Given the (1) lack of a standardized CVI screening and assessment method, (2) the retrospective nature of this study, and (3) integration of both neurology and neuro-ophthalmology medical record data, a custom CVI severity grading system was established. Overall visual function was graded: (0) no blink to light from a transilluminator; (1) blink to light; (2) fixate and follow; (3) objective vision, assessed via Teller Acuity Cards or optokinetic drum response (binocular testing with a manual striped drum rotated at 8–10 rpm); and (4) subjective visual acuity. Additional single points were given for positive visual field/preference noted on exam, response to kinetics/objects (other than OKN drum), light gazing, and color preference (up to 4 additional points altogether). The CVI grading score was totaled for each patient and grouped into three categories of Severe (0–2 points), Moderate (3–5 points), and Mild (6–8 points), with lower scores reflecting more impairment.

This study also sought to determine whether particular interventions targeted at improving visual function in patients contributed toward an improvement in our CVI grading score. A total of four interventions were assessed, which included: (1) Early Childhood Intervention (ECI) visual therapy services to help children and their families learn daily environmental adaptations that can maximize a child’s functional vision; (2) physical and occupational therapy (PT/OT) which involved motor development exercises that incorporated visuospatial training; (3) refractive correction with eyeglasses prescribed by cycloplegic retinoscopy according to the preferred practice patterns by the American Academy of Ophthalmology (AAO, 2021); and (4) a combination of all therapies. Of note, subjects undergoing refractive correction followed with their preferred provider and were not monitored for compliance or changes in their refraction by this study. The treatment effect of these interventions was assessed longitudinally via the CVI grading score on a per-encounter basis.

For descriptive analysis, categorical variables were reported as count and percentage. Chi-square test or Fisher’s exact test (for contingency tables with at least 20% expected frequency < 5) were used to test the associations between categorized CVI grade outcomes (mild, moderate, and severe) and categorical variables.

To assess the treatment effect of CVI interventions, a mixed-effects model was used to model the longitudinal CVI grade scores over time to account for the multiple measurements from each patient. Baseline characteristics were tested one at a time. Perinatal characteristics, history of cerebral palsy, genetic diagnosis, and abnormal fundoscopy findings were pre-specified as candidate confounders for the outcome. The confounders that were found to be significantly associated with CVI grade outcome (p < 0.05), and the longitudinal measured variables at each visit (age and presence of epilepsy) were then included in a multivariable mixed-effects model. The type of treatment was the main variable of interest and was kept in the model regardless of its significance. Those that reached a significance of p < 0.1 were retained in the final multivariable model. The treatments being separately evaluated were: (1) Therapy documented at each visit, (2) Therapy documented in any one visit throughout the follow-up period (patient-level variable), (3) PT/OT documented at each visit, and (4) Glasses/refractive error correction documented at each visit.

No adjustments for multiple testing were made due to the exploratory nature of the study. A 2-sided p value < 0.05 was used to determine the significance of variables in all analyses. All statistical testing was performed using SAS 9.4.

It is well known that CVI impacts vision on a clinical spectrum of mild visual disturbances to profound vision dysfunction. Despite devising a metric-based CVI grading system aimed at classifying mild, moderate, and severe instances of vision impairment, we found that no pediatric patients in our cohort met clinical criteria in their baseline ophthalmic examinations to be designated as having mild CVI. Rather, our cohort only describes moderate and severe cases of CVI and may reflect the fact that our larger patient population represents a tertiary referral pool with a skew toward having more severe underlying conditions and comorbidities requiring subspecialty evaluation. Conversely, pediatric patients who meet criteria for mild CVI may present with high-order visuospatial impairment that are underdiagnosed or unrecognized for appropriate workup, diagnosis, and referral (van Genderen et al., 2012; Chandna et al., 2021).

