Effects of Home-Based Antisuppression and Binocular Integration Training in Patients with Intermittent Exotropia: A Retrospective Study Authors: Ziqing Ding, Yulian Zhou, Jingrong Liu, Zhi Chen, Chen Zhao, Wen Wen Published: 2026 ABSTRACT ObjectiveTo evaluate the association between 6-month home-based antisuppression with binocular integration training with binocular sensory outcomes and the 2-year surgical rate in patients with intermittent exotropia (IXT).MethodsIn this retrospective cohort study, patients diagnosed with IXT were included and classified into training and control groups according to their treatment history. The primary outcome was improvement in fine stereopsis (FS) at 6 months. Secondary outcomes included improvements in coarse stereopsis (CS), dynamic stereopsis (DS), peripheral suppression (PS), foveal suppression (FoS) at 6 months, and surgical rate within 2 years. Propensity score matching and inverse probability of treatment weighting were applied to balance baseline characteristics. Relative risks were estimated using univariable and multivariable log-binomial regression. Subgroup analyses were performed for the primary outcome.ResultsA total of 252 patients were included. At 6 months, the training group showed significantly higher improvement rates than the control group in FS (61.54% vs 32.11%, P=0.0001), DS (49.65% vs 31.19%, P=0.0050), and CS (41.96% vs 22.94%, P=0.0025). The training group also had a significantly lower surgical rate within 2 years (8.39% vs 18.35%, P=0.0118). No significant differences were observed in PS or FoS. Subgroup analysis identified no significant interactions of FS improvements.ConclusionHome-based antisuppression with binocular integration training was associated with improved binocular sensory outcomes and a lower 2-year surgical rate in patients with IXT. These findings suggest that this intervention may represent a potential adjunctive non-surgical treatment option. Introduction Intermittent exotropia (IXT) is one of the most common types of strabismus, characterized by an intermittent outward deviation of one or both eyes that is often exacerbated by fatigue, visual inattention, or illness.1 The prevalence of IXT is approximately 3% among children in China.2 Beyond its cosmetic and psychosocial impact, IXT may impair visual function and quality of life through deficits in binocular vision, such as reduced stereoacuity, impaired sensory fusion, and decreased binocular stability.3 Although the exact pathophysiology of IXT remains incompletely understood, growing evidence suggests that it is not merely a disorder of ocular alignment but also involves abnormalities in binocular sensory processing.4,5 Patients with IXT frequently exhibit multiple sensory adaptations, including interocular suppression, anomalous retinal correspondence, impaired stereopsis, binocular imbalance and altered fusional function.3,6–8 Among these, interocular suppression develops as a compensatory mechanism to avoid diplopia and visual confusion during periods of ocular misalignment.6,7 However, persistent suppression may interfere with binocular integration and disrupt binocular fusion. Emerging evidence further suggests that increasing suppression may alter the balance of fusional convergence between the dominant and non-dominant eyes, leading to reduced control of ocular alignment.9 Consequently, these sensory abnormalities compromise binocular cooperation, weaken the control of stable ocular alignment, and may contribute to the progression of IXT. Therefore, restoring binocular sensory function has emerged as an important therapeutic target alongside correction of ocular alignment. Surgical correction remains the primary treatment for IXT. However, satisfactory postoperative motor alignment does not necessarily restore normal sensory function, and postoperative recurrence remains common.10,11 Furthermore, surgery is generally indicated for patients with deteriorating control. According to the Newcastle Control Score (NCS), surgical intervention is usually recommended for patients with poor control (NCS ≥3).12 Nevertheless, even patients with relatively good control (NCS <3) may eventually require surgery, with reported surgical rates of approximately 4.5%–7% within 3 years.13 In addition, nearly one-quarter of untreated patients experience deterioration in control during follow-up.14 Consequently, effective conservative interventions that preserve binocular function and potentially delay or reduce the need for surgery remain an important clinical objective. Among nonsurgical approaches, vision therapy aims to improve binocular coordination through perceptual and sensory rehabilitation. Based on current understanding of the neural mechanisms underlying IXT, antisuppression therapy seeks to reduce pathological suppression and restore simultaneous binocular perception, whereas binocular integration training promotes binocular cooperation by encouraging balanced use of both eyes. Previous binocular rehabilitation approaches, including dichoptic training, perceptual learning, and orthoptic therapy, have demonstrated encouraging improvements in binocular visual function, ocular alignment, and control in patients with IXT.15–19 These findings support the concept that binocular sensory plasticity can be enhanced through targeted rehabilitation. Among sensory outcomes, fine stereopsis represents one of the most sensitive indicators of binocular visual function because it depends on the detection of small retinal disparities and requires highly precise binocular integration.20,21 Nevertheless, dynamic stereopsis requires the integration of binocular disparity with motion cues and therefore reflects binocular visual performance under more natural viewing conditions, making it clinically relevant for evaluating functional binocular vision during everyday activities.22 Fine stereopsis has been shown to correlate with binocular control and functional visual performance in patients with IXT, making it a clinically meaningful outcome for evaluating sensory rehabilitation.23 In contrast to motor alignment alone, improvements in fine stereopsis may better reflect restoration of higher-order binocular function. Although