Assessing the Effects of Polychromatic Light Exposure on Mood in Adults: A Systematic Review Contrasting α-opic Equivalent Daylight Illuminances

ABSTRACT This is a systematic review of studies assessing the effects of polychromatic (white) ambient light on mood in healthy adults. We hypothesized that higher melanopic equivalent daylight illuminances (EDI) would be associated with better mood. The search identified 2,994 publications, of which 14 met inclusion criteria. The resulting database included the spectral power distributions, illuminances, mood measures, and methods. We used the CIE S026 toolbox to calculate the five α-opic EDI values to characterize the light exposures in each study. Regression models tested the associations between predictors α-opic EDI, duration, and timing, and the outcome, mood. The results showed that none of the five α-opic values were significantly associated with mood (all estimates <0.044, p > .490). A longer duration of light exposure (all estimates <0.154, p < .018) and timing of light exposure in the morning (all estimates <0.233, p < .001) were associated with better mood, but these effects did not persist after adjusting for the data being nested within studies. Average mood scores remained reasonably consistent across different α-opic EDI profiles, suggesting that exposure profiles might not influence mood. Notably, none of the studies included had an experimental condition with a peak for melanopic EDI greater than the other α-opic EDI values, therefore the hypothesized association between melanopic EDI and mood could not be tested. Future research designs should contrast selected light parameters to identify the photoreceptors involved in the processes of interest.


Introduction
In addition to stimulating visual perception, light influences many physiological and psychological processes, including cognition, sleep, and mood (Cajochen 2007;Duffy and Czeisler 2009;Golden et al. 2005).Light intensity, spectrum, duration, and timing are all thought to play a role in these effects (Prayag et al. 2019).Over recent years, researchers have revealed the prominent involvement of intrinsically photosensitive retinal ganglion cells (ipRGCs) in regulating circadian rhythms, melatonin secretion, and other biological processes.
ipRGCs express melanopsin, a photopigment.They also receive synaptic input from other retinal photoreceptors (S-, M-, L-cones and rods) and contribute to diverse physiological and behavioral functions (Belenky et al. 2003;Brown et al. 2021;Dacey et al. 2005;Jusuf et al. 2007;Schmidt and Kofuji 2010;Wong et al. 2007).Direct pathways linking ipRGCs and various brain regions involved in emotional processing have also been found in both animal (Fernandez et al. 2018;Hattar et al. 2006;Huang et al. 2019;Legates et al. 2014) and human (Vandewalle et al. 2010;Weil et al. 2022) studies.While many studies focused on the melanopsin system, all photoreceptor systems interact closely and their relative contribution to the effects of light on mood in healthy humans remains poorly understood.Furthermore, because indoor lighting is typically polychromatic (i.e., white appearing and inclusive of many wavelengths), there is a need to move beyond monochromatic light comparisons, which are often used in studies that examine fundamental processes.

Specifying light exposures
To synthesize the evidence gathered to date on the effects of polychromatic light on mood, the light specifications from past studies need to be converted to a standard scale to decipher the relative contribution of the different photoreceptors.It has become apparent that reporting light in photopic illuminance (based on the photopic spectral luminous efficiency function [V λ ] peak sensitivity at 555 nm) is inadequate for predicting the effects of a given light source on ipRGCinfluenced functions because melanopsin has a peak sensitivity at 490 nm.In order to improve comparability and replicability of studies, Lucas et al. (2014) proposed that effective irradiance of a light source should be calculated for each photoreceptor type.The units initially proposed were, however, not consistent with the Système International (SI) of weights and measures (CIE 2015), which would have complicated future scientific progress.
In 2018, the CIE published an international standard for the metrology of light exposures, known as CIE S026:2018 (CIE 2018).The CIE system defines physical quantities based on action spectra for the five known photoreceptor types, the quantities being given the adjectives S-cone-opic, M-cone-opic, L-cone-opic, rhodopic (rod sensitivity), and melanopic (melanopsinbased light sensitivity of ipRGCs).Collectively, these are known as α-opic quantities.CIE S026: 2018 provides definitions that are SI-compliant, which ensures measurement traceability.Using spectral irradiance data for a given light exposure and these action spectra, one can calculate the five α-opic irradiances that characterize that light exposure.One can relate these α-opic irradiances to the equivalent exposure of the CIE standard daylight illuminant (D65), and express the value as an illuminance, known as the α-opic equivalent daylight illuminance (EDI), in lux.Converging past results on a common scale using the CIE S026:2018 metrology system should enable more accurate conclusions and recommendations to be drawn about the properties of polychromatic light that support better mood and other effects.

