Exposure to Light at Night and Risk of Cancer: A Systematic Review, Meta-Analysis, and Data Synthesis

In 2019, cancer as a whole was the second leading cause of death globally, with 10.0 million deaths [1]. Although the most frequently diagnosed cancer type and the leading cause of cancer death vary widely between countries and regions of the world, the three cancers with the highest burden according to disability-adjusted life years for the period 2010 to 2019 were tracheal/bronchus/lung cancer, colorectal cancer, and stomach cancer [1,2]. Breast cancer was the cancer with the largest burden among women globally and fourth overall. The causes of cancer are multifactorial and include genetic, biological, and environmental factors depending on the location of the cancer. While the most common environmental risk factors of cancer include tobacco smoking, heavy alcohol consumption, obesity, and low fruit and vegetable intake [3,4,5], other modifiable and occupational factors have been implicated as carcinogenic [6,7,8,9,10].

Among the environmental risk factors for cancer, exposure to light at night (LAN) and sleep interruption have been of increasing interest. A dramatic increase in urbanization globally in the last several decades, with 80% of the world’s population now living in areas with exposure to LAN, points to an ever-growing need to understand the potential health implications that LAN and disrupted sleep rhythms may have on human health [11]. The International Agency for Research on Cancer in 2010 deemed shift work and LAN as “probably carcinogenic to humans” because of a possible link to multiple tumor types [12]. The exact mechanisms by which LAN exposure may increase an individual’s risk for cancer are unclear. Potential mechanistic factors proposed to date include circadian disruption, oxidative stress, and hormonal disruptions [13,14]. A link between LAN exposure and increased risk for lung cancers, leukemias, and hepatocarcinomas in mice models has already been established, but few observational studies in humans have been performed [15,16]. LAN is typically measured in human studies using satellite images or self-reported measures of indoor/outdoor light exposure. The majority of prior efforts to consolidate results from observational studies of LAN and cancer risk have focused on breast cancer, including three meta-analyses, with no prior systematic review and meta-analysis focused on other tumor types [17,18,19]. In the most recent meta-analysis of LAN and breast cancer risk by Urbano et al., including 17 studies published through September 2021, the highest level of LAN exposure had a positive association with breast cancer risk compared to the lowest level of LAN exposure (risk Ratio [RR], 1.11; 95% confidence interval [CI] 1.07–1.15) regardless of study design and method of LAN assessment [17].

Furthering our knowledge of the relationships between LAN exposure and cancer risk of multiple types is important in terms of public health policy, including cancer prevention and education. The aim of this study was to undertake a systematic review and, if possible, meta-analysis of observational studies of the association between LAN exposure and cancer risk.

This study was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [20] and is consistent with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines for the meta-analysis of observational studies. This study has not been registered [21].

The search strategy identified 8918 studies. Following the removal of 418 duplicates, 8500 studies underwent title and abstract review. Of these, 31 were identified for full-text evaluation. Of the 31 articles chosen for full review, we excluded 5 for having abstract only (n = 1), no usable primary data (n = 1), the wrong study design (n = 2), and the wrong study population (n = 1). Therefore, 26 studies were included in the final review (Figure 1).

The studies were published from 2001 to 2023. Geographically, the studies assessed populations in North America (n = 16) [25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40], Asia (n = 4) [41,42,43,44], Europe (n = 4) [45,46,47,48], and Australia (n = 2) [49,50]. The vast majority of these papers included breast cancer as an outcome (n = 19), while colorectal cancer (n = 2), prostate cancer (n = 1), thyroid cancer (n = 1), pancreatic cancer (n = 1), non-Hodgkin’s lymphoma (n = 1), liver cancer (n = 1), and endometrial cancer (n = 1) were represented in a total of eight studies. One study, Garcia-Saenz 2018, included data separately for both risks of breast and prostate cancers. Table 1 describes all studies and their associated study designs. Thirteen were case–control, and thirteen were cohort studies. Supplementary Table S3 contains additional details. Ten studies reported associations with indoor LAN, twelve with outdoor LAN, and four reported associations separately for both indoor and outdoor LAN.