previous studies have suggested that binocular training may improve sensory function in patients with IXT, most have evaluated only short-term changes in binocular vision or control ability. Whether these sensory improvements translate into better long-term clinical outcomes, particularly a reduced need for subsequent surgical intervention, remains unclear. In addition, evidence regarding structured home-based antisuppression combined with binocular integration training is still limited. Therefore, evaluating both sensory outcomes and subsequent surgical rates may provide a more comprehensive assessment of the potential clinical value of sensory rehabilitation. Accordingly, we retrospectively evaluated the effects of long-term home-based antisuppression combined with binocular integration training in patients with IXT. We primarily assessed changes in fine stereopsis, together with other measures of stereopsis and suppression, and further examined whether this intervention could reduce the surgical rate during a 2-year follow-up period. We hypothesized that long-term home-based binocular sensory rehabilitation would be associated with improved binocular sensory function and a lower subsequent surgical rate. Methods This retrospective study was approved by the Eye & ENT Hospital, Shanghai Medical School, Fudan University, Shanghai, China, and adhered to the tenets of the Declaration of Helsinki. Written informed consent was waived given the retrospective nature of this study and the use of anonymized patient data. Medical records of consecutive patients diagnosed with intermittent exotropia (IXT) between September 2020 and December 2022 at Eye & ENT Hospital, Shanghai Medical School, Fudan University, were reviewed. Treatment decisions were based on patient and family preference. Participants Eligible participants were younger than 20 years and had a baseline NCS of 1–2. Patients in the training group completed at least 6 months of home-based antisuppression with binocular integration training, whereas patients in the control group received no IXT-related treatment and were followed routinely. All included patients had at least 6 months of follow-up for sensory outcomes and 2 years of follow-up for surgical outcomes. The Newcastle Control Score was assessed at baseline as part of routine clinical evaluation to estimate the severity of ocular motor control. The NCS provides a standardized clinical assessment of IXT control, incorporating both observed control in daily life and clinical control during examination. To reduce heterogeneity in motor control, enrollment was restricted to patients with NCS scores of 1 to 2, representing mild to moderate levels of control. Patients were excluded if they were 20 years or older; had other forms of strabismus, including dissociated vertical deviation or esotropia; had ophthalmic, systemic, or neurologic conditions affecting visual function; had received prior IXT-related treatment (including surgery or vision therapy); had an NCS of 0 or ≥3; or had insufficient follow-up duration. Examinations All patients underwent comprehensive binocular visual function testing at baseline and at 6-month follow-up. Exodeviation angle at baseline was tested by Hirschberg Test (HT) by comparing the displacement of corneal reflections between the eyes at a distance of 33 cm from the patients. Surgical intervention was determined according to routine clinical decision-making. In general, surgery was considered for patients with deteriorating ocular control, increasing frequency or duration of manifest deviation, progressive exodeviation, impaired binocular function, or functional, cosmetic, or psychosocial concerns. The final decision was made jointly by the ophthalmologist and the patient’s family. Sensory assessments included: fine stereopsis (FS), coarse stereopsis (CS), dynamic stereopsis (DS), foveal suppression (FoS), and peripheral suppression (PS). Outcome assessors were blinded to the intervention status. All examinations used a 24-inch polarized display (LG2342p; 1920 × 1080 resolution) integrated with standardized examination software developed by the Guangzhou Medical Instrument Research Institute (Guangzhou, China). Participants were seated at a viewing distance of 80 cm with the screen center aligned at eye level and wore polarized three-dimensional glasses during testing.24 a)Fine stereopsis was evaluated using a random-dot stereogram containing a centrally embedded E optotype presented at disparities of 400, 300, 200, and 100 arcseconds within a 5° × 5° field. Surrounding elements were presented at zero disparity to provide a binocular reference. Participants identified the orientation of the optotype. Stereoacuity was considered present if the participant correctly identified disparities within the tested range.b)Coarse stereopsis was assessed using stimuli in which participants determined whether targets appeared protruded or concaved relative to the background plane. Only fully correct responses were recorded as normal performance.c)Dynamic stereopsis was examined using motion-based disparity stimuli in which background elements moved at different velocities. Participants identified the orientation of an embedded optotype under motion conditions. High-velocity stimuli required lower depth discrimination, whereas low-velocity stimuli required finer binocular processing. Only fully correct responses were recorded as normal performance.d)Peripheral suppression was assessed using dichoptic stimuli consisting of four shapes arranged in a 2 × 2 configuration at the center of the display. Each eye was presented with two diagonally positioned shapes under polarized viewing conditions. Participants were asked to report the number of shapes perceived. Reduced perception indicated the presence of peripheral suppression.e)Foveal suppression was evaluated using dichoptic central letter stimuli. Different letters were presented separately to each eye at the fixation point. Participants were instructed to identify the letter perceived at the center of the screen. Inability to perceive both inputs simultaneously was considered indicative of foveal suppression. a)Fine stereopsis was evaluated using a