Introduction

Cognitive deficits in cancer, often referred to as “cancer-related cognitive impairments” (CRCI), are commonly reported by breast cancer (BC) patients. CRCI are mainly subtle in nature and concern difficulties in memory, attention, executive functions, and processing speed1,2,3. Despite their subtle nature, these impairments can persist4 and negatively impact the quality of life of BC patients, hindering their ability to return to work and resume daily activities5,6.

Cognitive complaints are frequently reported by BC patients following chemotherapy7,8, and some studies also suggest an exacerbation of cognitive complaints among BC women undergoing endocrine therapy (ET)9,10. Besides cancer treatments, other factors such as age, psycho-affective and medical comorbidities, and sleep disorders can also contribute to the occurrence and persistence of cognitive decline in BC11,12,13,14,15,16. Notably, self-reported sleep difficulties were found to be strongly associated with CRCI12,17, and many BC patients express sleep difficulties13,18,19.

One of the most commonly reported symptoms of sleep disturbances in BC patients is sleep-disordered breathing (SDB)20. Specifically, SDB is a syndrome characterized by upper airway dysfunction during sleep, marked by snoring, heightened respiratory effort caused by increased resistance in the upper airway, and the tendency of the pharynx to collapse21,22. The primary consequences of SDB are nocturnal hypoxia and sleep fragmentation23. Interest in exploring the prevalence of SDB in BC patients has grown recently, marked by the initiation of a first study in 2021, which used polysomnography (PSG) in BC patients treated with chemotherapy, radiotherapy (RT) and/or ET24. This study found an 88% prevalence of SDB in BC patients. It is important to note, however, that the study did not compare its results to a control group. Individuals without cancer experiencing SDB often demonstrate cognitive difficulties25,26,27, particularly affecting memory, executive functions, and attention28. Additionally, numerous studies have confirmed an association between SDB and cognitive decline in older adults25,29,30,31,32.

Hypoxia refers to a condition characterized by a reduced intake of oxygen or an impaired ability of oxygen use33. Since the brain depends on an important oxygen supply to assure neuronal connectivity and synaptic activity34, oxygen deprivation can lead to cognitive dysfunction35. A recent literature review concluded that hypoxia is the primary contributor to cognitive dysfunction in elderly patients29. Similarly, a study by Yaffe and colleagues31 found that nocturnal hypoxia was more commonly associated with cognitive decline than other sleep-related disturbances. While sleep fragmentation and nocturnal hypoxia often co-occur in SDB, sleep fragmentation primarily impacts cognition through secondary mechanisms such as increased daytime sleepiness and reduced vigilance36. In the 2011 study by Yaffe and colleagues31, no association was found between sleep fragmentation and cognitive impairment, further supporting the dominant role of nocturnal hypoxia in cognitive decline.

Despite the presence of cognitive impairment in BC, and despite evidence linking hypoxia to cognitive impairment in the general population, this link has yet to be explored in BC. This highlights a significant gap in the available data, hindering the effective management of hypoxia and, as a result, limiting the potential to improve patients’ quality of life. The current study aims to characterize the presence of nocturnal hypoxia and its association with cognitive functioning in BC patients compared to HCs. By investigating the relationship between hypoxia and CRCI, we hypothesize that nocturnal hypoxia will be linked to cognitive impairment in BC patients.

Results

Clinical and demographic characteristics of the sample

The demographic and clinical characteristics of the participants are provided in Table 1. No significant differences were observed between BC patients and HCs in terms of demographic characteristics or depression (ps > 0.05). However, a significant difference was observed on the state anxiety as measured by STAI-A (p = 0.02).

BC patients reported significantly more sleep complaints (p = 0.034) and more severe insomnia symptoms (p = 0.012) compared to HCs (Table 1).

Table 1 Demographic (mean ± SD [range]) and clinical (n (%)) characteristics, and self-reported outcomes of HCs and BC patients.
Full size table
Table 2 Comparison of cognitive scores between BC patients and HCs.
Full size table
Table 3 Multivariable linear models measuring the relationship between nocturnal hypoxia and cognitive performance in the entire sample.
Full size table

Comparison of nocturnal hypoxia between BC patients and HCs

The nocturnal hypoxia scores are presented in Fig. 1. The composite hypoxia score differed significantly between the two groups, indicating more severe hypoxia, as assessed with the composite score, in BC patients compared to HCs (p = 0.028). Specifically, BC patients had lower minimal SpO2 than HCs (BC: 84.8 ± 5.5; HC: 88.1 ± 4.2; p = 0.028). However, there were no significant differences between the groups for ODI (BC: 13.5 ± 10.7; HCs: 11.7 ± 11; p = 0.48) or TST90% (BC: 3.6 ± 6.6; HCs: 0.6 ± 0.8; p = 0.26).

