Characteristics and long-term outcomes of childhood glaucoma: a retrospective-cohort study

Purpose: To evaluate the clinical characteristics and treatment outcomes of patients with childhood glaucoma. Methods: We retrospectively reviewed the data of patients with childhood glaucoma who visited the glaucoma clinics at the Queen Sirikit National Institute of Child Health and the King Chulalongkorn Memorial Hospital between January 2008 and January 2018. The diagnosis was based on the Childhood Glaucoma Research Network classification. We recorded their clinical characteristics and requirement of any glaucoma interventions. Results: A total of 691 eyes from 423 patients were included in this study. The patients predominantly comprised boys. The average follow-up duration was 71.3±63.8 months. The mean age at presentation was 3.9±4.4 years. Most patients presented with a high initial intraocular pressure (IOP). The average intial IOP of all patients was 28.5±11.2 mmHg. Glaucoma associated with non-acquired ocular anomalies (22.9%) was the most common subtype, followed by primary congenital glaucoma (20.8%). We recorded a family history of glaucoma in 6.4% of patients of the 234 patients with an available family history. Most patients had bilateral glaucoma (63.4%) and required at least one intervention (51.5%). The average IOP at the latest follow-up visit was 19.1±10.8 mmHg. All glaucoma types had significantly lower IOP, compared to that at their baselines (all p<0.001). Moreover, most patients had an unfavourable visual acuity (49.5%) at their latest visit. Conclusions: Secondary glaucoma associated with non-acquired ocular anomalies is the most common subtype of glaucoma. The majority of patients had unfavourable visual outcomes. These real-world findings are fundamental to acquire a better understanding of childhood glaucoma.


Introduction
Childhood glaucoma is a vision-threatening disorder with an incidence of 2.29 to 5.41 per 100,000 individuals 1,2 . The diagnosis of childhood glaucoma poses some challenges. This can be attributed to the variation in clinical presentations among different age groups. Despite high intraocular pressure (IOP) being the primary cause of glaucomatous damage, an accurate IOP measurement is not always obtained in children. Angle surgery is a common therapy but it is mostly associated with unfavourable outcomes in children, compared to that in adults 3 .
Childhood glaucoma encompasses several categories of glaucoma. The Childhood Glaucoma Research Network (CGRN) classification was proposed by an international consortium of glaucoma specialists in 2013 to standardise the definition of childhood glaucoma subtypes 4 . The prevalence of childhood glaucoma differs among ethnicities, ranging from 1:1,250 to 1:68,254 live births [5][6][7][8][9][10][11][12] . The incidence and clinical characteristics of childhood glaucoma in Thailand have not yet been reported. The Queen Sirikit National Institute of Child Health is one of the largest tertiary centres in Thailand. It is responsible for the treatment of a majority of the complex paediatric cases from all over the country. Paediatric glaucoma clinics have been established by the joint collaboration between the Queen Sirikit National Institute of Child Health and the King Chulalongkorn Memorial Hospital, a university-based hospital. These clinics aimed to treat all paediatric glaucoma cases that were referred to the aforementioned hospitals and have been operated by the same group of glaucoma specialists for more than 10 years. We aimed to describe the clinical characteristics and brief long-term treatment outcomes of the large paediatric glaucoma cohorts of the two major referral centres in Thailand.

Cohort selection
We retrospectively reviewed the medical records of all patients who had been examined at the paediatric glaucoma clinics of the Queen Sirikit National Institute of Child Health and the King Chulalongkorn Memorial Hospital between January 2008 and January 2018. The patient list was extracted from the hospital database to include all individuals that had at least one visit to the paediatric glaucoma clinic during the above period and/or subjects that had the ICD-10-CM diagnostic codes of Q15.0 and all H40 and H42 categories. The inclusion criteria were patients who aged <16 years at the time of the first clinic visit and met the CGRN glaucoma or glaucoma suspect definition 4 . Cases with incomplete medical record precluding the diagnosis were excluded. The CGRN definition of glaucoma and suspected glaucoma has been previously described 4 .

