Factors for Treatment Success in Anisometropic Amblyopia: Effect of Refractive Errors of the Amblyopic Eyes

Purpose: To investigate the factors for treatment success in anisometropic amblyopia according to the spherical equivalent (SE) type of amblyopic eyes. Methods: Medical records of 397 children with anisometropic amblyopia aged 3 to 12 years during 2010~2016 were retrospectively reviewed. Anisometropia was dened as ≥ 1 diopter (D) difference in SE, or ≥ 1.5 D difference of cylindrical error between the eyes. According to the SE of amblyopic eyes, patients were categorized into hyperopia (SE ≥ 1D), emmetropia (-1< SE <+1) and myopia (SE ≤ -1D) groups. Treatment success was dened as achieving interocular LogMAR visual acuity difference < 0.2. Multivariate logistic regression was used to analyze the factors for treatment success. Results: Signicant factors for the amblyopia treatment success in hyperopia group (n=270) were younger age [adjusted odds ratio (aOR) (95% condence interval, CI) = 0.60 (0.41-0.88)], larger astigmatism of sound eye [aOR (95% CI) = 0.04 (0-0.78)], better BCVA in amblyopic eyes at presentation [aOR (95% CI) = 0.01 (0-0.16)], longer follow-up period [aOR (95%CI)=1.1 (1.04-1.17)], and no comorbid strabismus [aOR (95%CI)= 0.25 (0.06-0.99)]. In myopia group (n=68), older age [aOR (95% CI) =0.45 (0.20-0.97)] and worse BCVA in amblyopic eyes [aOR (95% CI) = 0.04 (0-0.61)] were inversely associated with higher odds of treatment success. There was no signicant factor for treatment success in emmetropia group (n=59). Conclusions: The type of refractive error of amblyopic eyes at presentation affects the factors for treatment success of anisometropic amblyopia. then myopia group. This result is in accordance to our previous study, 13 even though this study included only the patients with sucient follow-up duration of more than six months, considering the duration for treatment success in our previous study (mean 6.0, 4.1, and 6.6 months for the hyperopia, emmetropia, and myopia group). The emmetropia group had the shortest duration to treatment success, and highest treatment success rate. The hyperopia group had higher treatment success rate during the follow-up than myopia group.


Introduction
Anisometropia, the difference in refractive errors between the two eyes, is the most common cause of amblyopia in children when it is large. Amblyopia has prevalence reported as 1-4% in preschool-aged children, 1-4 and it may lead to irreversible vision impairment if untreated during young ages. Extensive researches have been conducted to nd factors that in uence the outcome of amblyopia treatment, mostly in anisometropic and/or strabismic amblyopia, and various and con icting results have been reported. [5][6][7][8] There are also numerous reports of amblyopia treatment methods and their outcome using randomized controlled trials, however, strict compliance to one treatment protocol is not always possible in the real world clinics, therefore the exact treatment method and duration vary considering each patient and guardian at physician's discretion. Even with the diversity in treatment execution, treatment of anisometropic amblyopia is successful in most cases, and there seems to be clinical characteristics of patients that lead to treatment success more easily. Our clinical experience from pediatric ophthalmology clinic in a large eye hospital led to a hypothesis that anisometropic amblyopia treatment prognosis vary depending on refractive error type in amblyopic eyes. Factors associated with the refractive error that may affect the result of anisometropic amblyopia treatment may include the spherical equivalent (SE) of the amblyopic eye, the degree of astigmatism of amblyopic eye, and/or the degree of difference in refractive errors between the two eyes. To our knowledge, there have been a few papers that attempts to analyze the factors that in uence the success of amblyopia treatment according to the refractive error type of amblyopic eye. 5,9−12 In our previous study, we found that treatment success rate was the highest, and the duration to treatment success was shortest in the emmetropia group categorized by SE of amblyopic eyes. 13 . Our clinical impression was that SE type of amblyopic eye affects treatment success more than the SE difference, and hyperopic amblyopes behave differently from myopic amblyopes. In this study, we aimed to ascertain the effect of refractive error of amblyopic eye on the treatment success, and to con rm whether the factors affecting treatment success were different according to refractive error type of amblyopic eye in a usual pediatric ophthalmology practice.

