Whether anti-VEGF therapy can be safely suspended after a treat-and-extend (TAE) regimen in neovascular age-related macular degeneration (nAMD) remains controversial. Although treatment suspension is increasingly attempted in eyes achieving prolonged anatomical stability, reported recurrence rates vary widely and determinants of recurrence remain poorly defined, particularly across macular neovascularization (MNV) subtypes. We conducted a retrospective multicentre real-world study including consecutive nAMD eyes treated with anti-VEGF therapy in several French centres between 2014 and 2025. Eyes were eligible if they achieved anatomical stability—defined as absence of intraretinal or subretinal fluid on SD-OCT—at their maximal extension interval and were subsequently suspended from TAE (complete cessation of injections). Recurrence was defined as renewed need for anti-VEGF therapy due to clinical or SD-OCT signs of neovascular activity. MNV subtype was classified on multimodal imaging according to CONAN criteria. Because pure type 2 lesions and mixed type 1/2 lesions share a subretinal component and show overlapping exudative behaviour, we analysed them together as ‘MNV with subretinal component’, while acknowledging that fibrosis propensity may differ between pure and mixed phenotypes. Among 465 eyes managed with a TAE-based strategy, 88 eyes (18.9%) from 81 patients underwent treatment suspension. Mean age at baseline was 79.5 ± 8.5 years, and baseline BCVA was 59.1 ± 22.5 ETDRS letters. Patients received a median of 15.0 IQR: 11.0; 22.3 injections over a mean of 36.0 ± 19.8 months before suspension, with a median last treatment interval of 86.0 IQR: 63.0; 105.5 days. Compared with eyes remaining on TAE, suspended eyes showed a lower rate of type 1 lesions and a higher rate of lesions with a subretinal component. MNV subtypes among suspended eyes were 38.6% type 1, 43.2% MNV with a subretinal component, and 18.2% type 3. During a median post-suspension follow-up of 36.4 IQR: 25.9, 52.6 months, 38 of 88 eyes (43.2%) experienced exudative recurrence. Kaplan–Meier analysis estimated recurrence rates of 15.8% at 1 year, 42.1% at 2 years, and 78.9% at 4 years. The median time to recurrence was 5.5 IQR: 3.0–9.7 months, emphasizing that most relapses occurred early after treatment cessation. In multivariable Cox analysis, a longer duration from treatment initiation to TAE suspension was independently associated with recurrence (HR 1.03 per month, p = 0.018), suggesting that eyes requiring prolonged treatment before achieving stability remain biologically treatment-dependent. Recurrence rates did not significantly differ by MNV subtype (type 1: 44.1%; subretinal component: 47.4%; type 3: 31.2%; p = 0.59), and recurrence-free survival curves overlapped across subtypes (log-rank p = 0.61, Figure 1). However, exploratory subtype analyses revealed distinct risk profiles: recurrence in type 1 lesions was associated with longer treatment duration before suspension (HR per month, 1.03; 95% CI, 1.00–1.06; p = 0.028), while in lesions with a subretinal component, recurrence was more frequent in eyes with reticular pseudodrusen at baseline (p = 0.042). No significant predictor was identified in type 3 lesions. Functionally, mean BCVA remained globally stable after recurrence when prompt retreatment was initiated. Nevertheless, when stratified by subtype, eyes with type 1 MNV experienced a modest but significant decline in BCVA between TAE exit and final follow-up (−7.1 ± 18.3 letters, p = 0.042), whereas visual acuity remained stable in eyes with subretinal component lesions (p = 0.48) and type 3 MNV (p = 0.81). These findings suggest that late recurrence in type 1 MNV may carry a greater risk of cumulative functional damage despite retreatment. In summary, in real-world practice, only a minority of nAMD eyes reach a stage allowing TAE suspension, and recurrence after cessation is common and increases markedly over time. While MNV subtype alone did not significantly predict recurrence, treatment chronicity and structural biomarkers appear more informative for risk stratification. These data support cautious and individualized exit strategies, with close SD-OCT monitoring after suspension, particularly in eyes with long treatment histories or type 1 MNV. The study has received no funding. Geoffrey Zeender: No conflict of interest. Amandine Moulin: No conflict of interest. Pauline Rizzuto: No conflict of interest. Laurent Kodjikian: Consultant for AbbVie, Alimera, Bayer, Horus, Krys, Novartis, Roche, and Thea. Corinne Dot: Consultant for Bayer, Horus, Novartis and Thea. Sandra Elbany: No conflict of interest. Pierre Gascon: Consultant for Bayer and Novartis. Kevin Allaix: No conflict of interest. Audrey Feldman: No conflict of interest. Flore De Bats: No conflict of interest. Benjamin Wolff: No conflict of interest. Emmanuelle Bassouamina-Kinanga: No conflict of interest. Thibaud Mathis: Consultant for AbbVie, Bayer, GSK, Horus, Kol, Novartis and Roche.
Zeender et al. (Fri,) studied this question.