random-dot stereogram containing a centrally embedded E optotype presented at disparities of 400, 300, 200, and 100 arcseconds within a 5° × 5° field. Surrounding elements were presented at zero disparity to provide a binocular reference. Participants identified the orientation of the optotype. Stereoacuity was considered present if the participant correctly identified disparities within the tested range. Fine stereopsis was evaluated using a random-dot stereogram containing a centrally embedded E optotype presented at disparities of 400, 300, 200, and 100 arcseconds within a 5° × 5° field. Surrounding elements were presented at zero disparity to provide a binocular reference. Participants identified the orientation of the optotype. Stereoacuity was considered present if the participant correctly identified disparities within the tested range. b)Coarse stereopsis was assessed using stimuli in which participants determined whether targets appeared protruded or concaved relative to the background plane. Only fully correct responses were recorded as normal performance. Coarse stereopsis was assessed using stimuli in which participants determined whether targets appeared protruded or concaved relative to the background plane. Only fully correct responses were recorded as normal performance. c)Dynamic stereopsis was examined using motion-based disparity stimuli in which background elements moved at different velocities. Participants identified the orientation of an embedded optotype under motion conditions. High-velocity stimuli required lower depth discrimination, whereas low-velocity stimuli required finer binocular processing. Only fully correct responses were recorded as normal performance. Dynamic stereopsis was examined using motion-based disparity stimuli in which background elements moved at different velocities. Participants identified the orientation of an embedded optotype under motion conditions. High-velocity stimuli required lower depth discrimination, whereas low-velocity stimuli required finer binocular processing. Only fully correct responses were recorded as normal performance. d)Peripheral suppression was assessed using dichoptic stimuli consisting of four shapes arranged in a 2 × 2 configuration at the center of the display. Each eye was presented with two diagonally positioned shapes under polarized viewing conditions. Participants were asked to report the number of shapes perceived. Reduced perception indicated the presence of peripheral suppression. Peripheral suppression was assessed using dichoptic stimuli consisting of four shapes arranged in a 2 × 2 configuration at the center of the display. Each eye was presented with two diagonally positioned shapes under polarized viewing conditions. Participants were asked to report the number of shapes perceived. Reduced perception indicated the presence of peripheral suppression. e)Foveal suppression was evaluated using dichoptic central letter stimuli. Different letters were presented separately to each eye at the fixation point. Participants were instructed to identify the letter perceived at the center of the screen. Inability to perceive both inputs simultaneously was considered indicative of foveal suppression. Foveal suppression was evaluated using dichoptic central letter stimuli. Different letters were presented separately to each eye at the fixation point. Participants were instructed to identify the letter perceived at the center of the screen. Inability to perceive both inputs simultaneously was considered indicative of foveal suppression. Intervention The training group received a commercially available, game-based binocular training system developed by the National Engineering Research Center for Healthcare Devices (Guangdong, China). Training was performed at home for 30 minutes daily over 6 months. The program employed a red–blue or red-green dichoptic paradigm targeting antisuppression and binocular integration. Tasks were designed to enhance visual integration, stereoacuity, depth perception, contrast sensitivity, and hand–eye coordination. Although specific task selections were adjusted according to individual performance, all training modules were based on the same antisuppression and binocular integration framework. Representative examples of the training tasks are shown in Supplementary Figure 1. Task difficulty was adjusted according to individual performance. According to the software instructions, the task level generally increased after three correct responses and decreased after one incorrect response. Difficulty adjustment could involve changes in task level, stimulus characteristics, and response requirements, depending on the specific module. The system included a built-in monitoring function that recorded training duration and task performance, allowing clinicians to review training participation during follow-up. However, detailed quantitative adherence data were not retrospectively extracted for all patients and therefore were not included in the present analysis. Outcome Measures Stereopsis was recorded according to the predefined ordinal levels of the clinical stereotest. For FS, values worse than the largest measurable disparity were recorded as >400 arcsec. Changes in stereoacuity were evaluated as ordinal changes. The primary outcome was improvement in fine stereopsis (FS) at 6 months, defined as an improvement of at least one level from baseline. Secondary sensory outcomes included improvement rates of dynamic stereopsis (DS), coarse stereopsis (CS), foveal suppression (FoS), and peripheral suppression (PS) at 6 months. The clinical outcome was the surgical rate within 2 years of the initial visit. Statistical Analysis Descriptive analysis was performed for all participants. Categorical variables were expressed as numbers and percentages. Anisometropia was defined as a spherical equivalent difference of more than 2.0 dioptres(D) or a cylindrical error difference of more than 1.5 D. Continuous variables were expressed as mean ± standard deviation (SD). Continuous data were compared using an unpaired t-test or a Wilcoxon test. Pearson’s chi-squared test or Fisher’s exact test was used for categorical variables. A binary sex categorization