This was a multicenter, retrospective, observational cohort study including consecutive patients diagnosed with chronic CSC who underwent verteporfin PDT at 2 tertiary retina clinics: Department of Ophthalmology, IRCCS San Raffaele Hospital (Milan, Italy) and the Città della Salute e della Scienza Hospital, University of Turin (Turin, Italy). The study adhered to the Declaration of Helsinki and received approval from the institutional review boards of both centers. All patients provided written informed consent for clinical imaging and data use.

PDT was performed using verteporfin (Visudyne, Novartis, Basel, Switzerland) administered intravenously at either a half-dose (3 mg/m²) or half-fluence (25 J/cm²) regimen according to the treating physician’s preference. The laser wavelength was 689 nm, with a light intensity of 600 mW/cm². The PDT spot was positioned according to the area of choroidal hyperpermeability identified on mid- to late-phase ICGA (co-registered with the leakage site on FA), and not according to a predefined retinal location. Because all included eyes presented with active subfoveal SRF at the time of PDT, the treatment spot encompassed the subfoveal region in the vast majority of cases.

Information on whether each PDT session was performed using a half-dose or half-fluence regimen was not consistently available across all records and therefore was not included as a covariate in the primary regression models. Both regimens were applied according to standard clinical practice at the participating centers, and neither center reported a protocol-based preference.

Demographic and clinical data were collected for each patient, including the duration of symptoms for the episode leading to PDT, defined as the estimated time since the patient first noticed visual symptoms consistent with the current foveal detachment. This variable reflects the chronicity of the ongoing exudative episode, distinct from the overall disease duration. The date of first CSC diagnosis was also recorded for the study eye, indicating the initial identification of CSC in that eye, regardless of whether previous episodes had resolved. Information was also collected on previous or concomitant treatments for CSC, including oral eplerenone or acetazolamide, subthreshold micropulse laser, focal macular laser, and intravitreal anti-vascular endothelial growth factor (anti-VEGF) injections administered in the absence of MNV. BCVA was measured using standardized decimal charts and converted to logMAR (logMAR) units for statistical analysis; decimal values are reported once in the Results section for interpretative clarity.

All eyes underwent spectral-domain optical coherence tomography (SD-OCT; Spectralis HRA + OCT, Heidelberg Engineering, Heidelberg, Germany) at baseline (on the day of PDT or within 2 weeks prior) and at all subsequent follow-up visits documented in the medical record. Follow-up intervals were determined by the treating clinician based on individual clinical needs, and patients were advised to return for unscheduled visits in case of any new or worsening visual symptoms. A standardized imaging protocol was used across both centers. Each examination included 6 high-resolution radial B-scans centered on the fovea, providing detailed visualization of foveal microstructure and outer retinal layers, and a dense raster scan (49 B-scans) centered on the fovea for comprehensive assessment of subretinal and intraretinal fluid and other morphological features. The 6-radial-scan protocol was deliberately selected for this study because each radial B-scan passes through the foveal center, providing 6 independent foveocentric sampling lines at different angular orientations. For a grading system focused on the central 1 mm area, as the present one is, this yields high foveal sampling density. All B-scans were acquired in the high-resolution acquisition mode with automatic real-time (ART) frame averaging set to 25 frames per B-scan, with active eye-tracking. A prespecified scan quality threshold was applied: a foveal B-scan was considered acceptable for grading only if it had a Spectralis Quality score ≥20 dB and provided clear visualization of all retinal and choroidal layers relevant to grading (external limiting membrane, outer-retinal bands, retinal pigment epithelium, and sclerochoroidal interface) across the central 1 mm. In accordance with standard clinical practice at both participating centers, scans that did not meet these requirements were repeated within the same imaging session until adequate quality was achieved; no eyes were excluded from the final analysis on the basis of inadequate scan quality. All scans were acquired using the same device model and identical acquisition parameters at both centers to ensure intersite comparability.