Objectives/ hypotheses
This systematic review integrates existing data on the effects of polychromatic light exposure on mood in healthy human adults based on the contribution of the different photoreceptors as estimated by αopic EDI.Considering previous findings (e.g., Legates et al. 2012;Vandewalle et al. 2010), higher melanopic EDI was predicted to be the α-opic measure most strongly associated with positive mood.

Data sources
This report is part of a broader systematic review addressing the optimal light characteristics for mood, cognition, and physiological responses (other themes will be addressed in subsequent reports).The publication search was conducted in the databases PubMed, Web of Science, Scopus, PsycINFO, and in the journals Lighting Research & Technology, LEUKOS -Journal of Illuminating Engineering Society, and conference proceedings on January 15 th , 2020.Grey literature (conferences and other unpublished reports) was searched in APA PsycExtra on September 30th, 2020.The predefined search strategy used for each database can be found in the supplemental material (S1).Only studies published in English or French between 1990 and 2020 were considered.The review protocol was registered in PROSPERO (CRD42020149818).Although we had initially planned to do a quality and risk of bias assessment inspired by the Cochrane "Risk of bias" assessment tool (Higgins et al. 2011) for randomized controlled trials and the Risk of Bias Criteria for Effective Practice and Organization of Care (EPOC) reviews, such approaches turned out to be not optimally adapted for the studies included in the current review (e.g., particular challenges regarding blinding due to the nature of the interventions).Hence, we opted to focus on summarizing core methodological information relevant for light research as was done in a previous Cochrane review (Pachito et al. 2018).Nevertheless, all studies included in the final dataset were reviewed by the authors and deemed to be of sufficient methodological quality to be included in the current study.

Study selection
The collective search for mood, cognition, and physiological responses returned 2,994 potentially relevant publications, from which 1,113 duplicates were removed.Only studies conducted with healthy human adults (≥18 years old) were included.Studies focused on people with mental or physical health conditions were excluded from this review, but studies on individuals with myopia, hyperopia, or astigmatism were included.To be included, studies were required to provide sufficient light characteristics based on vertical light measurements to quantify α-opic equivalent daylight illuminance (α-opic EDI) values using the CIE toolbox (see Data Processing section for details) or the light source needed to be reported in S 026 or CIE TN 003 quantities.Studies reporting only α-opic equivalent daylight luminance (EDL) were excluded because this metric is not directly comparable with EDI values.Studies were excluded if they were based on non-ocular light perception (eyes closed), were limited to light stimulations that varied across the same condition (e.g., dynamic lighting), involved light exposure during sleep, used only monochromatic light, or only provided outcome data following a night of total sleep deprivation.Studies on shift work (real or simulated) were excluded because the focus of this project was daytime exposures and in shift work studies many extraneous variables in addition to light exposure can affect the outcomes of interest.If the full article was not available or no new empirical data was provided (e.g., previously published data or a review), the publication was also excluded.
The screening of the dataset was done in multiple steps (see Fig. 1).First, the above criteria were applied for screening the titles and abstracts (1,881 publications for mood, cognition, and physiological responses) which were completed twice by two research assistants using the Covidence software (https://covidence.org).This resulted in 467 publications being retained.These publications were then subject to a full-text evaluation using the criteria previously mentioned and again done twice by two research assistants.The resulting dataset, containing 111 publications, was then revised by the Scientific Advisory Board to ensure no key publications were missed.
In the last step for the present paper, all studies with outcomes related to mood were selected for inclusion in the current report.For the purpose of this review, mood is considered to reflect affect, hedonic tone, and/or happiness-  sadness.The final dataset for this report includes 14 publications with mood outcomes and for which light exposures could be calculated.The authors of selected articles were contacted to collect more information if all criteria were met but the article was lacking certain details.