Ethics approval
Ethical approval was obtained from the Research Ethic Committee of the Queen Sirikit National Institute of Child Health and Faculty of Medicine, Chulalongkorn University (REC.041/2562 and IRB.807/61). The requirement for written informed consent was waived due to the retrospective nature of the study.
We recorded the interventions during the follow-up course and final outcomes, including visual acuity (VA) and IOP at the latest available visit in eyes with a confirmed glaucoma diagnosis (diagnosis group 1 to 6). For glaucoma interventions, we reviewed the data to determine if the subjects had received any incisional surgeries (i.e. trabeculectomy, trabeculotomy, and glaucoma drainage device implantation), cyclodestructive laser procedures (i.e. diode transscleral cyclophotocoagulation, diode laser endoscopic cyclophotocoagulation), or a combination of both at any time point of the follow-up period.
The best-corrected VAs were determined using the LEA or Snellen chart at 10 feet or 20 feet, respectively. In contrast, VA was graded by the fixation patterns using a central, steady, and maintained (CSM) technique for patients who were too young to determine the pictures or numbers 14 . The LEA chart symbols were reproduced with permission from Good-Lite Co., Elgin, IL. According to Karr et al. 15 , the fixation pattern of CSM, CSUM, CUSUM and UCUSUM was estimated as the VA of ≥ 20/30, 20/30-20/100, ≤20/300, and ≤5/200, respectively. We extrapolated the Snellen acuity from the fixation grade with a modification from Karr et al.'s method and classified the VA into the following three groups: (1) favourable: best-corrected VA (BCVA) ≥20/70 or fixation grade of CSM, (2) moderately favourable: BCVA=20/70 to <20/400 or fixation grade of CSUM; and (3) unfavourable: BCVA≤20/400 or fixation grade of CUSUM or UCUSUM.

Statistical analysis
The categorical data were presented as counts and percentages. We conducted the Shapiro-Wilk test for the normality of continuous data distribution. The data were reported as means and standard deviations or medians and interquartile ranges, depending on the distribution. We used the analysis of variance to compare the initial IOP among the glaucoma subtypes and the paired t-test to compare the IOP during the initial and latest visit. Furthermore, the Stuart-Maxwell test for marginal homogeneity was used to compare the proportion of matched pairs of the VA during the initial and latest visit. The ordinal logistic regression was performed to identify the factors associated with less favourable VA outcome. Statistical analyses were performed using Stata 13.0 (Stata Corp, College Station, TX, USA). A P-value <0.05 was deemed statistically significant.
SCG-O was the most common subtype, accounting for 23% of the cohort or 29% of the glaucoma cases. In contrast, JOAG was the least common subtype. Figure 1 depicts the frequency of each glaucoma diagnosis.
While 157 glaucomatous eyes (29.5%) underwent an incisional surgery, 68 eyes (12.8%) underwent a cyclodestructive laser surgery. In contrast, 49 (9.2%) eyes required both incisional and cyclodestructive procedures at any time point during the follow-up period. Figure 3 and Table 3 presents the frequency of glaucoma intervention in each glaucoma type.
SCG-A had the highest proportion of favourable VA at the initial (57.5%) and latest (53.4%) visits. However, JOAG had the highest proportion of unfavourable VA at the initial (53.9%) and latest (73.3%) visits. We observed a higher proportion of unfavourable VA at the latest visit in the overall glaucoma cases (P=0.03), compared to that at the initial visit. A worsening of the VA was primarily observed in the SCG-O group. Despite an approach in the shift, it failed to attain a statistical significance (P=0.07). Moreover, the average IOP was 19.1±10.8 mmHg at the latest visit. All glaucoma types had significantly lower IOPs, compared to the baseline values (all p<0.001). Table 4 shows a comparison of the VA and IOP between the initial and latest visits for each glaucoma type.
The complete case analysis of 219 eyes showed that less favorable of initial visual acuity (p<0.001) and high last IOP (p<0.001) were significantly associated with poor visual outcome. In comparison to SCG-C, PCG (p=0.04), JOAG (p=0.01) and SCG-O (p=0.003) diagnoses were significantly associated with poor visual outcome. No significant risk of poor visual outcome was identified for the SCG-S and SCG-A groups when compared to the SCG-C group. (Table 5).
To elaborate more on the progression to unfavorable visual impairment, we further investigated the change of the VA group. Table 6 showed that most patients had no change in the VA group between the initial VA and the last VA. Most eyes with unfavorable last VA had already presented with unfavorable VA. The SCG-C tended to have an improvement of the VA. On the other hand, SCG-O had the highest proportion of worsening VA. None of JOAG had VA group change towards improvement.

