Real-world Outcomes among Eyes with Center-Involving Diabetic Macular Edema and Good Visual Acuity

Abstract

Purpose:

We aimed to determine long-term visual and anatomical outcomes among patients with center involving-diabetic macular edema and good vision and evaluate factors associated with visual and anatomic outcomes.

Materials and Methods:

In this retrospective study, all patients with type 2 diabetes aged ≥18 years had seen at the Wilmer Eye Institute between March 2015-June 2018 and with diabetic macular edema confirmed on spectral-domain optical coherence tomography imaging were included, provided they had visual acuity of 20/30 or better in ≥1 eye and a follow-up duration of ≥3 clinic visits. Change in logMAR visual acuity and central 1 mm foveal thickness from baseline, lines of visual acuity gained/lost for overall cohort stratified by treatment were analyzed.

Results:

Among 197 (243 eyes) participants, mean (± standard deviation) age was 63.4 ± 11.2 years, and half were males. Average duration of follow-up was 1.7 ± 0.7 years. One hundred and forty-six eyes (60%) received anti-vascular endothelial growth factor injections, at an average of 3.7 ± 2.9 injections/eye/year. Mean logMAR visual acuity at baseline was 0.1 ± 0.1 [Snellen 20/25] in both treatment and observation (no anti-vascular endothelial growth factor treatment received during and 3 months prior to the study inclusion period) groups. Final logMAR visual acuity was 0.2 ± 0.2 in the treatment group [Snellen 20/32] versus 0.1 ± 0.3 in observation group [Snellen 20/25]; (p = .23). Mean central foveal thickness changed from 333 ± 66 to 308 ± 45 microns in treatment group and 319 ± 41 to 308 ± 65 microns in observation group.

Conclusions:

After an average of 1.7 years of follow-up, there were no significant differences in final vision or central foveal thickness irrespective of whether patients received or did not receive treatment with anti-vascular endothelial growth factor injections.

Introduction

Diabetic macular edema (DME) is the leading cause of visual impairment in people with diabetes mellitus (DM).¹ According to global estimates, 422 million people are currently affected with DM,² with DME present in up to 7.5% of patients.³ In the United States, 1 in every 25 patients with DM aged ≥ 40 years is believed to have DME in at least 1 eye, which equates to an estimated 746,000 people.⁴

Over the past decade, treatment of center-involving DME (ci-DME), has evolved from laser photocoagulation to intravitreal anti-vascular endothelial growth factor (VEGF) agents. However, until the recently published results of Protocol V, a randomized, multi-center clinical trial by The Diabetic Retinopathy Clinical Research Network,⁵ data, on the efficacy of anti-VEGF injections, were limited to trials that enrolled patients with impaired vision (VA 20/32 or worse).^6–9 Protocol V found no significant differences in vision loss at 2 years for eyes with ci-DME and good visual acuity (20/25 or better), irrespective of whether they were initially managed with aflibercept, focal laser or observation.⁵ However, these findings may not necessarily reflect real-world outcomes. Most recently, the OBTAIN study reported real-world outcomes for ci-DME patients with good VA and like Protocol V, noted that on average, patients maintained good vision at 12 months, regardless of whether or not they received treatment.¹⁰ However, patients, even those who did not receive treatments, were followed very closely to identify a possible decline in vision that may warrant treatment. In Protocol V, the median cumulative visits over 2 years were 18 in the anti-VEGF group and 12 in the observation group.⁵ Such frequent follow-up visits can be burdensome both to the healthcare system as well as patients. This is especially true for patients with DM whose course is often complicated by the presence of other systemic complications, requiring frequent visits to physician offices.¹¹ Identification of risk factors associated with vision loss can, therefore, be helpful for two reasons. First, it allows for the possibility of personalizing follow-up plans and following those most at risk of vision loss more closely. Secondly, identifying and targeting modifiable risk factors can potentially help improve outcomes. In our study, we aimed to identify systemic factors and patient-related characteristics associated with visual and anatomic outcomes.

Specifically, we aimed to determine long-term visual and anatomical outcomes among this group of patients (i.e. with ci-DME and good vision), and evaluate factors associated with visual and anatomic outcomes.