Despite not capturing the milder clinical spectrum of CVI, our scoring system helped stratify moderate from severe disease. Additionally, more than half of our cohort had severely decreased vision evident on visual function testing with 23.3% (n = 58) demonstrating an absent BTL response and 34.5% (n = 86) with only a BTL response and no further visual capability. However, CVI can also present with additional ocular comorbidities including strabismus, ocular dysmotility, and optic nerve/retinal changes. Although we did not assess for strabismus due to the variability of the documented exam, the dilated fundus exam detected optic nerve abnormalities and chorioretinal lesions (scarring or lacunae) in our cohort. Optic nerve pallor was found in 17% (51/249) of our cohort, was noted to always present bilaterally, and ranged from isolated temporal to diffuse pallor. The presence of optic nerve pallor in these CVI children could result from either co-morbid anterior visual pathway dysfunction (e.g., congenital optic nerve abnormalities) or retrochiasmal pathology involving the optic radiation (e.g., periventricular leukomalacia) and resultant trans-synaptic degeneration (Dutton, 2003; Hoyt, 2003; Good, 2007; Lennartsson et al., 2018). Overall, our ophthalmic findings in this study are consistent with prior works that reported the prevalence of optic nerve pallor (16–42%) in CVI children (Huo et al., 1999; Hoyt, 2003; Khetpal and Donahue, 2007; Handa et al., 2018).

In the 160 children of our cohort born at term, 21.3% presented with neonatal seizures, 13.1% with HIE and 3.1% with IVH. Prior studies have reported that HIE is the most common cause of CVI in preterm and term children (Good et al., 1994; Huo et al., 1999; Fazzi et al., 2007; Khetpal and Donahue, 2007). Interestingly, our cohort reported a higher prevalence of neonatal seizures than HIE in both term and preterm children as an etiology of CVI. In addition, the incidence of neonatal seizures (n = 54) in the perinatal period in both preterm and term children had a higher propensity for developing into epilepsy (n = 51/54), and to a lesser degree, epileptic encephalopathy (n = 16/54). This observation suggests that neonatal seizures may be a potential prognostic sign for a systemic disease course and raises concerns for potential development of CVI.

The distinction of prematurity, defined as a child born before 37 completed weeks of gestation, is a critical feature of characterizing CVI in children. With 35.7% (n = 89) of our cohort meeting criteria for preterm birth, neurologic clinical assessment (clinical history, APGAR score, blood gas, Sarnat staging—classification scale for HIE) and/or initial imaging demonstrated co-morbid pathologic findings consistent with IVH (25.6%; n = 23/89), neonatal seizures (23.6%; n = 21/89), and HIE (21.3%; n = 19/89). As IVH and HIE are potential contributing precursors of periventricular leukomalacia (PVL), our findings are consistent with prior studies describing an association with preterm babies presenting with PVL on neuroimaging (Banker and Larroche, 1962; Jacobson and Dutton, 2000; Dutton, 2013). Finally, retinopathy of prematurity (ROP) was found in 21.3% (n = 19/89) of our cohort. Per medical record review, these children were monitored postnatally but none experienced any vision threatening sequalae.

While neuroimaging alone cannot diagnose CVI, it does provide valuable anatomical insights into developmental etiologies. A strength in our study was the fact that a majority of patients (211/249) had existing neuroimaging at the time of their initial ophthalmologic evaluation, and a vast majority had abnormal findings. Neuroimaging in CVI can reveal a spectrum of pathology ranging from focal to global brain involvement (Whiting et al., 1985; Eken et al., 1996; Ortibus et al., 2009). In our cohort, 93.8% presented with an abnormal finding on neuroimaging, the most common being structural anomalies suggestive of an underlying genetic etiology (e.g., migrational abnormalities) in 28.8%, followed closely by atrophy (unspecified cause, 28.2%). Outright presumed sequelae of prematurity (e.g., periventricular leukomalacia), and vascular insults in any location were also frequently reported, 22.7 and 23.7%, respectively (Bauer and Papadelis, 2019). In our study, no association was found between the type and location of identified neuroimaging findings on MRI, and the degree/severity of CVI at time-of-diagnosis, consistent with prior findings (Cioni et al., 1996; Chang and Borchert, 2020). This would suggest that, while structural neurologic injury plays a role in the pathogenesis of CVI, there may be other functional tests and characteristics of neuroimaging that better account for the severity of CVI and perhaps warrant more attention than baseline imaging (Fazzi et al., 2009; Ortibus et al., 2009). This observation is consistent with the notion that standard neuroimaging modalities are limited in their capability to correlate visual function with neuroanatomical changes in space and time. However, recent developments in novel techniques such diffusion tensor imaging and high angular resolution diffusion imaging (HARDI) will allow future investigators to closely examine white matter connectivity in the context of dysfunction and better elucidate potential pathways involved in CVI (Ortibus et al., 2009; Bauer et al., 2014; Martín et al., 2016).

