Amblyopia risk factors in congenital nasolacrimal duct obstruction: A longitudinal case-control study

Purpose To investigate longitudinal changes in risk factors for amblyopia in children treated with congenital nasolacrimal duct obstruction (CNLDO). Methods Retrospective observational case control study. A total of 446 children under 4 years of age who underwent probing and/or intubation for CNLDO between January 2004 and January 2018, and 446 age-matched controls were included. Cycloplegic refraction and ocular alignment were investigated at the time of treatment and after at least one year of symptom improvement. Children were classified as having amblyopia risk factors on the basis of the American Association for Pediatric Ophthalmology and Strabismus guideline in 2013. Main outcome measures The prevalence of amblyogenic refractive errors, and determinants associated with the presence of amblyogenic refractive errors in CNLDO patients. Results The prevalence of amblyogenic refractive errors in CNLDO patients (5.4%) was similar to that of the control group (6.5%) (P = 0.571). After one year of symptom improvement in CNLDO patients, the prevalence of amblyogenic refractive errors was 4.7%. There was no difference in the prevalence of amblyogenic refractive errors between unilateral and bilateral CNLDO patients. Multivariate analysis revealed that manifest strabismus was the only risk factor related with the presence of amblyogenic refractive errors (odds ratio = 6.383, confidence interval = 1.205–33.826, P = 0.029). Conclusions This study found no evidence to suggest that the prevalence of amblyopia risk factors is higher in CNLDO patients compared with normal controls. Manifest strabismus was the only determinant associated with the presence of amblyogenic refractive errors.


Introduction
Congenital nasolacrimal duct obstruction (CNLDO) appears in 5 to 15% of full-term infants [1,2]. CNLDO is characterized by constant tearing and intermittent discharge occurring in one or both eyes [1,2]. Several authors have reported a possible association between CNLDO and amblyogenic refractive errors, and it is generally accepted that the risk for amblyopia is higher in patients with CNLDO [3][4][5][6]. Persistent tearing induces blurred vision that interferes with emmetropization resulting in anisometropia which is a strong amblyogenic risk factor [3].
Among previous studies conducted on CNLDO patients, no large-scale study has evaluated the prevalence of risk factors for amblyopia in patients with CNLDO who underwent probing and/or intubation [3,4,[7][8][9][10][11]. A better understanding of the effects of CNLDO on amblyopia can be obtained by longitudinal observations of amblyopia risk factors, especially after symptom improvement. Furthermore, most of the previous reports did not include a control group [7,[9][10][11][12][13]. Only one study compared the prevalence of amblyopia between CNLDO and controls [6], however, CNLDO spontaneously resolved without treatment in most patients which leaves some doubt regarding the effect of CNLDO on the development of amblyopia.
Therefore, we aimed to investigate longitudinal changes in the prevalence of risk factors for amblyopia among patients with CNLDO after treatment, and to evaluate the clinical characteristics associated with amblyogenic refractive errors in this patient population.

Materials and methods
The institutional review board of Seoul National University Bundang Hospital approved the study (B-1802-453-102). This study was carried out in accordance with the recommendations of the Declaration of Helsinki for biomedical research involving human subjects. Informed consent was not given, as all patient records were provided to/accessed by the authors in a fully anonymized format. The institutional review board of Seoul National University Bundang Hospital waived the requirement for informed consent of parents.

Patients
We performed a retrospective review of patients with CNLDO under 4 years of age, who underwent either probing or intubation (by NK or HKJ) at Seoul National University Bundang Hospital, and Seoul Municipal Government-Seoul National University Boramae Medical Center between January 2004 and January 2018. A routine ophthalmologic examination including cycloplegic refraction and ocular deviation measurement was performed at the time of treatment and during annual follow-up examinations afterwards. Exclusion criteria were as follows; prematurity (< 37 weeks gestational age), low birth weight (1,500 to 2,499 g) [14], and any ocular deformity such as ptosis or keratopathy which could affect refractive errors. Children with systemic syndromes, severe mental disability [15], craniosynostosis [16], or hydrocephalus [16], which are known to be related to refractive errors were also excluded [17].
We randomly selected an age-matched control group from patients who visited the outpatient clinic of Seoul National University Bundang Hospital for routine eye examination. Children without CNLDO and with no ocular disease other than refractive errors were included. When both eyes of a patient met the eligibility criteria of age-matched controls, one eye was randomly selected for statistical analysis.