Risk of bias assessment (Supplementary Table S4) found eight studies (29.6%) to be at moderate risk of bias due to confounding because of a failure to control for other cancer risk factors or other environmental disruptors of circadian rhythm [27,42,43,44,46,47,48,49]. Examples of failure to control for other cancer risk factors include the failure to record estrogen or progesterone receptor positivity, family history of cancer, prior radiation, family history of breast cancer, postmenopausal hormone use, or smoking history. The risk of bias due to selection was judged to be moderate for eight studies due to small sample sizes, poorly generalizable populations, or unmatched cases [25,30,33,34,40,41,44,48]. Three studies (11.1%) were judged to be at critical risk of bias due to selection because of non-randomly selected controls or a collection of multiple moderate risk factors [42,43,49]. Regarding exposure measurement, 11 studies (40.7%) were judged to be at a moderate risk of bias. These studies were found to have utilized indirect measures of LAN, non-intuitive categories of light exposure, or solely self-report surveys [25,27,29,34,35,39,40,43,46,48,50]. One study (3.7%) was found to be at critical risk of measurement of exposure bias due to using a non-validated and non-expansive self-administered questionnaire [42]. All other domains were considered at low risk of bias in all studies. Regarding the study-level risk of bias assessment, three studies (11.1%) were determined to be at moderate risk of bias [43,48,49]. Only one study (3.7%) was deemed to be at critical risk of bias overall [42].

We conducted a systematic review and, for breast cancer risk, a meta-analysis of the published literature on the association between LAN exposure and the risk of cancer development. We identified 26 studies (13 case–control and 13 cohort) that directly addressed this issue. In the meta-analysis of 19 breast cancer studies, there was a modest increased risk associated with high LAN exposure. The association was similar whether LAN exposure was from indoor or outdoor sources. Except for breast cancer, there was a paucity of site-specific cancer studies.

There is increasing focus on the potential for environmental exposures to influence a person’s risk of developing cancer. Prior meta-analyses, for example, of air pollution [51,52] and smoking [53,54], have demonstrated a link with increased risk of cancer in these exposure groups. LAN exposure has been under increasing scrutiny as a potentially harmful environmental health exposure, with several meta-analyses noting its association with obesity [55], diabetes [56], and breast cancer [17,18,19]. In our systematic review, we focused on the important intersection between the environmental health exposure of LAN and cancer risk. In addition to looking at multiple cancer sites, our findings build on the most recent prior meta-analysis of breast cancer risk by incorporating two new studies. Since the most recent meta-analysis of LAN exposure and breast cancer risk was published in late 2021 by Urbano et al. [17], two new studies have been published regarding breast cancer risk and LAN exposure that fit the search criteria of our analysis [31,42]. The addition of these two studies further supports the conclusions of the previous meta-analyses that show a moderate positive association between LAN exposure and breast cancer risk in women. When stratified by indoor versus outdoor LAN exposure, our analysis further supports the findings of Urbano. That is, regardless of sources, LAN is associated with breast cancer risk. The low statistical heterogeneity between studies in our analysis emphasizes this positive relationship by indicating a low likelihood that the consistency of evidence found is due to chance. Nonetheless, there remains a relative paucity of studies assessing the important relationship between LAN exposure and breast cancer risk. The continued positive association between the two calls for further analysis of different study populations of interest as well as an investigation into the mechanism of this association.

The dearth of studies examining the association between LAN exposure and cancer risk extends beyond breast cancer to other cancer types. Only eight studies meeting our search criteria examined associations of LAN with non-breast cancers, with only colorectal cancer having been examined in more than one study. The wide range of results, such as studies of prostate and thyroid cancers showing a strong positive relationship with LAN exposure [38,46] and liver and endometrial cancers showing no appreciable association [36,37], demonstrates an interesting potential for LAN to also be positively associated with other (non-breast) cancer types. The extreme shortage of studies, however, prevents any definitive conclusions from being made. While some studies speculate the impairment of melatonin and estrogen release due to LAN exposure playing an important role in breast cancer [57], impaired sleep quality can also impact oxidative stress and inflammatory levels, cause metabolic and circadian disruptions, or affect other hormonal secretions [14,57,58,59]. Inflammatory states and metabolic and circadian disruptions are known carcinogens [60,61], so it follows that LAN may affect cancer types beyond breast cancer. Animal studies also further support the relationship between LAN exposure and other cancer types [61,62]. Therefore, more investigation is needed to assess if there is a similar increase in cancers such as liver, colorectal, or prostate cancer when exposed to increased LAN levels.