Materials and methods

Study design and population

This retrospective study was conducted among patients who were aged 18 years or older, had type 2 DM with DME, and presented to the Johns Hopkins Wilmer Eye institute between March 2015 and June 2018. The study adhered to the tenets of the Declaration of Helsinki and was approved by the Johns Hopkins University School of Medicine Institutional Review Board.

Cases of DME were initially identified using the International Classification of Diseases, Ninth and Tenth revision, Clinical Modification codes (ICD-9-CM and ICD-10-CM). Our case definition included any visits to the outpatient ophthalmology clinics between March 2015 and June 2018 that comprised the following diagnosis codes: DME (362.07), type 2 diabetes with unspecified diabetic retinopathy with macular edema (E11.311x), type 2 diabetes with mild non-proliferative diabetic retinopathy with macular edema (E11.321x), type 2 diabetes with moderate non-proliferative diabetic retinopathy with macular edema (E11.331x), type 2 diabetes with severe non-proliferative diabetic retinopathy with macular edema (E11.341x) and type 2 diabetes with proliferative diabetic retinopathy with macular edema (E11.351x).

Patients were included if they were ≥18 years at the time their diagnosis of type 2 DM and DME was first recorded in either the problem list or the visit diagnosis in the electronic medical record. Problem list refers to the patient’s consolidated medical history in the electronic medical record. We further restricted our search to include only those patients who also had optical coherence tomography (OCT) imaging done (Current Procedural Terminology codes 92134/92135) either on or after the date their DME was first diagnosed.

Inclusion criteria and exclusion criteria

Patients were included for final analysis if they had 1) a confirmed diagnosis of ci-DME by OCT (Spectralis, Heidelberg Engineering) after review of the OCT scan images by one of the co-authors (SZ), VA of 20/30 or better in at least one eye and a follow-up of at least three clinic visits. ci-DME was defined as any retinal thickening (≥305 microns for females and ≥320 microns for males)¹² and/or the presence of cysts suggestive of intraretinal fluid (IRF) and/or subretinal fluid (SRF) within 1 mm of the ETDRS center. Patients who did not meet the above criteria were excluded. Patient charts were also reviewed to confirm the absence of additional etiologies of macular edema including retinal vein occlusion and pseudo-phakic cystoid macular edema.

Data collection

The following data were extracted from the electronic medical record: age, sex, race, DME laterality, Snellen VA, clinical exam findings, diabetic retinopathy (DR) status at baseline and follow-up, duration of follow-up, history of ophthalmic procedures including cataract surgery, anti-VEGF injections, focal and pan-retinal photocoagulation (PRP) sessions, pars plana vitrectomy, and presence of systemic co-morbidities including diabetes, hypertension, dyslipidemia, obstructive sleep apnea. For VA, we considered the last line that was read correctly while wearing customary refractive correction. If improvement with a pinhole was documented, we used the VA with pinhole. OCT scans were reviewed by one of the co-authors (SZ) for scan quality, the accuracy of segmentation (using in-built manual calipers), and presence of ci-DME. The central 1 mm foveal thickness values (CFT), generated automatically by the device software, were recorded at baseline and final follow-up visits. Records with any incomplete or missing data were excluded.

Data analysis

Statistical analysis was performed using Stata version 14.1 (Stata Corp, College Station, TX). Snellen VA was converted into a logarithm of the minimum angle of resolution (logMAR) VA (rounded to the nearest decimal place) using the VA conversion chart.¹³ The VA outcome was expressed as 1) the mean change from baseline in logMAR VA, and 2) the proportion of eyes with no change in VA, gain or loss of 1 line and gain or loss of ≥2 lines of VA between baseline and last follow-up visit. We performed all analyses at the patient-eye level. Baseline characteristics were summarized with descriptive statistics. Mean values for patient demographics, baseline, and final logMAR VA and CFT, number of injections performed per eye per year and the duration of follow-up were calculated. These analyses were also performed after stratifying eyes based on 1) whether anti-VEGF treatment was or was not administered and, 2) whether cataract surgery was or was not performed during the study period. Chi-square and the Fisher exact tests were used to assess differences in categorical variables between the two study groups. The independent sample t-test was used to compare continuous variables.

