Higher HbA1c May Reduce Axial Length Elongation in Myopic Children: A Comparison Cohort Study

Abstract

Aims

To compare the annual axial length (AL) changes in myopic children with type 1 diabetes mellitus (T1DM) and those without diabetes.

Methods

There are two groups of myopic children in this retrospective cohort study. Group 1 consisted of myopic children with T1DM (44 eyes of 22 patients). Group 2 comprised age-matched myopic children without diabetes (44 eyes of 22 children). These two groups were compared with regard to their baseline clinical characteristics. A generalized estimating equations (GEE) model was also used to determine the most likely factor that contributed to the results.

Results

The average ages of group 1 and group 2 were 14.8 and 14.6 years, respectively. Children in group 1 had significantly slower annual AL changes (0.051 mm/year versus 0.103 mm/year; 50.5% slower, P = 0.011) and shorter baseline AL (23.97 versus 25.19 mm, P < 0.001) than those in group 2. GEE also showed that serum glycosylated hemoglobin (HbA1c) level (B = −0.023, P = 0.039) was the most important factor in reducing AL elongation in group 1 myopic children.

Conclusions

Long-term high HbA1c level may reduce AL elongation. A strict blood sugar control strategy in clinical practice is warranted to axial myopia progression in T1DM children.

Introduction

Type 1 diabetes mellitus (T1DM) is an early-onset disease resulting from destruction of pancreatic beta cells. Continuous measuring serum glycated hemoglobin (HbA1c) at three-month intervals is the most useful clinical index of the average blood sugar level in children with T1DM [1]. A strict blood sugar control strategy for these children can also prevent T1DM-related target organ damage such as retinopathy, nephropathy, and neuropathy [2–4].

Pathological myopia is defined as excessive axial myopia that leads to structural changes in the posterior segment of the eye (including posterior staphyloma, myopic maculopathy, and high myopia-associated optic neuropathy) and can lead to a loss in best-corrected visual acuity (BCVA) [5,6] that can impair quality of life. It is believed that high axial myopia is a progressive ocular disease that primarily results from excessive axial length (AL) elongation during childhood of school age and that usually stabilizes in adulthood [5,7]. Thus, continuous control of myopia progression in school-age children is crucial to prevent further BCVA loss related to pathological myopia [5–10]. During myopia monitoring, cycloplegic refractive error changes and AL changes are the two most important parameters that determine myopia progression and the effects of treatment [5,6,8]. Of the two, optical AL measurement can be more reliable and convenient because it is less influenced by accommodation in children (complete cycloplegia is not easy) [5,11]. It may also more directly contribute to pathological myopia [6,12].

In children with T1DM, profound short-term changes in refractive error can result from fluctuations in blood sugar level. These may progress toward myopia during hyperglycemia progression and toward hyperopia during recovery of hyperglycemia. The permanent refractive error can be influenced not only by axial myopia but also by some refractive myopia because of lens changes resulting from long-term higher blood sugar [5,13,14]. Therefore, the cycloplegic refractive error may not be a reliable measurement for myopia monitoring in schoolchildren with T1DM. Optical AL measurement is less likely to be influenced by blood sugar levels and can be a more accurate and appropriate way to follow up myopia progression in these children [11,13].

Thus, an ophthalmologist plays important roles in preventing vision loss from myopic and diabetic complications in children with T1DM. An extensive literature review failed to identify correlation between AL changes and long-term blood sugar levels. In addition, the correlations between myopia and blood sugar levels are controversial. Although some studies have found that hyperglycemia is associated with a trend to myopia [14–19], other studies demonstrated that refractive errors have little tendency to hyperopia or no relationship with increasing blood sugar levels [20–22]. It is also notable that these studies were all cross-sectional in design and used cycloplegic refractive error as the primary outcome measurement, which may contribute to the discrepancies in their conclusions. Therefore, a longitudinal study using optical biometry devices to measure AL may help resolve the controversy related to axial myopia [5,6,11].

