Commercially Available Phototherapy Devices for Treatment of Depression: Physical Characteristics of Emitted Light

We identified four SPD patterns (Figure 2). Fluorescent devices emitted one of two spectral signatures: a “warm” color-temperature with peaks of emitted energy (e.g., X3, X4, X9) (Figure 2a) and a “cool” color-temperature with similar peaks but increased energy in short and middle wavelengths (e.g., X1, X11, X13) (Figure 2b). Monochromatic devices included one blue-light LED device peaking around 413 nm (M4) and two green-light devices, one with an LED (peaking at 501 nm; V1) and one with a filtered fluorescent source (peaking at 503 nm; C1) (Figure 2c). White LED devices (X10, M1, M2, M3, V3; V2 was similar to V3) emitted a short-wavelength peak in the short (blue-appearing) wavelengths with a lower, broader peak in the middle wavelengths, generating light that appeared bluish-white (Figure 2d). Among these devices, there was a range of wavelengths at peak emission: 450 nm for X10, 461 nm for M1, 453 nm for M2, 445 nm for M3, and 463 nm for V3. The SPD patterns are shown relative to the sensitivities of the three cone-specific photoreceptors, the melanopic absorption curve, and the blue hazard curve (Figure 2e).

Table 1 lists illuminance at the MRD for each device. Most light boxes produced ≥7,000 lux at the MRD. Light intensity received by the observer decreased with increasing distance from the device (Figure 3a). The larger devices were brighter and produced ≥7,000 lux over a wide range of distances. For example, device X3 produced 12,000 lux at a distance of about 20 inches and 7,000 lux at a distance of approximately 28 inches. Smaller devices produced less intense light. For example, device X9 produced 7,000 lux at a distance of 13.5 inches, but a user any farther from this device would receive less illuminance than our minimum. Of the LED devices, one produced 3,750 lux at the MRD, but the other three produced substantially dimmer illuminance (Table 1). The three visor devices,V1, V2, and V3, produced 8,900, 13,800, and 5,900 lux, respectively, when assessed at 1.5 inches—the approximate exposure distance when worn. Our measurements generally agreed with manufacturer specifications where provided.

All but one of the large light boxes and all but one of the small boxes emitted ≥5,000 lux 6 inches radially off center in all directions (Table 1). Figure 3b shows the intensity distribution in the plane of the observer for a light box that met this criterion (X1); Figure 3c shows one that did not meet this criterion (X9); Figure 3d shows the narrow field produced by a beam device (M1); Figure S1 in the online supplement is a three-dimensional illustration of how illuminance decreased with movement off center of the plane of the observer for one large box (X4).

Mean±SD glare ratings by the volunteers are shown in Table 1. There was limited correlation between glare and illuminance; at a given illuminance, some devices produced much more glare than others. As a group, small light boxes were most likely to be perceived as glaring (i.e., score≥3.75), with three of these five devices considered uncomfortable to look at. We considered many devices to have inadequate diffusion with noticeable hot spots. None of the devices evidenced subjective flicker.

If melanopic lux represents an estimate of the therapeutic effect of a particular SPD, the efficacy ratio, or melanopic lux divided by photopic lux, represents the postulated therapeutic effect per unit of perceived brightness for a given source. Table 1 shows that the efficacy ratios of warm fluorescent devices were generally lower than those of cool fluorescent devices (mean=0.52, range 0.45–0.59, versus 0.86, 0.68–1.28, respectively). White LED devices preferentially emitted light with shorter wavelengths, resulting in higher efficacy ratios (mean=1.11, range 0.86–1.33). The two green-emitting devices had high efficacy ratios of 2.35 and 2.37, as did the one blue-emitting LED device (34.5). In this melanopic model, cool fluorescent sources appeared to have higher efficacy ratios than warm fluorescent devices, and the green and blue monochromatic sources had the highest efficacy ratios.

In Table 1, we report a protection ratio for each device, defined as the ratio of melanopic lux to blue light hazard, representing postulated therapeutic benefit per unit of hazard. Warm fluorescent devices had slightly higher protection ratios (mean=1.19, range 1.09–1.30) than cool fluorescent devices (mean=1.07, range 0.90–1.25). White LED devices (mean=1.09, range 0.83–1.41) showed a broader range, reflecting differing patterns of short- versus middle-wavelength energy, as illustrated in Figure 2d. The two green-emitting devices had high protection ratios of 5.39 and 6.35, as these devices emit very little short-wavelength light. The deep-blue-emitting LED device had the worst protection ratio, 0.13, which could be expected from its peak energy emission at 413 nm.