Generalized estimating equations were used to account for the correlation between eyes within patients. Forward step-wise linear and logistic regression models were created to explore the effects of covariates (age, sex, baseline VA, baseline CFT, duration of DM, DR status at baseline and follow-up, presence of systemic co-morbidities, anti-VEGF administration, pseudophakia at last follow-up and history of cataract surgery during study period) on 1) VA (final logMAR VA and proportion of eyes with ≥1 line gain/loss) and 2) anatomical (proportion of eyes with final CFT of <300 microns and with ≥10% decrease in CFT from baseline) outcomes. Statistical significance was set at p < .05.

Results

Of the 973 patients identified using ICD codes, 197 patients (243 eyes) met all study eligibility criteria and had complete imaging and clinical data (Figure 1). Of these, 46 patients (23.4%) had DME present in both eyes (46/243 eyes, 18.9%).

Figure 1.

Cohort selection of patients with center-involving diabetic macular edema (ci-DME) and good visual acuity (VA).

Study participants

Mean ± standard deviation (SD) age of the study participants was 63.4 ± 11.2 years, and 55.3% were male. Caucasians and African Americans made up 57.4% and 35% of the study cohort, respectively. Mean VA at baseline was 0.1 ± 0.1 logMAR (Snellen 20/25) and average CFT was 327.1 ± 57.6 microns (Table 1). DR status at baseline and last follow-up was as follows: non-proliferative DR (NPDR) (74.1% vs. 71.2%) and proliferative DR (PDR) (23% vs. 25.9%). Retinopathy status was missing for 7 (2.9%) eyes. Among eyes with NPDR, retinopathy severity at baseline and follow up was: mild (29.4% vs. 32.9%), moderate (46.1% vs. 52%), severe (16.1% vs. 13.9%) and unknown (8.3% vs. 1.2%).

Table 1.

Characteristics of overall cohort with center-involving diabetic macular (ci-DME) and vision 20/30 or better at baseline and during follow-up period.

Baseline
Age ± standard deviation [SD], range (years)	63.4 ± 11.2, 30-91
Gender, No. (%)
Female	88 (44.7)
Male	109 (55.3)
Race, No. (%)
Caucasian	113 (57.4)
African American	69 (35.0)
Asian	2 (1.0)
Other	13 (6.6)
Laterality, No. (%)
Right eye	128 (52.7)
Left eye	115 (47.3)
Lens status, No. (%)
Phakic	157 (64.6)
Pseudophakic	86 (35.4)
Duration of diabetes, range (years)	4.4 ± 5.8, 0-31.9
Diabetic retinopathy (DR) status, No. (%)
Non-proliferative DR	180 (74.1)
Mild	53 (29.4)
Moderate	83 (46.1)
Severe	29 (16.1)
Unknown	15 (8.3)
Proliferative DR	56 (23)
Missing	7 (2.9)
Prior treatments, No. (%)	69 (28.4)
Anti-VEGF a only	55
Laser only	8 (Panretinal photocoagulation [PRP] =7)
Pars plana vitrectomy only (PPV)	2
Anti-VEGF + laser	4
Anti-VEGF + PPV	0
All 3 treatments	0
Systemic co-morbidities, No. (%)
Hypertension	100 (50.8)
Hyperlipidemia	86 (43.7)
Obstructive sleep apnea	29 (14.7)
Mean visual acuity (VA), ± SD (logMAR)	0.1 ± 0.1 (Snellen 20/25)
Distribution of VA, No. (%)
Better than 20/20	2 (0.8)
20/20	69 (28.4)
20/25	120 (49.4)
20/30	52 (21.4)
Mean central foveal thickness (CFT) ± SD, range (microns)	327.1 ± 57.6, 234-620
Follow-up period
Variables	ci-DME (243 eyes, 197 Patients)
Mean duration of follow-up ± SD (years)	1.7 ± 0.7
Lens status, No. (%)
Phakic	137 (56.4)
Pseudophakic	106 (43.6)
DR status, No. (%)
Non-proliferative DR	173 (71.2)
Mild	57 (32.9)
Moderate	90 (52)
Severe	24 (13.9)
Unknown	2 (1.2)
Proliferative DR	63 (25.9)
Missing	7 (2.9)
Eyes receiving treatment during study period, No. (%)	155 (63.8)
Anti-VEGF injection only	126 (81.3)
Laser only	3 (1.9) (all PRP)
PPV only	3 (1.9)
Anti-VEGF injection + Laser	14 (9)
Anti-VEGF injection + PPV	6 (3.9)
All three treatments	3 (1.9)
Mean number of injections/eye/year, ± SD, range	3.7 ± 2.9, 0.4-13.6
Median number of injections/eye/year
Frequency of number of injections/eye/year, No. (%)	2
>0 - ≤3	72 (49.3)
>3 - ≤5	20 (13.7)
>5 - ≤7	35 (23.9)
>7 - ≤9	12 (8.2)
>9	7 (4.8)
Mean visual acuity (VA), ± SD (logMAR)	0.2 ± 0.2 (Snellen 20/32)
Distribution of VA, No. (%)
Better than 20/20	0 (0)
20/20	56 (23.1)
20/25	80 (32.9)
20/30	44 (18.1)
Worse than 20/30	63 (25.9)
Distribution of visual outcomes, No. (%)
Stayed Same	89 (36.6)
Gain of 1 line	36 (14.8)
Gain of 2 lines	7 (2.9)
Loss of 1 line	55 (22.6)
Loss of 2 or more lines	56 (23.1)
Eyes with VA drop below 20/30 during follow-up, No. (%)	144 (59.3)
Mean ± SD number of visits/eye/year where VA was noted below 20/30	1.4 ± 2.3
Mean CFT ± SD, range (microns)	307.8 ± 57.7, 204-653

