Nutritional Factors and Myopia: An Analysis of National Health and Nutrition Examination Survey Data

Abstract

Significance.

The rise in the prevalence of myopia, a significant worldwide public health concern, has been too rapid to be explained by genetic factors alone and thus suggests environmental influences.

Purpose.

Relatively little attention has been paid to the possible role of nutrition in myopia. The availability of the large National Health and Nutrition Examination Survey (NHANES) dataset, which includes results from vision examinations, offers the opportunity to investigate the relationship between several nutrition-related factors, including body metrics, and the presence and magnitude of myopia.

Methods.

Cross-sectional survey datasets with vision examination, demographic, body metrics and nutritional data, collected as part of NHANES over the years of 2003–2008, were extracted for analysis. Based on already published basic and epidemiological studies, the following parameters were selected for study: body height and mass index (BMI), demographics, serum Vitamin D and glucose/insulin levels, and caffeine intake, using multi-variable models and objectively measured refractive errors as the main outcome measure.

Results.

Data from a total of 6,855 ethnically-diverse Americans, aged 12–25 years were analyzed. In final multivariate models, female sex and age were the most significant factors related to myopia status and refractive error. In general, body metrics (BMI) nor nutritional factors (serum Vitamin D, glucose levels and caffeine intake) were found to be associated with refractive error or myopia status, however increased insulin levels was related to an increased odds of having myopia.

Conclusions.

These largely negative findings suggest that other environmental factors, such as those related to the visual environment, may contribute more to the development and/or progression of myopia and would argue for continued research in these areas in support of more evidence-based myopia clinical management.

Myopia has seen a rapid rise in its prevalence world-wide over the last generation, such that half of the world’s population is expected to be myopic by 2050.¹ Myopia carries a significant economic burden (approx. 268 billion US dollars),² as well as a sight-threatening ocular disease burden.³ While genetic factors play a role in the development of myopia,^4–8 they cannot explain the rapid increase in myopia prevalence.

Several environmental factors have been implicated in the development and progression of myopia, with signifcant research directed at the possible role of near work⁹ and more recently, of reduced time outdoors (discussed below¹⁰), as risk factors. However, in both cases, no general agreement has been reached on key contributing factors and the etiology of myopia will likely prove to be multi-factorial in nature. For example, early epidemiological studies of myopia noted an apparent link between the advent of formalized classroom education and the development of myopia in Inuit populations.^11,12 However, during this period of increasing Western influence, diets also underwent significant modification.^13,14

Overall, there has been limited investigation into the role of nutritional factors in myopia. In one early observational report, diets of high glycemic load were described as a risk factor for myopia by Cordain et al., who speculated that hyperinsulinemia might modify scleral growth factors, as a possible underlying mechanism.¹³ However, in apparent contradiction of this hypothesis, a cross-sectional study of Singaporean children (aged 7–9 years)¹⁵ found that children with shorter stature and increased body mass index (BMI) had less myopic refractions, assuming BMI reflects glycemic load. Nonetheless, studies using animal models of myopia have provided evidence linking insulin with enhanced eye growth,¹⁶ and also for protective effects of both caffeine and one of its metabolites, 7-methylxanthine, against myopia development^17–19 with choroidal and scleral targets as possible sites of action in the latter case.¹⁷

Significant attention has been paid recently to the protection afforded by outdoor exposure against myopia development, with four randomized control trials involving Asian schoolchildren reporting generally positive benefit from increased outdoor recess time.^20–23 While there is ongoing debate regarding the role of outdoor high light intensities,^24–27 it is also important to note that exposure to sunlight and serum Vitamin D levels are highly positively correlated. Nonetheless, although two relevant large cohort studies found an association between serum Vitamin D levels and refractive error,^28,29 their authors reached diverging interpretations of their results with respect to the role of low serum Vitamin D levels as a risk factor for myopia development, with one group noting the likely confounding effect of time outdoors on Vitamin D serum levels,²⁹ which is consistent with observations from a previous GWAS study.³⁰

The availability of the large US National Health and Nutrition Examination Survey (NHANES) dataset allowed for further investigation of the relationship between various nutritional and body metric factors and the presence and magnitude of myopia, which describes the scope of the study reported here. NHANES comprises a series of ongoing studies, which was initiated in 1960 by the U.S. Centers for Disease Control and Prevention (CDC), and aims to investigate the health and nutritional status of children and adults across the United States through the collection of data concerning demographic, socioeconomic and health-related variables (via physiological and laboratory measurements).