Conclusions/discussion
Childhood glaucoma comprises a group of eye disorders that affect children from their birth with a juvenile onset. Our study had an average follow-up of 6 years. SCG-O (22.9%) was the most common subtype, followed by PCG (20.8%) and SCG-A (18.9%). The condition mostly affected boys, with the majority being bilateral cases, similar to previously published results 16 . Primary glaucoma, both PCG and JOAG, and SCG-O commonly require at least one type of glaucoma intervention. Following the treatment, we observed significant IOP improvements in all subtypes. Nonetheless, half of the cases demonstrated unfavourable VA at the final visit.
Refraction, corneal diameter, and cup to disc ratio were the measurements taken at first visit.
* Data available in 234, 35,5,13,15,56,50, and 60 patients for total, PCG, JOAG, SCG-C, SCG-S, SCG-O, SCG-A, and GS, respectively. † Data available in 93, 1, 2, 16, 5, 5, 19, and 45      However, reports from Canada 6 , Egypt 9 , Great Britain and the Republic of Ireland 2 , and China 7 have found that the majority of cases comprised PCG. The prevalence of secondary glaucoma depends on the prevalence of its aetiology (e.g. childhood cataract, hereditary systemic disease), which can differ among regions and ethnicities. Furthermore, this variation can be explained by the diagnostic criteria and study design. The diagnostic criteria of the studies conducted before 2013 were not based on the CGRN classification. In addition, hospital-based studies tend to have a higher proportion of complex cases, such as patients with syndromic and systemic involvement than those conducted in population-based settings. Table 7 summarises the distribution of childhood glaucoma according to the CGRN classification 1,2,[5][6][7][9][10][11][12] .      Despite the association of PCG and JOAG with certain mutations 20 , our cohort revealed a positive family history in none of the PCG cases and in only one JOAG case. The true frequency of familial glaucoma, however, may be higher, as the data were available for only 55% of our patients. In addition, getting a family history without examining each family member tends to underestimate the actual occurrence of glaucoma in the family.