A striking majority of our CVI patients presented with either some form of developmental delay (DD) or cerebral palsy (CP) (231/249, 93.1%). We also found in our longitudinal cohort that CP at initial diagnosis was significantly associated with worsening CVI grade outcome over time. This is suspected to be due to multiple reasons. Both DD and CP are clinical diagnoses that may be comorbid, and children with CVI commonly have more than one underlying neurologic disorder (Castano et al., 2000; Fazzi et al., 2007; Handa et al., 2018). Additionally, DD and CP represent a large group of heterogeneous conditions that not only have highly variable etiologies, but in fact share many of them with CVI (e.g., prematurity, developmental and epileptic encephalopathies, and known genetic conditions).

As many as 26.1% of our patient cohort had a formal genetic diagnosis, including large chromosomal abnormalities (trisomy, deletions, duplications) and single gene pathogenic variants. The extent of our cohort’s abnormal genetic findings exceeded that of prior studies (Matsuba and Jan, 2006; Bosch et al., 2014). Despite this, we still believe this underestimates the true incidence of genetic anomalies among children with CVI. Most of our genetic testing was obtained through a separate neurology clinic, where the focus was on other comorbid neurologic conditions (CP, DD, epilepsy), rather than CVI. Instead, a dedicated rigorous approach in genetic testing has the potential to uncover a larger number of etiologic targets. Our study nonetheless identified genes that have previously been implicated in CVI pathogenesis (CDKL5, SLC35A2, LIS1) (Bosch et al., 2014), as well as those not described elsewhere in the literature (COXPD11, EIF2B5, FLAD1). Despite these findings however, our study ultimately did not reveal a significant correlation between CVI severity/improvement and underlying genetic diagnosis. Instead, a rigorous and unbiased approach that examines a more homogeneous patient population (i.e., no neurological comorbidities) may be valuable.

Most individuals of our cohort had a diagnosis of epilepsy at time of initial ophthalmologic evaluation (78.3%, n = 195/249). Among these, 33.8% had an identified form of epileptic encephalopathy (cerebral dysfunction related to often difficult-to-control epileptic activity), with infantile spasms/hypsarrhythmia being most common (83.3% of said group). This is likely related to the timing of initial ophthalmologic diagnosis vis-à-vis the most recent electroencephalographic studies. Nonetheless, these clinical findings are consistent with prior reports that have noted a significant seizure component in pediatric CVI patients, the majority of which were infantile spasms (Huo et al., 1999; Khetpal and Donahue, 2007; Handa et al., 2018).

With regards to seizure control, there is contradictory evidence in which earlier studies have reported improved visual development with improved seizure control, and a correspondingly poor prognosis for children with uncontrolled seizures (Wong, 1991; Good et al., 1994). However, it has also been recently reported that seizure control did not significantly improve vision, as well as evidence that seizure may not be contributory to CVI (Grant et al., 2008; Handa et al., 2018).