All baseline and follow-up scans in the full cohort (n = 139 eyes) were graded jointly by 2 retina specialists (A.B. and G.N.), by consensus, with adjudication by a senior author (E.B.) for any cases not converging during the consensus process. To formally quantify the reproducibility of the staging system, a separate subset of 50 eyes was additionally graded independently by the same 2 graders, without recourse to consensus; this independent subset was used exclusively to compute overall and stage-specific intergrader agreement metrics, including unweighted and quadratic-weighted Cohen’s κ (with 95% bootstrap CIs). Qualitative assessments were based on all 6 high-resolution radial B-scans and the dense macular scans. Quantitative measurements were specifically obtained from the horizontal (0°-180°) foveal B-scan included in the radial scan set, ensuring consistency across eyes and visits.

• 1.

  SRF: defined as hyporeflective spaces located between the outer boundary of the ELM and the inner boundary of the RPE and recorded as present or absent within the foveal region, defined as the central 1 mm zone centered on the fovea.

• 2.

  Intraretinal fluid (IRF): defined as hyporeflective cystic spaces located within the neuroretina, between the outer boundary of the internal limiting membrane (ILM) and the inner boundary of the ELM. IRF was recorded as present regardless of whether it involved the foveal region or was confined to extrafoveal areas.

• 3.

  Central subfield thickness (CST): measured on the horizontal B-scan as the distance between the ILM and the outer border of the RPE. In eyes without SRF, this measurement included only the neurosensory retina (possibly containing IRF if present); in eyes with foveal SRF, the measurement encompassed both the retina and the SRF space. This composite metric was used rather than separating retinal and SRF thickness because it corresponds to the standard CST value automatically provided by OCT devices and is more readily interpretable in clinical practice.

• 4.

  Subfoveal choroidal thickness (SCT): defined as the vertical distance between the outer border of the RPE and the sclerochoroidal interface, measured perpendicularly beneath the foveal center.

To assess outer retinal integrity, a morphologic staging system of the photoreceptor layer was developed based on the continuity, thickness, and reflectivity of the ellipsoid zone (EZ) and interdigitation zone (IZ) within the central 1 mm centered on the fovea. Grading was performed exclusively within the central 1 mm area centered on the foveal center. For each eye and eligible visit, the most advanced (worst) stage identified within the central 1 mm area centered on the foveal center was recorded as the representative grade. Photoreceptor alterations located outside the central 1 mm, which may be present in chronic CSC and may reflect previous, nonsubfoveal detachment episodes, were not included in the staging process and did not contribute to the assigned stage. The choice to restrict grading to the foveal 1 mm zone was motivated by the anatomical alignment between this region and the primary functional outcome of the study (BCVA), which is determined by the integrity of the foveal photoreceptors, and by the foveocentric sampling density provided by the 6-radial-scan acquisition protocol.

Photoreceptor morphology was categorized into 6 progressive stages ( Figure 1 ):

• 1.

  Stage 1A–Normal architecture:

  The EZ and IZ are separated and clearly distinguishable, with preserved reflectivity and normal layer thickness.

• 2.

  Stage 1B–Early fusion:

  The EZ and IZ bands are no longer individually resolvable as 2 separate hyperreflective lines, while the overall reflectivity and photoreceptor thickness remain preserved. The loss of distinct EZ-IZ separation is itself the defining feature of this stage, distinguishing it from Stage 1A.

• 3.

  Stage 2–Reactive thickening:

  The photoreceptor layer overlying the foveal subretinal fluid shows diffuse, homogeneous thickening with preserved reflectivity and no focal disruption, judged qualitatively as appreciably thicker than the corresponding photoreceptor layer of the immediately adjacent attached parafoveal retina of the same B-scan, after accounting for the mild physiological foveo-parafoveal thickness gradient.

• 4.

  Stage 3–Patchy EZ irregularity:

  The EZ shows irregular reflectivity with small focal incisions or notches above the ELM, producing a moth-eaten appearance without true loss of layer continuity; overall photoreceptor thickness is preserved or mildly increased.

• 5.