^avascular endothelial growth factor (VEGF)

The mean duration of follow-up was 1.7 ± 0.7 years. One hundred and fifty-five (63.8%) of the 243 included eyes received some form of treatment during the study period (Tables 2,3): intravitreal anti-VEGF therapy (146/155 eyes, 94.2%); PRP (15/155 eyes, 9.7%); vitrectomy (9/155 eyes, 5.8%); focal laser photocoagulation (5/155 eyes, 3.2%). The mean number of injections performed per eye per year was 3.7 ± 2.9 (range 0.4–13.6). The frequency of different anti-VEGF agents used was: aflibercept (68%), bevacizumab (26.4%) and ranibizumab (5.6%). No patient in our study was treated with steroid (dexamethasone) implants.

Table 2.

Baseline characteristics of cohort stratified by anti-vascular endothelial growth factor (VEGF) treatment.

	Anti-VEGF injections received (n=119 patients, 146 eyes)	No anti-VEGF injections (n=78 patients, 97 eyes)	P-value
Age ± standard deviation [SD], range (years)	61.3 ± 10.7	66.6 ± 11.2	<0.001
Gender, No. (%)			0.53
Female	51 (42.9)	37 (47.4)
Male	68 (57.1)	41 (52.6)
Race, No. (%)			0.58
Caucasian	72 (60.5)	41 (52.6)
African American	37 (31.1)	32 (41.0)
Asian	1 (0.8)	1 (1.3)
Other	9 (7.6)	4 (5.1)
Laterality, No. (%)			0.42
Right eye	80 (54.8)	48 (49.5)
Left eye	66 (45.2)	49 (50.5)
Lens status, No. (%)			0.01
Phakic	105 (71.9)	52 (53.6)
Pseudophakic	41 (28.1)	45 (46.4)
Duration of diabetes, range (years)	3.9 ± 5.0	5.1 ± 6.8	0.19
Diabetic retinopathy (DR) status, No. (%)			0.47
Non-proliferative DR	108 (73.9)	72 (74.2)
Mild	26 (24.1)	27 (37.5)
Moderate	54 (50)	29 (40.3)
Severe	19 (17.6)	10 (13.9)
Unknown	9 (8.3)	6 (8.3)
Proliferative DR	33 (22.6)	23 (23.7)
Missing	5 (3.4)	2 (2.1)
Prior treatments, No. (%)	61 (41.8)	8 (8.2)	<0.001
Anti-VEGF only	49 (80.3)	6 (75)
Laser only	8 (13.1)	0 (0)
Pars plana vitrectomy only (PPV)	0 (0)	2 (25)
Anti-VEGF + laser	4 (6.6)	0 (0)
Anti-VEGF + PPV	0 (0)	0 (0)
All 3 treatments	0 (0)	0 (0)
Systemic co-morbidities, No. (%)
Hypertension	61 (51.3)	39 (50)	0.86
Hyperlipidemia	48 (40.3)	38 (48.7)	0.25
Obstructive sleep apnea	19 (16)	10 (15.4)	0.54
Mean visual acuity (VA), ± SD (logMAR)	0.1 ± 0.1 (Snellen 20/25)	0.1 ± 0.1 (Snellen 20/25)	0.49
Distribution of VA, No. (%)			0.64
Better than 20/20	1 (0.6)	1 (1.1)
20/20	44 (28.4)	25 (28.4)
20/25	74 (47.7)	46 (52.7)
20/30	36 (23.2)	16 (18.2)
Mean central foveal thickness (CFT) ± SD, range (microns)	332.6 ± 66.1	318.8 ± 40.5	0.07