To explore the correlation of AL changes with long-term blood sugar levels, we conducted a longitudinal study to compare AL changes between myopic children with and without T1DM using an optical AL measurement device, and analyzed the factors possibly contributing to the results.

Methods

Subjects

This retrospective cohort study included two groups of subjects. Group 1 consisted of myopic children with T1DM; group 2 was composed of age-matched myopic children without DM. This study was approved by the Institutional Review Board of Chang Gung Memorial Hospital (Approval Number: 202001771B0) and followed the tenets of the Declaration of Helsinki.

Grouping and Inclusion Criteria

Subjects in group 1 were enrolled from the Chang Gung Juvenile Diabetes Eye Study (CGJDES) database. Detailed information about subject recruitment for CGJDES was described previously [4,23]. The inclusion criteria in the current study were: (1) continuous AL measurement at least twice; (2) age of all AL measurement time points < 18 years; (3) spherical equivalent (SEq) of refractive error between 0 and −12 diopter (D); (4) difference in refractive error between both eyes < 1 D and (5) both eyes with astigmatism < 2 D; (6) no history of DM retinopathy or any other ocular pathologies; and (7) no record of receiving any myopia control treatments (e.g., atropine, orthokeratology) [24].

Subjects in group 2 were enrolled from an open database of children undergoing myopia monitoring and control at Chang Gung Memorial Hospital, Keelung, Taiwan using an age-matching strategy. The inclusion criteria were the same as group 1 except for the presence of DM.

Myopia Monitoring Policy in The Database from Which Group 2 Was Recruited

As per the current practice for myopia monitoring in schoolchildren, we recorded the children’s baseline cycloplegic refractive error and AL at their first visit, then planned to follow up at 3 to 4-month intervals. If the AL of the children elongated slowly compared with that of children of the same age, and they had a low risk of developing high myopia, we kept observing them until the annual AL changes accelerated to within the treatment range [6,7]. If the patient’s AL elongated rapidly or with a high risk of progression to high myopia, we started atropine or orthokeratology treatment at the second follow-up visit after 3 to 4 months [6]. Therefore, the patients selected for group 2 were children with relatively stable myopia comparable with group 1 and with a long follow-up period (mean 26.7 and 12.9 months of follow-up in groups 1 and 2, respectively).

Ophthalmic Examinations

Ocular examinations included BCVA, refractive error (Auto Ref/Keratometer ARK-1a/ARK-1; Nidek Co., Ltd., Gamagori, Aichi, Japan), slit-lamp examination, and ophthalmoscopy. AL was measured by an optical AL measurement biometry device (IOLMaster 500, Carl Zeiss Meditec AG, Jena, Germany). Annual AL change was calculated as [(the difference between the last and the first AL, in mm) / (the number of days between the last and the first AL measurement)] × 365.25. In group 1, fundus photography was also performed to record and screen for diabetic retinopathy.

Physical and Biochemical Examinations

Physical characteristics such as age, gender, body height (BH), and body weight (BW) were also collected for analysis. For extended statistical analysis, body mass index (BMI) was calculated as a representative parameter for BW and BH.

In group 1, biochemical examinations such as HbA1c levels and lipid profiles were routinely repeated at three-month intervals, and urine analysis was performed at least annually or as necessary. Urine albumin to urine creatinine (A/C) ratio was calculated to evaluate proteinuria. The detailed protocol for examinations of group 1 has been described in two previous reports [4,23]. It is noted that there were no biochemical or urine examination data for group 2.

Statistical Analysis

The data were analyzed using SPSS (IBM SPSS Statistics for Windows version 23.0; IBM Corp, Armonk, NY, USA). Categorical variables were analyzed with the chi-square test, and continuous variables were analyzed with the independent t test for comparing the two groups. Fisher’s exact test was used when one or more of the cells in the chi-square test had an expected value < 5. For the analysis of two eyes in one patient, the generalized estimating equations (GEE) model was applied to account for the dependency of the outcome for the two eyes. The linking function was identity and distribution in the GEE was normal. Independent working correlation and robust standard error were adopted to obtain the significance of parameters with the lowest corrected quasi-likelihood under the independence model criterion (QICC). Univariate GEE was used to first determine the correlations between annual AL changes and each variable individually. Significant variables were put into a further multivariable GEE. A two-tailed P value < 0.05 denotes indicated statistical significance and no adjustment of alpha error was made in this study.

