hg in the IVB group (P = 0.41). During the course of this study, three eyes in the IVTA group had an elevation of the IOP to >25 mm Hg, and all were given antiglaucoma drug, which reduced the IOP to <20 mm Hg within 1 week. An increase in the IOP also was observed in IVTA group at 4 weeks and later, but it did not reach significance. There was no statistically significant difference in the IOP between the two treatment groups. 

Discussion

Our results showed that the SFCT of DME eyes was significantly reduced from the baseline by a single IVTA, but not by a single IVB. The CMT was decreased significantly by IVTA and IVB, but not after 8 weeks. The BCVA was significantly improved by both treatments. 

It is widely accepted that the pathologic changes in eyes with DR are present in the retina and choroid. Histologic studies have shown increased tortuosity, focal dilation or narrowing of the choroidal vessels, and the formation of sinus-like structures between the choroidal lobules. In some advanced cases, there is narrowing of the capillary lumens, capillary dropout, inflammation, interstitial edema, and other changes. ^6–8 Although the exact mechanism that causes the reduction of choroidal thickness by single IVTA, but not by IVB, was not determined, we suggest the following. First, steroid-responsive factors have a more important role in affecting the choroidal thickness in diabetic eyes than VEGF alone, because steroids have biological activities on tissues more broadly than VEGF. ^19–23 The most prominent effect of steroids would be an anti-inflammatory effect caused by preventing phospholipid release and decreasing the production of inflammatory cytokines. ^20,21 Corticosteroids are known to reduce tissue edema, and downregulate the release of prostaglandin and histamines. ²¹ Synthesis of endothelial nitric oxide synthase (eNOS), a potent vasodilator, also is inhibited by corticosteroids. ²³ All of these factors can cause vasodilation and/or edema, leading to choroidal thickening. Interestingly, the choroidal thickness was not affected by IVB. The VEGF also induces NOS, which may be a strong factor that would cause choroidal thickening. ²⁴ However, the VEGF-related factors might not be a major factor for the diabetic choroidal changes. 

A second possibility is a slower movement of bevacizumab to the choroid. However, a study on monkeys showed that bevacizumab can reach the choroid to reduce the fenestration of the choriocapillaris after a single intravitreal injection. This indicates that a single intravitreal anti-VEGF agent can affect the choroid. ²⁵ However, this phenomenon was not tested in diabetic human eyes. Considering the systemic vascular pathology in diabetes, it is unlikely that the pathological mechanisms of the retinal vessels differ greatly from those of the choroidal vessels. We found that the retinal thickness was reduced significantly by IVTA and IVB at 4 weeks, indicating that the retina was exposed to these agents more than the choroid after the intravitreal injection. So, it is not unreasonable to expect the choroid to change as did the retina. Additionally, VEGF is secreted more into the choroid than into the neural retina from the RPE cells. ^26,27 Thus, the level of free VEGF might be too high to be blocked by the anti-VEGF agents used. 

There are several contradictory reports on the choroidal thickness in eyes with DR Regatieri et al. ¹¹ reported that the decrease in choroidal thickness was related to the severity of the DR, and Adhi et al. ¹³ reported that the subfoveal medium choroidal vessel layer and choriocapillaris layer were significantly thinner. However, Kim et al. ¹⁰ reported that the subfoveal choroid was thicker in eyes with DME than in those without DME. More recently, a population-based study of more than 3000 eyes showed that patients with diabetes mellitus had a statistically significantly thicker subfoveal choroid, but the presence of DR was not associated with an abnormal choroidal thickness. ¹² In our study, the baseline average SFCT was 277 to 307 μm, which is as thick as that of normal controls. ¹⁷ The choroidal thickness decreased significantly associated with a reduction of retinal thickness after IVTA. 

According to the earlier studies, the present findings can be explained as follows. ^6–8 There are two types of pathologic changes in the diabetic choroid; the changes that decrease the choroidal thickness, such as the narrowing of the capillary lumens, capillary dropout, and the changes that increase the choroidal thickness, such as inflammation, dilatation of vessels, and interstitial edema. ^6–8 The choroidal thickness evaluated by OCT represents the overall results of these changes and the inconsistent balance of these opposite changes in each study may have caused the discrepancy of the previous reports. ^10–13 In this study, the baseline choroidal thickness was not different from the baseline nondiabetic choroidal thickness, although it was decreased after IVTA. Because steroids can reduce factors that could increase choroidal thickness, such as inflammation and tissue edema, ^19–24 this is a reasonable result. If the diabetic choroid did not have these factors or the factors are lost by treatment, then it should be thinner than the nondiabetic choroid, which is consistent with recent findings. ^11,13 No such change was observed in the IVB group probably because VEGF alone would not be the decisive factor that caused the change of choroidal thickness of diabetic eye or maybe because IVB was not effective enough to cause it. ^14,28 To reconcile these differences, qualitative analysis of factors other than the thickness would be helpful. 

The strengths of our study are its prospective nature and randomized enrollment. Thus, the effect of IVTA and IVB could be examined not only on the choroidal thicknesses, but also on other factors. There was a small and nonsignificant baseline difference in the subfoveal choroidal thickness between the two groups (P = 0.25). Because there is no officially approved anti-VEGF drug for DME in Japan during the study period, the Institutional Review Board (IRB) recommended that we use the least number of eyes as possible. Considering the present results, an estimated power analysis before the study was appropriate. Many studies have compared the effect of a steroid and an anti-VEGF agents on DME eyes, and the visual acuity and retinal thickness were examined. ^4,18,29–34 To our knowledge, our study was the first to evaluate the choroidal thickness, and our results may shed light on the role of choroid in diabetic retinopathy. 

A large randomized clinical trial showed that intravitreal ranibizumab is better than IVTA for DME. ⁴ The results of our study provided reliable information on the best treatment for DME at that time. Although the results of our study showed that IVTA decreased the choroidal thickness more than anti-VEGF drug, this does not necessarily mean that steroids are better than anti-VEGF drugs. Our study was designed to determine the pathologic mechanism associated with the thickness of the diabetic choroid, but not to compare the effectiveness of treatment on vision. Thus, we cannot make any conclusion about the treatment effect on DME from the results. We would like to stress this point strongly. 

The limitations of this study included the small number of subjects. A reduction of the choroidal thickness after IVTA was detected, but a significant improvement of BCVA was not detected compared to IVB treatment group. Recently, three consecutive monthly injections of ranibizumab were recommended for DME. A consecutive monthly injection protocol may have a stronger effect of anti-VEGF agents on the choroid. Above all, our results were obtained by bevacizumab, not by the officially approved ranibizumab. Thus, our results cannot be generalized to every clinical setting, where monthly injection of ranibizumab is recommended. In addition, although the inter-rater agreement was sufficiently high (Supplementary Table S1), manual segmentation always has potential bias. This should be remembered in interpreting the data. 

In conclusion, the choroidal thickness of eyes with DME is reduced by IVTA. These changes may be related strongly to steroid-sensitive factors rather than VEGF-related pathology in eyes with DME. It is difficult to conclude whether this change is the cause or a result of DR from the present data. However, the results suggest that a study of the choroid is important in determining the pathology of diabetic retinopathy in developing effective treatments. 

Acknowledgments

The authors thank Duco Hamasaki for providing critical discussions and suggestions to our study, and revision of the final manuscript. 

The authors alone are responsible for the content and writing of the paper. 

Disclosure: S. Sonoda, None; T. Sakamoto, None; T. Yamashita, None; H. Otsuka, None; M. Shirasawa, None; N. Kakiuchi, None; E. Uchino, None; H. Terasaki, None; H. Kawano, None