Distinct sleep and rest-activity profiles according to adjuvant treatment exposure in women treated for breast cancer

Abstract

Background

The associations between adjuvant treatment combinations and sleep-wake processes in women treated for breast cancer remain poorly understood, thereby limiting the development of individualized supportive care. This exploratory study aims to investigate this effect in women treated for breast cancer receiving chemotherapy (CT) combined or not with radio- (RT) and endocrine (ET) therapies on sleep-wake processes and cancer-related fatigue.

Patients and Methods

Sixty participants were evaluated (21 patients receiving CT+RT+ET [CT group], 23 patients receiving RT+ET only [non-CT group], and 16 healthy control women [HC group]). Participants completed questionnaires assessing cancer-related fatigue and sleep complaints and wore a wrist actigraph for 72 hours to extract sleep-derived and rest-activity rhythm (RAR) parameters. General linear model served to compare groups when controlling for age and cancer stage. Correlation matrices were computed using Spearman’s correlations to evaluate the links between variables in each patient’ group.

Results

Sleep complaints and cancer-related fatigue did not significantly differ between groups. The CT group had fewer but longer nocturnal awakenings and shorter sleep-onset latency compared to both non-CT and HC groups. The non-CT group had lower dichotomy index—reflecting robustness of the RAR—than the CT group but did not differ significantly from the HCs. Fatigue was negatively correlated with RAR amplitude in the non-CT group.

Conclusion

This exploratory study suggests that distinct patterns of sleep-wake parameters may be observed across the treatment exposure- groups (ie, RT and ET + or—CT). These results should not be interpreted as evidence that treatment exposure itself explains the observed sleep-wake differences and require further investigation to be confirmed.

Clinical trial registration number

NCT04391543 (https://clinicaltrials.gov/study/NCT04391543)

Implications for practice.

Sleep-wake processes may vary depending on adjuvant treatments for breast cancer, and actigraphy can be used to assess these differences. In this exploratory study, we observed differences in sleep-wake profiles across different adjuvant treatment exposure groups. Moreover, fatigue and rest-activity rhythm amplitude were associated, suggesting a link between both processes. These exploratory findings may help generate hypotheses regarding individualized supportive care strategies.

Introduction

Breast cancer represents the most frequent cancer in women, with 2.3 million new cases and approximately 670 000 deaths worldwide in 2022.¹ Significant advances in screening and early detection have contributed to the high survival rates² (up to 90%). Nevertheless, the global incidence of breast cancer continues to rise,³ especially in Western countries. As survival improves and more women live with or beyond breast cancer, increasing attention must be paid to long-term side effects associated with the disease and its treatments.

Among these side effects, disturbances in sleep and the rest-activity rhythm (RAR) are highly prevalent. Sleep complaints primarily refer to insomnia, occurring in almost 70% of patients with breast cancer.^4,5 RAR (day/night alternation of activity and rest periods, reflecting behaviors) changes, are often correlated with insomnia severity⁶ and constitute strong predictors of overall mortality in breast cancer.⁷ Recorded by actigraphy, it serves as an indirect proxy of circadian rhythms (ie, ∼24-hour biological oscillations in physiological and psychological functions driven by the biological clock in the suprachiasmatic nucleus in the hypothalamus), and its alterations may manifest at multiple stages of the disease trajectory, from pre-diagnosis through survivorship, in a substantial proportion of patients.⁸ These disruptions primarily manifest as a reduced amplitude (difference between the lowest and highest point of the fitted rhythm) and a lower mesor (mean value of the fitted rhythm). In a subset of cases, a phase delay is also observed, characterized by a delayed sleep onset and wake-up times.^9,10