METHODS

Participant Cohort

The study reported here was limited to participants during study years 2003–2008. As refractive error data were only available for participants aged 12 years or older, the current analyses were also limited to participants aged 12–25 years of age, with the upper limit taking into account the possibility of late-onset myopia.^31–33 Both univariate and multivariate analyses were applied to extracted data. Initially, participants with available refractive error data were extracted (n=6,855) if they met inclusion criteria of not having either previous refractive surgery (n=18) or the possibility of corneal disease, as indicated by keratometric readings of more than 50 D (n=21). Given that corneal power is, on average, approximately 48 to 50 D in full-term human infants^34,35 and only decreases with age,^36,37 corneas with 50 D or more in the steeper meridian was used as a potential biomarker for corneal disorders and used as a basis for excluding participants.

The demographic features (age, sex, ethnicity) of this final participant cohort (n=6855) are summarized in Table 1. In the case of the univariate analyses, participants with available refractive error data and the factor(s) of interest were included, with the sample size varying accordingly, as indicated below. However, only participants with complete datasets (n=1,974 (29%)) were used in multivariate analyses. Importantly, a sensitivity analysis demonstrated that participants with complete versus incomplete datasets had similar ethnicity distributions (<3% difference in all ethnicity categories), range of refractive errors (+8.5 D to −17.625 vs +6.00 D to −20.25 D), mean spherical equivalent refractive error (mean difference = .009 D) and proportion of participants with myopia (<0.3% difference). None of these differences were statistically significant (all comparisons, P >.05)

Table 1.

Summary of participant demographic characteristics, expressed in terms of number and percentage of participant cohort, in those participants aged 12–25 years of age with available refractive error data. The mean (SD) age of the participant cohort was 17.05 (3.63) years.

Participant Characteristics	TOTAL N=6,855
Female Sex	3,430 (50.04%)
Ethnicity
Non-Hispanic White	2,131 (31.09%)
Non-Hispanic Black	2,112 (30.81%)
Mexican American	1,942 (28.33%)
Other Hispanic	375 (5.47%)
Multi/Other	295 (4.30%)

All study sampling methods pertaining to the NHANES data set have been described elsewhere.³⁸ All study methods followed the Tenets of the Declaration of Helsinki, were approved by the National Center for Health Statistics Research Ethics Board and informed consent was documented prior to participant participation. Note also that data collection is limited to NHANES-trained health technicians, as a quality control measure. Pertinent to the analyses reported here, open-source data were extracted and used for analysis as described below.

Databases

The vision database includes non-cycloplegic objective auto-refractor (Nidek ARK-760) measurements. Recorded refractive errors (median of 3 repeated measures) were only included in analyses when a confidence rating of at least 5 (scale from 1 to 9) was achieved at the time of measurement. The mean (SD) confidence rating of the refraction data utilized in our analyses was 8.86 (0.43). For use in analyses, spherical equivalent refractive errors (SERs, i.e. average of the refractions in two principal meridians) were calculated from the data for the right eyes of all included participants, and SERs of −0.75 D or worse were classified as myopic. This more conservative definition of myopia was employed to avoid mis-classification of myopia, given that cyclopegic agents were not used in measuring refractive errors.

To monitor childhood growth and weight gain, NHANES collected a series of body metrics, including body mass index (BMI, kg/m²) and standing height (height, cm). From the NHANES laboratory database several nutritional factors were extracted for use in analyses. Total 25-(OH) Vitamin D levels (nmol/L) were measured in collected samples using a standardized liquid chromatography-tandem mass spectrometry method. Fasting (for at least 9 h) plasma glucose (hexokinase method, mmol/L) and serum insulin (ELISA method, pmol/L), were obtained for participants 12 yrs and older attending morning study visits as part of an ongoing effort by NHANES to estimate the prevalence of diabetes in the US. Information about caffeine intake was extracted from dietary interviews performed as part of the ‘What We Eat in America’ initiative. Using a validated US Department of Agriculture (USDA) survey method, participants were interviewed initially in-person during a study visit and subsequently by phone (3–10 days following the in-person interview, but not on the same day of the week). During both interviews, participants were encouaged to make use of a set of measuring guides (e.g. glasses/mugs, bowls, drink boxes and bottles, household spoons, measuring cups and spoons) to more accurately estimate the amounts of foods and liquids consumed. For the purpose of the current study, estimates of daily caffeine intake (in mg), from each of the two interviews were averaged for each participant.

