Journal Pre-proof

A bio-behavioral model of systemic inflammation at breast cancer diagnosis and

fatigue of clinical importance two years later

A. Di Meglio, J. Havas, M. Pagliuca, M.A. Franzoi, D. Soldato, C.K. Chiodi, E.

Gillanders, F. Dubuisson, V. Camara-Clayette, B. Pistilli, J. Ribeiro, F. Joly, P.H.

Cottu, O. Tredan, A. Bertaut, P.A. Ganz, J. Bower, A.H. Partridge, A.L. Martin, S.

Everhard, S. Boyault, S. Brutin, F. André, S. Michiels, C. Pradon, I. Vaz-Luis

PII: S0923-7534(24)01517-5

DOI: https://doi.org/10.1016/j.annonc.2024.07.728

Reference: ANNONC 1571

To appear in: Annals of Oncology

Received Date: 3 December 2023

Revised Date: 25 July 2024

Accepted Date: 29 July 2024

Please cite this article as: Di Meglio A, Havas J, Pagliuca M, Franzoi MA, Soldato D, Chiodi CK,

Gillanders E, Dubuisson F, Camara-Clayette V, Pistilli B, Ribeiro J, Joly F, Cottu PH, Tredan O, Bertaut

A, Ganz PA, Bower J, Partridge AH, Martin AL, Everhard S, Boyault S, Brutin S, André F, Michiels S,

Pradon C, Vaz-Luis I, A bio-behavioral model of systemic inflammation at breast cancer diagnosis and

fatigue of clinical importance two years later, Annals of Oncology (2024), doi: https://doi.org/10.1016/

j.annonc.2024.07.728.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition

of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of

record. This version will undergo additional copyediting, typesetting and review before it is published

in its final form, but we are providing this version to give early visibility of the article. Please note that,

during the production process, errors may be discovered which could affect the content, and all legal

disclaimers that apply to the journal pertain.

© 2024 The Author(s). Published by Elsevier Ltd on behalf of European Society for Medical Oncology.

1

ARTICLE TYPE: Original article

TITLE: A bio-behavioral model of systemic inflammation at breast cancer diagnosis and

fatigue of clinical importance two years later

Authors: A. Di Meglio1,*, J. Havas1

, M. Pagliuca1,2

, M. A. Franzoi1

, D. Soldato1

, C. K. Chiodi1

,

E. Gillanders1

, F. Dubuisson3

, V. Camara-Clayette4

, B. Pistilli5

, J. Ribeiro5

, F. Joly6

, P. H.

Cottu7

, O. Tredan8

, A. Bertaut9

, P. A. Ganz10, J. Bower10, A. H. Partridge11

, A. L. Martin12, S.

Everhard12, S. Boyault9

, S. Brutin4

, F. André1

, S. Michiels13, C. Pradon3,4, I. Vaz-Luis1,14

Affiliations:

1 Cancer Survivorship Program, INSERM U981, Gustave Roussy, Villejuif, France

2 Division of Breast Medical Oncology, Istituto Nazionale Tumori IRCCS “Fondazione G.

Pascale”, Naples, Italy

3 Department of Medical Biology and Pathology, Gustave Roussy, Villejuif, France

4 Biological Resource Center, AMMICa, INSERM US23/CNRS UMS3655, Gustave Roussy,

Villejuif, France

5 Medical Oncology Department, INSERM U981, Gustave Roussy, Villejuif, France

6 Centre Francois Baclesse, University UniCaen, Anticipe U1086 Inserm, Caen, France

7 Institut Curie, Paris, France

8 Centre Léon Bérard, Lyon, France

9 Centre Georges François Leclerc, Dijon, France

10 University of California, Los Angeles, CA, United States

11 Dana-Farber Cancer Institute, Boston, MA, United States

12 Unicancer, Paris, France

13 Service de Biostatistique et d’Epidémiologie, Gustave Roussy, Villejuif, France; Oncostat

U1018, Inserm, University Paris-Saclay, labeled « Ligue Contre le Cancer », France

Journal Pre-proof

2

14 Interdisciplinary Department for the Organization of Patient Pathways (DIOPP), Gustave

Roussy, Villejuif, France

*CORRESPONDING AUTHOR:

Name: Dr. Antonio Di Meglio

Address: Gustave Roussy, 114 Rue Edouard Vaillant, 94800, Villejuif, France

Email: antonio.di-meglio@gustaveroussy.fr

Phone: +33 (0)142114211

X (Twitter) handle: @dimeglio_anto Journal Pre-proof

3

HIGHLIGHTS

• • 34.4% of 1208 stage I-III breast cancer survivors had cancer-related fatigue of

clinical importance 2 years after diagnosis

• • High pre-treatment levels of the pro-inflammatory cytokine IL-6 were associated

with global fatigue 2 years later

• • Individuals with high levels of IL-6 had higher body mass index and were less

physically active than those with lower levels

• • Higher pre-treatment IL-2 and IL-10 were also associated with higher and lower

likelihood of global fatigue, respectively

• Higher C-reactive protein was associated with higher likelihood of cognitive

fatigue Journal Pre-proof

4

ABSTRACT

Background: We aimed to generate a model of cancer-related fatigue (CRF) of clinical

importance two years after diagnosis of breast cancer building on clinical and behavioral

factors and integrating pre-treatment markers of systemic inflammation.

Methods: Women with stage I-III HR+/HER2- breast cancer were included from the

multimodal, prospective CANTO cohort (NCT01993498). The primary outcome was global

CRF of clinical importance (EORTC QLQ-C30≥40/100) two years after diagnosis (year-2).

Secondary outcomes included physical, emotional, and cognitive CRF (EORTC QLQ-FA12).

All pre-treatment candidate variables were assessed at diagnosis, including inflammatory

markers (interleukin [IL]-1a, IL-1b, IL-2, IL-4, IL-6, IL-8, IL-10, interferon gamma, IL-1 receptor

antagonist, TNF-, and C-reactive protein), and were tested in multivariable logistic regression

models implementing multiple imputation and validation by 100-fold bootstrap resampling.

Results: Among 1208 patients, 415 (34.4%) reported global CRF of clinical importance at

year-2. High pre-treatment levels of IL-6 (Quartile 4 vs.1) were associated with global CRF at

year-2 (adjusted Odds Ratio [aOR]: 2.06 [95% Confidence Interval 1.40-3.03]; p=0.0002;

AUC=0.74). Patients with high pre-treatment IL-6 had unhealthier behaviors, including being

frequently either overweight or obese (62.4%; mean BMI 28.0 [SD 6.3] Kg/m2

) and physically

inactive (53.5% did not meet WHO recommendations). Clinical and behavioral associations

with CRF at year-2 included pre-treatment CRF (aOR vs no: 3.99 [2.81-5.66]), younger age

(per 1-year decrement: 1.02 [1.01-1.03]), current smoking (vs never: 1.81 [1.26-2.58]), and

worse insomnia or pain (per 10-unit increment: 1.08 [1.04-1.13], and 1.12 [1.04-1.21],

respectively). Secondary analyses indicated additional associations of IL-2 (aOR per log-unit

increment:1.32 [CI 1.03-1.70]) and IL-10 (0.73 [0.57-0.93]) with global CRF and of C-reactive

protein (1.42 [1.13-1.78]) with cognitive CRF at year-2. Emotional distress was consistently

associated with physical, emotional, and cognitive CRF.