  Stage 4–Structural discontinuity:

  The photoreceptor layer becomes thinned, with at least one true zone of full-thickness EZ interruption extending to an intact ELM.

• 6.

  Stage 5–Complete atrophy:

  Complete loss of the photoreceptor layer, with absence of both EZ and IZ signals and a residual ELM line suspended above the RPE overlying the SRF space, indicating complete photoreceptor degeneration.

Representative spectral-domain OCT images illustrating the 6 photoreceptor stages in chronic central serous chorioretinopathy (CSC), with Paired pre and post-treatment findings. For each stage, paired panels show the same eye at the pre-PDT visit (A1-F1, green-bordered foveal magnifications) and at the visit of complete subretinal fluid resolution after PDT (A2-F2, cyan-bordered foveal magnifications), with best-corrected visual acuity (BCVA) at both timepoints. Throughout the figure, yellow arrows indicate the external limiting membrane (ELM), magenta arrows the ellipsoid zone (EZ), and cyan arrows the interdigitation zone (IZ). (A) Stage 1A- Normal architecture. Distinct and continuous ELM, EZ, and IZ with normal reflectivity and preserved layer thickness, providing the architectural reference against which the subsequent stages are graded. Postresolution: full preservation of the ELM-EZ-IZ architecture, with BCVA unchanged. (B) Stage 1B- Early fusion. The EZ and IZ bands are no longer individually resolvable as 2 separate hyperreflective lines and fuse into a single hyperreflective band, while overall thickness and reflectivity remain preserved; this fusion is itself the defining grading feature, not an image-quality limitation. Postresolution: recovery of BCVA, with the EZ band clearly visible and the IZ band partially restored but not yet fully resolved. (C) Stage 2- Reactive thickening. The photoreceptor layer overlying the foveal SRF appears diffusely thicker than the photoreceptor layer of the immediately adjacent attached parafoveal retina of the same B-scan, with preserved reflectivity and no focal disruption— reflecting outer-segment elongation due to impaired RPE phagocytosis. Postresolution: normalization of photoreceptor thickness with restored EZ-IZ definition and recovery of BCVA, consistent with the still-reversible nature of this stage. (D) Stage 3- Patchy EZ irregularity (“moth-eaten” appearance). Focal discontinuities and irregular reflectivity of the EZ that do not reach the ELM, with overall photoreceptor thickness preserved or mildly increased. Postresolution: partial recovery of the EZ with residual irregularities and limited BCVA gain. (E) Stage 4- Structural discontinuity. Localized thinning of the outer retina with complete EZ interruption extending up to an intact ELM, consistent with partial photoreceptor loss. Postresolution: persistent focal EZ interruption despite anatomical reattachment, with limited or absent BCVA recovery. (F) Stage 5- Complete atrophy. Absence of both EZ and IZ signals with a residual ELM line suspended above the RPE, overlying persistent SRF, indicating end-stage photoreceptor degeneration. Postresolution: no reconstitution of EZ or IZ despite SRF resolution, with BCVA unchanged.

This grading was applied longitudinally at all visits where SRF was present to track dynamic changes in foveal photoreceptor morphology after PDT. Stages were assigned based on consensus-defined visual criteria without algorithmic assistance or automated segmentation. The resulting photoreceptor stage was used as the primary structural biomarker for correlation with visual acuity outcomes (BCVA) and anatomical response (foveal SRF resolution).

• 1.

  Baseline: defined as the visit corresponding to the first photodynamic therapy session (PDT1) within the study period.

• 2.

  Primary anatomical outcome: complete resolution of foveal SRF within 6 months after PDT1, as assessed on SD-OCT.

• 3.

  Responders: eyes that achieved complete foveal SRF resolution (SRF-free) within 6 months after PDT1.

• 4.

  Delayed responders: eyes without foveal SRF resolution within 6 months after PDT1 that subsequently achieved complete resolution. Delayed responders were further categorized as:

  • –

    With additional PDT: eyes that received ≥1 additional PDT before the first SRF-free visit.

