Sunlight exposure in infancy decreases risk of sporadic retinoblastoma, extent of intraocular disease

Abstract Background Prior ecologic studies suggest that UV exposure through sunlight to the retina might contribute to increased retinoblastoma incidence. Aims Our study objectives were (1) to examine the relationship between exposure to sunlight during postnatal retinal development (prior to diagnosis of sporadic disease) and the risk of retinoblastoma, and (2) to examine the relationship between sun exposure during postnatal retinal development, and the extent of disease among children with unilateral and bilateral retinoblastoma. Methods and results We interviewed 511 mothers in the EpiRbMx case‐control study about their child's exposure to sunlight during postnatal retinal cell division by examining three time periods prior to Rtb diagnosis coinciding with developmental stages in which outdoor activities vary. Weekly sun exposure was compared by age period, between unilateral (n = 259), bilateral (n = 120), and control (n = 132) children, accounting for two factors affecting UV exposure: residential elevation and reported use of coverings to shield eyes. For cases, association between sunlight exposure and clinical stage was examined by laterality at each age period. After adjusting for maternal education and elevation, sun exposure was lower in cases than controls in all three age periods especially during the first 6 months, and in children 12–23 months whose mothers did not cover their eyes when outdoors. In children diagnosed after 12 months of age, sun exposure during the second year of life (age 12–23 months) appeared inversely correlated (r = −0.25) with more advanced intraocular disease in bilateral Rtb children after adjusting for maternal education, residential elevation, and age of diagnosis (p < .09) consistent with effects of Vitamin D exposure on intraocular spread in earlier transgenic murine models of retinoblastoma, and suggesting potential chemopreventive strategies. Conclusion Sun exposure in early childhood is protective for retinoblastoma and may decrease degree of intraocular spread in children with bilateral Rtb.

whose mothers did not cover their eyes when outdoors. In children diagnosed after 12 months of age, sun exposure during the second year of life (age 12-23 months) appeared inversely correlated (r = −0.25) with more advanced intraocular disease in bilateral Rtb children after adjusting for maternal education, residential elevation, and age of diagnosis (p < .09) consistent with effects of Vitamin D exposure on intraocular spread in earlier transgenic murine models of retinoblastoma, and suggesting potential chemopreventive strategies.
Conclusion: Sun exposure in early childhood is protective for retinoblastoma and may decrease degree of intraocular spread in children with bilateral Rtb. Among pediatric tumors, Rtb has uniquely variable incidence with an approximately 50-fold geographic variation 5,6 suggesting environmental risk factors, 7 with higher incidence in the global south, in some ethnic groups such as US Native Americans, and in poorer subgroups in Latin America, 5,[8][9][10][11] with low-and middle-income countries having older ages at diagnosis and higher prevalence of more invasive disease. 12 Although mouse models have been critical for understanding aspects of Rtb tumorigenesis, mice with mutations in RB1-related genes develop retinal tumors that differ from those that develop in humans. 13 Hooper proposed that the lack of exposure to sunlight in laboratory mice might contribute to these differences, and that retinal exposure to sunlight might contribute to human retinal tumor formation. Specifically, he hypothesized that erythemal dose of ambient ultraviolet B radiation from sunlight increased the incidence of unilateral (but not bilateral) Rtb in humans, thus explaining the geographic variation in incidence. 14 Tropical climate, ethnic variation in UV susceptibility, and economic development have been suggested as confounders in Rtb's association with UV exposure, 15 and an ecological study suggested that increasing latitude (distance from the equator) decreased risk of Rtb (both disease forms), after adjusting for national level economic features. 16 One California-based study examining UV exposure during pregnancy (based on residential location, not individual exposure) found no association though the data suggested potentially decreased risk with the highest exposure (though the disease forms were not examined separately). 17 However, no studies examining ambient UV exposure in Rtb have accounted for individual differences in duration or age at time of UV exposure.
Perinatal exposure to various environmental exposures including parental diet and air pollution have been associated with Rtb risk.
Exposures impacting bilateral Rtb occur prior to conception while those impacting unilateral Rtb are during gestation or early childhood. [18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34] Few studies examined risk accounting for clinical stage or risk for disease progression, though in our study in Mexico (EpiRbMx), maternal education predicts extent of both intraocular (in bilateral) and extraocular (in unilateral) disease. 35,36 Sunlight exposure in temperate climates can serve as a source of UV. Although the American Academy of Pediatrics recommends that infants younger than 6 months should be kept out of direct sunlight and covered with protective clothing and hats, 37 globally mothers routinely take their infants outdoors, though the amount of time infants spend outdoors varies with multiple factors including social norms. Adherence to recommendations for using protective clothing is variable. Studies regarding sun protection behaviors have principally assessed adult behaviors, and data among caretakers and their practices with their children are limited, though one US-based study reported that less acculturated Hispanics were more likely to use sun protective clothing and hats instead of sunscreen. 38 Our study objectives were (1) to examine the relationship between exposure to sunlight during postnatal retinal development (prior to diagnosis of sporadic Rtb) and the risk of retinoblastoma, and (2) to examine the relationship between sun exposure during postnatal retinal development, and the extent of disease among children with unilateral and bilateral retinoblastoma.

| METHODS
This study includes participants in the ongoing IRB approved EpiRbMx (see Appendix S1 for additional details) study enrolled through February 2018 35 Consenting mothers of children with Rtb were interviewed at the time of diagnosis, while controls were interviewed at the time of recruitment. Mothers were interviewed about sun exposure, sociodemographic characteristics, and the pre-, perinatal, and early childhood home environment (for cases, prior to diagnosis) as previously described. 35,36 Mothers were the primary caretakers in the first 2 years of life.
We proposed to examine prediagnosis sun exposure by documenting time spent outdoors during postnatal retinal cell division, use of protective clothing, and clinical stage of Rtb. The sun exposure questionnaire was developed with mothers participating in the pilot phase case-control study where mothers reported routinely taking children outdoors to get sunlight. EpiRbMx mothers were thus queried on children's daily sunlight exposure during three age intervals: 0-5.9 months (corresponding to prior to sitting unassisted), 6-11.9 months (corresponding to children sitting but not walking independently), and 12-23.9 months (corresponding to walking independently but observed). Mothers were asked whether their child was outside during each of these age intervals, the number of minutes they were outside per day, as well as the number of days per week.
Mothers were then asked if they shielded their child's eyes from sunlight during their time outside and to identify the type of covering used. Case mothers were queried on the child's exposure to sun prior to the child's Rtb diagnosis. Sun exposure was calculated as minutes per week for each of the three age periods.

| Clinical stage
Intraocular disease was determined for each affected eye using the International Intraocular Retinoblastoma Classification (IIRC) criteria (from A to E, sequentially predicting lesser probability of eye salvage and generally greater retinal involvement) by the ophthalmologist but were only determined for patients from one recruiting hospital (the Hospital Infantil de Mexico) due to limited retcam availability. 39 For analyses, we used the classification for the more affected eye for bilateral cases. Extent of extraretinal spread using St Jude's (Pratt) staging 40 and International Staging System (ISS) 41 was assigned by the treating oncologist and study pathologist.

| Elevation
UV exposure does not vary appreciably with latitude within Mexico, but varies with time of day at which exposure occurs and geographic elevation. 42 We used residential postal codes to approximate geographic elevation. Postal codes corresponding to the family's primary residence at time of diagnosis (for cases) or of interview (for controls) 43 were entered into ArcGIS version 10.7 (ESRI, Redlands, CA), using the dynamic Terrain services that provide elevation values for use in analysis 44 (See Appendix S1 for more details).

| Statistical analysis
Descriptive analysis examined groups of controls, unilateral or bilateral cases, using Kruskal Wallis and Chi-square tests to detect group differences in quantitative and categorical variables, respectively, and Spearman correlation coefficient for bivariate associations. Linear models examined group differences in sun exposure (minutes per week) by age controlling for variables that were significant predictors of sun exposure in bivariate associations, namely maternal education, as well as elevation, where sun exposure with right skewed distribution was log-transformed to meet linear model assumptions, reduce impact of extreme values and improve model fitting. We calculated the geometric mean (GM) of sun exposure and derived covariateadjusted GM ratios with 95%confidence intervals (CI) for specific group comparisons using estimated model parameters. We assessed age specific association between duration of sunlight exposure and numerified disease stage (see Appendix S1) 35,36 for each laterality controlling for maternal education, elevation, and child's age at diagnosis using Spearman correlation coefficient.
All statistical tests were two-sided with significance level preset at .05. Data analysis used SAS 9.4.