was used in this study. In order to minimize the potential bias of treatment allocation and confounding, multivariable regression models, the propensity score matching (PSM) and inverse probability of treatment weighting (IPTW) analyses were performed. The following variables were selected a priori based on data in the published literature and biological plausibility to generate the propensity score: sex, age, spherical equivalent of the deviated eye, presence of anisometropia, exodeviation angle, baseline fine stereopsis and Newcastle Control Score. A standardized mean difference (SMD) was used to examine the degree of PSM and IPTW. A threshold of less than 0.1 was considered acceptable. We examined overlap plots of the propensity scores by treatment group to ensure that the area of common support was adequate. A 1:1 nearest neighbor matching without replacement was applied using a caliper width of 0.1. In the IPTW analyses, the predicted probabilities from the propensity score model were used to calculate the stabilized IPTW weights. For primary and secondary binary outcomes, we used univariable and multivariable log-binomial regression to estimate the relative risk, both with a two-sided 95% CI. No formal adjustment for multiple comparisons was applied to the secondary outcomes. As a sensitivity consideration, a Bonferroni-corrected significance threshold of 0.010 was considered for the five secondary outcomes. Subgroup analysis of the primary outcome was performed using the above log-binomial regression model analysis separately for each subgroup, including sex, age, spherical equivalent, presence of anisometropia, FS-baseline, DS-baseline, CS-baseline, exodeviation angle-baseline. All analyses were performed using the open-source statistical software R version 4.2.0 (R Foundation). A threshold of p < 0.05 (two-sided) was considered statistically significant. Results Characteristics of Population Of the 252 patients with IXT who were admitted to the EENT between September 2020 and December 2022, 143 (56.75%) received home-based antisuppression with binocular integration training, and 109 (43.25%) didn’t. During the study period, no training-related adverse effects, including eye fatigue, blurred vision, dry eye, ocular irritation, tearing, headache or treatment discontinuation related to adverse effects were reported. Figure 1 presents the eligibility criteria and flow chart of the study patients. After Propensity Score matching, 101 matched pairs were retained, and all covariates included in the propensity score model achieved acceptable balance, with absolute standardized mean differences below 0.10 (Table 1). Detailed baseline characteristics of the unmatched and matched cohorts are presented in Supplementary Table 1. Propensity score diagnostics were performed to assess common support, overlap, and covariate balance. The propensity score distributions before and after matching are shown in Supplementary Figures 2 and 3. Adequate overlap between the training and control groups was observed after matching, suggesting that comparable patients were available across groups. Covariate balance was assessed using standardized mean differences, and the SMD plot before and after matching is shown in Supplementary Figure 4.Table 1Baseline Characteristics of the Matched Cohort After Propensity Score MatchingCharacteristicControl (n=101)Training (n=101)P valueSMDSex0.7780.040 Male53 (52.48%)55 (54.46%) Female48 (47.52%)46 (45.54%)Age7.74 ± 2.497.85 ± 2.280.7460.046Spherical equivalent of the deviated eye, D0.1600.066 <−6.000 (0.00%)0 (0.00%) −6.00 to <−3.004 (3.96%)5 (4.95%) −3.00 to <−0.5037 (36.63%)26 (25.74%) −0.50 to +0.5031 (30.69%)41 (40.59%) >+0.50 to +3.0026 (25.74%)29 (28.71%) >+3.00 to +6.003 (2.97%)0 (0.00%)Anisometropia0.8000.036 No93 (92.08%)92 (91.09%) Yes8 (7.92%)9 (8.91%)Newcastle Control Score0.8880.020 149 (48.51%)48 (47.52%) 252 (51.49%)53 (52.48%)Baseline fine stereopsis0.6560.063 >400 arcsec65 (64.36%)68 (67.33%) ≤400 arcsec36 (35.64%)33 (32.67%)Baseline exodeviation angle, degree−10 (−15, −10)−10 (−15, −5)0.4260.009Notes: Data are presented as n (%), mean ± standard deviation, or median (Q1, Q3), as appropriate. An absolute SMD <0.10 was considered indicative of acceptable covariate balance after matching.Abbreviations: D, diopters; Q1, first quartile; Q2, median; Q3, third quartile; SMD, standardized mean difference. Figure 1Flow diagram of patient selection.The flowchart illustrates the patient selection process during a study. Initially, 1909 patients were screened. Of these, 1657 were excluded for reasons such as presence of amblyopia, other types of strabismus, other ophthalmic or systemic conditions, prior IXT-related treatment, other visual training, age over 20 years, lack of follow-up records, insufficient follow-up for surgical outcomes and baseline NCS criteria. This left 252 patients included in the final analysis. These patients were then divided based on whether they received training. Those who received training were placed in the Training Group, consisting of 143 patients, while those who did not receive training were placed in the Control Group, consisting of 109 patients.Patient selection flowchart: screening, exclusion, training and control groups. Table 1Baseline Characteristics of the Matched Cohort After Propensity Score MatchingCharacteristicControl (n=101)Training (n=101)P valueSMDSex0.7780.040 Male53 (52.48%)55 (54.46%) Female48 (47.52%)46 (45.54%)Age7.74 ± 2.497.85 ± 2.280.7460.046Spherical equivalent of the deviated eye, D0.1600.066 <−6.000 (0.00%)0 (0.00%) −6.00 to <−3.004 (3.96%)5 (4.95%) −3.00 to <−0.5037 (36.63%)26 (25.74%) −0.50 to +0.5031 (30.69%)41 (40.59%) >+0.50 to +3.0026 (25.74%)29 (28.71%) >+3.00 to +6.003 (2.97%)0 (0.00%)Anisometropia0.8000.036 No93 (92.08%)92 (91.09%) Yes8 (7.92%)9 (8.91%)Newcastle Control Score0.8880.020 149 (48.51%)48 (47.52%) 252 (51.49%)53 (52.48%)Baseline fine stereopsis0.6560.063 >400 arcsec65 (64.36%)68 (67.33%) ≤400 arcsec36 (35.64%)33 (32.67%)Baseline exodeviation angle, degree−10 (−15, −10)−10 (−15, −5)0.4260.009Notes: Data are presented as n (%), mean ± standard deviation, or median (Q1, Q3), as appropriate. An absolute SMD <0.10 was considered