Clinical evaluation
Data were collected on demographics and clinical characteristics, including patient's gestational age, birth weight, gender, age at diagnosis, number of probing and/or intubation, and age of treatment. Medical records were reviewed for ocular alignment and motility, cycloplegic refractive errors, and anterior and posterior segment findings. The initial refractive errors were recorded during the first ocular examination for all patients with cycloplegic refraction. Cycloplegic refraction was performed by waiting for 30-40 minutes after applying 1% cyclopentolate three times in each eye every 5 minutes. Ocular deviation was measured with the alternate prism and cover test (APCT), or the Krimsky test for subjects unable to perform the APCT. The final refractive errors were defined as the most recent cycloplegic refraction performed at least one year after symptom relief. We investigated the prevalence of risk factors for amblyopia based on the American Association for Pediatric Ophthalmology and Strabismus (AAPOS) guideline, revised in 2013 (Table 1) [18].

Data analysis and statistics
For statistical analysis, SPSS ver. 21.0 software (IBM Corporation, Armonk, NY, USA) was used. Demographic data (e.g., gender and age) were compared between CNLDO patients and controls using the chi-square test and the unpaired t test. Comparison between patients with CNLDO versus (vs.) controls and unilateral vs. bilateral CNLDO, were assessed using the independent t test or Pearson correlation test for continuous variables and chi-square test for categorical variables, as applicable. Bivariate logistic regression was performed in terms of the presence of amblyogenic refractive errors as the dependent variable to evaluate the relationship between related variables. Predictors with a p-value of 0.1 or less in univariate analysis were included as a candidate in multivariate analysis. A p-value < 0.05 was considered statistically significant. Data are presented as mean ± standard deviation unless stated otherwise.

Subject characteristics
This study initially enrolled 1037 patients with CNLDO who underwent treatment; 591 were excluded due to incomplete ocular examination (n = 526), out-of-age criteria (n = 27), Table 1

. Amblyopia risk factors based on the American Association for Pediatric Ophthalmology and Strabismus (AAPOS) original referral criteria in 2013.
Refractive amblyopia risk factor targets � Age, months Astigmatism Hyperopia Anisometropia Myopia Non-refractive amblyopia risk factor targets †

Comparison of CNLDO patients and age-matched controls
Overall, CNLDO patients had a similar prevalence of amblyogenic refractive errors compared with controls (5.4% vs. 6.5%, respectively; P = 0.571) ( Table 3). There was also no significant difference in the prevalence of each type of refractive errors between CNLDO patients and controls (all P>0.5) (

Comparison of unilateral and bilateral CNLDO
Among 446 patients with CNLDO, 96 (21.5%) had bilateral CNLDO. The mean age of treatment was not significantly different between patients with unilateral CNLDO and bilateral The interocular differences of spherical equivalent (SE) refractive errors and astigmatism were compared in 350 children with unilateral CNLDO. There was no significant difference in SE refractive errors (D, diopters) (P = 0.793) and astigmatism between the affected eye (2.00 ± 1.09 D) and unaffected eye (1.45 ± 1.03 D; P = 0.156) ( Table 4). In patients with unilateral CNLDO, 6 of 16 cases (37.5%) with amblyogenic refractive errors exhibited greater hyperopia in the affected eye compared with the fellow eye (Table 4). Table 5 summarizes factors associated with the presence of amblyogenic refractive errors in patients with CNLDO. In univariate analysis, performing intubation (odds ratio [OR] 4.264, 95% confidence interval [CI] 1.446-12.570; P = 0.009) and the presence of manifest strabismus (OR 6.162, 95% CI 1.215-31.248; P = 0.028) were significantly associated with the presence of amblyogenic refractive errors. In multivariate logistic regression models, in which no significant effects were eliminated, only the presence of manifest strabismus was significantly associated with the presence of amblyogenic refractive errors (OR 6.383, 95% CI 1.205-33.826; P = 0.029) ( Table 5).