Subtypes of Childhood Glaucoma
In line with the published literature, we found that surgical interventions were mostly required in the primary type of glaucoma, both PCG and JOAG. Moreover, medication was the mainstay of treatment for most secondary glaucoma cases 2 . In addition, SCG-O cases reported a high rate of surgical intervention 5 . The pathology of the above-mentioned subtype is related to angle dysgenesis, which usually makes it difficult to control the disease.
Although PCG, JOAG and SCG-O had the high rate of surgical intervention, a high proportion of eyes that had no glaucoma interventions was observed in our study. The primary reason for not receiving surgical intervention was that the IOP could be controlled with medications. All patients with evidence of high IOP were offered surgical intervention (trabeculotomy and goniotomy) for PCG. The option of surgical intervention was discussed with the families of those who had previously been treated with medication. After seeing the well-controlled IOP, some parents preferred to continue with the medication. Unlike PCG, the surgical intervention was not necessarily offered to all SCG-O and JOAG patients. The decision to perform surgery in SCG-O was mainly based on the IOP and visual prognosis. Conservative treatment with medications was preferred If there was very limited vision potential such as nystagmus or visual acuity of light perception. For JOAG, the approach was quite similar to that for adult glaucoma. The majority of cases that did not receive glaucoma intervention underwent selective laser trabeculoplasty and were able to achieve target IOPs, though with medications.
Ramkrishanan et al. reported a significant improvement of VA, which was sustained for at least four years of follow-up 21 . This was in contrast to the marginally significant worsening of VA observed in our study. This disparity could be attributed to a greater proportion of PCG cases in the study conducted by Ramkrishanan et al. (Ramkrishanan et al. 73.3% vs our study 20.8%). The improvement of VA in their study was attributed to an improved corneal clarity following surgery.
We found that the SCG-A cases had the most favourable VA at the latest visit and the best initial VA. The majority of the cases included steroid-induced glaucoma, trauma and uveitis. In general, acquired conditions might be more controllable than the subtypes related to congenital ocular malformations, such as PCG and SCG-O. However, we found an overall worse VA, compared to that reported in previous publications 6,11,21 . This discrepancy could be explained by the following aspects. First, we documented a high proportion of unfavourable VA during the initial visits. Khitri et al.
reported on an association between poor vision at diagnosis and visual impairment (<20/200) 22 . Second, our cases were diagnosed at an extremely young age, particularly in the PCG (median age 0.5 years) and SCG-O (median age 0.3 years) groups. Studies on the PCG subtype reported on final VA <20/200 in children diagnosed before the age of three months regardless of their IOP levels 23,24 . It was hypothesised that the earlier presentation reflected the poorer development of the angle. In other words, the disease was more severe. Nevertheless, the study design and definition of unfavourable VA differed among the studies. 6 (6.6%) 10 (11%) * We excluded patients who were glaucoma suspect or had an unknown diagnosis. † The diagnosis was reclassified from the original articles to follow the CGRN classification. PCG, primary congenital glaucoma; JOAG, juvenile open angle glaucoma; SCG-C, secondary glaucoma following cataract surgery; SCG-S, secondary glaucoma associated with non-acquired systemic condition; SCG-O, secondary glaucoma associated with ocular anomalies; SCG-A, secondary glaucoma associated with acquired conditions Our study found that a less favorable initial VA and a high last IOP were significantly associated with a poor visual outcome. Most patients had no change in the VA group between the initial VA and the last VA. In addition, the diagnosis of PCG, JOAG, and SCG-O were significantly associated with poor visual outcome. The underlying ocular abnormalities tended to limit the visual potential in SCG-O, the anterior segment dysgenesis in particular. Our SCG-O cases with an unfavourable level at the latest visit either initially presented with an unfavourable VA or had an underlying anterior segment anomaly. It is worth mentioning that the JOAG in our study had a poor visual outcome despite having fewer issues regarding corneal problems or amblyopia than the other glaucoma types. All JOAG cases presented at a very advanced stage of disease with a median C: D ratio of 0.9. Our data demonstrated that the VA and glaucoma stage at presentation were the main factors determining the unfavourable VA, suggesting that early diagnosis and treatment are necessary to prevent this unpleasant result.
It should be noted that unfavourable VA could be a result from a combination of factors other than glaucoma such as underlying ocular pathology, uncorrected refractive error or amblyopia. As all patients were concomitantly seen by pediatric ophthalmologists, we believe that inadequate orthoptic exercise and inappropriate refractive correction would be less the case. However, due to the limitations of a retrospective study, it would be difficult to clearly delineate the cause of poor visual outcome in each patient.
Our study had the strength of being a large cohort study with a long follow-up duration. Our data also represents the majority of childhood glaucoma cases in Thailand. However, it had several limitations. First, there were some incomplete data because of the retrospective design. Moreover, information on some clinical examinations, such as VA and IOP could not always be obtained in children at every clinic visit. Second, the long follow-up period resulted in a shift in the IOP measurement methods from a handheld contact tonometer (Tono-Pen; Reichert, New York, USA) to a rebound tonometer (iCare TAO1i, Tiolat Oy, Helsinki, Finland) in extremely young or non-cooperative children. Third, there was inadequate information to clearly identify the cause of unfavourable VA outcomes. Future research should explore this underlying issue.
In conclusion, data from the referral centres in Thailand showed a higher prevalence of secondary glaucoma than primary glaucoma. Using the CGRN classification, secondary glaucoma associated with non-acquired ocular anomalies was found to be the most common subtype. All subtypes, including primary glaucoma, were sporadic. A majority of the cases had unfavourable visual outcomes. These real-world findings are fundamental data and provide a better understanding of childhood glaucoma.