Table 3.

Patient characteristics stratified by anti-vascular endothelial growth factor (VEGF) treatment during follow-up period.

Variable	Anti-VEGF injections received (n=119 patients, 146 eyes)	No anti-VEGF injections (n=78 patients, 97 eyes)	P-value
Mean duration of follow-up ± SD, (years)	1.7 ± 0.7	1.6 ± 0.7	0.14
Lens status, No. (%)			<0.001
Phakic	99 (67.8)	38 (39.2)
Pseudophakic	47 (32.2)	59 (60.8)
Diabetic retinopathy (DR) status, No. (%)			0.33
Non-proliferative DR	104 (71.2)	69 (71.1)
Mild	28 (26.9)	29 (42)
Moderate	59 (56.7)	31 (44.9)
Severe	16 (15.4)	8 (11.6)
Unknown	1 (1)	1 (1.4)
Proliferative DR	38 (26)	25 (25.8)
Missing	4 (2.7)	3 (3.1)
Mean visual acuity (VA), ± SD (logMAR)	0.2 ± 0.2 (Snellen 20/32)	0.1 ± 0.3 (Snellen 20/25)	0.23
Distribution of VA, No. (%)			0.01
Better than 20/20	0	0
20/20	28 (19.2)	28 (28.9)
20/25	43 (29.5)	37 (38.1)
20/30	35 (23.9)	9 (9.3)
Worse than 20/30	40 (27.4)	23 (23.7)
Distribution of visual outcomes, No. (%)			0.74
Stayed Same	52 (35.6)	37 (38.1)
Gain of 1	20 (13.7)	16 (16.5)
Gain of 2 lines	3 (2.1)	4 (4.1)
Loss of 1 line	35 (23.9)	20 (20.6)
Loss of 2 or more lines	36 (24.7)	20 (20.6)
Eyes with VA drop below 20/30 during follow-up, No. (%)	105 (71.9)	39 (40.2)	<0.001
Mean ± SD number of visits/eye/year where VA was noted below 20/30	2.2 ± 2.7	0.3 ± 0.5
Mean CFT ± SD, range (microns)	307.7 ± 44.6	307.9 ± 65.3	0.98

Anti-VEGF treatments stratified by baseline VA

Twenty-nine percent (71/243) of eyes had baseline VA of 20/20 or better, of these 59% (42/71) were treated with anti-VEGF; 49% (120/243) of eyes had VA of 20/25, of these 57.5% (69/120) were treated with anti-VEGF injections; 21% (52/243) of eyes had VA of 20/30, of these 67% (35/52) were treated with anti-VEGF. For the remainder of the analysis, we considered eyes that received treatment with anti-VEGF as the “treatment” group and eyes that received no anti-VEGF treatment during and 3 months prior to the study inclusion period (i.e. March 2015) as the “observation” group.