Fig. 1
Fig. 1
Full size image

Comparisons of nocturnal hypoxia variables between BC patients and HCs. The hypoxia composite score differed significantly between the two groups, indicating more frequent hypoxia in BC patients compared to HCs. No significant difference was found for the Oxygen Desaturation Index ≥ 3% (ODI ≥ 3%). No significant difference was observed for Total Sleep Time with oxygen saturation ≤ 90% (TST90%). A significant between-group difference was found for Minimal Oxygen Saturation (SpO2).

Between-group differences for cognitive outcomes

No significant differences in scores were found across all cognitive domains assessed. However, a tendency towards a significant difference was observed for the episodic memory composite score (p = 0.06) (Table 2). For more details on raw neuropsychological test scores, please refer to Table S1 in the Supplemental document.

Relationship between nocturnal hypoxia and cognitive performance in the entire sample

We performed multivariable models to determine whether nocturnal hypoxia was associated with cognitive performance in BC. The results are presented in Table 3. No significant association were found between the composite hypoxia score and the composite scores of the different cognitive domains, nor when the predictor was the interaction between group and hypoxia (ps > 0.05). To minimize the risk of missing any specific effects, we examined the relationships between hypoxia and individual cognitive test scores (See Supplemental Online Content 1). These exploratory analyses did not reveal any significant associations (ps > 0.05).

Discussion

To our knowledge, this is the first study to explore the relationship between nocturnal hypoxia and cognitive performance in BC. All participants underwent full-night ambulatory PSG and the same cognitive assessments. A significant difference for nocturnal hypoxia was observed between the two groups, with BC patients experiencing more frequent hypoxia than HCs. However, no significant associations were found with cognitive scores.

One possible explanation for the differences between prior non-cancer studies reporting hypoxia-related cognitive impairment and our results relates to differences in the degree of nocturnal hypoxia. Actually, in our sample, nocturnal hypoxia is mild when compared to populations from SDB cohorts25,26,27,28,29,30,31,32 which might reduce the relationship with cognition. For example, chronic hypoxia in SDB was found to directly cause neuronal injury in the hippocampus and frontal cortex, leading to deficits in attention, memory and executive function37. In addition, cognitive impairment is common in BC and is primarily attributed to cancer treatments, inflammation and psychological factors38,39. These factors might overshadow nocturnal hypoxia’s effect on cognition in our sample.

In line with previous research (e.g., 40,41), our results revealed an increase in cognitive complaints, as measured by the “perceived cognitive abilities” subscale of the FACT-Cog, in BC patients compared to HCs. In this regard, BC patients reported a perceived decline in their cognitive abilities, whereas HCs did not perceive any significant reduction in their cognitive function. However, no significant differences were observed between the two groups in terms of objective cognitive scores, except for a tendency toward significance in episodic memory. Even-though this finding did not reach statistical significance, it suggests an area for additional research, as episodic memory has been identified as vulnerable in previous studies12,14,42,43. Our patients were primarily treated with RT and/or ET, with none receiving chemotherapy. This may explain why their cognitive performance was not impaired as expected. Previous research indicates that ET and RT generally have less impact on cognition compared to chemotherapy. For example, a survey of 2296 BC patients aged 34 to 82 revealed that 60% experienced cognitive symptoms following treatment, with more pronounced cognitive symptoms in patients treated exclusively with chemotherapy than in those who received only ET44. While rates of cognitive dysfunction in BC patients following ET vary between 32 and 64%45, the findings in the literature remain inconsistent46. In their literature review, Haggstrom and colleagues46 presented 72 studies on CRCI in BC patients treated with ET: some studies indicated an impact of ET on cognition, others found no effect, concluding that underdiagnosis of CRCI related to ET is common and should be better addressed. In the current study, patients treated with both ET and RT, or with RT alone, demonstrated comparable cognitive performance (see Supplemental Table S2).

