Clinical and imaging biomarkers of response to intravitreal dexamethasone implant in eyes with non-infectious uveitic macular oedema

Abstract

Objective

To investigate clinical and spectral-domain optical coherence tomography (SD-OCT) biomarkers correlating with pre-injection visual acuity (VA), post-injection VA, and the likelihood of macular oedema (MO) regression following dexamethasone (DEX) implant injection in non-infectious uveitic (NIU) patients.

Methods

Patient data from Uveitis Services in Milan, Paris, and Berlin were analysed. Eligible participants were NIU patients aged >18 years with MO as the primary indication for DEX treatment. SD-OCT scans and clinical data were collected at the time of DEX injection (pre-injection visit) and after 3 months (post-injection visit). Multivariable regression models, adjusted for pre-injection VA and lens status, were employed to explore associations. MO regression was defined as the absence of intraretinal/subretinal fluid at the post-injection visit.

Results

Our analysis comprised data from 173 DEX treatments, encompassing 103 eyes from 80 patients, with 38 eyes (37%) receiving repeated DEX injections. The absence of the ellipsoid zone (EZ) layer and disorganisation of the inner retinal layers (DRIL) were associated with worse pre- (+0.19 LogMAR, 95% CI 0.01–0.38, p = 0.06, and +0.10 LogMAR, 95% CI 0.02–0.21, p = 0.01) and post-injection VA (+0.33 LogMAR, 95% CI 0.08–0.57, p = 0.01, and +0.17 LogMAR, 95% CI 0.01–0.32, p = 0.04). EZ disruption and DRIL increased significantly (p = 0.01 and p = 0.04), and the chance of gaining ≥5 letters declined in eyes undergoing repeated DEX (p = 0.002). The rate of MO regression after each DEX was 67%. Prolonged MO duration (OR = 0.75/each year, p = 0.02) was associated with reduced likelihood of MO regression. Subretinal fluid was associated with higher rate of MO regression (OR = 6.09, p = 0.01).

Conclusion

Integrity of the inner and outer retina is associated with better visual response to DEX. Long-standing or recurrent MO is associated with less chance of both visual and anatomic response. Timely treatment is necessary to maximise the outcomes of MO in NIU patients.

Introduction

Macular oedema (MO) is the leading cause of visual impairment in cases of intermediate and posterior uveitis [1]. Its pathogenesis is multifactorial and includes inflammation and ischaemia, which converge to disrupt the integrity of the blood-retinal barrier (BRB). This disruption leads to the accumulation of intra- or extracellular fluid within the macular region. If not adequately treated, persistent MO may cause irreversible inner retina and photoreceptor atrophy, culminating in permanent vision impairment [2].

Both local and systemic corticosteroids, as well as immunosuppressive therapy (IMT), have emerged as efficacious interventions for managing MO secondary to non-infectious uveitis (NIU) [3]. The landmark HURON study demonstrated a six-fold increment in the probability of visual improvement among NIU MO patients treated with a 0.7-mg bioerodable intravitreal dexamethasone implant (DEX; Ozurdex; Abbvie, Chicago, Illinois, USA) compared to sham treatment [4]. However, subsequent real-world studies showed that functional and anatomic responses to DEX are not always favourable [5], with nearly 50% of cases manifesting persistent MO [6–10]. There is a knowledge gap concerning the clinical and morphologic factors determinants that underlie the variable responsiveness of NIU MO patients to intravitreal DEX therapy.

Optical coherence tomography (OCT) serves as an invaluable modality for quantifying central macular thickness (CRT) and assessing retinal layers’ integrity [11]. Previous studies showed only a moderate correlation between visual acuity (VA) and CRT [12]; thus, alternative biomarkers explaining VA variability in NIU MO patients must be identified. Ellipsoid zone (EZ) damage, hyperreflective foci (HRF), disorganisation of retinal inner layers (DRIL), cystoid spaces in the outer (ONL) or inner nuclear layer (INL), subretinal fluid (SF), and presence of vitreoretinal abnormalities, such as an epiretinal membrane (ERM), have been evaluated in other macular diseases [13]. The comprehensive assessment of these parameters in the context of NIU MO is lacking.