Conclusions: This study proposes a bio-behavioral framework linking pre-treatment systemic

inflammation with CRF of clinical importance two years later among a large prospective

sample of survivors of breast cancer.

Journal Pre-proof

5

KEYWORDS: breast cancer, cancer-related fatigue, inflammatory markers, health behaviors,

symptom management, survivorship

Journal Pre-proof

6

MANUSCRIPT

Introduction

Cancer-related fatigue (CRF) is almost universal during primary treatment for breast

cancer, however long-term CRF can be prevalent up to several years after treatment

completion.1–3 CRF is usually defined as more disabling than fatigue due to sleep deprivation

or overexertion, with a substantial proportion of patients (30-60%) describing this symptom as

moderate to severe, causing great detriment to quality of life (QOL).4–6

Extensive work has studied clinical, psychological, behavioral, environmental, sociodemographic, tumor-, and treatment-related correlates of CRF.1,7–11 Dissecting its

multidimensional characteristics, several studies have also suggested that the physical,

emotional, and cognitive manifestations of CRF may have distinct etiology, courses, and

determinants, including different pathophysiology, and a “subtype-specific” approach to better

tailor the management of CRF was proposed.12–14 Joining this effort to improve assessment

and management of CRF, our group previously developed and validated predictive models

building on clinical and behavioral characteristics15 and generated an online screening tool to

estimate individual risk of long-term CRF after breast cancer.16

Nevertheless, CRF remains a complex and multifactorial syndrome, and most of its

mechanisms and biological underpinnings are still elusive.1,9 Multiple pathways have been

studied in relation to CRF. Inflammation, dysregulation of the hypothalamic–pituitary–adrenal

axis, and/or activation of the autonomic nervous system have traditionally been advocated as

potential mechanisms of CRF, being able to influence each other and to activate additional

systems such as oxidative stress cascades, endocannabinoids, and gut microbiota.1,17 Over

the past few years, the link between cancer-related inflammation and CRF received the

greatest empirical attention. It has been suggested that CRF and other “sickness behaviors”,

including emotional distress, cognitive dysfunction, pain, and insomnia, may stem from central

stimulation resulting from peripheral activation of the inflammatory axis and the production of

pro-inflammatory cytokines.18–20 Carrol and colleagues also suggested that variation in

Journal Pre-proof

7

individual susceptibility to symptoms including CRF may be substantially driven by cellular

senescence and accelerated ageing resulting from tumor-host-treatment interactions.21 Preexisting factors including psychosocial and lifestyle traits, with mediation exerted by cancer

treatment, may influence inflammatory pathways, and facilitate the accumulation of cells

enriched with an inflammation-biased secretome.22–26 Particularly, several interleukins (IL;

such as IL-1 and IL-6), C-reactive protein, and tumor necrosis factor-alpha (TNF-) seemed

to have implications in orchestrating local and systemic effects leading to a wide spectrum of

host defense responses on energy levels.1,17 By signaling across the blood-brain interface,

increased inflammation can then impair several bodily systems, being a shared biological

substrate across oncologic, cardiovascular, and metabolic comorbidities.21 It is also plausible

that the effect of initial systemic inflammation keeps manifesting on CRF for years after cancer

diagnosis, as a combined result of physiological and accelerated aging, accumulation of

greater comorbidity, disrupted compensatory capacities, long-term treatment burden, and

perpetuating factors such as poor diet, physical inactivity and sleep disturbance.4,27–29

Taken together, previous research has generated evidence about inflammation,

neuroimmune interactions, and immune mechanisms for CRF. However, this evidence is not

always consistent and often limited by the cross-sectional nature and small sample size of

studies that are therefore sensitive to a number of biases.17,30 In the present analysis, we

evaluated the contribution of pre-treatment markers of systemic inflammation or inflammatory

axis activation (i.e., assessed at breast cancer diagnosis, before any treatment for breast

cancer) to models of CRF considered of clinical importance31,32 two years later, using a large,

prospective cohort of survivors of early-stage breast cancer. Our study builds on previous

knowledge suggesting an inflammatory basis for CRF and on existing evidence of the interplay

between health behaviors and CRF, also aiming to find potential interventional targets

PATIENTS AND METHODS

Study design

Journal Pre-proof

8

We used CANTO (CANcer TOxicity; ClinicalTrials.gov identifier: NCT01993498), a

large, prospective cohort of survivors of stage I-III breast cancer. Briefly, clinical, socioeconomic, behavioral, tumor and treatment characteristics, and patient-reported outcome data

are collected at diagnosis of breast cancer (pre-treatment, i.e., before any primary breast

cancer treatment including surgery, chemo-, or radiotherapy, as appropriate), then

longitudinally reassessed at approximately one, two, four, and six years after diagnosis.

Endocrine therapy and targeted therapies are allowed to be ongoing during the follow-up time

points. Patients experiencing disease recurrence, metastatic relapse, or death, exit the study

and do not contribute to the analyses from the event date forward. All patients provided written

informed consent. The study design was previously described (Ethics committee approval: IDRCB:2011-A01095-36,11-039).33

Cohort definition

This analysis included patients with hormone-receptor (HR)-positive, human epidermal

growth factor receptor (HER) 2-negative breast cancer diagnosed from 2012-2013, who had

a pre-treatment blood sample available for quantification of inflammatory markers. The final

analytic cohort included patients providing data on CRF at the year-2 time point (N= 1208 for

the primary outcome; the full study flowchart is presented in Supplementary Figure 1).

Outcome assessment

Our primary outcome of interest was global CRF at year-2 after diagnosis, assessed

using the three-item scale of the European Organisation for Research and Treatment of

Cancer (EORTC) Quality of Life Questionnaire (QLQ)-C30. A higher score on this scale

indicates a higher level of symptomatology. Scores were dichotomized using a threshold of

≥40/100, typically defining clinically important CRF.31,32

In addition, we assessed the physical, emotional, and cognitive dimensions of CRF at

year-2 after diagnosis as secondary outcomes, using the EORTC module measuring CRF

Journal Pre-proof

9

(EORTC QLQ-FA12), a multidimensional instrument to be used in conjunction with the core

EORTC QLQ-C30.34 The questionnaire includes five items for physical (Item 1-5), three for

emotional (Item 6-8), and two for cognitive CRF (Item 9-10). While there is no threshold for

clinical importance for the QLQ-FA12 scores, a cut-off value of 40/100 was used to identify

CRF dimensional symptoms of clinical relevance, consistent with a previous study.15

Additional information on the outcome assessment instruments and score calculation

is available in the Supplementary Methods.

Clinical and behavioral variables of interest

Based on clinical expertise and prior evidence of association with CRF1,7–11, we

included clinical, socio-demographic, behavioral, tumor- and treatment-related factors, and

symptoms (including pre-treatment fatigue) in our analyses. Variables were defined and

categorized as described in Table 1. These variables were collected at study entry (breast

cancer diagnosis, equivalent to pre-treatment in the CANTO study) during dedicated visits with

trained study nurses as per study protocol, and included the following: age at diagnosis of

breast cancer, Body Mass Index (BMI; objectively assessed during clinical study visits),

menopausal status, comorbidities (Charlson comorbidity index), previous mental health

problems, marital status, education, and income (ad hoc socio-economic questionnaire),

alcohol consumption, tobacco use, physical activity (Global Physical Activity Questionnaire16)35, breast cancer stage, axillary and breast surgery, receipt of chemotherapy, radiotherapy,

hormonal therapy, anxiety and depression (Hospital Anxiety and Depression Scale [HADS;

non-case, score 0-7; doubtful case, score 8-10; case, score 11-21])36, CRF, insomnia, pain

(EORTC QLQ-C30)37, and menopausal symptoms (i.e., hot flashes; Common Terminology

Criteria for Adverse Events [CTCAE v 4.0, Yes= any grade]).38

Biological variables of interest: serum inflammatory markers

Journal Pre-proof

10

Fasting blood samples were obtained pre-treatment, kept at room temperature for 30-

60 minutes, then centrifuged for 15 minutes at 1800g to obtain 200 to 400 uL of serum, which

was stored in two aliquots at -180°C for long-term preservation until analyzed.