Both sleep and RAR changes are intrinsically linked to cancer-related fatigue,^10,11 with higher levels of disturbances predicting higher cancer-related fatigue during treatments, although the chicken-and-egg question remains to be clearly answered.^12–15 Cancer-related fatigue is a distressing, persistent, subjective sense of physical, emotional, and/or cognitive tiredness or exhaustion related to cancer or cancer treatment¹⁶ and must be differentiated from sleepiness (higher need for naps with greater propensity to fall asleep while being inactive), which is a ubiquitous phenomenon, experienced not only as a symptom in a number of medical conditions but also as a normal physiological state by most.¹⁷ An important but often overlooked factor in understanding how sleep-wake disturbances and fatigue are modified depending on treatment exposure. Most studies have examined chemotherapy (CT)^{9–11,18–20} or radiotherapy (RT)^21–23 in isolation and both modalities have been associated with increased sleep disturbances^24,25 and cancer-related fatigue.^10,15,26,27 However, distinguishing their specific contributions is crucial. Notably, by controlling for other adjuvant modalities, Savard et al.²⁸ were able to attribute worsen insomnia specifically to RT finding no such independent effect for CT and endocrine therapy (ET).

Another study reported that patients undergoing CT or RT, but not ET, for breast cancer experienced higher levels of sleep disturbances compared to those not receiving adjuvant therapy,¹⁵ with increased sleep disturbances being significant predictors of greater cancer-related fatigue.¹⁵ The pattern of co-occurring sleep disturbances and RAR alterations also seems to differ between adjuvant treatment modalities. Amplitude (difference between the lowest and highest point of the fitted rhythm) and mesor (mean value of the fitted rhythm) have been repeatedly found to worsen from the first CT cycle and remain impaired at cycle 4.^10,18 Research on the effects of ET on sleep disturbances remains limited. Bhave et al.²⁹ reported no significant changes in sleep patterns with aromatase inhibitors (ie, a type of ET) administration. However, their findings revealed increased fatigue and lower daytime activity levels at ET initiation, which may contribute to a reduced mesor. Actigraphy-based studies have also identified lower sleep efficiency, increased nighttime awakenings, and higher nocturnal activity levels in breast cancer women receiving ET, consistent with their subjective complaints.³⁰ Recently, we demonstrated increased complaints of insomnia and cancer-related fatigue,³⁰ in patients receiving ET and RT following surgery compared to participants without cancer history. Notably, these patients had also lower daytime activity and sleep efficiency, more time awake, and higher activity levels at night. Therefore, while previous reports suggest different patterns of changes in sleep-wake processes following either CT, RT, or ET, few studies have investigated sleep-wake disruptions according to adjuvant treatment exposure.

Combined cancer treatments (CT, RT, and ET) induce persistent fatigue through synergistic mechanisms, notably related to systemic inflammation.³¹ For example, Garcia et al.³² demonstrated that adjuvant chemoendocrine therapy (CT+ET) significantly worsened fatigue and endocrine symptoms in patients suffering from early-stage breast cancer compared to ET alone. However, no indication on relationship between sleep, RAR and fatigue were given.

To expand upon these findings, we conducted a retrospective analysis of sleep and RAR characteristics derived from actigraphy in breast cancer women. The main objective was to quantify the effect of treatment exposure—that is, RT and ET treatments with or without CT exposure—on sleep-wake parameters and cancer-related fatigue. We hypothesized that patients exposed to CT in addition to RT and ET would show more severe sleep-wake alterations and fatigue than patients treated with RT and ET only. Moreover, based on previous reports, we expected relationship between rest-activity parameters and cancer-related fatigue. Because adjuvant treatment combinations are increasingly used in routine care, understanding whether they are associated with distinct sleep-wake profiles may help identify patients at risk for different supportive care needs.

Methods

Participants

Data for this exploratory study were drawn from 2 independent cohorts acquired in the context of 2 protocols conducted in line with the Declaration of Helsinki and approved by the ethics committee: the BIOCARE FActory project (ClinicalTrials.gov, NCT04391543, May 2020), a multicentric and longitudinal study of patients treated for breast cancer (CPP SUD EST VI, n°ID-RCB : 2019-A02525–52)^33,34; and the PROSOM-K project, a project dedicated to quantifying sleep and memory disturbances in patients treated by ET for breast cancer³⁵ (CPP Ile de France III, n°ID-RCB: 2017-A02778-45). Written informed consent was obtained from all participants.

Both cohorts included women diagnosed with non-metastatic breast cancer. Common inclusion criteria across cohorts included a diagnosis of localized breast cancer (stages I to IIIc), ongoing treatment, adult age (≥18 years), and a preserved general condition (ECOG performance status ≤2 or functional autonomy compatible with study procedures). Participants were also required to be able to understand and carry out the experimental procedures.