Outcomes stratified by anti-VEGF treatment

On average, patients in the treatment group (119 patients, 146 eyes) were younger than those in the observation group (78 patients, 97 eyes) (61.3 ± 10.7 vs 66.6 ± 11.2 years, respectively; p < .001). Distribution of baseline VA between the treatment and observation groups was: 20/20 (28% vs. 28%), 20/25 (48% vs. 53%) and 20/30 (23% vs.18%). A significantly higher proportion of eyes in the treatment group had received prior treatments within the last 3 months vs. those in the observation group (42% vs. 8.2%, p < .001). These prior treatments included: intravitreal anti-VEGF injections (87% vs. 75%), PRP (16% vs. 0%), focal laser (3% vs 0%) and vitrectomy (0% vs. 25%). There were no differences in terms of sex, systemic co-morbidities, duration of DM, baseline logMAR VA and baseline CFT between the two groups (Table 2). Final logMAR VA was 0.2 ± 0.2 in the treatment group [Snellen 20/32] versus 0.1 ± 0.3 in the observation group [Snellen 20/25]; (p = .23). Distribution of VA at the final follow-up visit in the treatment and observation groups was: 20/25 or better (49% vs 67%) and 20/30 or worse (51% vs 33%), respectively. There were no significant differences with respect to lines of VA gained or lost (Figure 2) and in follow-up duration between the 2 groups (Table 3). While a significantly higher proportion of eyes in the treatment group experienced a decline in VA below 20/30 compared to those in the observation group (72% vs. 40%, p < .001), the mean number of visits at which a decline in VA was noted per eye per year was low for both groups (2.2 ± 2.7 for treatment group vs. 0.3 ± 0.5 for observation group, p < .001). A statistically significant improvement in anatomical outcome between baseline and final follow-up visits was observed in both groups, with mean CFT decreasing by 7.5% in the treatment group (from 332.6 ± 66.1 to 307.7 ± 44.6) and by 3.2% in the observation group (from 318.8 ± 40.5 to 307.9 ± 65.3) (both p < .001). No statistically significant differences were found in CFT between both groups at the final follow-up.

Figure 2.

Visual outcomes stratified by anti-vascular endothelial growth factor (VEGF) treatment.

Supplementary Figure 1 shows the distribution of visual outcomes in terms of lines of VA gained and lost when stratified by the number of injections received per eye per year. Final mean VA (approximately Snellen 20/32) and mean difference in VA from baseline (+0.1 log MAR) were similar when stratified by a number of injections received (p = .6 and p = .5, respectively) (Supplementary Figures 2 and 3).

Outcomes stratified by cataract surgery

Twenty-one eyes of 17 patients underwent cataract surgery during the study period. Baseline patient characteristics between the two groups are summarized in Supplementary Table 1A. Despite comparable baseline VA, final logMAR VA was relatively worse in eyes that underwent cataract surgery during the study period compared to those that did not (0.3 ± 0.5 [Snellen 20/40] vs 0.2 ± 0.2 [Snellen 20/32]; p = .10) (Supplementary Table 1B). Approximately 29% of eyes that underwent cataract surgery received anti-VEGF treatments compared to 64% of those that did not (p < .01). VA (0.1 ± 0.1, Snellen 20/25 vs. 0.3 ± 0.6, Snellen 20/40) and foveal thickness (median CFT 275 vs. 294 microns) outcomes were noted to be better for eyes that underwent cataract surgery and were being treated with anti-VEGF injections (6/21 eyes) compared to eyes that had no history of anti-VEGF treatments (15/21 eyes).

Factors associated with visual outcomes

Proportion of eyes with lines gained/lost

On multivariable analysis, factors that were associated with a greater odds of gaining at least 1 line of VA included worse baseline VA (odds ratio [OR] 1.4, 95% confidence interval [CI] 1.2–1.6; p < .001) (evaluated as continuous variable) and male sex (OR 3.3, 95% CI 1.2–8.6; p = .02). In contrast, eyes treated with anti-VEGF injections were less likely to gain at least 1 line of VA (OR 0.3, 95% CI 0.1–0.8; p = .01) and a higher proportion lost at least 1 line of VA (OR 2.1, 95% CI; 1.0–4.3; p = .04) compared to eyes that did not receive anti-VEGF injections. Additional factors that were associated with losing at least 1 line of VA included older age (OR 1.04, 95% CI 1.0–1.1; p = .03) (evaluated as a continuous variable) and presence of PDR (OR 2.4, 95% CI 1.0–5.7; p = .05). Presence of PDR (OR 14.6, 95% CI 1.3–15.8; p = .03) and obstructive sleep apnea (OSA) (OR 2.9, 95% CI 1.1–7.5; p = .03) were also associated with a higher likelihood of losing at least 2 lines of acuity (Supplementary Tables 2–4).