This study investigates the clinical and spectral-domain OCT (SD-OCT) features correlating with pre-injection VA, post-injection VA, and the likelihood of MO regression after DEX in patients with NIU MO.

Methods

This was a retrospective, multicentre, noncomparative study on patients who attended the Uveitis Service of the Department of Ophthalmology, San Raffaele Hospital (Milan, Italy), the Department of Ophthalmology of Pitié Salpêtrière University Hospital (Paris, France), and the Department of Ophthalmology of Charité-Universitätsmedizin (Berlin, Germany). The Institutional Review Board (IRB) approved the study design at each centre. Patient’s written consent is exempted by the IRB due to the retrospective design of the study. All study procedures conformed with the tenets of the Health Insurance Portability and Accountability Act and the Declaration of Helsinki for research involving human subjects.

The initial pool of patients was retrieved from all those receiving the 0.7 mg DEX for NIU between October 2011 and July 2022. We included NIU patients older than 18 years, with VA better than counting fingers, and both clinical data and SD-OCT available for review at the time of DEX. All patients had MO as the main indication for treatment. We excluded patients previously treated with DEX and those with MO secondary to infectious uveitis (except for uveitis secondary to tuberculosis hypersensitivity), corneal or lens opacity affecting vision or impeding imaging assessment. We also excluded patients with MO secondary to other retinal diseases (e.g., retinal vein occlusion [RVO], diabetic macular oedema [DMO]), or vision-threatening comorbidities. In instances where eligible, both eyes of the same patient were included.

Patients’ data were collected from the visit immediately before the first DEX (pre-injection visit) with a maximum time gap of 6 weeks between the visit and the DEX procedure. Demographics, medical history, uveitis diagnosis, decimal VA values, and anterior- and posterior-segment findings were gathered. Patients who had undergone any form of local (periocular or intraocular) treatment prior to the first DEX were categorised as “non-naïve.” All patients were managed by experienced uveitis specialists, who made the diagnosis based on a combination of clinical, laboratory, and imaging findings and undertook treatment decisions. As per clinical practice shared between centres, DEX was administered in cases of uveitic MO unresponsive to systemic treatment, as a bridging therapy shifting to systemic IMT, or in patients with contraindications to systemic medications and no history of steroid-induced ocular hypertension or uncontrolled glaucoma.

Follow-up visits were scheduled within a 3-month timeframe after each DEX administration (post-injection visit), during which VA values were collected. Co-registered SD-OCT scans were used to assess MO regression, defined as the absence of intraretinal or subretinal fluid at that visit. Repeated DEX was permitted with a minimum gap of 4 months between successive treatments in case of MO persistence or recurrence, and data from repeated DEX administrations were collected at similar timepoints.

Optical coherence tomography analysis

All SD-OCT scans were acquired using a digital confocal scanning laser ophthalmoscope device (Spectralis HRA, Heidelberg Engineering; Heidelberg, Germany). A raster SD-OCT pattern centred on the fovea was obtained using an enhanced depth imaging technique.

The horizontal SD-OCT scan passing through the fovea acquired during the pre-injection visit was analysed for the presence of: (1) ERM; (2) DRIL, defined as the impossibility of distinguishing inner retinal layers boundaries in the central 1-millimetre area [14]; (3) HRF, defined as discrete and well-circumscribed dots of identical reflectivity as the RPE band [15]; (4) EZ disruption under the fovea; (4) SF; (5) cysts within the ONL; (6) cysts within the INL (Fig. 1). Quantitative parameters included CRT (μm) in the 1-mm foveal area, subfoveal choroidal thickness (CT, μm), and the vertical diameter of the largest intraretinal cyst (μm).

Fig. 1. Qualitative assessment of spectral-domain optical coherence tomography (SD-OCT).