The High Sensitivity Cytokine Custom array (CTK CST X, EV3881/EV3623), the

Metabolic Syndrome array I (METS I, EV3755), and the Metabolic Syndrome array II (METS

II, EV3759/A) were used for the quantification of the following serum inflammatory markers:

IL-1, IL-1, IL-2, IL-4, IL-8, IL-10, interferon gamma (IFN), IL-1 receptor antagonist (IL-1Ra)

(CTK), IL-6, TNF- (METS I), and C-reactive protein (METS II) (RANDOX Laboratories

Limited, UK). Previous studies suggested the implication of these markers in inflammatory

responses linked with CRF and other behavioral symptoms.1,13,17 All samples from a single

subject were assayed together on the same ELISA plate to minimize effects of inter-assay

variation, with internal quality controls applied twice, once at the beginning and once at the

end of each run, using the RANDOX Evidence Investigator™ Biochip Array technology.39

For the primary analysis focused on global CRF at year-2 (EORTC QLQ-C30),

concentrations were dichotomized as “low” vs “high” according to the lower limit of

quantification for the individual assay (sensitivity threshold). If the sensitivity threshold was

sufficiently low (i.e., ≤15th percentile), categories were defined according to the quartile (Q)

distribution as “low” (Q1), “middle low” (Q2), “middle high” (Q3), and “high” (Q4), to allow for a

more granular quantification. Similar approaches were previously used.40 C-reactive protein

values were categorized according to levels with clinical meaning (Normal/low [<1 mg/L],

moderately elevated [1 to <3 mg/L], and high [≥3mg/L]).41 Categorization is shown in Table 2.

Secondary analyses were performed using continuous log-transformed values of the

markers and including the ratio IL1Ra/IL-6, both for the primary outcome of global CRF

(EORTC QLQ-C30) and for outcomes of dimensions of CRF (EORTC QLQ-FA12).

Continuous values of the markers are available in Supplementary Table 1 and 2.

Statistical analysis

Journal Pre-proof

11

Cohort and outcome description. Descriptive statistics summarized distribution of

clinical and behavioral characteristics, and serum concentrations of inflammatory markers at

breast cancer diagnosis, as well as outcome distribution at year-2 after diagnosis.

Selection of factors to be included in the multivariable model. To assess clinical,

behavioral, and biological factors that could explain the variability in CRF of clinical importance

at year-2, we used multivariable logistic regression implementing automatic backward variable

selection under multiple imputation (MI), combined with bootstrapping, as described by

Heymans MW et al, to account both for variation linked to missing data and sampling.42 Missing

covariate data were handled generating 15 complete data replicates by MICE (Multivariate

Imputation by Chained Equations).43 The imputation model included (a) all variables that would

be part of the subsequent analytic model (as listed in Table 1, 2, and 3); (b) the outcome

variables (of note, missing outcome variables were not imputed, outcome data was used to

impute missing values in other covariates); and (c) auxiliary variables, included to help

minimize bias and improve precision of the estimates. These variables were identified using

domain knowledge or through association with incomplete variables, and included

menopausal status, health behaviors (BMI, physical activity, smoke, alcohol), presence of

concomitant medical conditions (i.e., history of previous cardio-circulatory, respiratory,

gastrointestinal, renal, hepatic, endocrine, muscle-articular, urologic, hematologic,

dermatologic, neurologic, allergic, gynecological disease), socioeconomic variables (marital

status, professional status, level of education), household income, and patient-reported health

(EORTC QLQ-C30 domains, continuous scores). For each replicate, we constructed 100

bootstrap sets by randomly drawing with replacement, therefore the total number of data sets

equaled 1500 (15-MI*100-bootstrap). For the primary analysis focused on global CRF

(EORTC QLQ-C30), we used strict selection criteria to define factors for the final multivariable

model, including categorical inflammatory markers. First, the automatic stepwise method used

a p-value cut-off of ≥0.05 to remove variables. Then, we calculated the proportion of times

each variable appeared in the models (i.e., the inclusion frequency) and retained in the final

Journal Pre-proof

12

multivariable model only variables for which the inclusion frequency exceeded 70%. These

criteria allow to exploit the wealth of data from the CANTO study without overfitting the models

and minimize the risk of including noise variables.42 Results obtained across imputed data sets

were combined using Rubin’s rules to produce Odds Ratio (OR) estimates and Confidence

Intervals (CIs) that incorporate uncertainty of imputed values.44

For secondary analyses, focused both on global CRF (EORTC QLQ-C30) and

dimensions of CRF (EORTC QLQ-FA12), we incorporated continuous log-transformed values

of the markers using a more lenient inclusion frequency of 50%, while keeping the probability

to remove variables strictly set at ≥0.05 in the stepwise procedure. This strategy would allow

us to obtain a less parsimonious model while limiting the inclusion of an excessive number of

non-informative variables.42,45

Model performance and bootstrap validation. The discriminative ability of the model

was assessed by c-index, equal to the Area Under the Receiver Operating Characteristic

Curve (AUC). Calibration was visually explored by plotting the observed and estimated

probabilities of clinically important CRF. The optimism-corrected AUC and calibration were

obtained by bootstrap.46

Sensitivity analyses. We aimed at testing whether different definitions, categorizations,

and selection criteria for variables of interest would impact modelling findings. Therefore,

several sets of sensitivity analysis were conducted for the primary outcome and additional

models were fit i) using pain and insomnia as categorical variables (i.e., dichotomizing

according to thresholds of clinical importance: a cut-off of 50/100 and 25/100 defined clinically

important insomnia and pain, respectively31,32; and ii) irrespective of statistical variable

selection (to correct for the potential confounding effect of age, health behaviors such as BMI,

tobacco smoke, and physical activity, and anxiety and depression). Finally, acknowledging the

importance of emotional distress in relation with CRF, iii) we fit models not including pretreatment CRF of clinical importance and including anxiety and depression either as (a)

categorical (HADS standard cut-off definitions) or (b) continuous variables.

Journal Pre-proof

13

Power considerations. With a binary outcome prevalence of approximately 35%, a

minimal sample size of 1022 patients was needed to minimize overfitting (expected shrinkage

of 10% or lower) and to ensure precise estimation of key parameters in a model with 15

variables (including an absolute difference of 0.05 in the model apparent and adjusted Rsquared value).47

Statistical analysis was performed using SAS statistical software Version 9.4.

Statistical significance was defined with a two-sided p-value<0.05.