Data Analyses

The NHANES analytical guidelines for 1999–2010 are available at www.cdc.gov/nchs/data/series/sr_02/sr02_161.pdf. Analyses were performed using Stata 14.2 (StataCorp, College Station TX USA). Statistical analyses (chi2 or Kruskal-Wallis) were performed to investigate the relationships between refractive errors and demographic and nutritional factors. Multivariable logistic and linear regression models were also created using myopia status (presence/absence) and magnitude of myopia (SER) as outcome variables. The inclusion of covariates in the final models was hypothesis-driven and collinearity was evaluated across all covariates. Ethnicity was included through the use of dummy variables and interaction terms were included in the final models if their coefficients differed significantly from zero, as determined by the Wald test. An alpha value of 0.05 was used in all data analyses, as an indicator of statistical significance. Summary statistics are reported as means, including standard deviations and/or 95% confidence intervals, unless indicated otherwise.

RESULTS

Demographic Factors and Myopia Status

Participant refractive error ranged from +8.5 D of hyperopia to −20.25 D of myopia, with a mean of low myopia (−0.86±1.90 D). Approximately thirty-five percent (35.3%), of the cohort were myopic, with the mean SER of this subgroup being −2.66±2.05 D. The mean SER of the remaining non-myopic participants was 0.12±0.74 D. Although there was a statistically significant increase in SER with increasing participant age (P=.001), the difference across age groups was small, although extreme myopia outliers existed in all age groups (Figure 1). Investigation into the possible influences of these outliers (dfbeta model diagnostics) and found no meaningful effects, for any age group. Both sex and ethnicity appeared to influence SER (Figure 2), although differences did not always reach statistical significance. Specifically, females were more likely to be myopic compared to males (37.8% vs 32.8%, P<.0001), and they also had a significantly greater magnitude of myopia compared to males (−2.78±2.04 vs. −2.54±2.06 D, P=.0002). In relation to ethnicity, except for the Multi/Other group which was on average more myopic (−1.38±2.31 D), the other ethnic groups had similar mean refractive errors (range: −0.75 to −0.89 D). Nonetheless, there were significant differences between these groups (based on ANOVA testing), both in terms of mean SER (P=.0001) and in the proportion of myopic participants (P<.0001), with Blacks having the lowest (33.1%) and the Other/Multi group, the highest (46.1%) values. Of those with myopia, mean SERs also differed significantly with ethnicity (P=.001), with Mexican Americans having the least myopic SER (−2.48±1.86 D) and the Other/Multi group, the highest (−3.15±2.21 D).

Figure 1.

Violin plot showing the significant relationship between SER and age for the participant cohort.

Figure 2.

Myopia presence (% of cohort) (A), as well as spherical equivalent refractive error (SER, D) for all participants (B) and for myopic participants (C), segregated by sex and ethnicity in all cases. Data presented as means with 95% confidence intervals.

Body Metrics and Nutritional Factors and Myopia Status

Body Metrics

Summary statistics for all body metric and nutritional factors, partitioned by participant demographic charateristics, including age (stratified into 12–18 and 19–25 yr age groups), are shown in Appendix Table 1, available at [LWW insert link] and described in more detail below. The relationships between each nutritional and body metric factor and age, stratified by ethnicity and myopia status are illustrated in Figure 3. Because the distributions for nutritional and body metric factors (except for standing height) proved to be quite skewed, significant outliers were identified for each factor via the interquartile range (IQR) method (outlier identified as falling outside of the lower and/or upper fences in a box and whisker plot) and removed from analyses. For any given factor, this method removed between 0.12% (Vitamin D) to 2.67% (serum insulin) of the data. Nonetheless, no significant differences were identified in a sensitivity analysis performed to determine if the removal of outliers influenced the results of linear regression analyses for the respective nutritional / body metric factors.

Figure 3.