Results

Subject Enrolment, Comparison of Demographic Data and Clinical Characteristics of the Two Groups

The flow diagram for case selection for final analysis is summarized in Figure 1. Finally, 44 eyes of 22 children were included in each group. The demographic data and clinical characteristics of the two groups are summarized in Table 1. The average ages of groups 1 and 2 were 14.8 years (range, 12.6 to 17.6 years; median, 14.7 years) and 14.6 years (range, 10.3 to 17.4 years; median, 15.2 years), respectively. The duration of T1DM in group 1 was 12.2 years (range, 10.6 to 15.3 years). Myopic children in group 1 had significantly slower annual AL changes (0.051 mm/year versus 0.103 mm/year; 50.5% slower, P = 0.011), shorter baseline AL (23.97 versus 25.19 mm, P < 0.001), and a longer total follow-up period (26.7 months versus 12.9 months, P < 0.001) than those in group 2.

Fig. 1.

A flow diagram for cases collection. AL, axial length; D, diopter; Seq, spherical equivalent; T1DM, type 1 diabetes mellitus.

Table 1.

Demographic data of studied patients

	Total subjects (n = 88)	Group 1 (T1DM) (n = 44)	Group 2 (Non-diabetes) (n = 44)	P value
Baseline Age, years (median)	14.7 ± 1.7 (14.9)	14.8 ± 1.3 (14.7)	14.6 ± 1.9 (15.2)	0.562a
Male sex, n (%)	44 (50.0%)	22 (50.0%)	22 (50.0%)	1.000b
Total Follow-up Periodc, months	19.8 ± 11.9	26.7 ± 11.7	12.9 ± 7.3	<0.001 a
Baseline SEq, diopter	−3.03 ± 2.58	−3.09 ± 2.32	−2.97 ± 2.83	0.829a
Baseline AL, mm	24.58 ± 1.56	23.97 ± 1.30	25.19 ± 1.57	<0.001 a
Annual AL changesd, mm/year	0.077 ± 0.097	0.051 ± 0.101	0.103 ± 0.087	0.011 a
Baseline Height, cm	160.3 ± 7.5	159.8 ± 7.8	160.9 ± 7.3	0.532a
Baseline Weight, kg	51.9 ± 8.1	52.4 ± 6.9	51.4 ± 9.2	0.566a
Baseline BMI, kg/m 2	20.1 ± 2.2	20.5 ± 2.1	19.7 ± 2.3	0.096a
Duration of T1DM, years		12.2 ± 1.4
Average HbA1c, %		9.1 ± 1.6
Average HDL, mg/dL		63.0 ± 11.1
Average LDL, mg/dL		107.2 ± 34.3
Average Cholesterol, mg/dL		181.4 ± 35.8
Average TG, mg/dL		76.7 ± 33.9
Average Urine A/C Ratio, mg/g		13.7 ± 19.8

Continuous data are presented as mean ± standard deviation.

The numbers in bold denote that the difference is statistically significant, P <0.05. T1DM, type 1 diabetes mellitus; AL, axial length; SEq, spherical equivalent; BMI, body mass index; HbA1c, glycosylated hemoglobin; HDL, high-density lipoprotein; LDL, low-density lipoprotein; TG, Triglycerides; A/C ratio, albumin to creatinine ratio.

^a.Independent t-test

^b.Pearson’s chi-square test.

^c.Total follow-up period, months, was calculated as [(the number of days between the last and the first AL measurement) / 30.44].

^d.Annual AL changes, mm/year, were calculated as [(the difference between the last and the first AL, mm) / (the number of days between the last and the first AL measurement)] x 365.25.