In the context of SDB, only one study has addressed its prevalence among BC patients. This study showed an 88% prevalence of SDB in this population following chemotherapy, RT and/or ET (21). However, the prevalence of SDB has been studied in other types of cancer, and particularly following RT. In a systematic review, Tawfik and colleagues47 showed a positive association between the occurrence of SDB and RT in 103 of 181 head and neck cancer patients (OR 1.16, 95% CI [0.52–2.56]; P = 0.718) with an overall prevalence of 63% (95% CI [0.36–0.85]; P = 0.343) and thus recommending the screening of all cancer patients treated with RT for early signs of SDB. In the current study, higher hypoxia in our sample of patients may be the result of the effects of RT. However, this issue remains understudied in BC patients and warrants further investigation. Additionally, it is worth mentioning that ET is a treatment that suppresses the production of estrogen and progesterone48. The absence of these hormones due to ET may lead to relaxation of the pharyngeal muscles, potentially resulting in greater airway obstruction and an increased risk of hypoxia. In the current study, no significant differences in nocturnal hypoxia were noted between patients treated with both ET and RT and patients treated with RT alone (See Supplemental Table S2). Therefore, ET may not be the primary factor influencing hypoxia in our sample. Finally, menopause is another factor that may explain the presence of hypoxia in our sample. Menopause leads to a reduction in estrogen and progesterone levels49. These hormones play a role in maintaining airway function by preserving the muscle tone of the throat50. Therefore, a reduction in these hormones could increase the likelihood of developing SDB by decreasing muscle tone, thus leading to more breathing difficulties51. While we cannot draw definitive conclusions at this stage, our results suggest that both treatment (probably RT, but not ET) and menopause may contribute to the presence of SDB and, consequently, nocturnal hypoxia in our sample. Higher presence of SDB is our patient sample is consistent with larger sleep complaints in this group. Indeed, BC patients reported more sleep disturbances and more severe insomnia symptoms compared to HCs. This finding is consistent with previous studies showing that sleep-related complaints are particularly prevalent in BC patients52,53,54,55.

This study has several limitations. First, the cross-sectional design limits the assessment of causal relationships between nocturnal hypoxia and cognition in BC. Additionally, the small effect size may have reduced statistical power. Yet, the tendencies toward significance are particularly noteworthy, as they align with previous findings and highlight the potential impact of BC on specific cognitive functions. Furthermore, the cognitive tests used in this study may not have been sensitive enough to detect subtle cognitive impairments that BC patients report in daily life. These tests were conducted in laboratory settings, which do not capture the daily life difficulties patients face. Future research with a longitudinal design is needed to clarify these findings. Studies with pre-treatment baselines and multiple follow-ups might help determine the evolution of this relationship in BC patients. Also, future studies must include a larger sample size in order to increase statistical power. Objective sleep assessment methods such as PSG and more sensitive cognitive assessments are needed to clarify these findings. Detailed cognitive testing to explore whether nocturnal hypoxia impacts cognition directly or through indirect pathways. This might allow us to conclude whether the lack of significant association demonstrated in our sample reflects treatment effects or a true absence of association.

In conclusion, nocturnal hypoxia is not associated with cognitive performance in our sample. Although hypoxia is more frequent during the night in BC patients compared to HCs, it may not be the primary factor contributing to cognitive difficulties in this population, and other sleep disturbances may play a role in explaining cognitive performance in this population. Future studies are needed in this field.

Materials and methods

Participants

Data were gathered from the PROSOM-K project56. All participants gave their written informed consent, and the study was approved by an ethics committee (CPP Ile de France III), with approval number ID-RCB: 2017-A02778-45. Patients were recruited at the Comprehensive Regional Cancer Centre, François Baclesse (Caen, Normandy, France). The study included two groups: 35 BC patients and 21 HCs.

Patient inclusion criteria were: (i) less than 70 years old, (ii) no metastatic BC, (iii) already undergone surgical or RT treatment, (iv) menopausal status since at least one year ago at the time of inclusion, (v) no personality disorder and progressive psychiatric disorder, (vi) no neurological sequelae, (vii) no drug use or use of alcohol, (viii) be a native French speaker, (ix) have at least a primary school level of education. Inclusion criteria of HCs were the same as for patients, including no history of cancer and no global cognitive impairment according to the Montreal Cognitive Assessment57. HC were matched in age and education with BC patients (Table 1).

Measures

Anxiety and depression assessments

Depression was assessed with the Beck Depression Inventory (BDI)58. Anxiety was measured with the State-Trait Anxiety Inventory (STAI-Y)59. BDI evaluates depression using a 21-item scale that provides information about the severity and nature of depression in the participants. The items range from 0 = the absence of symptoms to 3 = an intense level. The sum of the score will define the presence and severity of depression. A score of 1–10 = absence of depression, 11–16 = mild mood disturbance, 17–20 = borderline clinical depression, 21–30 = moderate depression, 31–40 = severe depression, and finally, a score over 40 = extreme depression. In regards to STAI-Y, the first 20 items assess state anxiety (STAI-A), or how the participant feels right now; the second 20 items assess trait anxiety (STAI-B), or how the participant generally feels. Higher scores indicate higher depression and anxiety symptoms, respectively.