Plots of standing body height (A), body mass index (B), Vitamin D level (C), caffeine intake (D), fasting serum insulin level (E) and fasting serum glucose level (F) against age, stratified by ethnicity group and myopia status; results of correlation analyses for myopic and nonmyopic subgroups also shown.

The average height and BMI of participants who had both refractive error and body metric data (n=6764, 99%), was 166.18±10.35 cm and 24.78±6.53 kg/m², respectively. Notably, the latter value is just outside the range for overweight, as defined by the United States Centers for Disease Control and Prevention (25.0 to <30 kg/m²; www.cdc.gov). Female participants had significantly higher BMIs compared to males (25.02±6.34 vs. 24.19±5.81 kg/m², P=.0001) and were significantly shorter than males (160.95±7.03 vs. 171.38±10.52 cm, P<.0001). Ethnicity also significantly influenced height (P<.0001) and BMI (P<.0001). In relation to height, Whites were the tallest (168.43±10.29 cm) and Mexican Americans, the shortest (162.93±9.52 cm). In relation to BMI, Blacks had the highest values (25.40±7.37 kg/m²), and the Other/Multi group, the lowest values (23.32±5.83 kg/m²). As expected, both BMI and height significantly increased with age, regardless of ethnicity or myopia status, except for height in Mexican American myopes (Figure 3A–3B). However, neither height nor BMI were significantly correlated with SER (Figure 4A–4B, both R²≤ .0004 and P ≥.12).

Figure 4.

Plots of spherical equivalent refractive error (SER, D) against standing body height (A) and body mass index (B); no significant correlation between either of the body metrics and SER observed.

Nutritional Factors

A total of 4,838 participants (71%) had both serum Vitamin D and refractive error data. The mean serum Vitamin D levels for these participants was 54.78±22.11 nmol/L, with a large percentage (~48%) classifiable as deficient in Vitamin D, as defined by the Vitamin D Council (below 50 nmol/L, www.vitamindcouncil.org). On average, females had a small but significant decrease in serum Vitamin D levels compared to males (54.08±23.44 vs. 55.20±19.89 nmol/L, P<.0001). There were also significant ethnicity-related differences (Figure 3C; P<.0001), with Blacks having the lowest levels (41.45±16.01 nmol/L) and Whites, the highest (71.28±22.02 nmol/L). Myopic and non-myopic groups had similar serum Vitamin D levels (54.13±21.58 vs. 54.92±21.84 nmol/L, P=.23), as also reflected in the non-significant correlation between serum Vitamin D levels and SERs for this cohort overall (Figure 5A, R²= .0002, P=.32).

Figure 5.

Plots of spherical equivalent refractive error (SER, D) and serum Vitamin D level (A), caffeine intake (B), fasting serum insulin level (C) and fasting serum glucose level (D); no significant correlation between any of the nutritional factors and SER observed.

For participants who had both caffeine intake and refractive error data (n=5,864, 86%), the mean daily caffeine intake was 44.07±53.60 mg, and increased significantly with age, regardless of ethnicity or myopia status, as might be expected (Figure 3D, all P≤.04). Overall, males had significantly higher caffeine intake compared to females (47.14±56.78 vs. 41.17±50.25 mg, P=.006) . There were also significant ethnicity-related differences in caffeine intake (P<.0001), with Blacks having the lowest intake (25.83±38.32 mg) and Whites the highest (64.30±64.48 mg). However, there was no significant difference in the mean caffeine intake of myopic and non-myopic participants (43.02±52.64 vs. 44.64±54.11 mg, respectively, P=.27) and no significant correlation between caffeine intake and SER for this cohort overall (Figure 5B, R²= .0000, P=.93).

The mean serum glucose and insulin levels for those participants that also had refractive error data (n=2,895, 42%) was 5.10±5.10 mmol/L and 70.20±48.29 pmol/L, respectively. In general, there was no effect of age on insulin or glucose levels, regardless of ethnicity or myopia status, except for an observation in White participants as shown in Figures 3E–F. Males had significantly higher glucose levels than females (5.22±0.45 vs.4.97±0.46 mmol/L, P<.00001), but lower insulin levels (64.89±48.87 vs. 76.04±49.17 pmol/L, P<.00001). Glucose levels varied significantly with ethnicity (P<.0001), with Blacks having the lowest levels (5.00±0.45 mmol/L) compared to all other ethnic groups, who had similar glucose levels (5.11 (White) to 5.19 (Other/Multi) mmol/L). Insulin levels also showed signifcant ethnicity-related differences (P<.0001), with the Other/Multi group having the lowest levels (60.43±39.89 pmol/L) and Other Hispanics, the highest (77.27±53.61 pmol/L). There was no statistically significant difference in glucose levels between myopic and non-myopic participants (P=.81), although insulin levels were significantly higher in myopic participants (P=.01). However, neither glucose levels nor insulin levels proved to be significantly correlated with SER (Figure 5C–5D, both R²≤ 0.004, P≥.30).

Multi-Variable Modeling

Results for all multivariate models are summarized in Table 2. An initial linear regression multivariate model was created to identify the factors associated with participant refractive error (SER). Given the collinearity of participant age and BMI with height and the fact that height was not related to myopia status in univariate analysis, height was not included in our final models. In addition, age was standardized by centering the variable on the mean age of the cohort (17.05 yr). While no significant associations between either ethnicity or any of the nutritional factors and SER were identified, the relationship between sex, age and SER proved to be statistically signifiant, with females having more myopic SERs than males (by −0.20 D [−0.380 to −0.015], P=.03) and older participants having more myopic SER (−0.04 D [−0.063 to −0.008] more myopia per 1 yr increase in age; P=.01). Although not significant, participants in the Multi/Other ethnicity group were also on average, slightly more myopic compared to Whites (by −0.41 D [−0.862 to 0.051], P=.08). We also explored potential non-linear effects of age by including age² in our model, however no increase in significance was found when all other model covariates were considered (combined age effect P=.16).

Table 2.

Results generated from multivariate models, with adjustment for demographic, nutritional and body metric factors in those participants with complete datasets (n=1,974). Statistically significant effects are shown in bold.

	Mean Refractive Error		Presence of Myopia
	Coefficient [95% CI]	P-value	Odds Ratio [95% CI]	P-value
Female Sex	−0.20 D [−0.380 to −0.015]	.03	1.27 [1.040 to 1.545]	.02
Age (per 1-yr increase)	−0.04 D [−0.063 to −0.008]	.01	1.03 [1.002 to 1.064]	.04
Ethnicity
Non-Hispanic White	Reference		Reference
Non-Hispanic Black	0.11 D [−0.162 to 0.379]	.43	0.81 [0.605 to 1.088]	.16
Mexican American	0.11 D [−0.132 to 0.350]	.37	0.84 [0.650 to 1.095]	.20
Other Hispanic	0.17 D [−0.322 to 0.655]	.50	0.77 [0.452 to 1.324]	.35
Multi / Other	−0.41 D [−0.862 to 0.051]	.08	1.13 [0.700 to 1.837]	.61
Total Vitamin D (per 1 nmol/L increase)	0.002 D [−0.002 to 0.007]	.33	1.00 [0.992 to 1.002]	.30
Daily Caffeine Intake (per 1mg increase)	−0.0008 D [−0.0009 to 0.0026]	.34	1.00 [0.997 to 1.000]	.16
Fasting Glucose (per 1 mmol/L increase)	−0.07 D [−0.272 to 0.138]	.52	1.05 [0.840 to 1.309]	.67
Insulin (per 1 pmol/L increase)	−0.0005 D [−0.0028 to 0.0019]	.70	1.003 [1.000 to 1.005]	.04
Body Mass Index (per 1 kg/cm2 increase)	0.002 D [−0.017 to 0.020]	.84	0.99 [0.971 to 1.010]	.35

Body height removed from models due to collinearity with age and BMI. Bold values are significant.

For participant myopia status, sex showed a significant effect such that the odds of having myopia was significantly greater for females than males (OR=1.27 [1.040 to 1.545], P =.02). While ethnicity had no significant effect, older age was significantly associated with a greater odds of having myopia (OR=1.03 [1.002 to 1.064], P =.04). In relation to body metrics and nutritional factor, only increased insulin levels were associated with a significantly increased odds of having myopia (OR=1.003 [1.0002 to 1.005], P=.04).