Most evidence supporting the efficacy of bright-light therapy is based on large 10,000-lux white (broad-spectrum) light boxes, such as X1–X4. Clinicians can inform their patients that these four devices are comparable to devices used in research and meet clinical guidelines for light therapy. X1–X3 produce 10,000 lux over a large area. X4 produces 10,000 lux at a shorter distance but has legs that permit easy use at this shorter distance and has been validated by research (6, 7). Clinicians may also indicate that light boxes such as X5, X6, and X11, are smaller options that may offer increased convenience but constrain user experience. There are free lux-measurement cell phone apps that permit patients to verify they are receiving 10,000 lux at a specific distance from the light source.

These devices emit light with energy relatively concentrated in the shorter (blue-appearing) and middle (green-appearing) wavelengths, to which the human circadian system is particularly sensitive (15). The retinal pigment melanopsin maximally absorbs such wavelengths and activates intrinsically photosensitive retinal ganglion cells, which project directly to the suprachiasmatic nucleus, the master clock of the circadian rhythm system (16). Recent research in mice suggests that the mood-enhancing effects of light may be due to a subset of such ganglion cells projecting to the perihabenular region (17). One small trial demonstrated therapeutic efficacy in seasonal affective disorder for bright green light compared with dim red light (18). Placebo-controlled trials of LED devices for patients with the disorder have shown efficacy for 398-lux narrow-band blue light (19), for 200-lux narrow-band blue light (20), and for 1,350-lux blue-enriched white light (21). A trial comparing 98-lux narrow-band blue light with 711-lux blue-enriched white light found similar therapeutic effect in patients with seasonal affective disorder (22), and a trial of 100-lux narrow-band blue light versus 10,000-lux bright white light found similar effect in patients with subsyndromal seasonal affective disorder (23). Given the absence of a placebo condition in the latter two studies, these findings should be considered preliminary. In general, however, the response rates found in these studies were similar to those of trials with 10,000-lux white light boxes. Together, these studies suggest that short and middle wavelengths are particularly effective for patients with seasonal affective disorder.

Of the devices we tested, M1 was found to be efficacious for the treatment of seasonal affective disorder in a randomized trial with 106 participants (24). The other three beam devices we tested provided much less intensity of exposure and were judged to be glaring or inadequately diffused. Device M4 emitted very short-wavelength light (413 nm peak), which not only increases blue hazard but is lower than the peak absorption wavelength of the melanopsin system. This device had the worst protection ratio of devices studied. Light in the deep-blue regions of the spectrum may offer a poor balance of therapeutic potential and retinal risk.

Certain patients nevertheless prefer beam devices (including M1 and other LED devices not included in this study) because of their portability and compact design, provided there is adequate diffusion to reduce glare. Clinicians may discuss a trial of such devices to offer certain practical advantages to patients, but patients need to know they must position these devices carefully to keep the light beam focused on the eyes, typically around 30 degrees from the line of sight. None of these devices produce 10,000 lux at MRD, and it is unknown whether different intensity requirements for efficacy apply to these small concentrated light sources than to large light box sources.

Early attempts at using light visors as therapy for seasonal affective disorder were unsuccessful, but the visors in published controlled trials emitted white incandescent light with most energy at long wavelengths (25). A promising open-label 4-week trial of an LED visor similar to V3 found full recovery (score <8 on the Structured Interview Guide for the Hamilton Depression Rating Scale–Seasonal Affective Disorder Version, a Hamilton Depression Rating Scale modified to reflect atypical neurovegetative symptoms of seasonal affective disorder) in 10 of 11 patients at the trial’s end. Although open-label, these preliminary data suggest that a visor device may provide adequate light to serve as a therapeutic option (26). We are unaware of any published placebo-controlled trial of an LED visor device. Of the three visor devices we studied (V1–V3), two provided ≥7,000 lux, and the third provided almost 6,000 lux. V1 emitted middle-wavelength, green-appearing light; V2 and V3 emitted bluish-tinted white light. All three visor devices were relatively glaring and warrant better diffusion. Because of their convenience, visor devices may interest certain patients. In discussing such devices with patients, the clinician should instruct the patient to adjust the visor device with a mirror to ensure direct illumination of the eyes, because patients are likely to reduce glare by misadjusting the device.