Final VA

We did not find any study variables that were significantly associated with the final VA on multivariable analysis (Supplementary Table 5).

Factors associated with anatomical outcomes

On multivariable analysis, eyes receiving anti-VEGF injections vs. those that did not (OR = 1.2, 95% CI 1.1–1.4; p = .002) and African Americans vs. Caucasians (OR = 1.2, 95% CI 1.0–1.4; p = .02) were more likely to achieve a final CFT of less than 300 microns. A greater proportion of eyes, with ci-DME and good VA, that were being treated with anti-VEGF agents also experienced at least a 10% decrease in CFT (OR = 1.2, 95% CI 1.1–1.4; p = .01) from baseline compared to eyes that did not receive any anti-VEGF treatment (Supplementary Tables 6–7).

Discussion

Among our study participants, more than 60% of the 243 eyes with ci-DME and VA of at least 20/30 were treated with anti-VEGF agents, at an average of 3.7 injections per eye per year. None of the patients received focal laser alone as the treatment for ci-DME. Although most patients with ci-DME and good VA were being treated with anti-VEGF agents, after 1.7 years of follow-up, there were no differences in final VA and CFT between patients who did and did not receive treatment.

Our study corroborates and further extends on findings from the OBTAIN study that found most eyes with ci-DME and very good VA to maintain their vision at 1 year of follow-up, irrespective of treatment status.¹⁰ However, some important differences exist between our study and the OBTAIN study. First, imaging in the OBTAIN study was performed using several different SD-OCT devices and ci-DME was defined as a CRT of >250 μm. In comparison, ci-DME in our study was defined using the gender-based thickness thresholds used by DRCR network’s Protocol T for Heidelberg SD-OCT imaging.¹² Moreover, our study provides additional information on factors associated with visual and anatomical outcomes in patients with ci-DME and good vision as well as the burden of visits to the physician’s office when managing these patients. Patients in the injection group were seen at an average frequency of nine office visits per year and received injections at less than half the visits (4 per year) in either eye. Even those who were not treated had an office visit of 4.8 times per year. This difference is even more marked in Protocol V where the number of visits in the observation group were 6.6 and 6.9 but the mean number of injections were 1.4 and 1.7 in years 1 and 2, respectively.⁵ These data suggest that perhaps there can be fewer office visits for some patients who do not need frequent treatments and are unlikely to have a decline in VA. In our study, the mean number of visits at which a decline in VA below 20/30 was noted per eye per year was low for both groups (2.2 ± 2.7 for treatment group vs. 0.3 ± 0.5 for observation group, p < .001). In our group, we identified that older age, presence of PDR and sleep apnea were associated with the highest risk of VA loss. Patients with these risk factors may be the ones who need more frequent follow-up and more aggressive anti-VEGF treatment. It may also be an opportunity to treat modifiable risk factors such as OSA and see if the treatment of OSA may decrease injection burden. Future studies are needed to confirm these findings.

We found that a larger proportion of eyes in the treatment group (42% vs. 8.3%, p < .001) had received treatments prior to being included in the study and that a significantly higher proportion of eyes in the treatment group experienced a decline in VA below 20/30 at some point during follow-up compared to those in the observation group (72% vs. 40%, p < .001). These two factors likely explain the criteria used by clinicians to treat were: 1) continuing to treat eyes that were previously receiving treatment to maintain their VA and 2) treating eyes when they experienced a decline in VA. When compared to the recently released results of Protocol V⁵ it seems that most clinicians were following the treatment strategy described for the observation group in that trial, in which anti-VEGF treatment was administered when there was a decline in VA.

Our study also gives the opportunity to evaluate the outcomes of 40% of patients in the observation group who had a decline in VA below 20/30 at some point but did not receive any treatments. The final VA of those who did not experience a decline in VA in the observation group was not significantly different from those in the observation group who did experience a decline (Snellen 20/25 vs 20/32) or compared to those in the treatment group (final Snellen VA: 20/32). However, this lack of difference could be due to our relatively small sample size or follow-up duration. Given that persistent edema is thought to cause photoreceptor damage and foveal atrophy and has been associated with poor long-term outcomes in several studies,^14,15 it is important to recognize a decline in VA and consider treatment at that point. It would be interesting to see how practice patterns change in light of the results from Protocol V, which used a decrease of ≥10 letters at 1 visit or 5–9 letters at two consecutive visits as criteria for anti-VEGF treatment in this group of patients.⁵

Although statistically significant, we observed that the mean decrease in CFT (−17 microns) in our patient population was much lower compared to other studies evaluating “real-world” outcomes of treating DME patients with anti-VEGF agents (mean decrease of at least 100 microns).^16–18 This finding can be attributable to the “ceiling effect” in that eyes with greater CFT at baseline are more likely to realize greater reductions in thickness over time.¹⁹ The mean baseline CFT in the previously mentioned studies ranged between 430 and 490 microns, with a SD of at least 100 microns. In contrast, the mean CFT in our study population was 327.1 ± 57.6 microns. Our finding that race seemingly affects anatomical outcomes is similar to that previously reported by Bressler et al.²⁰ who noted African American participants had greater reductions in central subfield thickness compared with the eyes of Caucasian participants.

Similar to prior studies^21,22 we found that eyes that underwent cataract surgery but did not receive anti-VEGF treatment had poor visual outcomes, confirming that this may be another indication to treat patients with ci-DME. Twenty-one eyes in our study underwent cataract surgery and visual and anatomic outcomes were noted to be better for eyes that underwent cataract surgery and were being treated with anti-VEGF injections (6/21 eyes, final VA of 20/25 and median CFT of 275 microns) compared to eyes that had no history of anti-VEGF treatments (15/21 eyes, final VA 20/40 and median CFT of 294 microns).

DME detection can be achieved using several different methods, including stereoscopic fundus photography or clinical examination with dilated biomicroscopy. Diffuse thickening or thickening in the form of retinal cysts as the initial sign of DME may go unnoticed with these examination methods, and OCT imaging is currently recognized as the reference standard for DME diagnosis and assessment.²³ However, it is important that clinicians as well as researchers remain aware of the differences in foveal thickness thresholds used for defining ci-DME between OCT devices. For instance, the DRCR Network’s defined criteria for diagnosing ci-DME on time-domain OCT imaging is a CFT of ≥250 microns. In contrast, their criteria for ci-DME detection on SD-OCT imaging are: a) ≥290 μm in women and ≥305 μm in men (Zeiss Cirrus) and, b) ≥305 μm in women and ≥320 in men (Heidelberg Spectralis). While Protocol V followed the DRCR Network-defined thickness criteria,⁵ the OBTAIN study in comparison defined ci-DME based on a CFT of ≥250 μm on any SD-OCT imaging device. Thus, to improve uniformity and comparability, it is important that future studies pay particular attention to the thickness values for OCT-based DME diagnosis.

Limitations of our study include its retrospective nature and the relatively small sample size. The strengths include a more detailed analysis of patient variables, allowing us to identify factors associated with visual and anatomic outcomes. Additional studies with larger patient cohorts and longer follow-ups are needed to establish which patients with ci-DME and good VA are at risk of VA loss and therefore need closer follow-up and treatments with anti-VEGF agents. Additionally, none of the patients in our study were treated with steroid (dexamethasone [DEX] implants). Therefore, we are unable to comment on the visual and anatomical outcomes among patients with ci-DME and good vision treated with DEX implants.

In summary, our study suggests that not all patients with ci-DME and good VA need anti-VEGF treatments. After an average of 1.7 years of follow-up, there was no statistically significant difference in final VA or retinal thickness irrespective of whether patients received or did not receive treatment with anti-VEGF agents. Being able to identify those patients who are likely to have a decline in VA may aid in the personalization of follow-up schedules for individual patients, potentially decreasing visit burden for some.