A Horizontal SD-OCT scan acquired before dexamethasone implant injection (DEX) in a patient with idiopathic panuveitis. The scan shows intraretinal fluid, with cystoid spaces in the outer (solid arrow) and inner nuclear layer (dashed arrow), and subretinal fluid (asterisk). B Horizontal SD-OCT scan acquired after DEX acquired with a follow-up mode. The scan shows complete regression of macular oedema. The ellipsoid zone (EZ) under the fovea is preserved, as shown in the enlarged panel. C Horizontal SD-OCT scan acquired before DEX in a patient with Behçet disease. The scan shows intraretinal fluid, with cystoid spaces in the outer and inner nuclear layer. There is disorganisation of retinal inner layers (DRIL) in the central 1-millimetre area. The EZ band is absent under the fovea. D Horizontal SD-OCT scan acquired after DEX acquired with a follow-up mode. The scan shows regression of macular oedema but persistence of DRIL. The EZ band is absent under the fovea.

Instances where features were indiscernible due to artefacts or poor visualisation were treated as missing data and left blank. A trained grader at each centre conducted all measurements while remaining masked to the demographic and clinical attributes of the study eyes. Any discrepancies among the graders were resolved by a senior uveitis specialist (MVC, ST, or DP).

Statistical analysis

All statistical analyses were performed utilising the open-source programming language R (version 1.2.5033). Inferential statistics yielding a p < 0.10 were deemed to hold clinical significance.

The sample size was calculated using the wp.mrt2arm function from the WebPower R package [16]. This function provided the sample size for a two-level hierarchical linear model, where the main covariate was the first level, and the subject/eye clustering was the second level. The effect size utilised for the calculation was drawn from the post hoc analysis of the VISUAL-1 trial, which examined the correlation between different SD-OCT parameters and VA in NIU MO. This analysis identified a mean difference of 0.17 LogMAR (95% confidence interval [CI] 0.07–0.27) between eyes with foveal DRIL and those without DRIL [17]. The between-subjects variance (σ = 0.0027) and the within-subjects variance (σ = 0.13374) were calculated in a sample of randomly chosen eyes (n = 50) from our data, with a linear mixed regression having pre-injection VA as the dependent variable and DRIL as the predictor. A sample size of 100 subjects, each contributing with 1.60 eyes, was powered enough to detect a difference of 0.17 LogMAR with a power of 0.8 and significance of 0.05 (assuming an ICC of 0.1).

Descriptive statistics are presented as means, medians, or proportions; group comparisons were undertaken with linear or logistic regression models.

Firstly, we explored the associations between pre-injection VA and clinical and morphologic biomarkers. We used a multiple linear regression model with LogMAR VA values as the dependent variable and demographic or SD-OCT features as covariates. Models were adjusted for the lens status (phakic vs. pseudophakic). Covariates were selected with a parsimonious approach, using a least absolute shrinkage and selection operator regression (LASSO) [18]. To address multicollinearity, the variance inflation factor of each covariate was computed. To account for bilateral treatment and repeated injections in certain eyes, nested random factors encompassing patient and eye identification numbers were included. Effect sizes were interpreted while holding all other independent variables constant. For missing data in regression predictors, multiple imputations were employed.

Secondly, we compared pre- and post-injection VA values. We investigated post-injection VA correlations with a multivariable regression model adjusted for pre-injection VA values, lens status (phakic vs. pseudophakic), and systemic treatment. We did not divide patients undergoing their first DEX from those receiving repeated DEX to avoid excluding a substantial portion of our data. To address potential confounding effects arising from the variability in the number of injections administered, we introduced nested random factors based on patient and eye identification numbers, enabling a clear statistical differentiation between initial and repeated injections. Regression estimates and their 95% CI were reported.

Finally, we examined the factors associated with MO regression after DEX using multivariable logistic regression models. The exponential logit of each factor was interpreted as the odds ratio (OR) concerning the binary outcome of MO regression vs. persistence.