Discussion
The prevalence of amblyogenic refractive errors in the general population ranges from 10.9% to 20.2% [19][20][21][22]. The present study demonstrated that 5.4% of children with CNLDO exhibited amblyogenic refractive errors, which was similar to the frequency (6.5%) in normal age-matched controls. Furthermore, this was the first study to determine longitudinal changes in the prevalence of amblyogenic refractive errors in CNLDO patients who were successfully treated. The prevalence of amblyogenic risk factors did not change after at least one year of symptom relief. Finally, the presence of manifest strabismus was the only factor associated with amblyogenic refractive errors in children with CNLDO. Previous studies have been inconclusive about the relationship between CNLDO and amblyopia. Some reported an increased risk for amblyopia in CNLDO patients [3-5, 9, 12], while others did not [6,11]. It has been suggested that persistent tearing and discharge leads to blurring of vision and interferes with clear focusing of images on the retina, which is critical for emmetropization [5]. In contrast, in a case-control cohort study, Ellis et al. [6] found no evidence to prove that persistent tear film interferes with emmetropization or the development of ocular alignment in CNLDO patients. A previous study examined visual acuity and refractive errors in CNLDO patients after more than 4 years of symptomatic improvement [6]. However, there was no detailed description of the severity of CNLDO and surgical treatment, as well as the prevalence of amblyogenic risk factors according to specific guidelines [6]. To overcome these limitations, in the present study, we limited subjects to children with persistent tearing that required probing or intubation.
Matta et al. [4] reported that 22% of children under 3 years of age with CNLDO exhibited risk factors for amblyopia as defined by the AAPOS vision screening guidelines revised in 2003 [23]. However, the guideline proposed by the AAPOS vision screening committee had not considered age-dependent changes of risk factors for amblyopia [18,23]. That is why older guidelines have a high over-referral rate and significant discrepancy between the rate of amblyopia and that of amblyopia risk factors [13,24,25]. Therefore, we used the newly adjusted guidelines incorporating age-related changes of amblyopia risk factors [18].
Unilateral CNLDO showed significant association with anisometropia compared with bilateral CNLDO in previous reports [5,7,10]. Clinically significant anisometropia is well documented as a critical cause of amblyopia [26]. A previous report revealed that anisometropic amblyopia developed in 5 out of 130 patients with unilateral CNLDO and all had amblyopia on the same side of CNLDO [27]. The study suggested that image deformation caused by mucopurulent discharge and excessive tear film in the early stage of vision development could cause a lack of proper emmetropization [7,27]. However, in the current study, the prevalence of anisometropia in both unilateral and bilateral CNLDO were similar to what is expected in the general population of children, which is contrary to previous reports [5,7,10]. This  [3-5, 9, 12], while our definition was based on the revised guideline in 2013 [18].
In the present study, manifest strabismus was the only significant factor associated with the presence of amblyogenic refractive errors. This is in line with previous reports that showed an association between esotropia and hyperopia/anisometropia [28][29][30]. Our finding supports the hypothesis that anisometropia and hyperopia, as well as pronounced astigmatism, may trigger the development of strabismus in susceptible patients owing to disturbances of fusion [30,31].
This study has a few limitations. Firstly, selection bias can occur in a retrospective study. Follow-up was determined at the discretion of the examiner, and compliance was dependent on the patient's caregiver. As a result, only 30.9% of the original CNLDO population completed follow-up examinations. Among the 19 patients with amblyogenic risk factors at the initial examination, only four patients performed the follow-up examination. This may have caused underestimation of the actual prevalence of amblyopia in this cohort. However, in order to reduce possible errors, comparison of prevalences between CNLDO patients and controls were determined if amblyogenic risk factors were present at least once at the initial and/ or final examinations. Nevertheless, the prevalence of amblyogenic risk factors revealed no statistical difference between patients and controls. Second, the control group was not recruited from the general population, but from those who were referred to our hospital for routine eye examination and had no amblyogenic risk factors including manifest strabismus. Thus, we do not know whether the prevalence of strabismus is higher in patients with CNLDO compared to those without CNLDO. However, the literature addressing the prevalence of childhood strabismus is substantial, [32][33][34] and our study population demonstrated similar values compared with previous studies. In addition, the risk factors for amblyopia in CNLDO patients are mainly related to refraction. Therefore, this does not change our qualitative conclusions. Finally, because all study patients were Koreans, extrapolating these results to other populations may be problematic.
In conclusion, in the present study, we demonstrated that there is no difference in the prevalence of amblyogenic refractive errors in patients with CNLDO compared with age-matched controls, and this tendency did not change even after one year of symptom resolution.