Additional linear regression modelling was undertaken using the data from those participants who were both myopic and had data for all other variables (n=703), with specific interest in associations between the magnitude of myopia and demographic, body metrics and nutritional factors (data not shown). While females had more myopic SERs than males on average (by −0.19 D), this difference was not statistically significant (P=.26). Older myopic participants also tended to be more myopic than younger participants, although this effect of age was also neither statistically nor clinically significant (−0.03 D more per 1 year increase in age; P=.24). In relation to ethnicity, the largest difference was between myopes in the Other/ Mixed compared to Whites groups (by −0.66 D), although this was not statistically significant (P=.10). None of the body metric and nutritional factors proved to have predictive value, as determinants of the magnitude of myopia (all P values >.20).

DISCUSSION

To our knowledge, the current study represents one of only three studies to exploit the NHANES database as a resource for investigating environmental contributions to the development of myopia,^39–41 with one of the other studies also investigating nutritional but not body metric factors.^40,41 The analyses used in our study revealed more females to be affected by myopia than males overall (12–25 yr cohort: 38% vs. 33%). Females were also found to have more myopic refractive errors and a greater odds of having myopia, as were older participants. While univariate analyses identified ethnicity-related differences in mean SERs and participant myopia, there was no significant association with either the presence of myopia or its magnitude, after controlling for other participant factors. In general, none of the nutritional factors examined (serum Vitamin D, plasma glucose and caffeine intake) proved to be significantly related to the presence of myopia in this participant cohort, although participants with increased insulin levels had a significantly increased odds of being myopic.

In relation to the influence of body metrics and consistent with the results of the current analysis, a number of population-based studies across the globe have reported relationships between increased eye length and/or myopia in taller individuals.^15,42–48 While these studies have generally involved older cohorts (≥40 years of age) with likely stable refractive errors, two Asian studies involving children (aged 7–9 years) provide exceptions.^15,48 In one study involving Taiwanese children⁴⁸ and another population-based study involving Chinese adults⁴⁹ height was found to be positively associated with longer eyes, but not with myopia, with the likely explanation for this apparent discrepancy lying in the other structural differences found in the eyes of taller individuals, namely deeper anterior chambers, thinner lenses, and flatter corneas. The latter findings are also generally consistent with sex differences identified in a systematic review of ocular biometry data, which found males to have longer eyes (by ~0.50 mm), flatter corneas (by ~0.50 D) and deeper anterior chamber depths (by ~0.16 mm) compared to females, except for those males of Asian ethnicity who had steeper corneas than females.⁵⁰ Further challenging the relevance of body metrics to myopia is one large Israeli cohort study (N=106,926) of conscripted males aged 17–19 yrs in which no relationship between myopia and either body height or mass index was found.⁵¹

As noted earlier and consistent with our results, Cordain et al., proposed a link between hyperinsulinemia with myopia development.¹³ Other studies have also reported links between diabetics and myopia.^52,53 However, that lenticular changes offer an explanation for the increased prevalence in myopia in diabetics is supported by results from a later, small study by some of the same researchers.⁵⁴ Nonetheless, that the glycemic profiles of populations worldwide might explain observed increases in myopia prevalence figures was suggested by the authors of a recent review of related epidemiological literature, which included speculation on possible mechanisms by which insulin could promote ocular growth.⁵⁵ However, evidence for the latter from studies involving animal models is equivocal; while insulin was found to promote myopia development in a few studies involving chicks, the pattern of axial elongation was atypical in that anterior segment changes contributed most to the overall increases in eye length.^16,56

Because increased time spent outdoors is recognized to be protective against myopia^21,23,57,58 and also tightly tied linked to serum levels of Vitamin D, there has been interest in whether Vitamin D alone might be protective. A number of related hypotheses concerning how low serum Vitamin D levels could increase the risk of myopia have been proposed, including up-regulation of scleral extracellular remodeling and synergistic interaction with retinoic acid, a recognized ocular growth regulator.⁵⁹ However, consistent with results of the current study, four large cohort studies, including one NHANES cohort analysis, failed to establish a link between low level of Vitamin D and myopia.^28,29,40,60 Furthermore, in the current study, Blacks had the lowest serum levels of Vitamin D, as has been reported in previous studies,⁶¹ yet Blacks also had the lowest proportion of myopes (35%, 1% below the average), with similar findings contained in other reports.^62–64 A recent study used Mendelian randomization to investigate the role of low serum vitamin D levels in myopia development, without the confounding effect of outdoor activity, also found no evidence of a causal relationship.³⁰ Interestingly, a study in tree shrews involving experimentally-induced myopia also failed to demonstrate a positive benefit from Vitamin D3 supplementation, although control animals were not deficient.⁶⁵