Results

The study encompassed a total of 103 eyes from 80 patients affected by NIU MO. Each eye underwent a median of 2 (interquartile range [IQR] 1–3) DEX administrations. Specifically, 65 eyes (63%) received a single DEX treatment, while 38 eyes (37%) underwent two or more implantations, up to a maximum of 8. In total, data from 173 pre- and post-injection DEX sessions were subjected to analysis. The median duration of follow-up extended to 13 months (IQR 5–24) after the first DEX.

Patients’ demographics

Demographic and clinical characteristics of the study patients collected at first DEX are shown in Table 1.

Table 1.

Demographic and clinical characteristics of the study patients at the time of the first DEX injection.

	Overall (N = 80)	Mean age (years)
Age (years)
Mean (SD)	59.4 (15.1)
Median [Min, Max]	61.0 [26.0, 95.0]
Gender
Female	41 (51%)
Male	39 (49%)
Anatomic location of uveitisa
Anterior uveitis	4 (5%)
Intermediate uveitis	25 (31%)
Posterior uveitis	20 (25%)
Panuveitis	31 (39%)
Aetiology of uveitis
JIA	1 (1.3%)	33.0 (NA)
Behçet	1 (1.3%)	41.0 (NA)
Birdshot chorioretinopathy	6 (7.4%)	54.7 (11.5)
Drug-induced uveitis	1 (1.3%)	71.0 (NA)
Crohn’s disease	1 (1.3%)	58.0 (NA)
HLA-B27	2 (2.5%)	45.0 (5.66)
Idiopathic	37 (46.2%)	59.7 (15.8)
Psoriasis	2 (2.5%)	60.5 (2.12)
Sarcoidosis	21 (26.2%)	63.2 (16.2)
Sympathetic ophthalmia	2 (2.5%)	70.5 (19.1)
TB-hypersensitivity uveitis	4 (5.0%)	55.8 (9.84)
Vogt-Koyanagi-Harada	2 (2.5%)	53.5 (6.36)
Duration of uveitis (months)
Mean (SD)	15.9 (11.0)
Median [Min, Max]	12.0 [0, 63]
Systemic IMT
No	36 (45%)
Yes	44 (55%)

N number, % percentage, SD standard deviation, Min minimum, Max maximum, JIA juvenile idiopathic arthritis, HLA human leucocyte antigen, TB tuberculosis, IMT immunosuppressive therapy.

^aThe location was given for the worse eye in patients with bilateral disease.

All patients shared Caucasian ancestry. The predominant uveitis subtype was panuveitis, observed in 31 patients (39%), with idiopathic uveitis being the most common diagnosis in 37 patients (46%), closely followed by sarcoid uveitis in 21 patients (26%). One patient had persistent uveitis and a positive QuantiFERON-gold test yet lacking active Tuberculosis manifestation. None had previously undergone vitrectomy.

Of the total patient cohort, 36 patients (45%) were undergoing systemic IMT, while 5 patients commenced systemic medication subsequent to their initial DEX intervention. Systemic treatments included oral corticosteroids (10 patients), methotrexate (9 patients), adalimumab (5 patients), mycophenolate mofetil (2 patients), or a combination of these agents (10 patients).

Twenty-seven eyes (26%) were pseudophakic. Sixty-four eyes (62%) were treatment-naïve, implying they had not undergone any prior therapeutic intervention. The remaining eyes had previously received periocular or intravitreal triamcinolone treatment for MO before embarking on the DEX regimen.

Clinical and SD-OCT characteristics associated with pre-injection VA

Supplementary Table 1 presents a comprehensive breakdown of SD-OCT characteristics, distinguishing between those assessed during the initial DEX treatment and those recorded in subsequent DEX sessions. NIU eyes had higher CRT (p = 0.05) and presented a higher rate of SF and HRF (p = 0.002 and p = 0.07) at their first DEX compared to subsequent DEX treatments. The proportion of eyes with disrupted or absent EZ considerably increased from the second treatment onwards (p = 0.01). DRIL, as well, was observed more frequently in eyes receiving multiple DEX treatments (p = 0.04). The prevalence of ERM remained relatively stable throughout the follow-up period (p = 0.7).