RESULTS

Cohort characteristics

In the overall cohort, mean age was 57.9 years (Standard Deviation [SD] 11.1), 191

patients (16.3%) had a history of previous mental health problems (i.e., mostly anxiety–

depressive disorders), 253 (21.2%) reported pre-treatment CRF of clinical importance, 425

(35.6%) and 82 (6.9%) had clinically suggestive symptoms of anxiety and depression at

diagnosis (cases), respectively. In addition, mean BMI was 25.9 Kg/m2 (SD 5.2), 204 patients

(17.3%) reported current smoking at diagnosis, and median total physical activity level was

14.0 Metabolic-equivalent of task-hour (MET-h)/week; Q1-Q3 0.0-40.0). Five hundred and

thirty three patients (44.1%) received (neo)adjuvant chemotherapy and 1109 (91.8%) were

treated with adjuvant hormonal therapy (Table 1).

Bio-behavioral model of global CRF at year-2

Four hundred fifteen patients (34.4%) reported CRF of clinical importance at year-2

post diagnosis of breast cancer.

The main model of CRF included a combination of clinical, behavioral, and biological

characteristics (inclusion frequencies are presented in Supplementary Table 3). IL-6 was

the only inflammatory marker selected with strict inclusion frequency >70% and therefore

tested in the main multivariable model of CRF at year-2. A larger proportion of patients among

those that reported CRF of clinical importance at year-2 had high levels of pre-treatment IL-6

Journal Pre-proof

14

compared to those that did not (30.8% and 21.9% respectively; Table 2), and this difference

was consistent across age groups and menopausal status (not shown). After correction for

clinical and behavioral characteristics, high pre-treatment IL-6 levels were associated with

CRF of clinical importance at year-2 (adjusted OR vs low IL-6: 2.06 [95% CI 1.40-3.03];

p=0.0002). The model AUC was 0.75 (95% CI 0.72-0.78; Optimism-corrected AUC=0.74;

Table 3; Supplementary Figure 2). Patients with high pre-treatment levels of IL-6 had

unhealthier behaviors compared to those with lower levels, including that they were frequently

either overweight or obese (62.4%; mean BMI 28.0 [SD 6.3] Kg/m2

), did not meet WHO

physical activity recommendations (53.5% reported <10 MET-hours/week; median time: total

activity 8.0 [Q1-Q3 0.0 to 28.0] MET-hours/week, transport and leisure-time activity 4.0 [0.0 to

18.3]), and several among them (20.1%) were current smokers (Supplementary Figure 3,

additional clinical characteristics are presented in Supplementary Table 4).

Pre-treatment clinical and behavioral factors that were associated with CRF of clinical

importance at year-2 in the main model included reporting pre-treatment fatigue of clinical

importance (OR vs no: 3.99 [95% CI 2.81-5.66]), younger age (per 1-year decrement: 1.02

[1.01-1.03]), being a current smoker (vs never: 1.81 [1.26-2.58]), and CRF-associated

symptom burden at diagnosis including worse insomnia (per 10-unit increment: 1.08 [1.04-

1.13]) and pain (per 10-unit increment: 1.12 [1.04-1.21]). Treatment-related factors and pretreatment physical activity were not retained in the model of global CRF.

Models using inflammatory markers as log-transformed continuous variables identified

several associations of CRF of clinical importance at year-2, including with pre-treatment IL-6

(OR per log-unit increment: 1.33 [1.11-1.60]), IL-2 (1.32 [95% CI 1.03-1.70]), and IL-10 (0.73

[95% CI 0.57-0.93]). Additional associations with clinical and behavioral factors were

consistent with the main model. Inclusion frequencies for this model are presented in

Supplementary Table 5, the full model is presented in Supplementary Table 6.

Models of CRF dimensions at year-2: physical, emotional, and cognitive CRF

Journal Pre-proof

15

At year-2 after diagnosis, 426 (35.1%), 268 (22.1%), and 165 patients (13.6%) reported

physical, emotional, and cognitive CRF of clinical importance, respectively. A descriptive

summary of clinical, behavioral and biological characteristics is provided in Supplementary

Table 7 and 8.

Focusing on the contribution of inflammatory markers to models of CRF dimensions,

significant associations were observed between increasing levels of C-reactive protein and

cognitive CRF (OR per log-unit increment in C-reactive protein 1.42 [95%CI 1.13-1.78]),

whereas there were no specific associations with physical or emotional CRF. Additional

relevant clinical and behavioral factors emerged from these models. Among these, pretreatment, clinically suggestive symptoms of depression were significantly associated with the

three dimensions of physical (OR case vs. normal 2.19 [95% CI 1.25-3.84]), emotional (OR

case vs. normal 1.97 [95% CI 1.13-3.42]), and cognitive CRF of clinical importance (OR case

vs. normal 1.92 [95% CI 1.05-3.49]), after multivariable correction that also included pretreatment CRF. In addition, pre-treatment, clinically suggestive symptoms of anxiety were

significantly associated with emotional CRF (OR case vs. normal 1.91 [95% CI 1.26-2.90]).

Finally, previous mental health problems were associated with physical (OR vs. no 1.72 [95%

CI 1.22-2.43]) and cognitive (OR vs. no 1.66 [95% CI 1.07-2.57]) CRF. Mastectomy was

associated with physical CRF (OR vs. conservative surgery 1.38 [95%CI 1.03-1.85). Full

models are presented in Supplementary Table 9.

Sensitivity analyses

Models indicated consistent associations between higher pre-treatment levels of IL-6

and CRF of clinical importance at year-2 across all sensitivity analyses, namely i) using

insomnia and pain as categorical variables (Supplementary Table 10); and ii) fitting a model

by including age, health behaviors, and emotional distress without variable selection

(Supplementary Table 11). Finally, iii) in a last set of sensitivity models not including pretreatment CRF, associations emerged also between pre-treatment depression and global CRF

of clinical importance at year-2 (OR for clinically suggestive case vs. non-case 2.06 [95% CI

Journal Pre-proof

16

1.24-3.43]; OR per unit increment in continuous depression score 1.07 [95% CI 1.03-1.11];

full models not shown).

DISCUSSION

This large, prospective study of women with HR+/HER2- breast cancer adds a

biological dimension focused on pre-treatment inflammation to clinical and behavioral models

of CRF. We found consistently increased likelihood of reporting global CRF of clinical

importance two years after breast cancer diagnosis among women with high pre-treatment

levels of IL-6, compared to those with low levels. Capitalizing on the role of behavioral factors,

we also report that pre-treatment excess weight and physical inactivity were prevalent among

women with high circulating levels of inflammatory marker. Additional biological associations

emerging from this study included those of pre-treatment IL-2 and IL-10 with global CRF as

well as of C-reactive protein with cognitive CRF. Clinical and behavioral factors associated

with global CRF included pre-treatment fatigue, younger age, current smoking, and worse

insomnia or pain. Models of physical, emotional, and cognitive dimensions of CRF were

significantly informed by metrics of pre-treatment emotional distress.