Each cohort had additional specific inclusion criteria. The BIOCARE FActory cohort included patients aged 18–80 years, treated at Clinique Victor Hugo—Center Jean Bernard or Center Hospitalier du Mans. The PROSOM-K cohort exclusively enrolled postmenopausal women (at least 1-year post-menopause), under 70 years of age, native French speakers, with a minimum educational level (level 3 on the Barbizet scale), and treated at the Center François Baclesse in Caen.

Exclusion criteria also overlapped to some extent, notably regarding neurological, psychiatric, or cognitive conditions that could interfere with the assessments, as well as the use of psychotropic medications. However, each cohort applied specific additional exclusions: the BIOCARE FActory cohort excluded patients with chronic illnesses associated with severe fatigue (eg, diabetes, cardiovascular diseases), or contraindications to physical activity. The PROSOM-K cohort, on the other hand, specifically excluded patients who had undergone CT or presented with paraneoplastic syndromes.

Data from 46 women diagnosed with breast cancer and 16 healthy controls (HCs) were initially considered for this study. The breast cancer sample was divided into 2 groups (n = 23 each): the CT group had received CT, RT, and ET, while the non-CT group had received RT and ET but no CT. Patients were respectively 5 and 9 months post-RT on average for the CT and the non-CT groups, and at 6 months post-CT for the CT group. In the CT group, most of the patients had received anthracycline + taxane-based CT (n = 16), 4 patients had received anthracycline + taxane + anti-HER2 therapy, 2 had received taxane + anti-HER2 therapy without anthracycline and 1 had received anthracycline-based CT alone. In the CT group, 4 had received aromatase inhibitor alone, 5 tamoxifen alone, 2 had received combined therapy, 1 patient was treated with Olaparib and information about molecules was missing for 5 patients. In the non-CT group, most of the patients had received aromatase inhibitor alone, 1 tamoxifen alone and 2 had received combined therapy. The HC group consisted of 16 women with no history of cancer.

Assessments

Self-report questionnaires

Cancer-related fatigue was assessed using the EORTC QLQ-FA12 questionnaire³⁶ for the BIOCARE FActory cohort. The FA12 consists of 12 items rated on a 4-point Likert scale (1-4), with higher scores indicating greater fatigue. Raw scores were summed (range 0-48) and linearly transformed to a 0-100 scale. For the PROSOM-K cohort, cancer-related fatigue was assessed using the multidimensional fatigue inventory questionnaire (MFI-20).³⁷ It consists of 20 items with a 5-point Likert scale designed to measure general fatigue, physical fatigue, mental fatigue, reduced motivation, and reduced activity. For the present analyses, only the 4 items of the general fatigue subscale were used, yielding a raw score ranging from 4 to 20. To enable comparability between instruments, a harmonized fatigue score was computed using the ones provided in the article by Friedrich et al.,³⁸ which applies correction coefficients to standardized scores derived from each questionnaire.

Subjective sleep quality was assessed using the Pittsburgh Sleep Quality Index questionnaire. It is a self-administered questionnaire comprising 19 items that evaluate the types and frequencies of sleep disturbances over a 1-month interval (subjective sleep quality, sleep latency, usual sleep efficiency, nocturnal disturbances, sleep duration, sleep medications, and daytime dysfunction).³⁹ The total score from the questionnaire was used as the variable for the analysis of subjective sleep.

Sleep-related And Rest activity Rhythm

Sleep-related and RAR parameters were assessed using actigraphy worn on the non-dominant wrist for 72 consecutive hours (including nights and to be removed only when washing). In the BIOCARE FActory cohort, participants wore an eTact® (BodyCap, Caen, France) and in the PROSOMK cohort, participant wore a MotionWatch 8®, (CamNtech Ltd, UK for PROSOM-K cohort). Device allocation differed across cohorts: the CT group was assessed using the eTact device, the HC group using the MotionWatch 8, and the non-CT cancer group using both devices (18 participants wore the MotionWatch 8 and 5 participants wore the eTact device).