indicative of acceptable covariate balance after matching.Abbreviations: D, diopters; Q1, first quartile; Q2, median; Q3, third quartile; SMD, standardized mean difference. Baseline Characteristics of the Matched Cohort After Propensity Score Matching Notes: Data are presented as n (%), mean ± standard deviation, or median (Q1, Q3), as appropriate. An absolute SMD <0.10 was considered indicative of acceptable covariate balance after matching. Abbreviations: D, diopters; Q1, first quartile; Q2, median; Q3, third quartile; SMD, standardized mean difference. Flow diagram of patient selection. Primary Outcome At the 6-month follow-up, improvement in FS was observed in 35 patients (32.11%) in the control group and 88 patients (61.54%) in the training group. In the multivariable log-binomial regression analysis, the training group showed a significantly higher rate of FS improvement than the control group, with an adjusted relative risk (RR) of 1.70 (95% CI, 1.27–2.28; P=0.0004). The results remained consistent in the propensity score analyses. After inverse probability of treatment weighting, the adjusted RR was 1.87 (95% CI, 1.36–2.58; P=0.0001), and after propensity score matching, the adjusted RR was 1.79 (95% CI, 1.32–2.45; P=0.0002) (Table 2).Table 2Association Between Home-Based Antisuppression and Binocular Integration Training and Improvement in Fine StereopsisAnalysisResult or Effect EstimateP valuePrimary outcome: improvement in fine stereopsis Control group35/109 (32.11%) Training group88/143 (61.54%) Absolute risk difference, percentage points29.43% Crude analysis, RR (95% CI)1.92 (1.44–2.64)<0.0001 Multivariable analysis, adjusted RR (95% CI)a1.70 (1.27–2.28)0.0004Propensity-score analyses 1:1 propensity-score matching, RR (95% CI)b1.79 (1.32–2.45)0.0002 Inverse probability of treatment weighting, RR (95% CI)c1.87 (1.36–2.58)0.0001Notes: Data are No./total No. (%) unless otherwise indicated. The absolute risk difference was calculated as the training group minus the control group. aThe multivariable model was adjusted for sex, age, spherical equivalent of the deviated eye, anisometropia, new castle score, baseline characteristics of fine stereopsis and exodeviation angle. b1:1 nearest-neighbor matching based on the propensity score. cInverse probability of treatment weighting based on the propensity score.Abbreviations: CI, confidence interval; FS, fine stereopsis; RR, relative risk. Table 2Association Between Home-Based Antisuppression and Binocular Integration Training and Improvement in Fine StereopsisAnalysisResult or Effect EstimateP valuePrimary outcome: improvement in fine stereopsis Control group35/109 (32.11%) Training group88/143 (61.54%) Absolute risk difference, percentage points29.43% Crude analysis, RR (95% CI)1.92 (1.44–2.64)<0.0001 Multivariable analysis, adjusted RR (95% CI)a1.70 (1.27–2.28)0.0004Propensity-score analyses 1:1 propensity-score matching, RR (95% CI)b1.79 (1.32–2.45)0.0002 Inverse probability of treatment weighting, RR (95% CI)c1.87 (1.36–2.58)0.0001Notes: Data are No./total No. (%) unless otherwise indicated. The absolute risk difference was calculated as the training group minus the control group. aThe multivariable model was adjusted for sex, age, spherical equivalent of the deviated eye, anisometropia, new castle score, baseline characteristics of fine stereopsis and exodeviation angle. b1:1 nearest-neighbor matching based on the propensity score. cInverse probability of treatment weighting based on the propensity score.Abbreviations: CI, confidence interval; FS, fine stereopsis; RR, relative risk. Association Between Home-Based Antisuppression and Binocular Integration Training and Improvement in Fine Stereopsis Notes: Data are No./total No. (%) unless otherwise indicated. The absolute risk difference was calculated as the training group minus the control group. aThe multivariable model was adjusted for sex, age, spherical equivalent of the deviated eye, anisometropia, new castle score, baseline characteristics of fine stereopsis and exodeviation angle. b1:1 nearest-neighbor matching based on the propensity score. cInverse probability of treatment weighting based on the propensity score. Abbreviations: CI, confidence interval; FS, fine stereopsis; RR, relative risk. The exploratory subgroup analyses are shown in Figure 2, and detailed numerical results are provided in Supplementary Figure 5. The direction of the association between training and FS improvement was generally consistent across the prespecified subgroups, and no significant interactions were observed for any baseline factors (P>0.10 for all comparisons). Given the small sample sizes and wide confidence intervals in some strata, particularly high refractive error and presence of anisometropia, the subgroup findings should be considered exploratory. Figure 2Forest plot of exploratory subgroup analyses for improvement in fine stereopsis at 6 months.Forest plot with columns: Subgroup, Adjusted RR (95% CI), P for interaction. X-axis ticks: 0.5, 1, 2, 4, 8; reference line at 1. Subgroups listed on Y-axis. All patients: 1.88 (1.39, 2.54). Sex interaction P=0.8098: Male 1.92 (1.24, 2.98); Female 1.84 (1.21, 2.79). Age interaction P=0.859: 3-9 years 1.87 (1.32, 2.64); 10-19 years 2.01 (1.09, 3.71). Refractive error: Low/moderate 1.88 (1.36, 2.60); High: dash. Anisometropia: No 1.77 (1.31, 2.38); Yes: dash. Fine stereopsis interaction P=0.4277: >400 arcsec 1.89 (1.28, 2.56); less than or equal to 400 arcsec 1.71 (0.94, 3.11). Dynamic stereopsis interaction P=0.8919: Abnormal 2.05 (1.29, 3.29); Normal 1.83 (1.23, 2.73). Coarse stereopsis interaction P=0.4073: Abnormal 1.64 (1.06, 2.54); Normal 2.15 (1.45, 3.20). Exodeviation angle interaction P=0.1605: -20 degrees to -10 degrees 3.65 (1.70, 7.84); -10 degrees to 0 degrees 2.00 (1.35, 2.97); 0 degrees 1.24 (0.72, 2.15).A forest plot of adjusted relative risk across subgroups, with most estimates above the null line.Notes: Subgroup analyses were exploratory and were not powered to detect subgroup-specific treatment effects. Some strata, particularly those defined by high refractive error or anisometropia, included relatively few patients and yielded wide 95% confidence intervals. No statistically