Baseline Age, Baseline AL and Presence of T1DM Are Negatively Correlated with Annual AL Changes in All Myopic Children (N = 88)

Univariate GEE showed that younger baseline age (B = −0.020, P = 0.009), longer baseline AL (B = 0.020, P = 0.008), and no presence of T1DM (B = −0.047, P = 0.049) were significantly associated with faster annual AL changes in all myopic children. Multivariable GEE including these three factors showed that only baseline age (B = −0.028, P < 0.001) and baseline AL (B = 0.028, P = 0.006) remained significantly associated with annual AL changes in all myopic children. These results are summarized as Table 2.

Table 2.

Associations between annual AL changes and age, gender, baseline AL, physical data and T1DM in all study subjects (n = 88)

Variables	Univariate			Multivariable
	Estimate	(95% CI)	P value	Estimate	(95% CI)	P value
Baseline Age, years	−0.020	(−0.034, −0.005)	0.009	−0.028	(−0.041, −0.014)	<0.001
Male Sex	−0.032	(−0.086, 0.022)	0.246
Baseline AL, mm	0.020	(0.005, 0.036)	0.008	0.028	(0.008, 0.048)	0.006
Total Follow-up Period, years a	−0.012	(−0.037, 0.013)	0.349
Baseline BMI, kg/m 2	−0.010	(−0.024, 0.003)	0.129
Presence of T1DM	−0.047	(−0.105, −0.001)	0.049	−0.013	(−0.072, 0.046)	0.671

Dependent variable: Annual AL changes (mm/year), were calculated as [(the difference between the last and the first AL, mm) / (the number of days between the last and the first AL measurement)] × 365.25.

The numbers in bold denote that the association is statistically significant, P <0.05, generalized estimating equations.

AL, axial length; T1DM, type 1 diabetes mellitus; CI, confidence interval; BMI, body mass index.

^a.Total follow-up period, years, was calculated as [(the number of days b

Mean Serum HbA1c Was Negatively Correlated with Annual AL Changes in Myopic Children in Group 1 (N = 44), While Baseline Age Was Negatively Correlated with Annual AL Changes in Group 2 (N = 44)

Because the presence of T1DM had a significant impact on the reduction in annual AL elongation (Table 2), we further investigated the correlation between annual AL changes and all available parameters in group 1 to determine possible contributory factors. Table 3 shows that in univariate GEE, HbA1c (B = −0.023, P = 0.039) was the only factor significantly correlated with AL annual change in group 1, while baseline age was the only significant factor in group 2.

Table 3.

Associations between annual AL changes and age, gender, baseline AL and physical/biochemistry data in subgroups (n = 44, 44)

Variables		Univariate, Group 1 (n = 44)			Univariate, Group 2 (n = 44)
		Estimate	(95% CI)	P value	Estimate	(95% CI)	P value
Baseline Age, years		−0.020	(−0.049, 0.008)	0.164	−0.018	(−0.032, −0.003)	0.015
Male Sex		−0.048	(−0.126, 0.030)	0.226	−0.016	(−0.083, 0.052)	0.646
Baseline AL, mm		0.028	(−0.008, 0.063)	0.125	0.009	(−0.006, 0.024)	0.257
Total Follow-up Period, years a		0.012	(−0.037, 0.062)	0.634	−0.012	(−0.081, 0.057)	0.732
Baseline BMI, kg/m 2		−0.004	(−0.028, 0.020)	0.730	−0.012	(−0.028, 0.003)	0.118
Average HbA1c, %		−0.023	(−0.045, −0.001)	0.039
Average HDL, mg/dL	/1000	−1.213	(−3.248, 0.822)	0.243
Average LDL, mg/dL	/1000	−0.099	(−1.203, 1.006)	0.861
Average Cholesterol, mg/dL	/1000	−0.345	(−1.407, 0.718)	0.525
Average TG, mg/dL	/1000	−0.773	(−2.055, 0.510)	0.238
Average Urine A/C Ratio, mg/g	/1000	−0.423	(−1.323, 0.477)	0.357

Dependent variable: Annual AL changes (mm/year), were calculated as [(the difference between the last and the first AL, mm) / (the number of days between the last and the first AL measurement)] × 365.25.