A 72-h recording enables computing sleep and RAR parameters as previously used in cancer population.^40–42 Data were processed using the proprietary software for each device (eTact analysis and Motionware 8). The e-tact, an emerging actimetry solution, has demonstrating high validity against Motionwatch 8 for total sleep time (r = 0.95), wake after sleep onset (r = 0.85) and sleep efficiency (r = 0.90). They both contain a tri-axial sensor detecting acceleration at a 0.01 g sensitivity threshold. Both devices were set on high sensitivity to allow best detection of wake episode during actigraphy analysis.

Sleep-related parameters were time in bed (minutes), total sleep time (minutes), sleep onset latency (minutes), wake after sleep onset (minutes), number of nocturnal awakenings, and sleep efficiency, defined as the percentage of total sleep time during the time in bed period. RAR parameters were estimated from extracted CSV actigraphy data (epoch 1 min) using Time Series Analysis Serial Cosinor 8.0 software (Expert Soft Technology, Richelieu, France). Single and Cosinor methods^43,44 allowed to evaluate the cosine mathematical function that best fits the data as a function of time. From this analysis, 2 parameters were used for this study, the mesor (rhythm-adjusted mean for a 24-h period) and the amplitude, which is the measure of one half the extent of the rhythmic variation in a cycle. For this study, the amplitude value was doubled to quantify the full amplitude of the RAR. These 2 parameters have been described as most representative of RAR disturbances in patients with breast cancer.⁴⁵ The dichotomy index I < O measures the percentage of time (ranging from 0% to 100%) that activity during in-bed periods falls below the median level of activity during out-of-bed periods. A higher I < O value, close to 100%, indicates a more robust and prominent RAR, whereas a lower or equal to 97.5% I < O values infer the presence of RAR disruption.⁴⁶

Statistical analysis

Due to the exploratory nature of the study, outliers were retained. Two patients from the CT group with missing data were fully removed from analyses and 3 participants had missing data for the PSQI but were retained for analyses.

Shapiro–Wilk tests were performed on the residuals of the linear models to assess the assumption of normality. Cancer-related fatigue, wake after sleep onset, number of nocturnal awakenings, Pittsburgh Sleep Quality Index, time in bed, total sleep time, age, amplitude, and mesor were considered normally distributed. sleep efficiency, sleep onset latency, and the I < O index significantly violated the normality assumption, suggesting their residuals deviated from a normal distribution. A Ln transformation was thus applied to these variables. Levene’s test was conducted to assess the homogeneity of variances across the 3 groups. A general linear model (GLM; GAMlj 3, JAMOVI), including age as a covariate, was applied to compare the 3 groups (both cancer groups and the control group). Another GLM was then applied including both ages and cancer stage between the 2 patient groups. Significant differences between groups were assessed first considering the overall model fit P-value. In case of significant model fit, group and age effects were considered. The level of significance was set a priori at 0.05 and FDR correction on the group effect considering actigraphy variable comparisons are reported for transparency. Pairwise comparisons were analyzed using a Tukey correction. Correlograms with Spearman’s correlations were computed in each group separately considering only variables that differed between groups, with the aim to evaluate the association of the different variables. Partial correlation with age as a covariate were also performed to verify the possible confounding effect of age on the relationship between variables.

Results

Demographical and clinical characteristics

The mean age of participants was 54.33 ± 1.61 years in the CT group, 64.9 ± 1.54 years in the non-CT group, and 62.19 ± 1.85 years in the + HC group (P < .01) (Table 1). Patients in the CT group are younger than in the non-CT (P < .001) and in the HC groups (P = .006). On average, participants had been undergoing ET for 4 ± 2 months in the CT group and 7 ± 3 months in the non-CT group. Regarding cancer stage distribution, the CT group included 1 patient at stage I, 14 at stage II, 7 at stage III, and 1 with missing data. The non-CT group included 8 patients at stage I, 14 at stage II, and 1 at stage III.

Table 1.

Clinical characteristics of the patients, as well as demographical characteristics and questionnaire’ results of all groups.

