Abstract

Nonvisual responses to light, such as melatonin suppression, are classically investigated by comparing illumination conditions that differ in intensity, spectrum, duration, and timing. When interpreting the data from such studies, and inferring the mechanisms involved, it is important to carefully consider the characteristics of the light stimulus used.
Data recently published by Nagare et al. (2019) highlight this point. Using a relevant within-design experimental protocol, the authors exposed 16 participants to 2 light spectra (3000 K vs. a modified 3000 K light-emitting diode source with a cyan gap) at 2 light intensities (low 400 lux vs. high 800 lux) for 1 h, starting at midnight, and they measured the magnitude of melatonin suppression induced by each light condition. Significant differences were found between high and low intensities, but there was no significant main effect of light spectrum. The authors interpreted the absence of a differential effect between the 3000 K and the cyan-gap spectra, at both intensities, as evidence that “cyan-gap light sources affect the human circadian system similarly to conventional light sources of the same CCT and photopic illuminance at the eyes.” This statement, however, has to be interpreted with caution, and we would like to underscore important contextual considerations.
First, the melanopic content of a light stimulus is the predominant predictor of melatonin suppression, and the sigmoidal irradiance response curve (IRC) linking the intensity of the stimulus and melatonin response (R² = 0.87 with melanopic intensity; Prayag et al., 2019) should be taken into account carefully in research designs. This IRC predicts that non-negligible levels of suppression can be observed at very low melanopic intensities (1.5 melanopic lux) and that saturation of the response can be reached with room light levels only (305 melanopic lux). This mathematical model is also characterized by a slope of 0.82% suppression per melanopic unit (lux/irradiance) between 8 and 55 melanopic lux. Quasi-identical thresholds were found very recently by Phillips et al. (2019). In their study, Nagare et al. (2019) used 2 low-light conditions (122 vs. 187 melanopic lux, from the cyan-gap source versus the 3000 K light, respectively) and 2 high-light conditions (265 vs. 436 melanopic lux), separated only by 0.2 log units between pairs in terms of melanopic intensity. Our model, as well as that of Phillips et al. (2019), predicts that such a small melanopic lux ratio could produce only marginal differences in melatonin suppression (3% difference between low-light condition pairs and 2% difference between high-light conditions). This prediction is in accordance with the results of Nagare et al., which report similar melatonin suppression levels at low (34% and 36%, difference = 2%) and high intensities (49% and 50%, difference = 1%). As a note of caution, the data used to derive the model by Prayag et al. (2019) were collected at approximately the same clock time as in Nagare et al. (2019) but with a dilated pupil and duration of light exposure of 90 min instead of 60 min. Accounting for those experimental differences would lead to a slight decrease in melatonin suppression because of the difference in duration (by ~7% according to Gronfier et al., Unpublished) and a further significant decrease due to the nondilated pupil in Nagare et al. (~12 times less light on the retina with nondilated pupil, according to our calculations based on Gooley et al., 2012). Altogether, however, this would not change the conclusion that the absence of a difference between the 2 light spectra, at the 2 intensities, is likely attributable to an insufficient contrast between the melanopic content of those light exposures and that caution must be exercised with the methodology used to estimate melatonin suppression (see below).
Second, methodologies used to estimate melatonin suppression are not insignificant. Using equation 1 below, Nagare et al. found a melatonin suppression of 34% and 36% under low light and of 49% and 50% under high light.
where Mn is the normalized melatonin concentration at each time point on the respective intervention nights and Md is the normalized melatonin concentration at each corresponding time point on the dim-light control night. When we recalculate the degree of melatonin suppression using our approach (equation 2 below; Prayag et al., 2019), with no prior normalization and with control adjustment, we find that melatonin suppression is 58% and 62% with high light intensity. The magnitude of this suppression is close to the saturation level we found (~72%) in our nonnormalized IRC, which decreases the likelihood of finding a difference between the 2 light spectra (3000 K and cyan-gap source).
where C0 is the melatonin concentration at each time point on the respective intervention nights and Ct is the melatonin concentration at each corresponding time point on the dim-light control night. From this score, the control-adjusted percentage suppression was obtained by subtracting the control (no light) condition percentage suppression score for each subject from that same subject’s light exposure score.
Taken together, with the relatively small range of melanopic content and the relatively high levels of melanopic illuminance used in Nagare et al.’s article, the conclusion that “cyan-gap light sources entertain similar effects compared to light sources of same CCT and photopic illuminance” is not cautious. Indeed, this result cannot be generalized to all cyan-gap light sources and all light intensities, and the conclusion can be applied only to the sources and intensity ranges used by the authors, at that particular time of day, and in that population (12 women, 4 men; 37.8 ± 13.9 years of age).
These results by Nagare et al. (2019) emphasize the limitation of contrasting light sources, including those with reduced power of about 480 nm corresponding to the peak sensitivity of the circadian timing system and the intrinsically photosensitive retinal ganglion cells (ipRGCs). The cyan-gap light spectra used by Nagare et al. have a reduced power (irradiance) between 450 and 545 nm by a factor of 2.5 compared with the original 3000 K light, which reduces melanopic lux content; they also have approximately 5 times more power (irradiance) at 440 nm compared with the original 3000 K, which adds melanopic content. As a result, the overall melanopic gap between the 2 spectra is modest (only 0.2 log units) and maintains the circadian effectiveness of this cyan-gap light source.
Finally, although our field has used melatonin suppression as a marker of the effects of light on the circadian system in the past, recent findings by Rahman et al. (2018) showed that circadian phase resetting and melatonin suppression are functionally decoupled in humans, such that one cannot be used as a proxy measure of the other. Therefore, the conclusion that “the results of the present study suggest that, for short-term exposures (≤1 h) ‘cyan-gap’ light sources affect the human circadian system similarly to conventional light sources of the same CCT and photopic illuminance at the eyes” cannot be drawn from melatonin suppression.