significant interactions were observed in the estimable prespecified subgroup analyses (all P for interaction >0.10). Some subgroup estimates could not be calculated because of insufficient data. Detailed effect estimates and 95% confidence intervals are presented in Supplementary Figure 5.Abbreviations: RR, relative risk; CI, confidence interval. Forest plot of exploratory subgroup analyses for improvement in fine stereopsis at 6 months. Secondary Outcome Several secondary outcomes were evaluated to explore additional visual changes associated with training. In the multivariable analysis (Table 3), the training group showed significantly higher improvement rates in DS and CS than the control group. The RR was 1.59 for DS improvement (95% CI, 1.15–2.21; P=0.0050) and 1.86 for CS improvement (95% CI, 1.24–2.78; P=0.0025). The training group also had a lower 2-year surgical rate than the control group, with an RR of 0.43 (95% CI, 0.22–0.83; P=0.0118), indicating a lower relative risk of undergoing surgery in the training group. The absolute risk reduction for the 2-year surgical outcome was 9.96%, corresponding to an estimated number needed to treat (NNT) of 11. Clinically, this suggests that training approximately 11 patients would be associated with one fewer surgical procedure within 2 years compared with observation alone. As a sensitivity consideration, DS and CS improvement remained statistically significant when a Bonferroni-corrected threshold of 0.010 was considered for the five secondary outcomes, whereas the association with the 2-year surgical rate remained directionally consistent but did not meet this conservative threshold. No significant between-group differences were observed in the improvement rates of FoS or PS, despite the antisuppression component of the training program. Potential explanations for this finding are discussed in the Discussion section.Table 3Association Between Home-Based Antisuppression and Binocular Integration Training and Secondary Outcomes and the 2-Year Surgical RateImprovement in Dynamic StereopsisControl group34/109 (31.19%)Training group71/143 (49.65%)Absolute risk difference, percentage points18.46%Crude analysis, RR (95% CI)1.59 (1.15–2.20)0.0049Multivariable analysis, adjusted RR (95% CI)a1.59 (1.15–2.21)0.0050Improvement in coarse stereopsisControl group25/109 (22.94%)Training group60/143 (41.96%)Absolute risk difference, percentage points19.02%Crude analysis, RR (95% CI)1.83 (1.23–2.72)0.0027Multivariable analysis, adjusted RR (95% CI)a1.86 (1.24–2.78)0.0025Improvement in peripheral suppressionControl group19/109 (17.43%)Training group33/143 (23.08%)Absolute risk difference, percentage points5.65%Crude analysis, RR (95% CI)1.32 (0.80–2.20)0.2776Multivariable analysis, adjusted RR (95% CI)a1.50 (0.91–2.49)0.1153Improvement in foveal suppressionControl group30/109 (27.52%)Training group50/143 (34.97%)Absolute risk difference, percentage points7.45%Crude analysis, RR (95% CI)1.27 (0.87–1.85)0.2144Multivariable analysis, adjusted RR (95% CI)a1.25 (0.87–1.81)0.2318Surgery within 2 yearsControl group20/109 (18.35%)Training group12/143 (8.39%)Absolute risk difference, percentage points−9.96%Crude analysis, RR (95% CI)0.46 (0.23–0.89)0.0223Multivariable analysis, adjusted RR (95% CI)a0.43 (0.22–0.83)0.0118Notes: Data are No./total No. (%) unless otherwise indicated. The absolute risk difference was calculated as the training group minus the control group. The analyses included all 252 patients. aThe multivariable log-binomial model was adjusted for sex, age, spherical equivalent of the deviated eye, anisometropia, new castle score, baseline characteristics of fine stereopsis and baseline exodeviation angle.Abbreviations: CI, confidence interval; CS, coarse stereopsis; DS, dynamic stereopsis; FoS, foveal suppression; PS, peripheral suppression; RR, relative risk. Table 3Association Between Home-Based Antisuppression and Binocular Integration Training and Secondary Outcomes and the 2-Year Surgical RateImprovement in Dynamic StereopsisControl group34/109 (31.19%)Training group71/143 (49.65%)Absolute risk difference, percentage points18.46%Crude analysis, RR (95% CI)1.59 (1.15–2.20)0.0049Multivariable analysis, adjusted RR (95% CI)a1.59 (1.15–2.21)0.0050Improvement in coarse stereopsisControl group25/109 (22.94%)Training group60/143 (41.96%)Absolute risk difference, percentage points19.02%Crude analysis, RR (95% CI)1.83 (1.23–2.72)0.0027Multivariable analysis, adjusted RR (95% CI)a1.86 (1.24–2.78)0.0025Improvement in peripheral suppressionControl group19/109 (17.43%)Training group33/143 (23.08%)Absolute risk difference, percentage points5.65%Crude analysis, RR (95% CI)1.32 (0.80–2.20)0.2776Multivariable analysis, adjusted RR (95% CI)a1.50 (0.91–2.49)0.1153Improvement in foveal suppressionControl group30/109 (27.52%)Training group50/143 (34.97%)Absolute risk difference, percentage points7.45%Crude analysis, RR (95% CI)1.27 (0.87–1.85)0.2144Multivariable analysis, adjusted RR (95% CI)a1.25 (0.87–1.81)0.2318Surgery within 2 yearsControl group20/109 (18.35%)Training group12/143 (8.39%)Absolute risk difference, percentage points−9.96%Crude analysis, RR (95% CI)0.46 (0.23–0.89)0.0223Multivariable analysis, adjusted RR (95% CI)a0.43 (0.22–0.83)0.0118Notes: Data are No./total No. (%) unless otherwise indicated. The absolute risk difference was calculated as the training group minus the control group. The analyses included all 252 patients. aThe multivariable log-binomial model was adjusted for sex, age, spherical equivalent of the deviated eye, anisometropia, new castle score, baseline characteristics of fine stereopsis and baseline exodeviation angle.Abbreviations: CI, confidence interval; CS, coarse stereopsis; DS, dynamic stereopsis; FoS, foveal suppression; PS, peripheral suppression; RR, relative risk. Association Between Home-Based Antisuppression and Binocular Integration Training and Secondary Outcomes and the 2-Year Surgical Rate Notes: Data are No./total No. (%) unless otherwise indicated. The absolute risk difference was calculated as the training group minus the control group. The analyses included all 252 patients. aThe multivariable log-binomial model was adjusted for sex, age, spherical