Variables	CT	Non-CT	HC	P-values of the model (η²p)	Post hoc results
	n = 21	n = 23	n = 16
Age	54.33 ± 1.61	64.9 ± 1.54	62.19 ± 1.85	<.01* (.265)	CT < non-CT (P < .001); CT < HC (P < .006)
Fatigue score	53.56 ± 2.14	58.34 ± 1.95	52.97 ± 2.24	.073**
Sleep complaints	8.50 ± 1,5	7.99 ± 0.97	5.04 ± 1.08	.080**
Endocrine therapy duration (months, mean ± SD)	4 ± 2	7 ± 3
Cancer stage distribution	I = 1	I = 8
	II = 14	II = 14
	III = 7	III = 1
	NaN = 1

Data for fatigue and sleep complaints are presented as adjusted means ± SE. All variables, except age, were tested with age as a covariate.

Abbreviations: η²p, indicates partial eta square reflecting effect size as follows: 0.01 = small, 0.06 = medium, 0.14 = large; CT, patients receiving chemotherapy, radiotherapy and endocrine therapy; HC, healthy control women; non-CT, patients receiving radiotherapy and endocrine therapy.

^*Indicates P < .05 for group effect.

^**Indicates the P-value of the model fit overall when model fitting was not significant. Post hoc results are reported in case of P-values of the model <.05, otherwise, not reported.

Self-report questionnaires

No significant group differences were observed for cancer-related fatigue (model fit F[3, 56] = 2.45, P = .073) (Table 1). Subjective sleep complaints (PSQI) did not differ significantly between groups (F[2, 53] = 3.06, P = .055). When comparing both cancer groups, no significant difference was observed for fatigue (P = .369), nor for sleep complaints (P = .334).

Sleep-related And Rest-Activity Rhythm

Results reported in Table 2 and Figure 1 refer to the 3 group comparisons with age as a covariate. Additional results related to comparison between only the 2 patient groups considering cancer stage as a covariate are reported in the text at the end of each paragraph.

Table 2.

Sleep-related and rest-activity rhythm parameters results as assessed with actigraphy across the 3 groups.

Parameters	CT (n = 21)	Non-CT (n = 23)	HC (n = 16)	P-values of the model (η²p)	P-values group effect FDR corrected	Post hoc results
Time in bed (min)	525.00 ± 14.67	530.86 ± 14.44	517.52 ± 15.33	.697**	.799	–
Total sleep time (min)	434.59 ± 15.87	445.17 ± 14.44	439.10 ± 16.58	.585**	.799	–
Sleep efficiency (%)	82.91 ± 1.66	83.51 ± 1.51	84.60 ± 1.74	.521**	.799	–
Duration of wake after sleep onset (min)	88.51 ± 7.47	62.67 ± 6.80	57.81 ± 7.81	<.018* (.133)	.054	CT > HC (P = .021); CT > non-CT (P = .051)
Number of nocturnal awakenings	18.99 ± 2.52	33.24 ± 2.29	36.61 ± 2.63	<.001* (.303)	.0045*	CT < non-CT (P < .0001); CT < HC (P < .001)
Sleep onset latency (min)	2.90 ± 3.02	14.54 ± 2.75	10.51 ± 3.16	<.001* (.259)	.0045*	CT < non-CT (P < .001); CT < HC (P = .002)
I < O index	95.33 ± 0.69	96.82 ± 0.63	97.82 ± 0.72	.055 (.099)	.099
Amplitude (counts)	151.28 ± 11.75	138.20 ± 10.68	175.79 ± 12.27	.067 (.137)	.1
Mesor (counts)	79.26 ± 5.59	74.14 ± 5.08	90.32 ± 5.84	.106	.13	–

Data are presented as adjusted means ± SE and pairwise group comparisons were performed using estimated marginal means with Tukey correction.

Abbreviations: η²p, indicates partial eta square reflecting effect size as follows: 0.01 = small, 0.06 = medium, 0.14 = large; CT, patients receiving chemotherapy, radiotherapy and endocrine therapy; HC, healthy control women; non-CT, patients receiving radiotherapy and endocrine therapy.

^*Indicates P < .05 for group effect.

^**Indicates the P value of the model fit overall when model fitting was not significant. Post hoc results are reported in case of P values of the model <.05, otherwise, not reported.

Figure 1.

Violin plots showing the distribution of values across groups.