The numbers in bold denote that the association is statistically significant, P <0.05, generalized estimating equations.

AL, axial length; T1DM, type 1 diabetes mellitus; CI, confidence interval; BMI, body mass index; HbA1c, glycosylated hemoglobin; HDL, high-density lipoprotein; LDL, low-density lipoprotein; TG, Triglycerides; A/C ratio, albumin to creatinine ratio.

^a.Total follow-up period, years, was calculated as [(the number of days between the last and the first AL measurement) / 365.25].

Discussion

To our knowledge, this is the first cohort study to compare the annual AL changes in myopic children with and without T1DM. In addition, it demonstrated the benefit of using optical AL measurement as an outcome measurement to avoid bias from refractive error fluctuations in patients with DM. The study showed shorter baseline AL and slower annual AL elongation in group 1 compared with group 2. Higher average HbA1c levels in group 1 may be a factor in this difference.

In the current study, annual AL changes were 50.5% slower in group 1 (0.051 mm/year) than in group 2 (0.103 mm/year). The presence of T1DM could be a factor in the reduced axial elongation and the effect was even seen with current myopia treatments, such as low dose atropine and orthokeratology [5,8,10]. Further analysis showed that mean HbA1c level was significantly negatively correlated with the annual AL changes in group 1 [22]. These results are consistent with those of a 9-week experiment in diabetic rabbits, which showed slower axial progression and hyperopia accompanying pancreatic beta-cell impairment and chronic hyperglycemia [25].

The myopic children in group 1 had a shorter AL at baseline than those in group 2. This may have resulted from the accumulated reducing effects of long-term higher blood sugar on AL elongation before the first AL measurements were taken. The current strategy for controlling T1DM is usually conservative with the prescription of lower doses of insulin/insulin analogs for T1DM children than those prescribed for adults, because children are more vulnerable to severe sequelae of hypoglycemia that may cause impaired development of the neurologic system [26]. Therefore, the average HbA1c level in children with T1DM tends to be higher than in nondiabetic children; this could cause the slower annual AL elongation compared with the nondiabetic group, hence resulting in the shorter baseline AL in group 1. It is also notable that the shorter baseline AL of group 1 with similar refractive errors to those of group 2 may result from increasing long-term changes in lens power in children with T1DM (Table 1) [13,14]. A future prospective study focusing on this issue is necessary to explore these observations.

The negative correlation between annual AL changes and long-term blood sugar levels may be explained by the effect and level of insulin-like growth factor 1 (IGF-1). IGF-1 can stimulate the growth of the eyeball (AL elongation) [27,28] in a manner similar to its stimulation of body weight gain and growth [29]. Low IGF-I is frequently seen in patients with T1DM [30] which may result in a smaller annual AL elongation. To test this hypothesis, the correlation between IGF-1 levels and annual AL changes requires further investigation with a prospective large-scale study. If the hypothesis is supported by the results of such a study, local suppression of IGF-1 can become a candidate new treatment for control of myopia. There were no statistically significant correlations between BW, BH, and BMI changes and the average HbA1c in group 1 (data not shown). AL may be affected by many factors other than IGF-1, such as the genetic and environmental risk factors related to myopia in children [5], future large studies focusing on BW, BH, AL, myopia risk factors, and IGF-1 should better identify their relationships.

A large proportion of patients were excluded from this study because most of the children were receiving myopia control treatments. This may have led to the enrollment of only patients with relatively stable AL changes, which suited our study purpose to determine the correlation between AL changes and blood sugar levels in myopic children without myopia treatment. Moreover, because the same selection criteria were applied to both groups, and the mean follow-up times for both groups were more than 1 year, the bias during intergroup analyses may have been minimal. Furthermore, because the analysis showed a negative correlation between HbA1c and annual AL changes in group 1 (Table 3), we could conclude that higher long-term blood sugar levels may be correlated with a greater reduction in AL elongation. However, a large-scale prospective design study is necessary to confirm these results.