  • –

    Without additional PDT: eyes that achieved SRF resolution without receiving additional PDT before the SRF-free visit.

• 5.

  Nonresponders: eyes that never achieved foveal SRF resolution either within or beyond 6 months up to the last available visit.

• 6.

  Effective PDT: for delayed responders with additional PDT sessions, the last PDT performed before the first SRF-free visit.

• 7.

  Time variables: all times were referenced to PDT1 (months from baseline). For delayed responders without additional PDT, the visit immediately prior to resolution was identified to characterize the photoreceptor stage when SRF was still present.

• 8.

  Primary visual acuity outcome: change in best-corrected visual acuity (ΔBCVA, logMAR), defined as BCVAbaseline − BCVAfollow-up; positive ΔBCVA values indicate visual improvement.

• 9.

  For 6-month outcome analyses, the follow-up timepoint was defined as the first visit showing complete SRF resolution for responders (SRF-free visit), and as the last available visit within 6 months after PDT1 for eyes without SRF resolution.

All statistical analyses were performed using R software (version 4.4.0; R Foundation for Statistical Computing, Vienna, Austria). Continuous variables are reported as mean ± SD (SD) or median [IQR, IQR], and categorical variables as counts and percentages. Statistical significance was set at a 2-sided P <.05. The primary anatomical endpoint was the time to complete foveal SRF resolution within 6 months after PDT1, and the primary functional endpoint was ΔBCVA among responders. All other analyses were considered secondary or exploratory. No data imputation was performed.

Group comparisons were performed using the Mann–Whitney U test or Wilcoxon signed-rank test for continuous variables, and chi-square or Fisher’s exact tests for categorical variables. Differences in baseline features across photoreceptor stages were assessed using the Kruskal–Wallis test with Dunn’s post hoc comparisons (Holm correction). Intergrader agreement for photoreceptor staging was evaluated in the 50-eye independent reliability subset using exact concordance, unweighted and quadratic-weighted Cohen’s κ with 95% bootstrap CIs, and stage-specific agreement metrics.

Time-to-event analyses were conducted using Cox proportional hazards regression. Eyes not achieving SRF resolution within 6 months were censored at their last visit within this interval. Eyes receiving an additional PDT before resolution were right-censored at the date of the subsequent PDT so that the model reflected the effect of the initial PDT only. Proportional hazards assumptions were assessed using Schoenfeld residuals.

For visual acuity outcomes among responders, linear regression models were used to evaluate predictors of ΔBCVA, and ANCOVA models were applied to final BCVA (logMAR). Covariates in the ANCOVA included baseline BCVA, photoreceptor stage, age, symptom duration, CST, SCT, and IRF. Photoreceptor stage was modeled as an ordinal predictor in the Cox regression, reflecting its monotonic structural progression, whereas it was modeled as a categorical factor in visual outcome analyses to allow for nonlinear functional differences across discrete stages. Sensitivity analyses comparing linear and categorical parameterizations yielded consistent results. Multicollinearity among predictors of multivariable models was assessed using variance inflation factors (VIF), with values >5 considered indicative of problematic collinearity.

Because treatment regimen (half-dose vs half-fluence PDT) was incompletely documented, it was not included in multivariable models. A sensitivity analysis including treatment center (Milan vs Turin) as a covariate was performed for the Cox, ΔBCVA, and ANCOVA models. Center was not significant and did not alter effect estimates. Given that a minority of patients contributed both eyes, all primary regression models were repeated with patient-clustered robust SEs; results were unchanged.

For delayed responders, BCVA was compared across baseline, effective PDT, and SRF-resolution visits using the Friedman test with Wilcoxon posthoc tests (Holm correction). For nonresponders, BCVA and photoreceptor stage were evaluated between baseline and last visit using paired Wilcoxon tests.

To examine robustness to variable follow-up completeness, all primary models (Cox, ΔBCVA, ANCOVA) were repeated in eyes with at least 1 follow-up visit ≥4 months after baseline. Effect directions and magnitudes were unchanged.