Methods
This study retrospectively reviewed medical records of patients aged 3 to 12 years who were diagnosed with anisometropic amblyopia at Strabismus & Pediatric Ophthalmology Center in Kim Eye Hospital between the period of January 1, 2010, and December 31, 2016. The patients who had any other ocular pathology or systemic condition that might affect visual acuity or those who had developmental delay were excluded from this study. Any patient with a follow-up period of less than 6 months was also excluded from this study to rule out the cases without su cient treatment duration to reach success. Amblyopia was de ned as the best corrected visual acuity (BCVA) difference of more than two logarithm of the minimum angle of resolution (logMAR) lines between the two eyes. Anisometropia was de ned as spherical equivalent difference of more than 1 diopter (D), or cylinder difference of more than 1.5 D regardless of meridians, between the two eyes. Age at presentation, sex, BCVA at the rst visit, refraction of both eyes, ocular alignment and type of manifest strabismus, previous amblyopia treatment history, and BCVA at the last visit were collected retrospectively using electronic medical records. Visual acuity was measured using a Snellen chart and converted to LogMAR values for analyses. Amblyopia treatment was started with a full spectacle correction after cycloplegic refraction. Patients were followed usually 1 or 2 months after spectacle prescription, and occlusion with patching was prescribed if there was no vision improvement in amblyopic eye. Occlusion dosage was determined considering the degree of amblyopia and response to treatment, usually from one to six hours per day. Occlusion was continued aiming for equal visual acuity on both eyes, as long as the patient and caregiver cooperate. Subjects were followed up every 2-4 months.
Subjects were categorized according to the SE of amblyopic eye by cycloplegic refraction at the rst examination, into the hyperopia group (SE ≥ 1 D), emmetropia group (-1 < SE < + 1), and myopia group (SE ≤-1 D). The treatment success was de ned as achieving BCVA difference of less than two logMAR lines between the two eyes. Treatment success during the follow-up was primary outcome for univariate and multivariate logistic regression analysis. All Statistical analysis was performed using R 3.6.0 Statistical Software (R foundation for statistical computing, Vienna, Austria). To compare the three SE groups, Chi-squared test and Kruskal-Wallis test were performed using R Statistical Software. Post-hoc analysis was conducted using Mann-Whitney test, and p-values < 0.017 were considered statistically signi cant according to the Bonferroni's adjustment. Univariate and Multiple logistic regression analyses were performed to determine the in uence of each factor on the success of amblyopia treatment. Factors which was statistically signi cant in univariate logistic regression analysis were included in multivariate logistic regression. Factors of interest (SE of amblyopic eye, cylinder of amblyopic eye, cylinder of sound eye, and SE difference between amblyopic and sound eyes at presentation) were also included in multivariate logistic regression analysis regardless of their signi cance in univariate analysis. Signi cant multicollinearity was encountered between the cylinder of the amblyopic eye and the difference of cylinder between the amblyopic eye and sound eye (Pearson's correlation coe cient − 0.887, p = 0.001, data not shown), and we analyzed each factor separately. However, putting only one variable out of them did not change signi cant factors in multivariate analysis results in all of the SE groups, and this paper presents results that included sound eye cylinder values. Age at the last visit were calculated in univariate analyses, but excluded for multivariable analyses due to multicollinearity with age at rst exam. All of the refractive errors used in this study were determined in minus cylinder form, thus an increase of cylinder value means decrease of absolute value of cylinder. This study was performed in accordance with the tenets of the Declaration of Helsinki. Approval to conduct this study was obtained from the Institutional Review Board of Kim's Eye Hospital (IRB 2018-01-012). Informed consent was waived by the Institutional Review Board because this study was conducted retrospectively using medical records without identi able private information.

Results
A total of 397 children (193 male and 205 female) were included in this study. According to the SE of amblyopic eye at the rst exam, there were 270 subjects (68.01%) in the hyperopia group, 59 subjects (14.61%) in the emmetropia group, and 68 subjects (17.13%) in the myopia group. Overall, as for the treatment of amblyopia, both glasses and patching were used in 365 (94.81%), prescribing glasses only in 20 (5.19%), and glasses and atropine penalization in 12 (3.12%) patients. Mean follow-up duration was 33.4 months (Standard deviation, SD = 18.7). Detailed demographics and clinical characteristics of the study population are shown in Table 1. Age at presentation, follow-up duration, comorbid strabismus, and previous amblyopia treatment history were not signi cantly different among the SE groups. The emmetropia group had the smallest interocular BCVA difference and SE difference, and the largest interocular cylinder difference between the two eyes at presentation among the three SE groups. Treatment success rate during the follow-up was signi cantly different among the SE groups (96.61% in emmetropia group, 91.48% in hyperopia, and 82.35% in myopia group, respectively, p = 0.016 by Chi-square test, Table 1). BCVA of the amblyopic eyes were not signi cantly different among the SE groups when they achieved the treatment success (p = 0.343).