There is convergence between our findings and those from previous studies supporting

an inflammatory basis for CRF and other behavioral symptoms among cancer patients and

suggesting that perturbation of immune system homeostasis may help structure elevations in

circulating inflammatory markers in survivors with persistent CRF.1,20,48 Such alterations may

manifest in multiple ways, including reactivation of latent infections, deregulation of

glucocorticoid signaling and alterations in lymphocyte subsets.1,17 Several authors reported, at

various points across the survivorship trajectory, associations between pro-inflammatory

cytokine activity with CRF, especially in the post-treatment period. Our study validates and

extends the findings of many smaller samples. Noteworthy, high levels of IL-6 across the

distribution were associated with CRF in our main models, and additional identified markers

included C-reactive protein, IL-10, and IL-2. While previous studies usually assessed panels

Journal Pre-proof

17

of inflammatory markers17, several reported associations between IL-6

49–52 and/or C-reactive

protein30,53–55 with CRF. IL-6 is a pleiotropic cytokine produced by a variety of cell types,

involved in antigen-specific immune responses and inflammatory responses, mediation of

acute phase reactions, and in host-defense mechanisms56–58, whereas C-reactive protein is a

sensitive downstream indicator of inflammation, and its concentrations have clinical meaning

in several settings.59,60 Some evidence also indicated associations of CRF with IL-1061

,

suggested to increase in response to pro-inflammatory cytokines and to exert an antiinflammatory effect by modulating their levels.62 This is consistent with our findings indicating

a reverse relationship between higher levels of IL-10 and lower likelihood of CRF. Finally, the

association between the pro-inflammatory cytokine IL-2 and CRF was observed less

frequently in previous studies, and its increase in fatigued survivors could be put in relation

with elevation of multiple cytokines involved in the inflammatory process rather than of a single

marker.17,63 Previous literature has also examined the IL-6 gene single nucleotide

polymorphisms (SNPs) and CRF. Associations with CRF were described before, during, and

after cancer treatment, extending also to SNPs in genes encoding IL-1, IL-10, and TNF-a, and

expression of Type I interferon genes.64–68 Furthermore, evidence is available linking

circulating levels of IL-6 and its genetic polymorphisms with other behavioral symptoms such

as depression and memory problems, which are common in severely fatigued survivors.69

While inflammation is among the most discussed mechanisms of fatigue, including in

the non-cancer realm (e.g., for fatigue associated with autoimmune, neurological, and

musculoskeletal diseases)70, scenarios have been proposed where its contribution is less

clear. Results of several studies, either with longitudinal or cross-sectional designs, were

controversial or failed to demonstrate a relationship between inflammatory markers and CRF

independently of other clinical, behavioral or socio-demographic risk factors.71,72 Of note, these

studies are heterogeneous in terms of design, analytic methods, and CRF outcomes,

assessment instruments, and time-points.17 Cluster analyses may help further dissect

subgroups of CRF driven by concomitant clinical conditions (e.g., psychological disturbance,

Journal Pre-proof

18

sleep dysregulation and pain, obesity), from those associated with inflammatory or different

biological patterns. By looking at a post-diagnosis survivors population, Schmidt et al.13

delineated different subgroups of CRF, including one occurring along a history of depression,

a second associated with inflammation (i.e., with high levels of inflammatory markers including

IL-6, but also TNF-, IL-1, resistin, VEGF-A and GM-CSF), and a third characterized by

metabolic markers such as leptin. The inflammation-associated cluster also expressed high

levels of pain and high BMI, findings that echo the associations with other symptoms (e.g.,

insomnia, pain) and bio-behavioral factors (e.g. increased BMI, inactivity) observed in our

cohort. These authors have therefore proposed targeting fatigue phenotypes that are driven

by different mechanisms with specific interventions, modeling fatigue management upon

patient features and biological hallmarks of CRF subtypes.8,13 Nevertheless, the contributing

mechanisms and biological substrates of the different CRF dimensions still need further

elucidation.70 For example, there were higher manifestations of physical CRF in the cluster

associated with inflammation13, consistent with other studies.53,73,74 Conversely, in our cohort

with a different design, the only dimension that seemed associated with inflammation was

cognitive CRF, for which a relationship with pre-treatment C-reactive protein was evident. In

the CANTO-Cog sub-study focused on cognition, we also reported associations between high

levels of C-reactive protein assessed at diagnosis of breast cancer and overall cancer-related

cognitive impairment, processing speed and episodic memory impairments two years later.75

Others recently suggested longitudinal relationships between C-reactive protein and cognitive

complaints in older breast cancer survivors.76 Such findings can help contextualize the

associations between C-reactive protein and the feelings of “having trouble thinking clearly”

and “confusion”34 reported by survivors with elevated cognitive CRF in our cohort.

Development and validation of a model of clinical and behavioral risk factors for CRF

and related discussion was the focus of a previous manuscript15, however the associations

between emotional distress and CRF reported herein merit further discussion. Clustering of

emotional distress with fatigue was reported by numerous previous studies7,8,13,77–80, including

Journal Pre-proof

19

ours6,15, and CRF characterized by depressive symptoms/anxiety was suggested to manifest

with more severe and persisting physical, emotional, and cognitive symptoms compared to

CRF not linked with depression. In our main model, emotional distress variables were not

retained in association with global CRF two years later. This association might have been

masked by the inclusion of pre-treatment CRF of clinical importance in the models, as also

suggested by sensitivity analyses. However, pre-treatment CRF is one of the strongest and

most consistent predictors of post-treatment CRF and may set the stage for elevated CRF

years after treatment completion, therefore including this metric in a pre-treatment model is

particularly relevant. Some authors also suggested that most studies looking at predictors of

CRF did not control for pre-treatment CRF and therefore the independent contribution of

depression (and other clinical factors) above pre-existing CRF might be not entirely clear.1

In

general, models including a baseline measurement of the outcome yield better performance

and more accurate estimates.81–84 As opposed to models of global CRF, our models of the

physical, emotional, and cognitive dimensions of CRF had very consistent associations with

symptoms suggestive of anxiety or depression at the moment of diagnosis or with mental

health problems that pre-existed breast cancer diagnosis. These disorders may determine

poorer psychological adjustment to cancer and increase vulnerability to long-term CRF.8

Beyond impairing coping capacity, emotional vulnerability was also linked to increased

inflammatory responses to stress.19,85,86 Confirming what previously suggested, our findings

underpin the role of emotional distress in defining a CRF phenotype with multidimensional

manifestations (e.g., slowing down, having trouble getting things started, helplessness,

frustration, impaired thinking ability) rather than just general feelings of weakness and need to

rest.8

In terms of other clinical and behavioral factors, receipt of more extensive breast

surgery was previously associated with CRF87, a finding that we also report. It is important to

highlight that the timing of outcome assessment in our study may be responsible for the lack

of observed associations between CRF and other treatment-related factors. CRF was

Journal Pre-proof

20

assessed at least one year after primary treatment completion in our cohort. We previously

reported an early association of chemotherapy with global and physical CRF at approximately

3-6 months after treatment.15 Consistently, previous studies had pointed at chemotherapy as

a precipitating factor for CRF (i.e. arising during treatment), with weaker effects on long-term,

persistent CRF.7 On the contrary, several studies, including ours, had identified hormonal

therapy as a factor perpetuating CRF later during follow-up.15,27 Associations with hormonal

therapy might not be evident in the present HR+/HER2-cohort, as the overwhelming majority

received adjuvant hormonal therapy. Further research should look at interactions between

cancer treatment and inflammation on long-term symptoms, as these may not be manifested

until several years after initial treatment, including until longer term exposures to hormonal

therapy. In addition, we did not find associations between pre-treatment levels of physical

activity and CRF, which might partly be explained by the observational nature of the cohort

with self-reported activity and subject to over-reporting, and by the time elapsed between

exercise exposure and outcome assessment.