The initial DEX treatment was associated with worse VA compared to subsequent treatments (0.48 ± 0.34 vs. 0.44 ± 0.28 LogMAR, p = 0.001). Pre-injection VA displayed an association with CRT (p = 0.008). Notably, the correlation between VA and CRT demonstrated a more optimal fit when modelled exponentially rather than linearly (Fig. 2A).

Fig. 2. Correlation between pre- and post-injection visual acuity (VA).

A Scatterplot showing the association between pre-injection VA (expressed as LogMAR) and central retinal thickness (CRT). The relationship between VA and CRT was best explained by an exponential fit (blue line) rather than a linear one (red line). The grey shadow shows the confidence interval of the interpolating line. B Scatterplot showing the association between pre-injection and post-injection VA (expressed as LogMAR), according to the presence of disorganisation of retinal inner layers (DRIL) in the central 1-millimetre area. Eyes with DRIL had worse pre-injection and post-injection VA. No interaction is seen. The grey shadow shows the confidence interval of the interpolating line. C Scatterplot showing the association between pre-injection and post-injection VA (expressed as LogMAR), according to the presence of ellipsoid zone (EZ) under the fovea. Eyes with EZ had worse pre-injection and post-injection VA. No interaction is seen. The grey shadow shows the confidence interval of the interpolating line. D Scatterplot showing the association between pre-injection and post-injection VA (expressed as LogMAR), according to the presence of epiretinal membrane (ERM). A possible interaction is noticeable, suggesting that the effect of ERM is negligible in eyes with good pre-injection VA and greater in eyes with worse pre-injection VA. The grey shadow shows the confidence interval of the interpolating line.

Employing a multiple linear regression analysis, we discerned that eyes with absent EZ layer and those with DRIL exhibited worse pre-injection VA, with estimated differences of +0.19 (p = 0.04) and +0.10 (p = 0.01) LogMAR, respectively, relative to eyes with intact EZ and no DRIL. Additionally, patients not undergoing systemic IMT for NIU displayed poorer pre-injection VA values (+0.11 LogMAR, p = 0.07) (Supplementary Table 2). The model accounted for 23% of the variance in pre-injection VA values (marginal R² = 0.23).

Clinical and SD-OCT characteristics associated with post-injection VA

Mean post-injection VA was 0.34 ± 0.27 LogMAR. A significant association emerged between post-injection VA and pre-injection VA, with an increment of 0.42 LogMAR for every unitary increase in pre-injection VA values (p < 0.001).

Following each DEX intervention, a notable improvement in VA was observed, with an average improvement of 0.13 ± 0.25 LogMAR (p = 0.003). In terms of letter gains, the proportions of 5-, 10-, and 15-letter improvements amounted to 21 (12%), 12 (7%), and 38 (22%) across the 173 DEX instances subjected to analysis. Importantly, the likelihood of achieving a minimum of 5-letter enhancement was higher after the initial DEX treatment in comparison to subsequent DEX sessions (50/101 [52%] vs. 19/72 [26%], p = 0.002).

The multiple linear regression analysis disclosed significant associations between post-injection VA and specific SD-OCT findings, namely the presence of DRIL, EZ disruption, and ERM. Specifically, post-injection VA was on average 0.17 (p = 0.04), 0.33 (p = 0.01), and 0.12 (p = 0.04) LogMAR worse than eyes with no DRIL, EZ disruption, or ERM, respectively (Fig. 2B, C). Notably, a discernible trend emerged where the impact of ERM was more pronounced in eyes presenting poorer pre-injection VA (Fig. 2D and Supplementary Table 3). The model elucidated 50% of the variability in post-injection VA values (marginal R² = 0.50).

Clinical and SD-OCT characteristics associated with MO regression

Three-month follow-up SD-OCT was available after 134 DEX. Of these, 90 cases exhibited MO regression, accounting for a substantial rate of 67%. Eyes displaying MO regression showcased notably improved post-injection VA when contrasted with eyes harbouring persistent MO (0.29 ± 0.27 vs. 0.49 ± 0.48 LogMAR). The difference was −0.15 LogMAR (p = 0.002), indicating a tangible clinical impact.