equivalent of the deviated eye, anisometropia, new castle score, baseline characteristics of fine stereopsis and baseline exodeviation angle. Abbreviations: CI, confidence interval; CS, coarse stereopsis; DS, dynamic stereopsis; FoS, foveal suppression; PS, peripheral suppression; RR, relative risk. Discussion This retrospective study evaluated the association between 6-month home-based antisuppression with binocular integration training and binocular sensory outcomes as well as the 2-year surgical rate in patients with IXT. The training group showed higher improvement rates in fine, dynamic, and coarse stereopsis than the control group, whereas no significant between-group differences were observed in foveal or peripheral suppression. In addition, patients in the training group had a lower 2-year surgical rate than those managed with observation. These findings suggest that home-based binocular sensory rehabilitation may have potential value as an adjunctive conservative management strategy for IXT. Stereopsis is a key component of binocular sensory function and reflects the integrity of cortical binocular processing.23 Fine stereopsis represents the ability to detect small binocular disparities and therefore reflects precise binocular integration, whereas coarse stereopsis is based on larger disparities and represents a more fundamental level of binocular depth perception. Dynamic stereopsis additionally requires the integration of binocular disparity with motion information and may more closely reflect binocular processing under natural, time-varying visual conditions.22 Previous studies have demonstrated the clinical relevance of stereoacuity in patients with intermittent exotropia. Better preoperative stereoacuity has been associated with more favorable postoperative sensory outcomes and higher surgical success rates,25–27 whereas a worse preoperative distance stereoacuity has been reported to predict a lower likelihood of postoperative stereoacuity improvement.28 Stereoacuity is also correlated with the control grade of exotropia, suggesting a close relationship between binocular sensory function and the ability to maintain ocular alignment.19 In addition, poorer stereopsis has been associated with lower Intermittent Exotropia Questionnaire scores in affected children and their parents, indicating poorer disease-specific health-related quality of life.29,30 Beyond clinical assessments, reduced stereoacuity has been linked to impaired performance in depth-dependent visuomotor activities, including reaching, grasping, and fine hand–eye coordination, as well as poorer reading and academic performance in children.31–34 Therefore, such improvements may have clinical implications beyond better performance on stereoacuity tests, including more effective binocular control, greater stability of ocular alignment, better depth-dependent visuomotor performance, and improved health-related quality of life. Nevertheless, because these functional outcomes were not directly evaluated in the present study, their relationship with the observed sensory improvements should be confirmed in future prospective studies using objective functional measures and patient-reported outcomes. Our findings are broadly consistent with growing evidence supporting nonsurgical binocular rehabilitation in IXT. Office-based therapy with home reinforcement has been reported to improve distance and near control of exodeviation, while office-based vergence and antisuppression therapy has shown beneficial effects in children with small-to-moderate angle IXT.15,35 Other approaches have also been investigated. For example, a randomized controlled trial of early postoperative orthoptic therapy in children with IXT reported improvements in stereopsis, fusional convergence amplitude, and surgical outcomes, whereas virtual-reality-based dichoptic visual training has been associated with improvements in stereoacuity and ocular alignment.36,37 In related binocular vision disorders, flicker-glass therapy has been reported to improve visual acuity, contrast sensitivity, and stereoacuity, while binocular integration training may facilitate the combination of visual signals from the two eyes and enhance stereoscopic perception.19,38 Collectively, these training-related improvements suggest that binocular sensory processing remains plastic and may be enhanced through targeted visual training. Compared with many office-based programs, the home-based format used in the present study may provide a more accessible option for long-term binocular sensory rehabilitation, although direct comparisons between office-based and home-based approaches remain limited. A noteworthy finding of the present study was that improvements in stereopsis were observed without significant between-group differences in foveal or peripheral suppression. This result may appear inconsistent with the antisuppression component of the training program, but several explanations should be considered. First, suppression was assessed using qualitative clinical tests, which may be less sensitive than quantitative methods for detecting subtle changes in suppression depth or suppression scotoma size.39,40 Antisuppression training may reduce the intensity or spatial extent of suppression without producing a categorical change from “suppression present” to “suppression absent”. Previous studies have shown that interocular suppression in IXT can be measured as a graded imbalance rather than a purely binary phenomenon, supporting the idea that categorical clinical tests may underestimate partial sensory improvement.41 Another possible explanation is that suppression represents a spatially organized cortical adaptation rather than a simple, all-or-none loss of input from the deviated eye. A study investigating the neural basis of suppression demonstrated that patients with exotropia retained perceptual access to signals from the deviated eye despite complementary suppression scotomas and adaptive shifts in visual direction.42 These findings suggest that residual binocular signals may remain available for enhanced stereoscopic integration, whereas established patterns of cortical suppression may be more resistant to change and require a longer treatment duration to produce measurable improvement. This interpretation is partly supported by a study in which 12 months of visual perceptual training significantly improved stereoacuity, while the reduction in suppression approached statistical significance.37 Therefore, suppression may improve more gradually than stereopsis, and measurable changes may emerge only after longer training or follow-up. In addition, baseline suppression status may have limited the detectable improvement in suppression outcomes. Patients without measurable suppression at baseline had little room for categorical improvement, whereas patients with more persistent or severe suppression may have required longer training or greater treatment exposure before a change could be detected by qualitative clinical tests. The lower 2-year surgical rate observed in the training group is clinically relevant and suggests that binocular sensory rehabilitation may have potential long-term value. One possible explanation is that improved stereopsis may reflect more effective binocular integration and cooperation, which could help support ocular control and alignment stability. Previous surgical studies have reported associations between binocular sensory status and motor outcomes in IXT, including relationships between preserved sensory fusion or better stereoacuity and more favorable postoperative alignment.43,44 However, although the improved stereopsis may reflect enhanced binocular function, surgical decision-making is influenced by multiple factors, including deviation control and magnitude, symptoms, clinical judgment, individual expectations, socioeconomic considerations and parental concerns. Accordingly, the lower surgical rate suggests an exploratory and potential longer-term clinical benefit of training for patients with IXT. Neural plasticity may be one possible explanation for the observed improvements in binocular sensory function after training. Previous studies have reported abnormal cortical activity and connectivity in visual regions involved in binocular fusion and disparity processing in patients with IXT.42,45–47 Increased activation of the right V3A and left intraparietal sulcus has been observed in patients with IXT who showed improved stereopsis after ocular realignment, suggesting functional plasticity within cortical disparity-processing networks.48 However, the present study did not directly assess cortical function or neural adaptation. Therefore, the role of neural plasticity remains speculative. Future studies using task-based fMRI, functional connectivity analysis, or eye-specific EEG/VEP paradigms are needed to determine whether stereopsis improvement after training is accompanied by measurable neural changes. There are several limitations of our study. First, the retrospective design introduces potential bias because treatment decisions were based on patient preference rather than randomized allocation. Although PSM and IPTW were applied to reduce baseline imbalance, unmeasured confounding factors, such as socioeconomic status, parental involvement, and disease characteristics, may still exist. In addition, because of the retrospective nature of the study, no formal a priori sample size or power calculation was performed. Therefore, the study may have been underpowered to detect differences in some secondary outcomes and subgroup analyses. Moreover, multiple secondary outcomes were evaluated without adjustment for multiple comparisons, thus, the possibility of type I error cannot be excluded. Accordingly, the findings for secondary outcomes and subgroup analyses should be viewed as exploratory. Second, although the training system recorded training duration and task performance, detailed quantitative adherence data were not retrospectively extracted for all patients. Therefore, the influence of treatment adherence on binocular sensory improvement could not be fully evaluated. Third, the Newcastle Control Score was used only as a baseline inclusion criterion to reduce heterogeneity of ocular motor control among participants but was not analyzed as a longitudinal outcome measure or stratified variable. Future studies are expected to integrate sensory outcomes, quality of life and motor control measures, such as the NCS, which may provide a more comprehensive evaluation of treatment effects. In addition, prism cover test measurements were not consistently documented in the retrospective medical records, and precise deviation magnitude in prism diopters therefore could not be reliably analyzed. Fourth, certain clinical characteristics of intermittent exotropia, such as vergence fluctuation,49 precise magnitude of deviation and detailed classifications of IXT subtypes, were not systematically evaluated in this study, which may affect binocular sensory function and the interpretability of the findings. Also, the exact magnitude of stereoacuity improvement could not be reliably calculated because stereoacuity was measured using ordinal clinical levels, and values worse than 400 arcsec were recorded as beyond the measurable range rather than as exact thresholds. Finally, this study included only patients with relatively good control, and the findings may not be generalizable to patients with poorer control or more severe intermittent exotropia. Prospective randomized controlled trials with standardized training protocols, broader and comprehensive inclusion criteria and longer follow-up periods are therefore needed to confirm these findings. In conclusion, among patients with intermittent exotropia, 6 months of home-based antisuppression with binocular integration training was associated with greater improvements in fine, dynamic, and coarse stereopsis, whereas no significant improvement was observed in suppression. The training group also showed a lower 2-year surgical rate, suggesting potential longer-term clinical relevance. These findings support the potential role of this training paradigm as a nonsurgical binocular sensory rehabilitation strategy for patients with intermittent exotropia. Prospective randomized controlled trials are needed to validate these findings and optimize training protocols for clinical application.