The presence of T1DM was significantly negatively correlated with annual AL changes in all myopic children; however, it lost significance after adjustment for baseline age and AL, with only a trend remaining (Table 2). The reason may be that the presence of DM is a general parameter, which does not represent the true blood sugar level of each subject; that is, some children with T1DM who had an AL elongation rate similar to that of children without DM may have controlled their blood sugar well. Fortunately, the group 1 analysis further supported the significant negative correlation between HbA1c level and annual AL elongation (Table 3), which may directly support our hypothesis.

In addition to T1DM, baseline age was a significant factor that was negatively associated with annual AL changes in all subjects and in group 2 myopic children (Tables 2 and 3). This finding is consistent with those of previous studies [5,6]. The loss of its significance in group 1 may be related to the extremely slow axial elongation in children with T1DM. In the current study, we matched the baseline age of the two groups, but the total follow-up period was longer for group 1 than for group 2 (26.7 months versus 12.9 months). To evaluate a possible age-related slowing in the effect on AL elongation (a concave curve of AL growth over time) [31], we compared the annual AL elongation in the two groups with respect to follow-up time and baseline age. The average age ranges of subjects during the follow-up period were 14.7 to 16.8 years in group 1 and 14.1 to 14.9 years in group 2. A previous study reported that the average annual AL elongation for these two age intervals was 0.19 and 0.23 mm, respectively [31]. This means that the different follow-up periods may have contributed 0.04 mm (0.19 versus 0.23 mm; 17%) [31], of the overall difference between the two groups (0.051 versus 0.103 mm; 50.5% difference) (Table 1). Furthermore, GEE showed no correlation between total follow-up period and annual AL changes in all myopic children (Table 2); therefore, the slower AL elongation in group 1 may be not only related to the longer follow-up period in group 1, but may also involve other factors.

Baseline AL is another significant factor that was positively associated with annual AL changes in all myopic subjects (Table 2). However, a loss of significance but the same trend was noted in the subgroup analysis (Table 3). The loss of significance in subgroup analysis may have been due to the lower statistical power because of the smaller numbers of subjects. However, baseline AL shows correlations with many factors including baseline age, degree of myopia, corneal power, lens power, anterior chamber depth, and annual AL changes in an individual. These factors modulate baseline AL in different directions and have complicated interactions, and there are no established criteria to define normal or accelerated annual axial elongation in a given individual [5,6]. In the current study, the evidence suggested that a possible reason for the shorter baseline AL in group 1 is a slower average annual AL elongation (Table 1 to 3).

There are several limitations to the current study, including its small sample size, retrospective design, and lack of some laboratory and ocular examination data, such as serum IGF-1, corneal power, lens power, lens thickness, anterior chamber depth, and cycloplegic refractive error. We enrolled two eyes from each patient in this study and applied generalized estimating equations (GEE) model to account for the dependency of the outcome for the two eyes. In addition, when we used right eye from each patient for analysis, the results were exactly the same in three tables (please see Supplementary Table 1 to 3). Because refractive error, anterior chamber depth, and lens power may be influenced by short- and long-term changes in blood sugar level [13,14], changes in AL may remain the most direct and wide-ranging parameter for monitoring axial myopia progression in children with T1DM [5,6,8]. This may compensate for the lack of the ocular biometry data listed above. To further explore the role of serum IGF-1 level in blood sugar and annual AL changes, a prospective long-term cohort study enrolling more subjects without these limitations is necessary in the future.

In conclusion, the current study showed that high blood sugar is associated with a reduction in the annual AL elongation in T1DM children, and that this effect was in proportion to the HbA1c level. These results indicate that a strict blood sugar control strategy is warranted to axial myopia progression in T1DM children. Moreover, further investigation based on the results of the current study may provide the rationale for a new treatment for myopia progression.

Data availability

The datasets analyzed during the current study are available from the corresponding author on reasonable request.