Results from single-variable comparisons are shown in Supplementary Table 4. The multiple logistic regression analysis showed that advancing age (OR = 0.95 for each decade, p = 0.049) and a prolonged duration of MO (OR = 0.75 for each year, p = 0.02) were associated with a lower chance of MO regression after DEX. A higher pre-injection CRT (OR = 0.58 for each 100 μm, p = 0.03) and the presence of HRF (OR = 0.12, p = 0.004) were also independent risk factors of persistent MO. In a notable contrast, the presence of SRF was associated with higher rates of MO regression (OR = 6.09, p = 0.01) (Table 2).

Table 2.

Factors associated with macular oedema (MO) regression 3 months after DEX.

Characteristic	OR	95% CI	p value
Age (for 10 years)	0.95	0.90, 0.99	0.049
Duration of MO (for 1 year)	0.75	0.53, 0.91	0.02
CRT (100 μm)	0.58	0.36, 0.94	0.03
Subretinal fluid	6.09	1.69, 22.0	0.01
Hyperreflective foci	0.12	0.03, 0.50	0.004

OR odds ratio, CI confidence interval, CRT central retinal thickness.

Discussion

This study assessed the associations between pre-injection VA, post-injection VA, and clinical and SD-OCT characteristics of eyes affected by NIU MO treated with intravitreal DEX. We identified EZ damage, the presence of DRIL, and increased macular thickness as biomarkers signifying poorer pre-injection VA. Furthermore, EZ disruption and DRIL emerged as predictive factors associated with unsatisfactory visual outcomes, their prevalence escalating notably in cases necessitating multiple treatment interventions. Additionally, advanced age, prolonged MO duration, higher CRT, and the presence of HRF were identified as risk factors linked to persistent MO following each DEX treatment. This analysis may assist general ophthalmologists and uveitis specialists in forecasting the prognosis of NIU patients undergoing intravitreal treatment for MO.

According to randomised clinical trials and real-life data, the degree of visual improvement in NIU eyes undergoing treatment for MO exhibits a substantial range of variability. The HURON trial, for instance, reported a 42% rate of 15-letter gainers at 3 months and 38% at 6 months following intravitreal DEX treatment [4]. Furthermore, real-world investigations have revealed that up to 40% of treated eyes fail to achieve an improvement of three Early Treatment Diabetic Retinopathy Study (ETDRS) lines after DEX administration [5, 19, 20]. While the patients enrolled in our study exhibited an overall increase in visual acuity post-DEX, the extent of these gains was heterogeneous, with 60% of eyes displaying improvement of less than five letters. It is noteworthy that there remains a scarcity of comprehensive data elucidating morpho-functional correlations within NIU MO eyes [21]. Existing studies probing the impact of single or repeated DEX interventions on NIU have predominantly been based on limited samples and only a select few have investigated the morphological predictors influencing both visual and anatomical outcomes [6, 8, 22, 23]. Within this context, our study stands as a distinctive contribution, harnessing data encompassing nearly 200 DEX treatments in NIU patients, who underwent serial SD-OCT imaging in the endeavour to pinpoint the factors influencing the response to DEX therapy.

Ciulla et al. investigated the relationship between VA and CRT in eyes with NIU MO [12]. The authors found a weak association between VA and CRT, with the latter explaining only a minimal fraction of the variability in VA scores [12]. Our findings, however, identified an exponential relationship between pre-injection VA and CRT values, signifying that both extremely low (e.g., <300 μm) and high (e.g., >600 μm) CRT values were associated with suboptimal VA. While elevated CRT values typically suggest aggravated MO, reduced values could potentially indicate macular atrophy—a common end-stage manifestation of exudative macular disorders [24]. Nevertheless, our investigation revealed that CRT values, irrespective of employing linear or quadratic fits, only accounted for less than 7% of the variance in pre-injection VA within our cohort. This observation underscores the existence of additional structural biomarkers contributing to the complex tapestry of VA outcomes.