Open Peer Review
the "Conclusion" that "The condition mostly affected boys, with the majority being bilateral cases, similar to previously published results."b)：In the "Conclusion" that "In addition, SCG-O cases reported a high rate of surgical intervention." In the "Conclusion", you described that "This disparity could be attributed to a greater proportion of PCG cases in the study conducted by Ramkrishanan et al." Please Please provide the PCG ratio of this research and the comparison with that of your study to compare the difference between the two more intuitively.

5.
In the description SCG-A had the most favourable VA at the latest visit and the best initial VA, you described that "The majority of the cases included steroid-induced glaucoma, which might be more controllable than the subtypes related to congenital ocular malformations, such as PCG and SCG-O." But from Figure 2, steroid-induced glaucoma accounts for 29%, which is not so high. Does the authors mean "The majority of the cases such as steroidinduced glaucoma, which might be more controllable than the subtypes related to congenital ocular malformations…"? 6.

If applicable, is the statistical analysis and its interpretation appropriate? Partly
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Yes 4. There are several places in the article that need to be supported by the literature: a)：In the "Conclusion" that "The condition mostly affected boys, with the majority being bilateral cases, similar to previously published results."b)：In the "Conclusion" that "In addition, SCG-O cases reported a high rate of surgical intervention." Response: Thank you for the comment. The following references have been added.
To ( 5. In the "Conclusion", you described that "This disparity could be attributed to a greater proportion of PCG cases in the study conducted by Ramkrishanan et al." Please Please provide the PCG ratio of this research and the comparison with that of your study to compare the difference between the two more intuitively.

Response:
We appreciate the reviewer's comment. We have included the proportion of Ramkrishanan's study in comparison with our study. Conclusions/Discussion -7 th paragraph: "This disparity could be attributed to a greater proportion of PCG cases in the study conducted by Ramkrishanan et al. (Ramkrishanan et al. 73.3% vs our study 20.8%)." 6. In the description SCG-A had the most favourable VA at the latest visit and the best initial VA, you described that "The majority of the cases included steroid-induced glaucoma, which might be more controllable than the subtypes related to congenital ocular malformations, such as PCG and SCG-O." But from Figure 2, steroid-induced glaucoma accounts for 29%, which is not so high. Does the authors mean "The majority of the cases such as steroidinduced glaucoma, which might be more controllable than the subtypes related to congenital ocular malformations…"?