Hyperopia group
In the univariate analysis of the hyperopia group, SE of the sound eye at the rst exam, BCVA of the sound eye, and follow-up duration signi cantly increased the odds of treatment success, whereas age at presentation, SE of the amblyopic eye at the rst exam, cylinder of the sound eye at the rst exam, SE difference at the rst exam, BCVA of the amblyopic eye at the rst exam, BCVA difference at the rst exam, and concomitant strabismus signi cantly decreased the odds of treatment success (Table 2).   (Table 2). However, both were not signi cant factors in multivariate analysis in this group (Table 3).

Myopia group
Univariate analysis of the myopia group found that the SE of amblyopic eye at the rst exam signi cantly increased the odds of treatment success, whereas age at presentation, SE difference at the rst exam, and BCVA of the amblyopic eye at the rst exam decreased the odds of treatment success (  (Table 3).

Discussions
In the present study, 397 patients who were treated > 6 months for anisometropic amblyopia in a secondary referral eye hospital were analyzed. The factors affecting treatment success in anisometropic amblyopia were different according to the SE type of the amblyopic eyes. The age and the BCVA of the amblyopic eye at presentation were factors affecting the odds of amblyopia treatment success in hyperopia and myopia groups. Smaller astigmatism of sound eye at presentation, longer follow-up period, and the absence of comorbid strabismus were signi cant factors increasing the odds of treatment success only in hyperopia group. Furthermore, there was no signi cant factor affecting treatment success in the emmetropia group.
There are diverse reports about treatment outcome of anisometropic amblyopia. However, as the hyperopic difference in refractive error are more amblyogenic than myopic differences 14  Treatment success rates during the follow-up were signi cantly different according to SE type in amblyopic eye at the rst exam; the highest success rate in emmetropia group, followed by hyperopia group and then myopia group. This result is in accordance to our previous study, 13 even though this study included only the patients with su cient follow-up duration of more than six months, considering the duration for treatment success in our previous study (mean 6.0, 4.1, and 6.6 months for the hyperopia, emmetropia, and myopia group). The emmetropia group had the shortest duration to treatment success, and highest treatment success rate. The hyperopia group had higher treatment success rate during the follow-up than myopia group.
These different treatment results according to SE type in amblyopic eye group made us questioned which of the refractive error factors, including the amount of SE or cylinder of the amblyopic eye, or SE or cylinder difference between the two eyes, are more important for treatment success in anisometropic amblyopia, and if there are differences in treatment success factors by the SE groups. In our univariate analyses results, the signi cance of SE in amblyopic eye, cylinder of the amblyopic eye, and SE difference were different in each SE group. In univariate analyses of hyperopia group, SE of amblyopic eye, SE of sound eye, cylinder of sound eye, and difference of SE between two eyes at rst exam were signi cant factors. In the univariate analyses of emmetropia group, SE of sound eye and cylinder of amblyopic eye, while in the myopia group, SE of amblyopic eye and difference of SE between two eyes were signi cant factors. However, the signi cance disappeared as the other factors were adjusted in each SE group, which implies that SE type was more important than other refractive error-related variables and other factors unrelated to refractive error were more important for treatment success than the refractive error itself within each SE group.
Previously, Hussein et al. reviewed the record of 104 children aged 3 to 8 years with anisometropic amblyopia, and found that neither the type or amount of refractive error nor the difference in the refractive power between the two eyes was a signi cant risk factor for treatment failure 15 . Myopic anisometropia were only 22% in their study population, and rest were classi ed as hyperopic anisometropia. They reported age above 6 at the onset of treatment, and worse than 20/200 initial BCVA of amblyopic eye as failure risk factors. These results are in accordance with our negative results about refractive factors, also with the age and BCVA of amblyopic eye at presentation as signi cant factors for treatment success in hyperopia and myopia groups. On the contrary, Cobb et al. reviewed 112 children with anisometropic amblyopia who treated with spectacles and patching, and reported that the age at presentation had no effect on the nal visual outcome, while the amount of refractive error and degree of anisometropia do correlate strongly with nal visual acuity 16 . It is notable that most (87%) of their study population were hyperopic and the myopic anisometropia were only 12.5%. Kirandi et al. reviewed 64 children aged 7-9 years with anisometropic amblyopia who were treated with spectacles and patching, and reported that refractive error of SE > + 3D in the amblyopic eye was a risk factor for treatment failure 6 . The study population of Kirandi et al. was also mostly comprised of hyperopic anisometropic subjects (n = 60, 93.7%). The results of Cobb et al. and Kirandi et al. may be comparable to our results of hyperopia group. In our