The presence of DRIL and EZ disruption was associated with worse VA, both prior to and post-DEX intervention. DRIL, indicative of the loss of inner retinal layers’ laminar arrangement on SD-OCT scans, serves as a surrogate marker for irreparable damage to amacrine, bipolar, and horizontal cells [17]. Conversely, the integrity of the EZ band mirrors the health of photoreceptor outer segments. DRIL has been associated with inferior treatment response in eyes with DMO [25, 26], RVO [14, 27], and idiopathic ERM [28]. Moreover, studies have highlighted that the presence and extent of DRIL correlate with adverse visual outcomes in NIU patients managed with systemic IMT (Adalimumab) [17]. On the other hand, post hoc analyses of the PEACHTREE and AZALEA trials, investigating the efficacy of suprachoroidal triamcinolone acetonide in NIU, underscored the potent association of EZ status with baseline and post-treatment VA, explaining up to a noteworthy 25% of the total variation [29]. Our data robustly reaffirm DRIL and EZ disruption as independent, potent prognostic biomarkers with negative implications for VA within the context of real-world clinical practice.

A multicentre French study revealed a distinct trend wherein treatment-naïve patients at baseline (n = 6/22, 27%) exhibited a higher chance of experiencing visual improvement after DEX compared to those with prior treatment history (n = 8/46, 17%) [30]. This observation underscores the potential influence of long-standing MO history on DEX response. In our study, the duration of MO did not exhibit a direct correlation with VA, but a notable pattern emerged wherein the probability of achieving at least a 5-letter enhancement was greater following the initial DEX treatment as opposed to subsequent sessions. While regression towards the mean could contribute to this observation, there is also the possibility of patients developing reduced responsiveness (tachyphylaxis) to repeated DEX treatments. However, it’s important to note that previous studies have not definitively demonstrated a decrease in the effectiveness of repeated intravitreal steroids. Nonetheless, our hypothesis suggests an alternative explanation: the recurrence of MO might contribute to cumulative retinal damage over time [31]. This notion gains support from the fact that the prevalence of DRIL and EZ disruption increased with repeated DEX treatments. This observed trend could potentially indicate a progressive retinal deterioration associated with recurrent MO.

Furthermore, our investigation hinted at a potential inverse association between systemic IMT and pre-injection VA, suggesting compromised vision in subjects not subjected to systemic IMT. Prolonged inflammation and potential undertreatment are likely culprits behind the cumulative retinal damage in these cases.

The impact of DEX exhibited a diminished effect in eyes harbouring ERM. Epiretinal membranes are relatively common in NIU, with an estimated prevalence of 41% [32]. Eyes with ERM tend to have worse VA than their counterparts [32], and their capacity to achieve significant visual gains following surgical ERM removal remains limited [33]. Investigations by Munk et al. and Khurana and Porco, probing treatment outcomes in NIU-related MO with coexisting ERM, similarly underscored the restricted visual effects of intravitreal interventions in these cases [6, 34]. Our data align with these findings, showcasing a potential interplay between VA and ERM. Specifically, eyes with ERM and poor pre-injection VA exhibited more compromised post-injection vision than eyes devoid of ERM. We hypothesise ERM might exert additive deleterious impact on macular structures, a phenomenon not rectified through medical treatment.