Self-reported sleep quality assessments

The Insomnia Severity Index (ISI)60 is a self-report questionnaire assessing the nature, severity, and impact of insomnia. It evaluates the following dimensions: severity of sleep onset, sleep maintenance, early morning awakening problems, sleep dissatisfaction, interference of sleep difficulties with daytime functioning, noticeability of sleep problems by others, and distress caused by sleep difficulties. A score ≥ 7 is considered to indicate the presence of insomnia.

The Pittsburgh Sleep Quality Index (PSQI)61 is a self-report questionnaire that is used to assess efficiency and quality of sleep over a 1-month time interval. It measures the following components: subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleep medication, and daytime dysfunction. A score > 5 is considered to indicate a significant sleep disturbance.

Neuropsychological assessments

Subjective cognitive complaints were assessed using the Functional Assessment of Cancer Therapy Cognitive Scale (FACT-Cog) in its French version62.

Episodic memory was assessed with the RL/RI-16 test63, a French adaptation of the Grober and Buschke’s procedure. We calculated the sum of the three immediate free recall scores, the sum of the three immediate total (i.e., free and cued) recall scores, and we used the delayed free recall and the delayed total recall scores.

Working memory was assessed with the Digit Span forward and backward test64. Episodic buffer was assessed with a binding task, the multimodal integration task65: participants had to mentally associate four colored letters and the location of the cross of the same color randomly distributed over a grid. Then a grid appeared with one black letter and the participants indicated if it was correctly located. The score corresponded to the number of correct trials.

The assessment of executive functions focused on three domains, each evaluated through specific cognitive tests. Planning was assessed using the score of accuracy in part 1 of the Zoo Map Test66. Flexibility was evaluated using the Trail Making Test (TMT67, by the following score [TMTB time – TMTA time]/TMTA time. Inhibition was measured with the Stroop test68, [Interference time – Color time]/Word time.

Processing speed corresponded to the average mean time taken to name the colors of the Stroop test and to perform the TMT Part A.

To minimize the issue of multiple statistical testing, we computed composite scores for each cognitive domain. Performance on various cognitive tests were z-transformed and averaged. Before averaging, Z-scores derived from reaction times, percentages/number of errors were reversed so that increasing values always indicated better performance. Accordingly, we categorized four cognitive composite scores: Executive Function, Processing Speed, Working Memory and Episodic Memory (See Supplement).

Hypoxia characterization

All participants underwent a full night recording of sleep using ambulatory PSG at home with monitoring of EEG, EOG, chin, diaphragm and EMG, nasal pressure, oro-nasal thermistor signals, sound (snoring), thoracic and abdominal plethysmography and pulse oximetry routinely used in PSG to measure minimal oxygen saturation (SpO2). All sensors were connected to the Compumedics Siesta sleep system and placed by an EEG technician. Sleep scoring was conducted according to American Academy of Sleep Medicine criteria (AASM)69.

Three variables were extracted to reflect hypoxia as previously proposed70,71. As defined by Berry and colleagues72 following The AASM Manual for the Scoring of Sleep and Associated Events:

  • Oxygen Desaturation Index ≥ 3% (ODI) is defined as the number of times per hour of sleep that blood oxygen levels drop by 3% or more from baseline levels.

  • Total Sleep Time with oxygen saturation ≤ 90% (TST90%) represents the cumulative amount of time during which the blood oxygen saturation levels remain at or below 90% throughout the sleep period.

  • Minimal Oxygen Saturation (SpO2%) reflects the lowest recorded oxygen saturation level during the sleep period. A lower SpO2 is a marker of severe oxygen desaturation episodes.

We also computed a composite score for nocturnal hypoxia. The hypoxia composite scores correspond to the mean of z scores of SpO2%, ODI and TST90%70. Z-scores derived from ODI and TST90% were reversed so that increasing values always indicated better respiratory efficiency.

Statistical analysis

Demographical and clinical characteristics were described using mean, standard deviation and range for quantitative variables and using frequencies and percentage for qualitative variables. The Shapiro-Wilk test was used to assess the normality of the data. Since the normality test was rejected, we applied the non-parametric Wilcoxon Mann-Whitney test to compare demographic, sleep, cognitive, and quality of life characteristics between the groups. Finally, we conducted multivariable models to determine whether hypoxia was associated with cognitive functioning in BC. These models were, adjusted on age, anxiety (STAI-B), depression (BDI), and the group effect, as these factors are known to influence cognitive functioning and performed on the whole sample. Analyses were conducted using R software, version 4.3.0 with statistical significance set at p < 0.05.