A relevant question is whether we can implement our -or similar- models to improve

patient stratification, targeted provision of behavioral interventions, and monitoring strategies

to prevent CRF deterioration in clinical practice. In addition to actionable clinical factors such

as baseline symptom burden, excess adiposity and inactivity were common among individuals

with increased markers of inflammation in our analysis, suggesting potential modifiable targets

for interventions. As previously indicated, an expanded, reprogrammed, and metabolically

active adipose tissue may alter systemic physiology of individuals with obesity, by enhancing

secretion of cytokines, including IL-6, and adipokines, and activate key pathways that trigger

or precipitate the CRF cascade.1,88 Analogously, dysregulation in systemic inflammation is

observed in sedentary individuals.1 The link between inflammation, behaviors, and CRF

symptoms is further reinforced by data from interventional research. Although there is still

uncertainty about whether a shared inflammatory substrate is consistently present or not,

several behavioral interventions, including those focused on exercise, psychological support,

Journal Pre-proof

21

mindfulness, and yoga, have shown a positive impact on CRF and the associated spectrum

of behavioral symptoms that include cognitive disturbance, emotional distress, social

withdrawal, and sleep troubles, suggesting a common biological basis.24–26,48,89–95 By acting on

modifiable host-specific factors such as sedentary behavior, but also on poor sleep,

psychosocial stress and catastrophizing, behavioral interventions can target inflammation,

disrupt its feed-forward pathways, and mitigate symptoms.96 It is evident that one-size-fits-all

approaches may not be sufficient facing the complexity and multidimensionality of CRF

subtypes and its entanglement with other behavioral symptoms. Our models are not intended

to point at a single (or few) diagnostic biomarker(s) of CRF. They rather provide a biobehavioral framework for thorough evaluation of potential risk factors for CRF since diagnosis,

including screening for the presence of pre-treatment CRF and past medical history with a

focus on previous mental health problems, and addressing acute emotional vulnerability and

additional symptoms such as insomnia and pain. Behavioral factors should also be thoroughly

assessed and options to improve an unhealthy lifestyle (e.g., excess weight, inactivity, tobacco

smoking) offered as appropriate. When moderate-to-severe pre-treatment CRF is present, it

should be treated. These recommendations are reflected in current CRF guidelines for

patients and survivors across the disease spectrum96–99 and many of them represent

cornerstones of optimal survivorship care beyond the management of CRF.100 There is still no

evidence demonstrating that “intercepting the risk” of post-treatment symptoms early along

the cancer trajectory and correcting modifiable factors may alter the course of CRF. However,

refined knowledge of symptom science, including leveraging bio-behavioral models of CRF,

paves the way and sets out the rationale for prospective studies testing these hypotheses.101

Examples of such studies, assessing feasibility, acceptability, and effectiveness of risk

stratification and personalized supportive care pathways are ongoing or in the pipeline at our

institution102, and similar efforts are in place elsewhere, aiming at implementing CRF screening

in routine care to provide early treatment options.103

Journal Pre-proof

22

We acknowledge that the model we propose is only “a” bio-behavioral model that may

explain CRF. It focuses on pre-treatment factors and on outcomes reported two years later.

Some factors such as childhood trauma, or biochemical parameters such as cortisol and

hormonal levels could not be investigated. We included a relatively homogeneous population

of breast cancer survivors from the CANTO study without a healthy control group that could

have allowed comparisons with individuals that did not receive cancer treatment. We also

acknowledge a potential impact of variability linked to stability of the analyte in the samples

and measurement error that is intrinsically linked to the evaluation methods implemented in

any biological analysis. However, all analytic procedures underwent strict quality control at a

reference center (Gustave Roussy, Villejuif, France), ensuring collection, storage, and profiling

standards are met104, and that quantifications are consistent with studies using similar

assays.39,105 Despite measures taken to optimize variable selection, some may have not been

retained due to inclusion thresholds. Nevertheless, we used an established methodology42

and results are strengthened by several sensitivity analyses. Additional specific strengths

include a prospective and longitudinal design, and the wealth of available data, including a

baseline, pre-treatment evaluation of serum inflammatory markers and of the outcome.

In conclusion, we generated a bio-behavioral model of CRF incorporating pretreatment clinical, behavioral, and biological factors. While our results build and strengthen an

evidence base about the inflammatory biology involved in CRF, future studies should be

encouraged to unveil additional mechanisms underlying symptom onset and evolution. Novel

technologies such as wearable biosensors could be exploited to test the clinical utility of

devices allowing for a continuous monitoring of systemic inflammation and behaviors,

providing data regarding potential “dynamic markers”.106–108 An agnostic approach, delving into

multi-omics, may also unlock a more comprehensive understanding of alternative biological

pathways and substrates that extend beyond cancer-related inflammation, providing further

actionable mechanistic targets.101

Journal Pre-proof

23

ACKNOWLEDGEMENTS: The authors are grateful to the Gustave Roussy laboratory staff,

particularly to Inès Simouh for the preanalytical processing of stored samples and the

management of reagents.

FUNDING: This work was supported by Conquer Cancer, the American Society of Clinical

Oncology (ASCO), and Rising Tide Foundation for Clinical Cancer Research [Career

Pathway Grant in Symptom Management to ADM]; Foundation ARC [grant number

ARCPGA2022010004401_4882 to ADM]; Breast Cancer Research Foundation [grant number

not applicable] to IVL; Susan G. Komen [grant number Career Catalyst Research grant

CCR17483507 to IVL]; Foundation Gustave Roussy [grant number not applicable to IVL]; and

the French Government under the “Investment for the Future” program managed by the

National Research Agency (ANR) [grant number ANR-10-COHO-0004 (CANTO); grant

number ANR-18-IBHU-0002 (PRISM); grant number ANR-17-RHUS-008 (MyPROBE) to FA].

Funders had no role in collection, analysis or interpretation of data.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Antonio Di Meglio - Expert Testimony: Kephren, Techspert (Personal, none related to this

manuscript)

Maria Alice Franzoi - Speaker: Novartis (Inst); Research funding: Resilience Care (Inst.)

Barbara Pistilli - Consulting or Advisory Role: Puma Biotechnology, Pierre Fabre, Novartis,

Myriad Genetics, AstraZeneca, Daiichi Sankyo/UCB Japan; Research Funding: Pfizer (Inst),

Puma Biotechnology (Inst), Merus (Inst), Daiichi-Sankyo (Inst); Travel, Accommodations,

Expenses: Pfizer, AstraZeneca, MSD Oncology, Novartis, Pierre Fabre

Florence Joly - Consulting or Advisory Role: AstraZeneca, Janssen, Ipsen, Pfizer, MSD

Oncology, Bristol Myers Squibb, GlaxoSmithKline, Astellas Pharma, Clovis Oncology, Amgen,

Seattle Genetics, Bayer; Travel, Accommodations, Expenses: Janssen, AstraZeneca, Ipsen,

GlaxoSmithKline, BMS

Paul H. Cottu - Honoraria: Pfizer, Novartis (Inst), Roche, NanoString Technologies (Inst), Lilly

Journal Pre-proof

24

Consulting or Advisory Role: Pfizer, Lilly; Research Funding: Pfizer (Inst); Travel,

Accommodations, Expenses: Roche, Pfizer

Olivier Tredan - Consulting or Advisory Role: Roche, Pfizer, Lilly, AstraZeneca, MSD

Oncology, Daiichi Sankyo Europe GmbH, Eisai Europe, Sandoz-Novartis, Seattle Genetics,