In the epilepsy group, 75% of patients had optimal seizure control at time-of-diagnosis of CVI. The mean CVI grade was more severe for patients with uncontrolled epilepsy (1.19) than controlled epilepsy (1.54). A trend toward mean CVI grade improvement was seen with improved seizure control. Interestingly, the use of multiple antiseizure medications (3 or more antiseizure medications) correlated with an increased baseline severity of CVI (Table 2). This may be indicative of the severity of the underlying epilepsy syndrome, and possibly a more reliable marker of epilepsy disease burden than subjective reporting of optimal or poor control. Additionally, the use of multiple antiseizure medications was also negatively associated with longitudinal CVI outcome, and suggests that seizure activity reflects a cumulative burden on CVI prognosis over time. Of additional interest, a number of patients with a history of infantile spasms or other early epilepsy syndromes may have received medications known to impair visual function (e.g., vigabatrin); however, given the lack of usual related findings (e.g., visual field loss), this may not sufficiently explain the difference in CVI grade across groups. Practically, it is important for the vision care provider to consider the number of seizure medications and whether if it may be affecting the child’s performance during a CVI vision assessment. This would importantly pair up with the increasing tendency toward early aggressive seizure/epilepsy management among neurologists, including offering earlier curative or palliative epilepsy surgery options and few medication trials (Prideaux et al., 2018; Roth et al., 2021; Perry et al., 2022).

Presently, there is no standardized treatment for CVI. Given its heterogeneous nature and the broad range of visual impairment, it has been difficult to design and execute studies determining if a particular therapy or intervention is effective. Additionally, the paucity of evidence on the clinical effectiveness of CVI-based therapy makes our study unique in that our cohort was longitudinally followed for treatment effect after receiving multimodal interventions. It has been previously reported that improvements in vision in CVI patients were not associable to any unique etiology (Huo et al., 1999; Handa et al., 2018). Accordingly, our study cohort received therapies targeted at CVI visual improvement, including ECI vision therapy, visual rehabilitation (PT/OT), refractive error correction, and combination therapy.

Out of the therapy modalities that were offered to patients, we found that refractive correction with glasses was associated with the greatest improvement in CVI grade outcome over time. This is compared to ECI, PT/OT, or a combination of all therapies, which did not demonstrate significance with longitudinal CVI grade. Correcting refractive error is pursued for multiple pediatric conditions, and our study provides evidence that it is viable therapy for moderate and severe pediatric CVI as well. Prior studies have also identified treatable associated ophthalmic conditions (e.g., refractive error, accommodative insufficiency, cataract) in a significant proportion of CVI patients (Philip and Dutton, 2014; Pehere et al., 2018). Taken together, our evidence suggests that patients ought to pursue management for treatable ophthalmic conditions that co-present with CVI.

As an aggregate, having therapy documented at any visit was strongly associated with an average improvement in CVI grade of 0.14 per month, compared to no therapy. This was additionally supported at the patient level, whereby having at least one therapy recorded at any visit during a patient’s follow-up period was also associated with a mean improvement in CVI grade of 0.25 per month. While refractive error correction as mentioned above may represent the majority contribution toward this improvement, the natural development of the visual system in pediatric patients may also play a role. In our multivariate model, age played a small but significant positive contribution toward CVI grade outcome, but the overall effect size is insufficiently large to explain the total improvement afforded by receiving CVI therapy during follow-up.

Our study has multiple limitations that are common in retrospective studies. The CVI grading system described in this study was an attempt to incorporate subjective and objective clinical data, but may not have encompassed a true and robust quantification of visual function and behavior. A grading system developed after patient data collection may help better refine our CVI scale for future prospective studies. Given the retrospective nature of this study, certain aspects of data collection may be incomplete (information bias). Particularly, our data from genetic testing arose from workup targeted at separate underlying conditions (e.g., epilepsy), and may not represent the optimal context to identify causative genes associated with CVI. Due to the retrospective nature of the study, therapy modalities were not implemented in an evenly distributed manner. Additionally, our unbalanced sample sizes for our moderate vs. severe CVI groups (n = 52 vs. 197) reduces the statistical power.