Data processing
Following data extraction, light information for the publications that did not report light characteristics as EDI values based on the CIE S026:2018 metrology system were converted (toolbox available at https://doi.org/10.25039/S026.2018.TB ;CIE 2018) and scaled to the photopic illuminance reported in the study.Several articles provided spectral data in 1 nm bands.For other publications that reported spectral power distribution (SPD) as a figure, these figures were digitized to extract numerical values for each 5 nm band (WebPlotDigitizer, Version 4.4, California, USA).Visual inspection and manual adjustments were implemented to reduce error in all digitized data extraction.Correa et al. (2016) reported the Lucas et al. (2014) "melanopic illuminances."We converted this value into the CIE S026 melanopic EDI using a conversion factor (Schlangen and Price 2021).No conversion factor exists for other α-opic EDIs; therefore, only melanopic irradiance was analyzed for this study.Viola et al. (2008) reported that the light exposures included a combination of daylight and electric light but there was sufficient information available to estimate melanopic EDI for the conditions.
Subjective mood outcomes included measures such as the Positive and Negative Affect Schedule (PANAS; Watson and Clark 1988) or other Likert scales, visual analogue scales, and checklists pertaining to mood.Means and standard deviations or standard errors were transcribed from each article; outcomes reported in figures were also digitized to extract central tendencies and variance indices.Only scales with higher scores reflecting better mood were included.Raw scores were converted to standardized scores using the following formula: When mood measures were collected at multiple time points within a single light condition, the time point showing the strongest mood effect, the largest sample size, or the longest elapsed time between light onset and the mood measure was selected (in that order).
Additional study details, such as light exposure duration and time of day of light exposure, were also aggregated into the database.These were considered as possible factors influencing light effects.

Statistical analyses
Standard meta-analytic approaches were not appropriate for this review because the lighting conditions used in the reviewed studies varied considerably and the main focus of the current review was the contribution of the five α-opic EDIs on mood outcomes rather than global effect sizes.
Multiple linear regression using the maximum likelihood estimates approach was applied to determine how standardized mood outcomes (dependent variable) were associated with each of the α-opic EDIs (continuous in lux), while adjusting for duration (continuous in minutes) and time of light exposure (categorical: morning (07:00 to 11:59), afternoon (12:00 to 16:59), night (17:00 to 06:59), or uncontrolled times).Distinct regressions were fit for each α-opic EDI because of the small sample size and multicollinearity (within any one light condition, the five α-opic EDI values are inherently correlated).The α-opic EDIs and duration of exposure were log 10 -transformed to improve normality.Additional exploratory analyses with interaction terms between α-opic EDIs, time and duration of exposure were also conducted (see supplemental material S3).Interaction terms did not yield any significant effects, and therefore, this report focuses on the simpler regression without interaction terms.
A second set of analyses was conducted to account for the fact that these data stem from light exposure conditions overlapping across multiple studies integrated in a meta-analysis.Linear mixed-effects regressions were conducted using the same design as the multiple regression described above with the addition of the study identifier as a random-effect parameter.
All regressions were performed using the "lm" function or the "lme4" package (Bates 2020) of the R software, version 4.1 (R Core Team 2021).The Akaike Information Criterion (AIC) was used to compare the multiple linear regression to the linear mixed-effects regression, with lower AIC values indicating a better balance between goodness-of-fit and parsimony of the model (Akaike 1987).
Similar to approaches used in previous reviews on light (Souman et al. 2018b), data were separated by light characteristics and visually presented to highlight patterns in light profiles.

Study characteristics
Table 1 summarizes the characteristics of the included studies.Overall, these studies included a total of approximately 475 participants (339 female; 57% female) exposed to a total of 52 light conditions following a crossover (n = 31 light conditions), cross-sectional (n = 12 light conditions), or mixed (n = 9 light conditions) design.None of these studies used pharmacological pupil dilation.Two publications were based on field studies, seven on simulated office environments, and five on laboratory studies.Eight studies had a controlled lighting period to adapt the eyes prior to the experimental light conditions.Other methodological characteristics of the included studies are reported in Table 1.