Pierre Fabre, Gilead Sciences; Research Funding: Roche (Inst), Bristol Myers Squibb (Inst),

MSD Oncology (Inst), AstraZeneca (Inst), Novartis (Inst), Bayer (Inst); Travel,

Accommodations, Expenses: Roche, Novartis, Pfizer, Lilly, AstraZeneca, MSD Oncology

Patricia A. Ganz - Leadership: Intrinsic LifeSciences (I); Stock and Other Ownership

Interests: Xenon Pharma (I), Intrinsic LifeSciences (I), Silarus Therapeutics (I), Teva,

Novartis, Merck, Johnson & Johnson, Pfizer, GlaxoSmithKline, Abbott Laboratories;

Consulting or Advisory Role: InformedDNA, Vifor Pharma (I), Ambys Medicines (I), Global

Blood Therapeutics (I), GlaxoSmithKline (I), Ionis Pharmaceuticals (I), Akebia Therapeutics

(I), Protagonist Therapeutics (I), Regeneron (I), Sierra Oncology (I), Rockwell Medical

Technologies Inc (I), Astellas Pharma (I), Gossamer Bio (I), American Regent (I), Disc

Medicine (I), Blue Note Therapeutics, Grail; Research Funding: Blue Note Therapeutics

(Inst); Patents, Royalties, Other Intellectual Property: Related to iron metabolism and the

anemia of chronic disease, Up-to-Date royalties for section editor on survivorship (I); Travel,

Accommodations, Expenses: Intrinsic LifeSciences (I)

Ann H. Partridge - Patents, Royalties, Other Intellectual Property: I receive small royalty

payments for coauthoring the breast cancer survivorship section of UpToDate; Open

Payments Link: https://openpaymentsdata.cms.gov/physician/835197

Fabrice André - Stock and Other Ownership Interests: Pegacsy; Research

Funding: AstraZeneca (Inst), Novartis (Inst), Pfizer (Inst), Lilly (Inst), Roche (Inst), Daiichi

(Inst); Travel, Accommodations, Expenses: Novartis, Roche, GlaxoSmithKline,

AstraZeneca

Stefan Michiels - Consulting or Advisory Role: IDDI, Sensorion, Biophytis, Servier, Yuhan,

Amaris Consulting, Roche

Journal Pre-proof

25

Ines Vaz-Luis - Speaker honoraria: from Amgen, AstraZeneca, Pfizer/Edimark, Novartis,

Sandoz (Institutional); Writing engagement: from Pfizer/Edimark (Institutional); Research

funding from Resilience Care (Institutional); Travel: Novartis

All remaining authors have declared no conflicts of interest.

Journal Pre-proof

26

REFERENCES

1. Bower JE. Cancer-related fatigue--mechanisms, risk factors, and treatments. Nat Rev Clin

Oncol. October 2014;11(10):597–609.

2. Al Maqbali M, Al Sinani M, Al Naamani Z, Al Badi K, Tanash MI. Prevalence of Fatigue in

Patients With Cancer: A Systematic Review and Meta-Analysis. J Pain Symptom Manage.

January 1, 2021;61(1):167-189.e14.

3. Kang YE, Yoon JH, Park N hyun, Ahn YC, Lee EJ, Son CG. Prevalence of cancer-related fatigue

based on severity: a systematic review and meta-analysis. Scientific Reports 2023 13:1.

August 7, 2023;13(1):1–11.

4. Abrahams HJG, Gielissen MFM, Schmits IC, Verhagen CAHHVM, Rovers MM, Knoop H. Risk

factors, prevalence, and course of severe fatigue after breast cancer treatment: a metaanalysis involving 12 327 breast cancer survivors. Annals of Oncology. June 2016;27(6):965–

974.

5. Schmidt ME, Hermann S, Arndt V, Steindorf K. Prevalence and severity of long‐term physical,

emotional, and cognitive fatigue across 15 different cancer entities. Cancer Med. November

1, 2020;9(21):8053.

6. Vaz-Luis I, Di Meglio A, Havas J, El-Mouhebb M, Lapidari P, Presti D, et al. Long-Term

Longitudinal Patterns of Patient-Reported Fatigue After Breast Cancer: A Group-Based

Trajectory Analysis. Journal of Clinical Oncology. March 15, 2022;JCO2101958.

7. Schmidt ME, Chang-Claude J, Seibold P, Vrieling A, Heinz J, Flesch-Janys D, et al. Determinants

of long-term fatigue in breast cancer survivors: results of a prospective patient cohort study.

Psychooncology. 2015;24(1):40–46.

8. Schmidt ME, Blickle P, Steindorf K. Cancer-related fatigue: Identification of hallmarks to

enable refined treatment approaches. Psychooncology. December 1, 2022;31(12):2169–2176.

9. Sleight AG, Crowder SL, Skarbinski J, Coen P, Parker NH, Hoogland AI, et al. A New Approach

to Understanding Cancer-Related Fatigue: Leveraging the 3P Model to Facilitate Risk

Prediction and Clinical Care. Cancers (Basel). April 1, 2022;14(8).

10. Wang Y, Tian L, Liu X, Zhang H, Tang Y, Zhang H, et al. Multidimensional Predictors of CancerRelated Fatigue Based on the Predisposing, Precipitating, and Perpetuating (3P) Model: A

Systematic Review. Cancers (Basel). December 1, 2023;15(24).

11. Rosas JC, Aguado-Barrera ME, Azria D, Briers E, Elliott R, Farcy-Jacquet MP, et al.

(Pre)treatment risk factors for late fatigue and fatigue trajectories following radiotherapy for

breast cancer. Int J Cancer. November 1, 2023;153(9):1579–1591.

12. Schmidt ME, Wiskemann J, Schneeweiss A, Potthoff K, Ulrich CM, Steindorf K. Determinants

of physical, affective, and cognitive fatigue during breast cancer therapy and 12 months

follow-up. Int J Cancer. March 15, 2018;142(6):1148–1157.

Journal Pre-proof

27

13. Schmidt ME, Maurer T, Behrens S, Seibold P, Obi N, Chang-Claude J, et al. Cancer-related

fatigue: Towards a more targeted approach based on classification by biomarkers and

psychological factors. Int J Cancer. March 15, 2024;154(6):1011–1018.

14. Thong MSY, Mols F, van de Poll-Franse L V., Sprangers MAG, van der Rijt CCD, Barsevick AM,

et al. Identifying the subtypes of cancer-related fatigue: results from the population-based

PROFILES registry. Journal of Cancer Survivorship. February 1, 2018;12(1):38–46.

15. Di Meglio A, Havas J, Soldato D, Presti D, Martin E, Pistilli B, et al. Development and Validation

of a Predictive Model of Severe Fatigue After Breast Cancer Diagnosis: Toward a Personalized

Framework in Survivorship Care. J Clin Oncol. January 21, 2022;JCO2101252.

16. Programme Interval - Risk Prediction of Severe Cancer-Related Fatigue in Survivors of Breast

Cancer | Gustave Roussy [Internet]. [cited January 5, 2022]. Available at:

https://www.gustaveroussy.fr/fr/interval-breast-cancer-related-fatigue-calculator

17. García-González D, Medino-Muñoz J, Romero-Elías M, García-Foncillas J, Ruiz-Casado A.

Biological mechanisms of cancer-related fatigue in breast cancer survivors after treatment: a

scoping review. Journal of Cancer Survivorship. November 6, 2023;1–31.