| RESULTS
Enrollment to EpiRbMx was high (97.9%). A total of 511 children (259 with unilateral Rtb, 120 with bilateral Rtb, and 132 controls) who met eligibility criteria for EpiRbMx were included (see Figure S1). Table 1 shows demographic and geographic characteristics. Only age at diagnosis differed between unilateral and bilateral cases as expected, while all other characteristics including those of birth and geographic distribution did not differ between unilateral or bilateral cases or controls. The distribution of residential elevation is shown in Figure S2. Maternal age at interview and at child's birth, child's birth weight, and residential elevation were all unrelated to the amount of sun exposure at any age period.
For age 0-6 months, 500 mothers reported sun exposure, 458 answered whether they covered their child's eyes while outdoors, and 395 reported the types of coverings they used to cover their children's eyes. For age 6-12 months, 444 mothers reported on sun exposure, with 427 responding on whether they used a covering to shield their children's eyes from sunlight. In the second year of life, 371 mothers reported sun exposure, with 367 responding on using a covering to shield their children's eyes. As expected, the proportion of mothers reporting using a covering to shield their child's eyes decreased with the child's age. In the first 6 months of life, 84% (384) reported using a covering to shield their child's eyes when in the sun.  Table 2 presents the mean exposure during each age period, with and without excluding mothers reported that they did not take their child outside. Sunlight exposure was further stratified by whether or not mothers reported using a covering to shield their child's eyes from sunlight when taking them outdoors. As expected, sun exposure increased with age in cases and controls. In the first 6 months of life, controls spent significantly more weekly minutes in the sun than the cases. When examined separately by whether eyes were covered while outdoors, the pattern in sunlight exposure was similar among the younger infants (<6 months) regardless of eye covering. However, while group differences between cases and controls seemed larger in those with eyes uncovered during sun exposure, this difference was not statistically significant.
In the older age groups, controls had a higher mean sunlight exposure than cases.
Among older children whose wore a covering to shield their eyes from sunlight, mean sunlight exposure in controls was either between those of the two case groups (6.1-11.9 months) or were similar to the cases (12-23.9 months), while among children without a covering to shield the child's eyes from sunlight, mean sunlight exposure in controls was much higher than in cases. Together, our data suggest a protective effect of sun exposure modified by wearing a covering to shield the child's eyes from sunlight.
To further examine the group differences in sunlight exposure suggested by Table 2, we used linear regression models to assess covariate-adjusted group differences for the children during the first age period and to assess those with eyes uncovered in the older age periods. Maternal education was inversely related to sunlight exposure with higher correlation in older age periods: specifically, the Spearman correlation coefficient was r = −0.0627 (n = 499, p = . 16) for age < 6 months; r = −0.0976 for age 6-11.9 months (n = 443, p = .0400); and r = −0.1676 for age 12-23.9 months (n = 370, p = .0012). Table 3 shows the covariates-adjusted geometric mean (GM) for weekly positive sun exposure for cases and controls, as well as the geometric mean ratio (GMR) comparing cases of unilateral and bilateral each to controls. After controlling for maternal education and geographic elevation, weekly sun exposure increased with age for children without eye covering, and in each age, cases had less sunlight exposure than controls, indicated by a GMR below one. The difference in covariate adjusted sun exposure was greatest in the <6 month olds whose eyes were not covered, appearing greatest in bilateral cases with a more than 2.2-fold difference between controls and bilateral cases (GMR =0.39; 95% CI: 0.16-0.96). In children with eyes uncovered at age 6-11.9 months, the nearly twofold control to case sun exposure difference was similar between the two disease forms, though the group difference between bilateral cases and controls was not statistically significant due to smaller group sizes.
For children not wearing any covering to shield their eyes in their second year of life, cases again had significantly lower sunlight exposure than controls with again an apparently larger difference for bilateral cases. In summary, our results yielded evidence supporting that sunlight exposure was lower in cases than controls for children with eyes uncovered during exposure, and also for infants under 6 months regardless of whether or not they were reported to wear a covering to shield their eyes.
To explore whether the group differences in sunlight exposure among infants under 6 months depended on the type of eye covering, we examined the types of eye coverings that mothers reported using. Of the 384 children whose mothers reported using a covering to shield their eyes, 319 reported using either a hat or a blanket. A total of 163 mothers reported using a hat to shield (cover) the child's eyes, while 156 used a blanket which covered the back of the head (hood-like) and most of body. Other less frequent "coverings"   Figure S1). Not all cases had stage rating by the three scales. For each rating scale, we calculated mean (SD) of the (numerified) staging score, age of diagnosis and age period-specific sunlight exposure calculated as shown in Table 5 for unilateral and bilateral cases.    Vitamin D deficiency increases risk of multiple chronic diseases 50 and is highly prevalent globally, even in countries with lower latitudes, where it was previously assumed that UV radiation was adequate to prevent deficiency.
T A B L E 6 Spearman partial correlation coefficient (r) for covariate-adjusted association between clinical stage and weekly sunlight exposure, by type of retinoblastoma and time period of exposure  Limitations of our study included potential differential recall biases between case and control mothers, though it is unclear that they would preferentially influence the amount of time reported as being outdoors or the reported use or type of covering. Control children were also older than case children, though this would not differentially affect their reported sunlight exposure. Importantly, our survey findings reflected a time prior to case parents noting symptoms. The association with intraocular clinical stage of disease would not be influenced by such bias as parents would rarely be aware of the granularity of intraocular grading of their child's tumor. Additionally, UV exposure varies diurnally, and we had exact time of day for the sun exposure in only a small subset of our data, though this subset overwhelmingly suggested that most exposure occurs during peak sunlight as expected given cultural norms.

| CONCLUSION
We report apparent differences in sunlight exposure during postnatal retinal development that may contribute to formation and progression of unilateral and bilateral forms of retinoblastoma. Contrary to prior global ecologic studies, sun exposure during infancy and toddlerhood appears protective for retinoblastoma development and appears associated with lesser progression of intraocular disease in bilateral retinoblastoma. Sun exposure may exert a protective effect that is specific to eye exposure in early infancy (a more local effect), or to whole body exposure, suggesting a more systemic effect, that impacts later infancy and the second year of life. Contributions from early life exposure to sunlight, may impact disease progression differently in the two forms of retinoblastoma. Together with data from prior transgenic rodent models, 46,47 these protective effects of sun exposure suggest a vitamin D-related mechanism with potential avenues for chemoprevention and therapy in bilateral retinoblastoma.