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Partly visual acuity (VA) and high last IOP were significantly associated with poor visual outcome. In comparison to SCG-C, PCG, JOAG and SCG-O diagnoses were significantly associated with poor visual outcome. No significant risk of poor visual outcome was identified for SCG-S and SCG-A when compared to SCG-C.
As all patients were primarily seen by pediatric ophthalmologists (please refer to the response to reviewer 1's query No.3), we believe that inadequate orthoptic exercise and inappropriate refractive correction would be less the case. However, we could not assess the adherence to the treatment and follow-up with the pediatric ophthalmologists.
To elaborate more on the progression to unfavorable visual impairment, we further investigated the change of the VA group. Table 6 in the revised manuscript showed that most patients had no change in the VA group between the initial VA and the last VA. Most eyes with unfavorable last VA had already presented with unfavorable VA.
The SCG-C tended to have an improvement of the VA. On the other hand, SCG-O had the highest proportion of worsening VA. There was no improvement in the VA group in any of the JOAG.
There were 12 eyes of SCG-O that showed worsening of the VA to the unfavourable VA at the latest visit. Almost all cases had limited visual potential due to cornea and/or anterior segment anomalies (Peters anomaly, microcornea and sclerocornea). These patients had their initial visit at a young age (all < 6 years) and the initial VA was recorded as 'fix and follow', which was classified in this study as 'favourable VA'. However, the final VA ranged from HM to FC 3 ft, which could be explained by their cornea and/or anterior segment conditions. There were 2 cases of congenital ectopia lentis. One of them underwent cataract surgery without IOL implantation. The other developed retinal detachment after trabeculectomy. Both had the final VA of hand motion.
Among the 13 JOAG with available initial visual acuity and latest visual acuity, there were 2 eyes that showed worsening of the VA group. The first case presented with a C:D ratio of 0.9 in the right eye. The contralateral eye had no light perception at the presentation. The patient had successful trabeculectomy in the right eye but lost to follow-up afterwards. She came back with the vision of PL and full disc cupping. The second case had an initial C:D ratio of 0.95 and an initial VA of 20/100. Although the patient responded well to the treatment and the IOPs were in the low teens, the VA slowly got worse. After approximately 12 years, the last VA was hand motion. Regarding 7 eyes with unfavourable final VA and unfavourable initial VA, all C:D ratios at presentation were at least 0.9.
Our data suggested that the initial VA and glaucoma stage at presentation were the main factors determining the unfavourable VA outcome. The underlying ocular abnormalities tended to limit the visual potential in SCG-O, the anterior segment dysgenesis in particular. We now mentioned this point in the 9 th and 10 th paragraph of the Discussion.
2. As the authors mentioned, surgical interventions were mostly required in the primary type of glaucoma, both PCG and JOAG, and especially SCG-O. However, from this study, no glaucoma intervention was performed to 46.7, and 42.3% of the patients of JOAG and SCG-O, respectively. Explain the reason why almost a half of the patients did not need the surgical intervention.
Response: We thank the reviewer for the comment. Although PCG, JOAG and SCG-O had the high rate of surgical intervention, a high proportion of eyes that had no glaucoma interventions was observed in our study. The primary reason for not receiving surgical intervention was that the IOP could be controlled with medications. All patients with evidence of high IOP were offered surgical intervention (trabeculotomy and goniotomy) for PCG. The option of surgical intervention was discussed with the families of those who had previously been treated with medication. After seeing the well-controlled IOP, some parents preferred to continue with the medication. Unlike PCG, the surgical intervention was not necessarily offered to all SCG-O and JOAG patients. The decision to perform surgery in SCG-O was mainly based on the IOP and visual prognosis. Conservative treatment with medications was preferred If there was very limited vision potential such as nystagmus or visual acuity of light perception. For JOAG, the approach was quite similar to that for adult glaucoma. The majority of cases that did not receive glaucoma intervention underwent selective laser trabeculoplasty and were able to achieve target IOPs, though with medications.

You may analyze the factors associated with poor visual outcome.
Response: Thank you for the astute comment. As suggested by the reviewer, we performed the analysis with an ordinal logistic regression to explore the factors associated with visual outcome. The complete case analysis of 219 eyes showed that less favorable of initial visual acuity and high last IOP were significantly associated with poor visual outcome. In comparison to SCG-C, PCG, JOAG and SCG-O diagnoses were significantly associated with poor visual outcome. No significant risk of poor visual outcome was identified for the SCG-S and SCG-A groups when compared to the SCG-C group. We have updated the Methods and Results sections to include the description of the statistical analysis and the table in the revised manuscript. (new Table 5).

5.
Given the fact that PCG needs surgical management, the proportion of nonsurgically managed eyes among PCG cases was unexpectedly high. What were the reasons for nonsurgical management in those patients? Please address this issue in the discussion section.
Response: Thank you for the comment. The primary reason for not receiving surgical intervention was that the IOP could be controlled with medications. All patients with evidence of high IOP were offered surgical intervention (trabeculotomy and goniotomy) for PCG. The option of surgical intervention was discussed with the families of those who had previously been treated with medication. After seeing the well-controlled IOP, some parents preferred to continue with the medication.
For additional information of the patients who did not have glaucoma intervention, please refer to the response to reviewer 2's query No.2. We have now included this explanation in the 6 th paragraph of the Discussion.
6. The authors concluded that all subtypes of glaucoma were sporadic. Although the patient-reported or family-reported 'family history of glaucoma' was uncommon, the actual familial occurrence may be more than can be assessed by history taking. Unless each family member was examined for glaucoma, you may not be certain that all subtypes of childhood glaucoma were sporadic.