Relative contribution of α-opic EDIs
Figure 2 shows plots of standardized mood metrics as a function of the five α-opic EDIs.A sub-division of these plots by duration (<30 minutes, 30-60 minutes, and >60 minutes; minimum = 15 minutes, maximum = 495 minutes) can be found in the supplemental material (S2).
Based on multiple linear regression, higher mood ratings were significantly associated with longer light exposure (all estimates <0.154, p < .018)and with morning light exposure compared to uncontrolled timing of light exposure (all estimates <0.233, p < .001)for all α-opic EDIs (see Table 2).None of the α-opic EDIs were significantly associated with mood outcomes (all estimates <0.044, p > .490).
Table 3 reports statistics from the linear mixed-effect regression in which the study identifier was added as a random effect.Significant random effects of study identifiers (all F > 12.6, p < .001)were found for each α-opic EDI regression.The time (all estimates <|0.153|, p > .128)and duration of exposure (all estimates <|0.020|, p > .484)effects did not survive this adjustment for the study identifiers.None of the α-opic EDIs were found to be significantly associated with mood outcomes (all estimates <0.004, p > .849).
The AIC was significantly lower in the linear mixed-effect models (all AIC < −95.2) compared to the multiple linear regression models (all AIC > −33.4;Chi-squared > 67.3, p < .0001),suggesting that the linear mixed-effect regression models had a better balance between goodnessof-fit and parsimony than the multiple linear regression models.

Mood outcomes across distinct α-opic EDI profiles
Figure 3 depicts α-opic EDI values for each light condition in each study included in this review separated by exposure duration (<30 minutes, 30-60 minutes, and >60 minutes).Overall, αopic EDIs ranged from 0 to 4,905 lux and different EDI profiles emerged where certain photoreceptors had higher EDI values depending on the light source spectrum.Specifically, in two of the light conditions the S-, M-, and L-coneopic EDIs were higher than the rhodopic and melanopic EDIs (Knaier et al. 2018;Leichtfried et al. 2015).Five light conditions had higher  4 reports standardized mood outcomes averaged across these different EDI profiles.Although the small number of light conditions included in each EDI profile precluded statistical analyses by EDI profile, mood outcomes were observed to be highly comparable across the different EDI profiles.A figure depicting the EDI profiles of light conditions with a peak in the short-wavelength section of the spectrum can be found in the supplemental material (S4).As can be seen, these studies did not have higher melanopic EDI values as compared to the other photoreceptors.A figure of all SPDs with a peak in the short-wavelength range and their associated EDI profile can also be found in the supplemental material (S5).

Synthesis of individual study findings
Nine studies out of the 14 included in this review did not report any significant effect of light intensity or spectral composition on mood (Borragán et al. 2017;Correa et al. 2016;Huiberts et al. 2015Huiberts et al. , 2017;;Iskra-Golec et al. 2012;Knaier et al. 2018;Smolders et al. 2016;Veitch 1997).Of the five studies reporting a significant effect, all observed that better mood was associated with higher illuminance light conditions (Chellappa et al. 2011;Leichtfried et al. 2015;Mason et al. 2022;Ru et al. 2019;Smolders and de Kort 2014;Viola et al. 2008) and/or higher correlated color temperature (CCT; Chellappa et al. 2011;Leichtfried et al. 2015;Viola et al. 2008), except for one study reporting an association with lower CCT (Ru et al. 2019).Most of the studies with short light exposures had the highest α-opic EDI values (see Fig. 3).