The one column device (C1) we studied—one of only two green-light devices in the study—produced minimal glare and had a small tabletop footprint. Prior research found that a similar device phase shifted circadian rhythms similarly to a 10,000-lux light box (27) and potentiated the antidepressant effect (28). These previous studies, however, used two columns, so patients would need to purchase two columns or a double column unit to replicate those effects. In discussing this option with patients, clinicians should indicate that further research is needed with this type of device.

We tested only a portion of the commercially available devices, and failure to include a device should not be taken as a negative recommendation. Some manufacturers declined to provide devices for testing, and others submitted only part of their product line. Additionally, given the apparent trade-off in intensity and duration (11), dimmer devices may be effective if used for longer periods, but the required treatment duration may be prohibitively inconvenient. Conversely, satisfying our three clinical criteria may suggest but does not guarantee efficacy for a device: such a guarantee would require an adequate clinical trial.

An unavoidable limitation to our study was that inadequate research has been done regarding optimal wavelengths for bright-light therapy. Most devices in this study emitted broadband white or bluish-white light with protection ratios near 1. One device emitted very deep-blue light resulting in a much lower protection ratio because of increased blue light hazard. There is no evidence that treatment efficacy is greater with short-wavelength light than with adequate levels of broad-spectrum white light, and devices that emit high levels of short-wavelength light may pose unnecessary risk to the retina. Two devices in the study emitted green light with notably higher protection ratios. Figure 2e illustrates the comparative sensitivities of the cyanopic (blue), chloropic (green), and erythropic (red) cone systems, the melanopic system, and blue light hazard estimated from primate data. If melanopic lux predicts antidepressant response, then green-appearing light in the 500–540 nm range may provide the best balance of therapeutic efficacy and safety. Research on the therapeutic efficacy of light in the green wavelengths is urgently needed. Our results echo the conclusions of Baczynska and Price (12): marketed light therapy devices represent a wide range of intensities, directionalities, and spectral properties, and only limited research is available to clarify optimal parameters for light treatment.

Clinicians should note that the U.S. Food and Drug Administration has yet to approve any light therapy device for depression and has not regulated such devices. Despite evidence on its antidepressant efficacy, bright-light therapy has not been studied to the same extent as pharmacological approaches. A properly diffused, conventional large light box exposes the eye to less light than being outdoors on a sunny day, and research suggests such boxes are ophthalmologically safe (29). However, there are limited data about the long-term effects of bright-light therapy in general, and no data regarding the safety of newer device types (30). Caution is needed in treating individuals with retinal disease, medical illness such as diabetes associated with retinal disease, and concurrent use of photosensitizing medications (11).

Of the 24 devices tested, only four were similar (or identical) to the large light boxes used in most research on bright-light therapy (X1–X4). Three devices were smaller light boxes that met our three criteria (X5, X6, and X11). Five devices were found to have properties supported by at least some research (M1, V1–V3, and C1). These devices are identified in Table S1 of the online supplement. Descriptions and photographs of these devices are provided on our institutional website (https://medicine.yale.edu/psychiatry/research/programs/clinical_people/winter.aspx).

Traditional light boxes, such as X1–X4, are generally preferred as a first-line approach to bright-light therapy, given that they are best supported by research. After response to the therapy is established, however, a trial of a smaller device that offers certain advantages, such as convenience, may be attractive to some patients. Clinicians should inform their patients that many devices marketed on the Internet do not produce 10,000 lux at a reasonable distance, over a reasonable area, with a reasonable degree of comfort. Clinicians should only recommend light boxes that specify the distance at which they produce 10,000 lux.

Our results should enable clinicians to understand what types of light devices are best supported by current research and to discuss options knowledgeably with patients. We urge clinicians to familiarize themselves with clinical guidelines (Box 1) for bright-light therapy and to increase their use of this underutilized treatment for depression.