hyperopia group, however, the age at presentation was a signi cant factor, and neither the amount of SE nor the difference of SE were a signi cant factor for treatment success. These different results may be due to different de nition of treatment outcome, and method of analyses. Our study has its merits to analyze the effect of SE amount or the SE difference on the amblyopia treatment success in the hyperopia group, and adjusted factors other than refractive errors, such as follow-up duration or concomitant strabismus. Pang et al. prospectively analyzed myopic anisometropic amblyopia patients, and found that the nal VA in the amblyopic eye was associated with the VA in the amblyopic eye at baseline and the amount of anisometropia 10 . The improvement in VA with patching was inversely associated with patients age. In the similar manner, the age at presentation and the VA of amblyopic eye at rst exam were signi cant factor for treatment success in our myopia group. However, the difference of SE was not a signi cant factor in our study.
Despite different treatment success rate according to SE type of amblyopic eye in our study, the SE amount of amblyopic eye at rst exam was not a signi cant factor in all of the SE groups after adjusting potential confounding variables. Similarly, the difference of SE between the two eyes was not signi cant factor in all SE groups in the multivariate analysis, neither. Although the amount of SE difference is a well-known risk factor for developing amblyopia 17 , our results showed that after adequate optical correction and amblyopia treatment, it is not a signi cant factor for treatment success in each SE group. Usually we prescribed glasses after cycloplegic refraction at initial visit, and waited for 1 to 2 months to see the effect of optical adaptation. If there was no sign of vision improving, patching or atropine penalization was initiated. Twenty subjects (5.19%) achieved treatment success without patching or atropine penalization.
The cylinder value of the amblyopic eye was also not a signi cant factor for all SE groups in multivariate analyses. The emmetropia group had a larger cylinder value than those of the hyperopia and myopia group. The emmetropia group subjects in this study may have been classi ed as astigmatic or mixed astigmatic amblyopia in other studies. Our emmetropia group showed higher treatment success compared to the myopia group, and comparably similar to the hyperopia group. This may imply that SE is more important than the cylindrical amount itself, in other words, even if the astigmatism is large, treatment outcome may be good, as long as the SE is close to emmetropia. There are con icting studies on the impact of astigmatism for the treatment success in amblyopia. Hussein et al. reported that eyes with signi cant astigmatism were less likely to achieve successful outcomes in cases of anisometropic amblyopia 18 , while others reported that the degree of astigmatism is not a signi cant factor for treatment outcome 6,19 .
It is interesting that the odds of treatment success in the hyperopia group were associated with the cylinder value of the sound eye.
There was a moderate correlation between the cylinder of sound and amblyopic eyes in all SE group (Pearson's correlation coe cient 0.555, p = 0.001, data not shown). However, the cylinder value in the amblyopic eye was not a signi cant factor for treatment success in any of the three SE groups. We speculated that the larger cylinder value of sound eye would make the cylinder difference between two eyes smaller, however, there was only weak correlation between cylinder value of amblyopic and sound eye (Pearson's correlation coe cient − 0.163, p = 0.001, data not shown). In the multivariate analysis including the cylinder difference between two eyes instead of the cylinder in the amblyopic eye considering multicollinearity, the cylinder difference was not a signi cant factor for treatment success in any SE group as well. Further studies are needed to interpret these results.
Age at presentation was a signi cant factor for amblyopia treatment success in hyperopia and myopia group of our study. The younger at presentation, the more likely it was to have the greater odds of treatment success in anisometropic amblyopia. This result agrees with many previous reports, which have found the better visual outcomes in younger than older patients 9,15,20,21 . However, some authors insisted that the age at presentation had no effect on the nal visual outcome. 7,16,22 These various results might be due to the different de nition of treatment success. It is notable that the age at presentation was not a signi cant factor for treatment success in our emmetropia group. One study about astigmatic amblyopia reported that the age at presentation did not in uence nal visual acuity 19 . Even though our emmetropia group may comparable with usual astigmatic anisometropia group from other study due to its large cylinder value, but our emmetropia group only included subjects who had SE is close to emmetropia. We may speculate that emmetropic SE is a strong prognostic factor that can overcome the age.
BCVA of amblyopic eye at the rst visit was also a signi cant factor in both the hyperopia and myopia groups. This study showed that the chance of achieving treatment success in both groups depended on the visual acuity of the amblyopic eye at the rst visit. These results are consistent with those of previous studies. 7,18−20 On the contrary, in the emmetropia group, BCVA of the amblyopic eye was not a signi cant factor affecting treatment success. This result might be due to the fact the initial BCVA of amblyopic eye in emmetropia group was better than the other two SE groups.