18. Dantzer R. Cytokine-induced sickness behavior: where do we stand? Brain Behav Immun.

2001;15(1):7–24.

19. Dantzer R, O’Connor JC, Freund GG, Johnson RW, Kelley KW. From inflammation to sickness

and depression: when the immune system subjugates the brain. Nat Rev Neurosci. January

2008;9(1):46–56.

20. Bower JE, Lamkin DM. Inflammation and cancer-related fatigue: mechanisms, contributing

factors, and treatment implications. Brain Behav Immun. March 15, 2013;

21. Carroll JE, Bower JE, Ganz PA. Cancer-related accelerated ageing and biobehavioural

modifiers: a framework for research and clinical care. Nature Reviews Clinical Oncology 2021

19:3. December 6, 2021;19(3):173–187.

22. Miller AH, Ancoli-Israel S, Bower JE, Capuron L, Irwin MR. Neuroendocrine-Immune

Mechanisms of Behavioral Comorbidities in Patients With Cancer. J Clin Oncol. February 2,

2008;26(6):971.

23. Campisi J. Senescent cells, tumor suppression, and organismal aging: good citizens, bad

neighbors. Cell. February 25, 2005;120(4):513–522.

24. Schmidt ME, Meynköhn A, Habermann N, Wiskemann J, Oelmann J, Hof H, et al. Resistance

Exercise and Inflammation in Breast Cancer Patients Undergoing Adjuvant Radiation Therapy:

Mediation Analysis From a Randomized, Controlled Intervention Trial. Int J Radiat Oncol Biol

Phys. February 1, 2016;94(2):329–337.

25. van Vulpen JK, Schmidt ME, Velthuis MJ, Wiskemann J, Schneeweiss A, Vermeulen RCH, et al.

Effects of physical exercise on markers of inflammation in breast cancer patients during

adjuvant chemotherapy. Breast Cancer Res Treat. April 1, 2018;168(2):421–431.

Journal Pre-proof

28

26. Hiensch AE, Mijwel S, Bargiela D, Wengström Y, May AM, Rundqvist H. Inflammation

Mediates Exercise Effects on Fatigue in Patients with Breast Cancer. Med Sci Sports Exerc.

March 1, 2021;53(3):496–504.

27. Ganz PA, Petersen L, Bower JE, Crespi CM. Impact of adjuvant endocrine therapy on quality of

life and symptoms: Observational data over 12 months from the mind-body study. Journal of

Clinical Oncology. March 10, 2016;34(8):816–824.

28. Ferreira ARR, Di Meglio A, Pistilli B, Gbenou ASS, El-Mouhebb M, Dauchy S, et al. Differential

impact of endocrine therapy and chemotherapy on quality of life of breast cancer survivors: a

prospective patient-reported outcomes analysis. Annals of Oncology. October 8,

2019;30(11):1784–1795.

29. Di Meglio A, Havas J, Gbenou AS, Martin E, El-Mouhebb M, Pistilli B, et al. Dynamics of LongTerm Patient-Reported Quality of Life and Health Behaviors After Adjuvant Breast Cancer

Chemotherapy. J Clin Oncol. April 1, 2022 [cited July 25, 2024];40(27). Available at:

https://pubmed.ncbi.nlm.nih.gov/35446677/

30. Bauml J, Chen L, Chen J, Boyer J, Kalos M, Li SQ, et al. Arthralgia among women taking

aromatase inhibitors: is there a shared inflammatory mechanism with co-morbid fatigue and

insomnia? Breast Cancer Res. June 28, 2015;17(1).

31. Giesinger JM, Kuijpers W, Young T, Tomaszewski KA, Friend E, Zabernigg A, et al. Thresholds

for clinical importance for four key domains of the EORTC QLQ-C30: physical functioning,

emotional functioning, fatigue and pain. Health Qual Life Outcomes. 2016;14:87.

32. Giesinger JM, Loth FLC, Aaronson NK, Arraras JI, Caocci G, Efficace F, et al. Thresholds for

clinical importance were established to improve interpretation of the EORTC QLQ-C30 in

clinical practice and research. J Clin Epidemiol. February 1, 2020;118:1–8.

33. Vaz-Luis I, Cottu P, Mesleard C, Martin AL, Dumas A, Dauchy S, et al. UNICANCER: French

prospective cohort study of treatment-related chronic toxicity in women with localised breast

cancer (CANTO). ESMO Open. September 1, 2019;4(5):e000562.

34. Weis J, Tomaszewski KA, Hammerlid E, Ignacio Arraras J, Conroy T, Lanceley A, et al.

International Psychometric Validation of an EORTC Quality of Life Module Measuring Cancer

Related Fatigue (EORTC QLQ-FA12). J Natl Cancer Inst. May 1, 2017;109(5).

35. Global Physical Activity Questionnaire Analysis Guide GPAQ Analysis Guide Global Physical

Activity Questionnaire (GPAQ) Analysis Guide.

36. Zigmond AS, Snaith RP. The hospital anxiety and depression scale. Acta Psychiatr Scand. June

1983;67(6):361–70.

37. Aaronson NK, Ahmedzai S, Bergman B, Bullinger M, Cull A, Duez NJ, et al. The European

Organization for Research and Treatment of Cancer QLQ-C30: a quality-of-life instrument for

use in international clinical trials in oncology. J Natl Cancer Inst. 1993;85(5):365–376.

Journal Pre-proof

29

38. Cancer Institute N. Common Terminology Criteria for Adverse Events (CTCAE) Version 4.0.

2009.

39. Evidence Investigator - Randox Laboratories [Internet]. [cited June 15, 2023]. Available at:

https://www.randox.com/evidence-investigator-2/

40. Henry NL, Pchejetski D, A’Hern R, Nguyen AT, Charles P, Waxman J, et al. Inflammatory

cytokines and aromatase inhibitor-associated musculoskeletal syndrome: a case-control

study. Br J Cancer. July 27, 2010;103(3):291–296.

41. Sabatine MS, Morrow DA, Jablonski KA, Rice MM, Warnica JW, Domanski MJ, et al. Prognostic

significance of the Centers for Disease Control/American Heart Association high-sensitivity Creactive protein cut points for cardiovascular and other outcomes in patients with stable

coronary artery disease. Circulation. March 2007;115(12):1528–1536.

42. Heymans MW, Van Buuren S, Knol DL, Van Mechelen W, De Vet HCW. Variable selection

under multiple imputation using the bootstrap in a prognostic study. BMC Med Res

Methodol. July 13, 2007;7(1):1–10.

43. White IR, Royston P, Wood AM. Multiple imputation using chained equations: Issues and

guidance for practice. Stat Med. February 20, 2011;30(4):377–399.

44. Campion WM, Rubin DB. Multiple Imputation for Nonresponse in Surveys. Rubin DB (ed):

Journal of Marketing Research. June 9, 1989;26(4):485.

45. Jr HGG. PARSIMONY AND EFFICIENCY. Scientific Method in Practice. March 12, 2002 [cited

July 23, 2024];269–326. Available at: https://www.cambridge.org/core/books/scientificmethod-in-practice/parsimony-and-efficiency/F8327BE6D73C3D00214B65AC04D350F2

46. Harrell FE, Lee KL, Mark DB. Multivariable prognostic models: Issues in developing models,

evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med.

February 28, 1996;15(4):361–387.

47. Riley RD, Ensor J, