Overall, results from our study failed to identify factors that substantially predict urinary measures of aMT6s. The only factors that were significantly related to aMT6s levels were age, length-of-night, and SES and together they explained very little of the variance in aMT6s levels. Nevertheless, our study does provide some important findings relevant to future investigations of health outcomes related to melatonin and/or LAN exposures.
There is increasing interest in the use of satellite imagery data to evaluate potential effects of outdoor environmental light pollution in wildlife and in human populations, including a number of studies that have investigated potential links to cancer incidence[20–22]. Our analyses demonstrated that the widely-available low-dynamic range annual satellite imagery data are insufficient for distinguishing areas with differing LAN values within suburban and urban areas. Since many health outcomes, including many cancers, are more common in urbanized areas[23–27], use of the low-dynamic range satellite data will likely be inadequate for investigating LAN exposures within such areas. While the high-dynamic range data are currently only available for 2006, efforts are underway to develop such high-dynamic range data for other years. Researchers interested in this topic would be prudent to pursue such data.
The other novel finding generated by our analyses was the significance of neighborhood SES. Our analysis showed that aMT6s concentrations tended to be low among women living in the lowest SES areas and increased with greater neighborhood SES up to a point after which the concentration leveled off and possibly declined slightly at the very highest levels of SES. Furthermore, this relationship, while statistically significant in the full study population, appeared to be of lesser importance in the women over age 55. Overall, very little is known about how melatonin varies by SES. Similar to our findings, a study of approximately 200 Canadian women reported that urinary measures of aMT6s were lower among the lowest educated women (i.e. high school diploma or less), highest in the middle category of education, with a slight decrease in the highest educated group. In contrast, Burgess and colleagues in their evaluation of predictors of salivary measures of melatonin found education was not a significant predictor.
Interpretation of our findings on SES can only be speculative. It is possible that the neighborhood measure of SES is serving as a proxy for a number of behavioral and/or anthropomorphic factors that influence melatonin levels. We did, however, incorporate variables in our model selection process for many of the most important of these factors including alcohol consumption, smoking, BMI, use of exogenous hormones and other medications, coffee consumption, and physical activity – none of which were independent and significant predictors of aMT6s. Another possibility is that low SES neighborhoods may have environmental stressors (e.g., noise, crime) other than light pollution that we did not measure in our study but could disrupt circadian rhythms and lead to lower levels of aMT6s. Given that a number of the health outcomes of high interest with respect to circadian disruption display strong SES-related risk gradients, research into other features of the built environment represents an important area for future inquiry.
In addition to SES, age and length-of-night were the only other significant predictors of aMT6s identified in our analysis. These findings are generally consistent with the limited body of literature on this topic. Two studies have reported seasonal differences in melatonin levels with higher levels associated with greater length-of-night[15, 17] and lower levels in the summer months, although two other studies reported no relationship with month or season. Most studies have reported declines in melatonin levels with increasing age, albeit most have reported this relationship to be linear[15, 17, 18, 31–33]. Consistent with our results, two studies noted more dramatic age-related declines among older individuals[34, 35]; conversely, a few other studies suggested that the age-related declines in melatonin are greater, or are limited to, early adulthood and then level-off[18, 36–38]. Measurements of melatonin in older adults living under controlled light–dark conditions in one study reported no age-related declines in melatonin levels, leading the authors to suggest that changes in sleep behaviors might be responsible for the age-related declines reported in other observational studies. In our study, however, we observed no differences in sleep duration by age group (data not show). Our results, in the context of the somewhat conflicting literature on this, underscore the need to carefully consider the shape of the age-related risks when conducting health-related research on this topic. The fact that we observed a shift in the relationship between age and aMT6s around age 55 suggests this may be especially important when evaluating health risks in women, whose risks often change in their early- to mid-50s after menopause. This may be particularly relevant to breast cancer which exhibits a different constellation of risk relationships in pre- and post-menopausal women and for which LAN exposures have been postulated to be a potential risk factor.
Our evaluation of outdoor LAN, which showed a very modest inverse, but not statistically significant, relationship to aMT6s concentrations, was hindered by a number of obstacles. While we used the best satellite imagery data available to objectively estimate outdoor LAN values, because the high-dynamic range data was only available for 2006, the LAN estimates were not temporally congruent with the urine in which the aMT6s was measured (collected in 2000). An examination of the low-dynamic range data for all years spanning this time period (2000–2006), however, suggested relative stability in LAN values over this time for the study area. Furthermore, the Spearman rank correlation between the 2000 low-dynamic range data and the 2006 high-dynamic range data was 0.96.
Another reason for the lack of an association seen between outdoor LAN and urinary measures of aMT6s could be that while the satellite imagery data may be an adequate predictor of outdoor ambient light, it may not reflect light exposures experienced at night when participants are likely to spend the majority of their time indoors. Intervening factors, such as the use of blinds/curtains, time spent indoors, other sources of indoor light, etc. all are likely contributors to LAN exposures. The importance of this is underscored by findings from a recent study that compared estimates of LAN from satellite imagery data to calibrated photometric measurements obtained from personal monitors and reported that satellite imagery data did not correlate with personal photometric measurements. Thus, it is important that studies aimed at evaluating outdoor LAN exposures do so with full consideration of these other factors.
Finally, there are a number of limitations in the estimates of our melatonin levels that are worth noting. The urine specimens were collected over a 24-hour time period. This precludes our ability to examine the timing of peak melatonin concentration which is likely an important factor in circadian disruption. Furthermore, while there is a good deal of evidence that first morning urine captures peak night-time melatonin excretion, the degree to which 24-hour urine specimens reflect this is not well-documented[13, 31]. The aMT6s assay, however, is a well-validated biomarker, demonstrating good correlation with serum melatonin levels, and sufficient stability over time[13, 15, 31, 41, 42] to serve as a reasonable estimate of chronic levels.