Follow-up duration had a signi cant positive effect on treatment success in the hyperopia group only. The longer follow-up duration made the odds of treatment success increase by 1.1 times. It can be suspected that the longer follow-up duration is associated with good compliance and selection bias, therefore yield good treatment results. However, follow-up duration is not a signi cant factor for the emmetropia and myopia groups whereas their follow-up durations were not signi cantly different, therefore different SE group demonstrated different treatment response by follow-up duration.
Strabismus was a signi cant poor prognostic factor for treatment success only in the hyperopia group. The hyperopia group subjects with strabismus had odds of treatment success decreased by 0.25 times than the hyperopia group subjects without strabismus. In myopia group only one subject had vertical strabismus (1.47%). Fifty-four percent of emmetropia group had strabismus, but strabismus was not a signi cant factor for treatment success. There were contradictory reports about the effect of strabismus on amblyopia treatment results. Some reported strabismus is not a signi cant risk factor 7,8,18 , while others found it as a signi cant failure factor 16,23 .
Further study with larger number of subjects is warranted to clarify the in uence of strabismus on each refractive type of anisometropic amblyopia treatment result.
The results that there were no signi cant factors related to treatment success in the emmetropia group might need some interpretation. We speculated that the relatively small number of patients with very high success rates made lack of diversity to predict success and failure in this group.
We focused on achieving treatment success any time during the follow-up, not the success at the last visit in our study. De ning treatment success at the last visit will re ect uctuation of visual acuity such as recurrence with or without recovery after recurrence. In our study, treatment success rates at the last follow-up were also signi cantly different among the SE groups (96.61% for emmetropia, 86.30% for hyperopia, and 77.94% for myopia group, respectively, p = 0.001 by Chi-square test, data not shown), whereas treatment success rate during the follow-up and at the last exam were not signi cantly different whithin each SE group (p = 0.055, 1.0, 0.519 for the hyperopia, emmetropia, and myopia group respectively, p-values by Chi-squared test, data not shown). There were 14 patients (5.19%) in the hyperopia group, and 3 patients (4.41%) in the myopia group who had once achieved treatment success, but failed to maintain it at the last exam. In a prospective follow-up study after treatment cessation in children with successfully treated amblyopia due to anisometropia, strabismus or both, the risk of amblyopia recurrence was reported to be 24% within the rst year off treatment 25 .
In other retrospective study also reported recurrence of amblyopia after a cessation of occlusion therapy was 27% within the rst year off treatment 26 . Relatively lower recurrence rate of our result than previous studies may be due to given active treatment during the whole follow-up period in our subjects, such as re-occlusion therapy. It is also interesting that the emmetropia group had no recurrence at all, while the other two SE groups had some. Due to the small number of recurrent cases and possibility of visual acuity uctuation during the follow-up, we de ned our primary end-point as the treatment success rate during the follow-up.
There are some limitations in our study. Due to the retrospective study design, treatment method and duration were not able to controlled. Also, there was a limitation to evaluate the compliance of patching and wearing glasses. However, this study has its merit that a large scale study population recruited from a secondary referral eye hospital which would re ect the general population better than the tertiary referral hospitals. The goal of this study was to compare the treatment outcome depending on the SE type of amblyopic eye in anisometropia amblyopia in real world pediatric ophthalmology clinic. Also this is one of a few studies to evaluate the factors affecting amblyopia treatment success according to the refractive error type of amblyopic eye.

Conclusions
In conclusion, the amount of refractive error and astigmatism of the amblyopic eyes, and the difference in refractive errors between the two eyes were not signi cant factors affecting the amblyopia treatment success after adjusting other factors in all SE groups, in anisometropic amblyopia patients who visited a secondary referral eye hospital in South Korea. However, the treatment results and the factors affecting amblyopia treatment success were different according to the refractive error type of amblyopic eyes, therefore the SE type of the amblyopic eyes is the prognostic factor.

Declarations
Author contributions : The authors con rm contribution to the paper as follows: study conception and design: SHB, ESK, DHK; data collection: SHB, ESK, DHK; analysis and interpretation of results: SHB, DHK, DDC; draft manuscript preparation: DDC; All authors reviewed the results and approved the nal version of the manuscript.References