It is important to note the inherent differences in mechanisms of exposure to indoor and outdoor light. Of the 16 studies using outdoor light as a method of LAN exposure, all used satellite data as the source of outdoor LAN exposure; 13 of 16 studies utilized the U.S. Defense Meteorological Satellite Program’s Global Radiance Calibrated Night-time Lights high-dynamic range data, which had the benefit of assigning radiance units (nW/cm²/sr) to geocoded addresses. Two of the 16 studies, both in Spain, utilized ISS images to predict luminance and blue light exposure for each pixel, which was then geocoded. The final study assessing outdoor LAN utilized the New World Atlas of Artificial Night Sky Brightness to measure luminance on a grid to assign to resident addresses [11]. The use of satellite-gathered information has many advantages, including the ability to standardize luminance levels across large swaths of land, geocode the luminance to specific residence addresses, and easily analyze data. However, satellite data cannot fully reflect the true amount of LAN exposure an individual at a certain address may receive. Tree cover, building height, the ability of outdoor light to actually penetrate an individual’s sleeping environment, and local lighting levels undetectable by satellite imagery can greatly affect LAN exposure. Urban saturation of LAN also may distort the ability of satellite imagery to accurately measure LAN based on the limited resolution of images.

Indoor measurements of LAN are also an imperfect assessment of complete LAN exposure. Indoor LAN can reflect the true exposure of an individual to LAN exposure to a greater degree than outdoor LAN, as the vast majority of individuals are indoors at night. However, accurately measuring indoor LAN can be difficult. The 14 studies in our analysis of both breast and non-breast cancers all administered a questionnaire to assess indoor LAN exposure. Some of the questionnaires used an iteration of a three- or four-point scale to quantify the level of indoor LAN by asking the participant questions, such as if they could see their hand in front of their face, if they could see across the room, or if they could read a book comfortably. Other yes and no questions were often included in the questionnaire, such as if the participant used a nightlight or had a TV or other strong source of light turned on during the night. The reliability of using subjective recall of light at night exposure, especially when participants were asked to remember their exposure for the preceding decade, is questionable, and it is unclear if this was able to be standardized across studies. Validation studies comparing with the objective measurements above are needed before definitive conclusions can be drawn from observational studies of LAN exposure.

This review of the effects of LAN on cancer risk has several strengths and limitations. Our study validates and supports previous studies [17,18,19] exploring the overall relationship between LAN exposure and breast cancer risk while also being the first study to synthesize the evidence around the association of LAN exposure with the risk of other cancer types. We provide a solid framework to continue investigations into the effects of LAN on cancer types other than breast cancer. As discussed before, the inherent limitations regarding the measurement of LAN exposure present fluctuations in an unclear direction depending on whether the exposure was indoor or outdoor. Indoor measurements of LAN were more reflective of true LAN exposure but were subject to recall and measurement bias, while outdoor measures of LAN are more accurately measurable but are subject to wide variability in their ability to reach participants at night. Next, while the association with breast cancer risk was found with both case–control and prospective cohort studies, better evidence for causality comes from prospective cohort studies [63]. At the moment, the Bradford Hill criteria for causality [64] between LAN exposure and breast cancer that are well discussed in the literature are consistency, specificity, and the biological gradient. However, biological plausibility and coherence could be strengthened through biochemical research on sleep, hormonal, and cancer pathways. Temporality can be further investigated through future studies ascertaining LAN exposure prior to cancer development as well as exposure to LAN across the lifetime. Our meta-analysis was also limited by the fact that the combined effect of LAN was calculated by incorporating different maximum exposure definitions, some categorical and some with a numerical percentile cutoff. There are also potential confounders that were unaccounted for among some or all of the studies, such as additional environmental exposures like noise or occupational status. However, most studies controlled for the major risk factors for breast cancer like parity, obesity, age of menopause, and smoking status. For non-breast cancer studies, the scarcity of analyses prevents any strong conclusions from being made regarding the impact of LAN on other cancer risks and requires additional effort in the future to elucidate the nuances of this relationship. This also limits the generalizability of our observations regarding non-breast cancer studies. More studies investigating the association between LAN exposure and cancer risk need to be performed for other cancer types, specifically for cancers such as endometrial, ovarian, and prostate that are also highly hormonally regulated, like breast cancer [65]. In line with this, biochemical research that further elucidates mechanisms of action between sleep, hormonal axis regulation, and oncogenesis can work in tandem with population-based studies such as ours to further paint a clearer picture of the role sleep has on cancer incidence.

Increasing urbanization and the ubiquity of LAN across more and more of the global population present a growing public health concern regarding cancer incidence. Our study found a positive association between both indoor and outdoor LAN exposures and breast cancer risk. We also highlight the lack of information regarding the association between LAN exposure and the risks of non-breast cancers. We suggest more investigation be conducted regarding the specific mechanisms between LAN exposure and cancer risk as well as analysis on specific groups that may be at risk for other cancer types. Further elucidation of these relationships may inform public health and governmental policy regarding LAN to reduce the global burden of cancer.