A number of recent animal model studies have yielded strong supporting data for the potential therapeutic benefits of caeffine and one of its metabolites, 7-methylxanthine (7-MX), a non-selective adenosine receptor antagonist.^17–19 These results are also consistent with a relative reduction in axial elongation and myopia progression observed with oral 7-MX in an earlier pediatric clinical trial in Denmark.⁶⁷ Longer-term clinical trials are on-going in Denmark, the only country to have approved oral 7-MX tablets for myopia control to-date. Unfortunately, analyses reported here were necessarily limited to NHANES survey-based data covering caffeine intake, as data covering caffeine metabolites in urine are only available from 2009, when refractive error measurements were discontinued. Related to other nutritional factors, to our knowledge there have only been two other systematic analyses of this open-access NHANES dataset with respect to myopia development and/or progression and nutritional factors to-date.^40,41 One of these studies examined the association of total zinc intake and myopia in NHANES participants aged 12–19 years and found no association.⁴¹

The large ethnically-diverse NHANES participant cohort combined with the analyses used in the current study represent its major strengths, with the availability of objective refractive error data representing an additional strength, given that axial length data were not available. In addition, that all measurements were performed in a standardized way by trained technicians, according to well-defined protocols, across all NHANES sites, adds further value to this dataset. Nonetheless, there are several limitations to consider, the most significant of which relates to the ethnic categories utilized during the study years analyzed. Given the high prevalence of myopia in Asian populations, both in Asian countries and in the US, the lack of an ‘Asian’ category represents a major limitation. Nonetheless, the fact that the Other/Mixed ethnic category, albeit small, had the highest proportion of myopes (46%; 11% more than the cohort average) and who were also relatively more myopic (−3.15 D; 0.67 D more than the myopic cohort average), likely reflects the fact that Asian participants were included in this ethnic category. In addition, it is important to note that the refractive error data were limited to children ≥12 years of age, after the typical age of onset of childhood myopia and as reflected in the presence of myopia in many of the participants in our study cohort. This may in part explain why age, which has been shown to influence refractive error in many other studies,^62,67–69 proved to be of only borderline statistical significance in our multi-variate modeling. In addition, the nutritional data captured from adolescents are likely to be different from that representing younger children. Finally, while it is important to disentangle the likely confounding effects of outdoor activity when considering the relationship between serum Vitamin D levels and myopia; neither comprehensive measures of outdoor activity nor season of data collection are available in this dataset, and only minimal data concerning sunlight exposure is available.

CONCLUSIONS AND CLINICAL RELEVANCE

The key risk factors related to myopia development and/or progression remain to be identifed and are very likely multi-factorial. However, clinicians are frequently called on to make recommendations about behavioral modifications that might reduce myopia development and progression. While sex and age appear to be most closely tied to the presence and magnitude of myopia, the results presented here suggest that insulin levels may also be an important factor. No other nutritional or body metric factors appear to be closely tied to the presence or magnitude of myopia. These largely negative findings suggest that other environmental factors, such as those related to the visual environment, may contribute more to the development and/or progression of myopia and would argue for continued research in these areas in support of more evidence-based myopia clinical management.

APPENDIX

Appendix Table A1.

Summary statistics (Mean±SD) for participants with available body metric and nutritional factors data, stratified by sex and age group.