Each violin represents the kernel density of the raw data, with the embedded boxplot indicating the median and interquartile range. Individual observations are displayed as jittered points. Dots and error bars represent the estimated marginal means (EMMs) and their 95% confidence intervals obtained from the ANCOVA, adjusted for age. Pairwise group comparisons were performed using estimated marginal means with Tukey correction, and significant differences are indicated above the plots. CT, patients receiving chemotherapy, radiotherapy and endocrine therapy; HC, healthy control women; non-CT, patients receiving radiotherapy and endocrine therapy. *P < .05.

Sleep-related parameters

No significant group difference was found for time in bed (model fit F[3, 56] = 0.2483, P = .697), total sleep time (model fit F[3, 56] = 0.65, P = .585) and sleep efficiency (model fit F[3, 56] = 0.76, P = .521). For the duration of wake after sleep onset, a significant group effect was detected (F[2, 56] = 4.30, P < .018). Patients in the CT group had longer duration of awakenings during night than the HC group (P = .021), while difference with the non-CT group was not significant (P = .051). For the number of nocturnal awakenings, a group effect was found (F[2, 56] = 11.04, P < .001). Post hoc analyses revealed significantly less awakenings in the CT group compared to both the non-CT (P < .0001) and HC groups (P < .001), with no significant difference between non-CT and HCs (P = .587). Ln transformed sleep onset latency significantly differed between groups (F[2, 56] = 9.74; P < .001). Post hoc tests indicated that the CT group had significantly shorter sleep onset latency than both the non-CT (P < .001) and the HC groups (P = .002), while the non-CT and the HC groups did not differ (P = .931).

When comparing only both patient’ groups, and when cancer stage was added as a covariate, time in bed (P = .176), total sleep time (P = .471), sleep efficiency (P = .214), did not differ between CT and non-CT groups. The duration of wake after sleep onset (P = .05) was higher and the number of nocturnal awakenings was lower in the CT group compared to the non-CT group (P = .022). Sleep onset latency was lower in the CT group compared with the non-CT (P = .01).

RAR parameters

For amplitude, despite a significant model fit (F[3, 56] = 2.95; P = .024), no significant group effect was detected (F[2, 56] = 2.83, P = .067), For mesor, no significant group effect was detected (F[2, 56] = 2.34, P = .106). For the Ln I < O index, the model fit was not strictly significant (F[3, 56] = 4.79, P = .055).

When comparing only both patient’ groups, and when cancer stage was added as a covariate, the overall GLM was not significant for amplitude (P = .163) and mesor (P = .740). The I < O index was significantly higher in the non-CT than in the CT group (P = .025).

Relationships between actigraphy parameters and fatigue scores

Results from correlograms display in Figure 2 refer to correlation without age as a covariate. Additional partial correlation using age as a covariate are reported in the end of this paragraph. In the CT group, the I < O Index was negatively correlated with both the number (r = −0.32; P = .039) and the duration of awakenings (r = −0.31; P = .001) and positively correlated with the amplitude of the rhythm (r = 0.61; P = .002). In the non-CT group, the I < O Index was positively correlated with the amplitude of the rhythm (r = 0.64; P = .001) and negatively with the sleep onset latency (r = −0.55; P = .006). The amplitude of the rhythm was positively correlated with the duration of awakenings (r = 0.47; P = .024) and negatively with fatigue (r = −0.51; P = .013). Partial correlation using age as a covariate led to the following results. In the CT group, only the correlations between I < O Index and both the number (r = −0.51; P = .023) and the duration of awakenings (r = −0.63; P = .003) remained significant. In the non-CT group, only the correlation between fatigue and amplitude of the rhythm remained significant (r = −0.54; P = .010).

Figure 2.

Correlograms illustrating relationships between sleep-wake parameters and fatigue with heatmap correlation in each patients group.

CT, patients receiving chemotherapy, radiotherapy and endocrine therapy; HC, healthy control women; non-CT, patients receiving radiotherapy and endocrine therapy. *P < .05; **P < .01. Only significant correlations (P < .05) are displayed. Colors and circle sizes refer to the correlation coefficients. Blue circles indicate positive correlations and red circles indicate negative correlations. Circle size and color intensity are proportional to the correlation coefficient (r), and the color scale below each correlogram indicates the corresponding correlation values.

Discussion

Our findings support the observation that different adjuvant treatment exposure group may be associated with distinct sleep-wake profiles. Although exploratory, these observations should be interpreted as associations observed within this specific dataset rather than as effects of treatment exposure itself. Our results show that patients who received CT (CT group) had fewer but longer nighttime wakefulness and had a shorter sleep-onset latency compared to both the non-CT patients and the HCs. In contrast, the non-CT group did not significantly differ from other groups. Fatigue did not significantly differ between groups but significantly correlated with the amplitude of rhythm in the non-CT group. These results may reflect differences in sleep-wake characteristics across treatment-defined groups and could partly explain the heterogeneity observed across previous studies investigating sleep disturbances in breast cancer survivors. However, these profiles should not be interpreted as evidence of treatment-specific effects and require confirmatory studies.

The profile of patients that did not receive CT (non-CT group) here reflects alterations in the RAR, with significantly lower robustness of the rhythm, compared to the CT group. This profile is in line with previous animal studies reporting RAR alterations following doxorubicin administration. Results showed reduced both running wheel activity and time spent using the running wheel for over 5 weeks after treatment, suggesting persistent reduced daytime activity—and thus RAR alteration—following CT.⁴⁷

The profile of a lower number of awakenings–although longer in duration—combined with shorter sleep onset latency in the group of patients having received CT may reflect increased sleep propensity in this group, although alternative explanations remain possible. In support, previous results showed a disturbed homeostatic process (ie, sleep pressure) following breast cancer and its treatments both in animal models⁴⁸ and in patients.⁴⁹ One may hypothesize that this higher sleep propensity could be due to either higher fatigue or possibly higher sleepiness. Unfortunately, the current study did not evaluate sleepiness; we thus can only suggest that these results could be in line with distinct fatigue/sleepiness profiles across treatment-defined groups. Furthermore, both concepts are often mixed in the oncological population and even within clinical assessment tools themselves. For instance, widely used fatigue scales frequently include items directly addressing sleep propensity (eg, “Did you feel sleepy during the day?”), despite the fact that both terms refer to different physiological mechanisms. Fatigue is a subjective sense of physical, emotional, and/or cognitive tiredness or exhaustion related to cancer and/or cancer treatment that is not proportional to recent activity and interferes with usual functioning⁵⁰ while sleepiness reflects sleep propensity and difficulty staying awake while being inactive. While fatigue induces limitation in activities, sleepiness is associated with higher need for naps with greater propensity falling asleep. Given that the CT group had shorter sleep-onset latency and lower number of awakenings at night than the non-CT group, and considering the definition of fatigue and sleepiness, we can hypothesize about greater daytime sleepiness in the CT group, however, this hypothesis cannot be confirmed here and requires to be further tested using dedicated questionnaires (eg, Epworth Sleepiness Scale).

Exploratory correlation analyses performed in the non-CT group revealed a significant association between fatigue and the RAR amplitude, with lower amplitude being associated with higher fatigue levels (Figure 2). This result raises the hypothesis of a relationship between RAR alterations and cancer-related fatigue in breast cancer survivors who did not receive CT, expanding previous reports in patients treated with CT.¹⁰ Such significant association between fatigue and RAR amplitude in the non-CT group should be interpreted with caution given the limited sample size of this subgroup. Although this finding is consistent with previous evidence linking RAR disruption and cancer-related fatigue, it remains hypothesis-generating, preclude from causal relationship and requires replication in larger cohorts specifically designed to investigate these relationships while accounting for potential confounding factors.

Previous reports have shown larger feelings of fatigue both during and following treatments for breast cancer. The duration of ET appears as a significant factor leading to greater fatigue⁵¹ and cumulative effects of CT and ET were associated with early greater fatigue (ie, 3- and 6-months post baseline evaluation) but not later.³² In the current study, in contrast to our hypothesis, we observed no significant differences in fatigue levels between patient groups.³² This lack of difference in fatigue between patient groups may be related to the time of evaluation following treatments which may have limited our ability to distinguish effects of CT from those of other treatments on fatigue. In the current study, evaluations were done respectively 5 and 9 months in mean post-RT for the CT and the non-CT groups, and at 6 months post-CT for the CT group. This timeline may have prevented from highlighting severe fatigue due to either RT or CT in the current study. Seemly, given that ET had been initiated less than 1-year prior, the acute effects of CT had likely waned,³² while the duration of ET was arguably insufficient to induce severe fatigue.^51,52 Therefore, time since treatment is another potential source of heterogeneity and future studies should investigate the long-term effects of treatment exposure on fatigue, which may lead to higher levels of fatigue than in the current report.

The current report suffers from limitations and requires further investigation to be confirmed. First, several relevant variables are missing in this study, such as sleepiness, insomnia, chronotype, sleep diary, vasomotor symptoms, pain, anxiety/depression. Second, residual confounding cannot be entirely excluded, despite statistical adjustment for age and cancer stage. Although these variables were included as covariates, the groups remained markedly different in both age and cancer stage, and statistical adjustment may not fully account for their potential effects. With regard to age, current findings do not support an effect of age, as older individuals are typically more prone to longer sleep onset latency and more frequent awakenings—unlike the observations in the present sample. Regarding cancer stage, higher stages are often associated with more intensive treatments, greater psychological stress, distinct medical trajectories, and an increased risk of insomnia symptoms.⁵³ Importantly, treatment exposure, time since end of treatment, cohort, assessment device, and fatigue questionnaire were structurally confounded in the present study, and these factors cannot be statistically disentangled. Consequently, the observed associations between treatment-defined groups should be interpreted as hypothesis-generating rather than as treatment-specific effects. Replication in prospective cohorts, using standardized questionnaires and identical actigraphy devices, will be necessary to determine whether these distinct profiles truly reflect biological differences induced by treatment combinations. For instance, fewer but longer awakenings may stem from actigraph algorithm limitations, detection threshold issues, or differences between micro-awakenings and prolonged awakenings, distinctions that could not be resolved in this study.

Conclusion

Overall, this exploratory study suggests that different treatment-defined group may show different sleep-wake profiles rather than a single pattern of sleep disturbance. In the current dataset, women previously exposed to CT exhibit alterations predominantly affecting nocturnal sleep continuity in contrast to women treated with RT and ET alone. Importantly, these observations are hypothesis-generating and require confirmation in larger prospective studies using standardized assessments, as they describe associations observed in this specific dataset and should not be interpreted as evidence that treatment exposure itself explains the observed sleep-wake differences.

Author contributions

Alicia Leclercq (Data curation, Formal analysis, Investigation, Resources, Validation, Writing—original draft, Writing—review & editing), Abderrahmane Rahmani (Funding acquisition, Supervision, Writing—review & editing), Baptiste Morel (Funding acquisition, Writing—review & editing), Antoine Chatrenet (Investigation), Hugues Bourgeois (Investigation, Writing—review & editing), Oana Cojocarasu (Investigation, Writing—review & editing), Clara Mathie (Investigation, Writing—review & editing), Caroline Fonsegrive (Investigation, Writing—review & editing), François Gernier (Investigation, Writing—review & editing), Marie Fernette (Investigation, Writing—review & editing), Jean-Michel Grellard (Project administration, Writing—review & editing), Florence Joly (Funding acquisition, Supervision, Writing—review & editing), Mylène Duivon (Data curation, Investigation, Writing—review & editing), Bénédicte Giffard (Funding acquisition, Supervision, Validation, Writing—review & editing), Joy Perrier (Data curation, Formal analysis, Funding acquisition, Supervision, Validation, Writing—original draft, Writing—review & editing), and Tristan Martin (Formal analysis, Investigation, Writing—original draft, Writing—review & editing)

Funding

BIOCARE FActory study was funded by the “Région Pays de la Loire” and “La Ligue Contre Le Cancer.” PROSOMK study was supported by the Fondation ARC pour la recherche sur le cancer (2017–2020), the French Sleep Society (SFRMS), the Région Normandie (Réseaux d’Intérêts Normands, RIN), the Cancéropôle Nord-Ouest, and the Ligue contre le Cancer—Normandie.

Conflicts of interest

None declared.

Data availability

Data will be provided upon reasonable request to the corresponding author.