Discussion
Results from this systematic review of polychromatic light exposures could not reveal any significant effects of α-opic EDI values on mood outcomes within the  profiles of α-opic EDI tested to date.Accordingly, the inspection of individual study results revealed that a large proportion of the studies included in this review did not report any significant effect of light intensity or spectrum on mood outcomes.Overall, these observations could not support the proposed hypothesis that the effects of polychromatic light on mood would rely more heavily on melanopic photoreceptors.Better mood was associated with morning light exposure and longer durations of exposure, as would be expected based on previous studies (e.g., Smolders et al. 2012Smolders et al. , 2013)).This suggests a potential dose-response curve more dependent on the timing and duration of polychromatic light exposure than on α-opic differences in EDI values for the ranges covered in the current review.However, these effects of timing and duration were no longer significant in the more complex models adjusting for the data being    nested within studies.The conclusions that can be drawn from the current review regarding the lack of association observed between mood and melanopic EDI are limited, notably because of the restricted range of EDI profiles in the studies identified for this review.Aside from phenomena tied to retinal responses (including ipRGC stimulation), other mechanisms, including visual processes, may contribute to the mood enhancing effects of polychromatic light.This highlights the need for further studies to assess more complex interactions not only between EDI, duration and timing of light exposure, but also other factors such as brain responses and spatial distribution of light.

Associations between mood and α-opic EDIs
Current findings depart from those of previous controlled studies focused on narrowband (close to monochromatic) light on mood (e.g., Vandewalle et al. 2010).Our systematic review could not confirm that variations in light intensity and spectrum within polychromatic white light have a short-term effect on mood in healthy individuals.Some of the previous research has shown that short-wavelength light, daytime light with higher melanopic EDI, and general lighting with a higher CCT can improve mood and well-being (Borisuit et al. 2015;Mills et al. 2007;Münch et al. 2020;Viola et al. 2008;Xiao et al. 2021).In the current review, after rescaling light on comparable units to estimate its effects on individual photoreceptors, there was no clear evidence that the intensity of light exposure predicted mood and no evidence that stimulation of any one photoreceptor was more influential on mood than others.One reason for this might be the small differences between the α-opic EDIs for any one light condition within studies (seen in the overall relatively flat profiles, presented in groups by exposure duration, in Fig. 3).
Through visual inspection of light data from individual studies, we sought to identify profiles of peaks across α-opic EDIs.This aimed to determine whether the different light sources in this combined dataset could differentiate exposures to the different photoreceptors (particularly the ipRGCs) and to provide insight into which photoreceptors could be linked to mood outcomes.None of the polychromatic light conditions from the included studies had a higher peak in melanopic EDI compared to other α-opic EDIs.This prevented any insights on α-opic EDI profiles, which may activate ipRGCs more strongly than other photoreceptors.Some studies did use conditions in which polychromatic light resulted in higher S-, M-, and/or L-cone-opic EDIs, but the mood outcomes were rather similar across these light conditions.There might have been a slightly higher mood rating on average in conditions with a higher peak in S-cone-opic EDI (see supplemental material, S2), but this remains to be confirmed through larger datasets enabling statistical comparisons.
Overall, the current observations are limited by the α-opic EDI distribution of the specific polychromatic lights used in the reviewed studies.Despite using light sources with an SPD peak in the short-wavelength range ("blue-enriched light"), many studies used light where these peaks were combined with peaks in longer-wavelength light (red light).This creates a light which may be deemed to be more visually natural, which was likely an important goal for these studies, but would not necessarily lead to a major peak in melanopic EDI relative to other α-opic EDIs.For instance, beyond the increase in ipRGC activations, this type of light could also lead to increased L-cone activation.It was expected that the EDI profiles of study conditions with an SPD peaking in the short-wavelengths would have higher melanopic EDI.This was, however, not observed.Importantly, the light stimulus in many of these studies was not designed to preferentially activate the ipRGCs.

Mood, duration, timing of light exposure, and melanopic EDI
The primary goal of including duration and time of exposure in these analyses was to control for these factors because they are known to influence the effects of light on mood (Bedrosian and Nelson 2017;Chang et al. 2012;Chellappa et al. 2011;Khalsa et al. 2003;Rüger et al. 2013;St Hilaire et al. 2012;Vandewalle et al. 2013).Specifically, it has been suggested that improvements in mood are likely to manifest over longer durations of bright light exposure or when repeated over several days (Partonen and Lönnqvist 2000), although, to our knowledge, no dose-response curve has been created yet.A large longitudinal study also reported that individuals with more daytime spent outdoors have a reduced risk of lifetime depression (Burns et al. 2021).Furthermore, greater improvements in mood have been reported following light exposure in the morning compared to other times of the day (Bedrosian andNelson 2013, 2017;Leichtfried et al. 2015).It is also understood that high melanopic EDI during the day and low melanopic EDI light exposure during the night are supportive for alertness, circadian rhythms, and sleep (CIE 2019), which can indirectly influence mood (Baron and Reid 2014;DeWeerdt 2022;Finan et al. 2015;Kahn-Greene et al. 2007).This being said, the evidence on an association between light, sleep/ circadian rhythms, and mood in healthy individuals remains conflicted (Böhmer et al. 2021).In the current review, the effects of duration and time of exposure did not survive in more complex models accounting for data being nested within studies.Considering that in the articles covered in the current review, the variations in light exposure durations and time of exposures were limited (n = 16 morning, n = 12 afternoon, n = 11 night, uncontrolled time = 12), no strong conclusion should be drawn about these parameters.However, it is unlikely that variations in time of exposure were significant confounders affecting the main α-opic results at the center of the current review's objectives since we observed no significant interaction between time of exposure and αopic EDIs (see supplemental material S3).

Study population
Mood is a complex phenomenon that can be influenced by many contextual factors, which are bound to vary from one study to the other, such as other properties of the physical environment, social interactions, and a vast range of situational and personal factors.These factors could not be measured or quantified in this review and are very likely to have created significant noise in the data.
One may also postulate that α-opic EDIs may be a stronger predictor of positive mood in the context of mood disorders than for general mood in mentally healthy individuals.Firstly, in the context of mood disorders, poorer mood creates a greater potential for mood improvement whereas, in mentally healthy people, general mood might plateau, leaving less room for potential improvements.Accordingly, one study (Smolders et al. 2012) included in the current review highlighted that their participants felt on average happy and without tension or sadness prior to light exposure, which left little room for improvement.Secondly, tools to measure general mood (often based on a single question) might not have enough sensitivity as compared to scales used to measure combinations of specific symptoms linked to mood disorders.Thirdly, individuals with mood disorders could respond differentially to light than healthy individuals.
Indeed, there are indications that light may have a different effect in individuals with mood disorders compared to controls (Laurenzo et al. 2016).This notion is supported by previous observations of variations in the human peak sensitivity of melanopsin to short-wavelength light (Bailes and Lucas 2013), the contribution of ipRGCs to mood regulation processes in rodents (Legates et al. 2012(Legates et al. , 2014)), and the abnormal sensitivity of the circadian clock to light in the context of mood disorders (Bullock et al. 2019;McGlashan et al. 2019;Roecklein et al. 2013).Furthermore, studies have shown that light (monochromatic and polychromatic) can elevate mood in the context of mood disorders (Geoffroy et al. 2019;Glickman et al. 2006;Meesters et al. 2018Meesters et al. , 2016;;Nixon et al. 2021) and fMRI studies also suggest that it can increase brain responses to emotional stimuli (Vandewalle et al. 2010).

Methodological refinement
Studies conducted prior to the publication of action spectra on melatonin suppression (Brainard et al. 2001;Thapan et al. 2001) and the identification of ipRGCs (Berson et al. 2002) necessarily could not be designed to target these photoreceptors, nor could they use reporting standards for light that had not yet been established.The literature landscape spreading over these different periods thus necessarily holds methodological inconsistencies and gaps.Similarly, field studies typically face significant methodological challenges, both in terms of controlled light manipulations and measurements, yet they provide critical input that will enable practical applications.With the benefit of more knowledge, guidance documents have been developed to ensure that future investigations can be more readily compared (CIE 2020;Spitschan et al. 2019).The authors are recommended to report α-opic EDIs (or at least illuminance and light source SPD) measured at the eye in the direction of gaze in addition to horizontal light measurements.
The luminance distribution in the physical space also affects what the retina receives and by extension the ipRGC-influenced response to light (CIE 2020;Glickman et al. 2003;Hannibal et al. 2017;Lasko et al. 1999;Rüger et al. 2005;Visser et al. 1999).Differences in luminance distribution between studies in this review could have added to differences in light exposures and could have obscured differences in mood.Where possible, spatially-resolved spectral measurements could be used to more precisely characterize light exposures (Knoop et al. 2019).
In addition, spatial distribution influences mood through its effects on visual appraisals; for instance, in office settings a direct-indirect light distribution contributes to a more pleasant mood (Hsieh 2015;Veitch et al. 2013).Therefore, researchers also should describe, in detail, the physical space used in the study.Relevant parameters include surface reflectance, temporal light modulation characteristics, color fidelity, and, where applicable, intensity/spectrum measurements from different possible viewing perspectives; see (CIE 2020).
Designing light source spectra to target specific photoreceptors is more complex than increasing the exposure to specific wavelengths (e.g., adding a spectral peak around 490 nm to target ipRGCs) because the results could differ in color appearance or color rendering or could also stimulate other photoreceptors more than intended.Therefore, studies using metameric spectral tuning may be required to fully differentiate the relative influence of distinct α-opic profiles of polychromatic light on mood.Metameric lighting, which has matching light source color appearance with different SPDs, is a quickly growing field, with studies already exploring its use on alertness and melatonin production (Allen et al. 2018;Souman et al. 2018a).Using this method, a recent study employed a multi-channel LED system to find chromaticity areas in the CIE xy color space that are relevant for melanopic efficacy (Zandi et al. 2021).Keeping in mind that increased color fidelity can induce better mood (Cajochen et al. 2019), more studies investigating mood in combination with this lighting approach optimized for ipRGC activation are needed.
Considering that light exposure can change the sensitivity of ocular photoreception to subsequent exposures (Chang et al. 2011;Hebert et al. 2002;Jasser et al. 2006;Kozaki et al. 2015;te Kulve et al. 2019;Zeitzer et al. 2011), controlling for and reporting light history would be important.However, there is currently no consensus on how this should be done.As such there is considerable variability in the methods applied to control for previous light history, for example with different lengths of adaptation in the dark or under various levels of dim light, sometimes preceded by saturating white flashes.Consensus on standard protocols for controlling light history would help enhance comparability across studies.However, this may not be as feasible in the context of naturalistic studies.

Limitations
The vast methodological inconsistencies across studies hinder firm conclusions.The variety of mood measures used across the studies also limit their comparability by reducing construct validity.Mood, a difficult construct to measure that may be conceptualized in different ways (Larsen and Diener 1992), was integrated as a secondary measure in several of the studies in this review.Mood can be influenced by many other factors, some of which are also sensitive to the effects of light, such as alertness and performance.Further work is required to disentangle the potential interactions between the effects of light on alertness, performance, and mood.
This review was limited by the light conditions used in published studies, which were not designed from the outset to cover distinct α-opic profiles and were confounded by other factors.For instance, most studies which had higher α-opic EDIs were also the ones where the duration of light exposure was the shortest.Considering that in mice, there are direct effects of light on mood driven by a direct projection of ipRGCs to the perihabenular (PHb) and lateral habenula (LHb; Aranda and Schmidt 2021;Fernandez et al. 2018;Legates et al. 2012), the involvement of ipRGCs in mood responses to light could possibly emerge more clearly from studies comparing well-controlled polychromatic light conditions that clearly differ in α-opic EDI with repeated measurements over a moderate to long exposure duration.
In contrast to other reviews, which have included monochromatic light exposures (Brown 2020;Siraji et al. 2022), this review focused on the effects of polychromatic light, which is more common in naturalistic living environments.This perhaps reduced the α-opic EDI differences between experimental conditions relative to the contrasts typically attained when comparing monochromatic light conditions.However, limited α-opic EDI differences increased the generalizability of the findings.
The lack of reported or controlled preexposure history also created an important limitation in the interpretation of results.Without consistent information concerning prior light exposure, it was not possible to control for differences between studies on this variable.Most studies had a multiday exposure design which may have additional uncontrolled factors that could account for the resulting mood outcomes.Visual comfort and glare have also been found to affect mood (Borisuit et al. 2015;Cajochen et al. 2019).Many of the included studies neither controlled for nor studied these variables; therefore, mood effects might have been obscured.
Considering representativeness, although the samples were overall balanced for sex, many of the studies included participants from a limited age range, often young adults.There was also limited cultural variability, as most were performed in Europe or North America.

Conclusions
Overall, this systematic review could not support the hypothesis that higher polychromatic light exposure intensity detected by any one of the five ocular photoreceptors described in CIE S026:2018 is associated with better mood, within the narrow range of EDI profiles that were included in the current review.The review could not separate effects of spectrum and intensity from the strong effects of light exposure duration and betweenstudy differences in procedures and context.As such, more polychromatic light studies are warranted to systematically test specific combinations of α-opic EDI values optimally targeting each photoreceptor while controlling for exposure duration.Consistent use of the CIE metrology system to characterize light exposures will enable more accurate conclusions about the potential involvement of different photoreceptor systems in the effects of light in humans.Such information could enable revised guidance for interior lighting design and for daily light exposure patterns supportive of well-being, health, and performance.

Fig. 1 .
Fig. 1.PRISMA flow diagram.Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA; Page et al. 2021).This review is part of a larger project assessing the effects of light not only on mood, but also on cognition and physiology.The top components of this PRISMA Flow Diagram represent the larger scope of the review.The isolation of the mood-specific studies from those focused solely on cognition or physiology was done in the last step (depicted in the lowest box on the right of the diagram -"Not related to mood").

Fig. 3 .
Fig. 3. Study conditions can be characterized by their patterns (profile) of α-opic EDI values.Each panel shows studies of a particular exposure duration: (a) under 30 min; (b) 30-60 min; (c) >60 min.Lines of the same color are different conditions from one study, the conditions being differentiated by type of line (e.g., dotted, dashes).

Table 1 .
(Veitch 1997;Viola et al. 2008) light condition across all studies.Time of exposure categorized into AM (6:00-11:59), PM (afternoon; 12:00-16:59), night (17:00-6:59), standard deviation (SD), light-emitting diodes (LED), Positive and Negative Affect Schedule (PANAS), visual analogue scale (VAS), University of Wales Institute of Science and Technology (UWIST), equivalent daylight illuminance (EDI).*Studydesignrelevant to light conditions included in this review, **Mixed design in which all participants were exposed to all light conditions in a crossover manner, but where most participants were allocated to either a morning or an afternoon session cross-sectionally.***Mixeddesign in which participants were exposed to different illuminance levels in a crossover manner but were allocated to either a high or low correlated color temperature.****Notamultiday light exposure study design.†Counterbalancedorder of light exposure.‡Randomizedgroups.†‡Randomlyassigned to counterbalanced order.All light measurements taken at eye level or converted to vertical(Veitch 1997;Viola et al. 2008).F: females.§Custom-made mood measure.Spectral data was provided in 1 nm bands.
(Borragán et al. 2017;Iskra-Golec et al.unction of each α-opic EDI.Results did not differ when the two studies with low EDI values were removed, therefore the two studies were kept in the analyses.S-cone-opic EDIs relative to all other EDIs(Borragán et al. 2017;Iskra-Golec et al.

Table 2 .
Multiple linear regression models assessing the relationship between α-opic EDIs and mood.Estimates (Est.; calculated for fixed effects per one unit of log 10 -transformed values for continuous variables, and relative to morning for time of exposure (categorical: morning (6:00 to 11:59), afternoon (noon to 16:59), night (17:00 to 6:59), or uncontrolled times), standard error (SE), and p values.

Table 3 .
Linear mixed-effect models assessing the relationship between α-opic EDIs and mood.F values for Study Identifier and Estimates for all other parameters (Est.; calculated for fixed effects per one unit of log 10 -transformed values for continuous variables, and relative to morning for time of exposure (categorical: morning (6:00 to 11:59), afternoon (noon to 16:59), night (17:00 to 6:59), or uncontrolled times), standard deviation (SD) for Study Identifier and standard error (SE) for all other parameters, and p values.

Table 4 .
Correa et al. (2016)the different relative α-opic profiles.Correa et al. (2016)was not included in profile analyses because only melanopic EDI could be retrieved.