A delayed anatomic response has been associated with suboptimal visual recovery following treatment with suprachoroidal triamcinolone acetonide [29]. In our study, the persistence of intraretinal or subretinal fluid emerged as a distinctive marker of compromised post-injection VA. Several studies have reported the rate of persistent or recurrent MO in NIU [5, 7, 8, 10]. A retrospective case series of 18 eyes treated with DEX found MO resolution in 72% of cases [6], findings closely aligned with our own observed rate of 67%. Our investigation illuminated key variables that influence MO regression. Advanced age, prolonged MO duration, and a higher pre-injection CRT collectively contributed to a diminished likelihood of MO regression. The presence of HRF also emerged as a risk factor for decreased MO responsiveness. While the definitive underpinning of HRF remains elusive, they may arise from lipoprotein extravasation due to compromised inner blood-retinal barrier integrity. Alternatively, HRF could signify microglial activation, potentially indicative of ongoing intraretinal inflammation [35]. Our analysis further unveiled that the presence of SF was associated with a remarkable six-fold increase in the probability of MO regression. This aligns with a US retrospective study of 101 eyes with uveitic MO [36]. SF tends to occur in MO of relatively shorter duration and potentially signifies intact connections between Müller cells and foveal cones [34]. Consequently, SF could potentially serve as a positive indicator of favourable responses to local or systemic treatments.

We recognise certain limitations inherent in the retrospective design and the presence of missing data within this study. It’s important to acknowledge that patients under the care of tertiary uveitis centres might exhibit anticipated outcomes skewed towards the more challenging end of the spectrum, potentially impacting the generalisability of our findings. Disparities in post-injection visit scheduling across different centres introduced heterogeneity in patients’ follow-up periods. In fact, post-injection SD-OCT scans were available in 75% of the cases. We cannot exclude data were not missing at random; patients not returning for follow-up SD-OCT could be those with either very robust or minimal responses to DEX.

Furthermore, our categorisation of persistent MO relied on the presence of intraretinal or subretinal fluid in follow-up SD-OCT scans. The timing of our post-injection assessments, set at 3 months, prevented the differentiation between eyes harbouring persistent fluid and those exhibiting early MO recurrence. Our regression analyses did not incorporate certain clinical variables, such as anterior segment inflammation, cataract grading, and vitritis severity—each potentially impacting VA. Consequently, a notable portion of the variance in VA remains unaccounted for within our models.

In conclusion, this study assessed the relationship between clinical and SD-OCT biomarkers and VA in NIU patients with MO treated with DEX. Our comprehensive analysis underscored the significance of both inner and outer retinal integrity in dictating favourable visual responses to treatment, irrespective of the extent of macular thickening. Furthermore, the cumulative impact of prolonged and recurrent MO on retinal health is undeniable, associating with reduced chance for both visual and anatomical improvement. We conclude that implementing local and systemic IMT is necessary to maximise the therapeutic outcomes of MO in NIU patients.

Summary

What was known before

• OCT allows a repeatable evaluation of central macular thickness and retinal layer’s integrity.

• There is only a moderate correlation between visual acuity and central retinal thickness.

• Various retinal biomarkers have been evaluated in other macular diseases, but not in uveitic macular oedema.

What this study adds

• Integrity of the inner and outer retina is associated with a better visual response to intraocular dexamethasone implants.

• Longstanding macular oedema is associated with less chance of both visual and anatomical response.

Author contributions

All the authors contributed to the conception or design of the work, the acquisition, analysis, and interpretation of data, drafting the work, and revising it critically for intellectual content. Each coauthor has seen and agrees with how his or her name is listed.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Financial disclosures: CG, PS, EM: no financial disclosures. MVC, ST, DP, AG, AR previously received travel grants from Allergan as International Retinal Panel members. FB consultant for: Allergan Inc (Irvine, California, USA), Bayer Schering-Pharma (Berlin, Germany), Hoffmann-La-Roche (Basel, Switzerland), Novartis (Basel, Switzerland), Sanofi-Aventis (Paris, France), Thrombogenics (Heverlee, Belgium), Zeiss (Dublin, USA), Boehringer-Ingelheim, Fidia Sooft, NTC Pharma, Sifi. AL consultant for: Allergan, Bayer health care, Beyeonics, ForSight Labs, Notal Vision, Novartis, Roche, WebMD, Syneos, Xbrane, Nanoretina, Ocuterra, Ripple Therapeutics, Annexon, MJHEvents, Iveric Bio, Biogen, Johnson & Johnson, Ophtimedrx, Ocuphire Pharma, Iqvia.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41433-023-02802-7.