FACTORS			ETHNICITY					REFRACTIVE ERROR STATUS
	Sex	Age	White	Black	Mex. Am.	Other Hisp.	Multi/Other	Myope	Non-myope
Standing Height (cm) ^+#*	Male	12–18 y 19–25 y	170.91±11.09 178.70±7.05	170.49±10.65 178.35±7.12	166.93±10.13 170.90±7.75	167.58±10.02 172.29±6.35	167.67±11.89 175.56±8.79	170.03±10.65 176.98±7.73	168.95±10.83 175.76±8.14
	Female	12–18 19–25	162.13±6.79 164.58±6.24	161.59±6.89 162.99±6.90	157.84±6.43 158.40±6.37	158.54±6.331 60.82±7.55	159.49±7.35 160.42±6.02	160.65±7.00 162.96±6.79	160.22±6.95 161.48±7.06
BMI (kg/cm2) ^+#	Male	12–18 19–25	23.07±5.34 25.74±5.60	23.25±5.84 26.56±6.54	23.78±5.67 26.44±5.29	23.34±5.15 26.19±5.77	22.98±5.81 25.15±5.20	23.91±5.78 26.44±6.19	23.08±5.52 25.99±5.59
	Female	12–18 19–25	23.12±5.27 26.08±6.30	25.36±6.76 28.38±7.38	24.01±5.67 27.25±6.21	24.19±6.08 26.24±5.92	22.17±5.10 24.12±5.94	24.42±6.08 25.88±6.17	23.94±5.96 27.53±6.79
Vitamin D (nmol/L) ^+#	Male	12–18 19–25	70.40±17.93 67.33±20.27	45.41±16.56 37.28±14.13	55.18±15.57 54.14±15.64	57.42±19.18 48.70±15.91	56.30±17.95 45.76±15.38	55.14±19.07 52.40±20.76	56.09±19.51 54.44±21.26
	Female	12–18 19–25	71.03±22.94 76.46±26.47	39.37±14.86 39.15±16.70	50.17±15.02 51.16±18.88	58.76±19.24 57.71±21.27	47.89±16.36 48.96±20.92	50.85±20.43 60.32±27.04	53.10±22.16 56.51±26.33
Caffeine (mg) ^+#*	Male	12–18 19–25	60.08±61.32 88.59±76.55	20.98±33.23 39.38±49.47	41.89±46.74 67.44±59.55	38.45±48.70 53.90±56.40	44.18±52.12 62.64±75.78	38.34±48.80 61.84±65.07	40.36±50.74 69.11±68.80
	Female	12–18 19–25	51.63±54.01 71.72±67.93	23.01±34.19 35.78±46.53	33.29±39.64 48.88±50.51	27.62±33.13 62.67±70.43	28.49±33.56 53.15±53.19	34.80±43.72 54.29±60.04	34.64±43.69 55.70±59.64
Glucose (mmol/L) ^+#*	Male	12–18 19–25	5.26±0.41 5.19±0.45	5.11±0.43 5.15±0.38	5.28±0.47 5.37±0.49	5.32±0.44 5.25±0.52	5.32±0.35 5.32±0.45	38.34±48.48 61.84±65.07	40.36±50.74 69.11±68.80
	Female	12–18 19–25	5.04±0.42 4.89±0.44	4.87±0.45 4.82±0.46	5.06±0.46 5.01±0.52	5.12±0.42 4.98±0.63	5.12±0.40 4.95±0.32	5.03±0.45 4.93±0.48	4.98±0.45 4.91±0.48
Insulin (pmol/L) ^+#*	Male	12–18 19–25	63.36±46.05 55.28±43.74	67.36±48.94 54.29±48.16	73.14±46.60 67.03±45.97	74.74±51.42 61.50±47.60	61.21±32.66 46.47±26.36	72.85±49.15 61.31±48.96	66.07±46.24 55.88±43.21
	Female	12–18 19–25	67.15±36.86 59.60±46.12	83.09±52.80 81.46±54.17	82.94±50.38 79.93±47.10	85.73±55.20 80.33±58.31	69.04±43.81 60.59±51.90	81.89±51.50 72.23±53.32	75.86±46.17 72.05±48.76

^{# =}significant sex effect in those aged 12–18 (data pooled across ethnicity & myopia status)

^{* =}significant sex effect in those aged 19–25 (data pooled across ethnicity & myopia status)

^{^ =}significant ethnicity effect in those aged 12–18 (data pooled across sex & myopia status)

^{+ =}significant ethnicity effect in those aged 19–25 (data pooled across sex & myopia status)

Note: no significant effect found related to myopia status for any factor in either sex or age group. Unit abbreviations: centimeters (cm), kilogram per centimeter (kg/cm), millimole (mmol), milligram (mg), nanomoles per liter (nmol/L), millimoles per liter (mmol/L), picomoles per liter (pmol/L)