Retinal phototoxicity and the evaluation of the blue light hazard of a new solid-state lighting technology.

Exposure Limit Values (ELV) for artificial lighting were defined in order to prevent light-induced damage to the retina. The evaluation of the lighting devices include the correction of their spectra by the B(λ) function or blue light hazard function, representing the relative spectral sensitivity of the human eye to the blue light. This weighting function peaks between 435 and 440 nm. In this study we evaluate a new generation of light emitting diode (LED), the GaN-on-GaN (gallium nitride on gallium nitride) LED, that present an emission peak in the purple part of the spectrum. Wistar rats were exposed to GaN-on-GaN and conventional diodes at different retinal doses (from 2.2 to 0.5 J/cm2). We show that GaN-on-GaN diodes are more toxic than conventional LED for the rat neural retina and the rat retinal pigment epithelium, indicating that the BLH (blue light hazard) weighting is not adapted to this type of diodes. One of the reasons of this increased toxicity is the effects of shorter wavelengths on mitochondria polarization. We also show that the threshold of phototoxic retinal dose in the rat (fixed at 11 J/cm2, BLH weighted) is overestimated, suggesting that the values used for regulations, calculated in primates using the same methods than in rats, should be revised.

All the LEDs used in the device were previously seasoned for at least 100 h and installed after reaching a stable light output. The light levels and light spectra were measured at the location of the cage (center of the device). Before and after the experiments, the illuminance levels were measured to check the stability of the device.
The transmission spectrum of the cage was not measured. However, the light spectra and irradiance levels were measured using sensors placed within the transparent cage. As the cage was made of a neutral uncolored transparent material, it did not introduce any spectral distortion of the light emitted by the device.

Concerning the light doses used in this paper:
IEC 62471 is a safety standard for light sources. It defines a classification of light sources based on risk groups. The risk groups are defined by the exposure time required to exceed the internationally accepted limit values set by ICNIRP. By definition of IEC 62471, a light source classified in risk group 0 (no risk) does not exceed the ICNIRP exposure limit in 10 000s, at 20 cm from the source.
Since the IEC 62471 is a standard on light sources, it expresses the exposure limit in terms of light source quantities, not retinal quantities. For the retinal blue light hazard, the ICNIRP exposure limits are the following: For exposure time between 0.25 s and 10 000 s : Exposure limit DELB = 10 000 J/m²/sr : this a dose of BLH-weighted radiance of light source For exposure time exceeding 10 000 s : Exposure limit LELB = 100 W/m²/sr : this a BLH-weighted radiance of light source (not a dose anymore) The ICNIRP guidelines give a formula to compute retinal irradiance as a function of source radiance (equation 2 in the guidelines). Using this formula allows us to express the exposure limit in terms of retinal irradiance dose. With the human eye parameters used in the guidelines (transmittance = 0.9, effective focal length = 17 mm, pupil diameter 3 mm), the exposure limit in terms of retinal irradiance dose is exactly 2.2 J/m² , for exposure times between 0.25 s and 10 000 s.
The retinal exposure limit of 2.2 J/cm² for blue light is also explicitly given as a "basic restriction" in another ICNRIP publication: The paper of Van Norren (reference 2, figure 1 right panel) recapitulates several studies in photo toxicology, and shows a photosensitivity threshold for blue light for the monkey of 22 J/cm² .This is higher by a factor of 10 compared with the ICNIRP basic restriction, which is a logical safety factor.
The threshold value for the monkey is 22 J/cm². But for the rat, this value is lower: 11 J/cm². This is not found in the ICNIRP guidelines but in the paper of van Norren (figure 1, left panel). If we impose to this value the same protection factor that was used for the monkey. This gives a complete "safe" dose of 1.1 J/cm². This was the reason for using 1 J/cm².
It is important to note that at 1 J/cm² the damage of the retina is still important. Therefore we made new experiments at 0.5 J/cm², half of the dose, to see if this dose is toxic.

Concerning the BLH function:
The BLH function (action spectrum) takes into account the Ham's damage to the retina, that is the so called type II damage. The type I damage described by Noell affects mostly photoreceptors and particularly rods. The use of the BLH function to evaluate phototoxicity in humans not considering type I damage comes from the fact that human retina is rich in cones and thus, only the damage of RPE (type II) seems relevant.
However, it has to be considered that the exclusive presence of cones in the retina only involves the macula. The rate of cones/rods decreases as we go from the macula to the periphery. In periphery, type I damage could be significant. This is important for vision: people presenting retinitis pigmentosa lose their peripheral vision and keeps only a tunnel vision, making their life very difficult. They are legally blind. In addition, the shrinking of the visual field is a known feature in human ageing. We agree with the fact that our results are not transposable to human and even less to human macula but they can give some idea of the light toxicity in human peripheral retina.
Moreover, the use of the BLH discards the opposite part of the visual spectrum, this is the red part. This is very important since protective effects for different cells including retinal cells retina have been described in the literature. These effects are called photo-modulation. So that, it would be probably not the same thing to be exposed to a light with an balanced spectrum or to a light poor in red wavelength. This was discussed in the ANSES report of 2019 (https://www.anses.fr/fr/system/files/AP2014SA0253Ra.pdf). In addition, the use of BLH function assumes that wavelength below 425 nm are efficiently absorbed by the lens and the cornea. This is true for adults but not for children.
In conclusion, the BLH function is very useful to evaluate photochemical damage to the macula but it skips the effects induced by other wavelengths (green and red) that can have synergic or antagonic effects in retinal cells. The data that we show here shows this fact experimentally, using commercially available lamps.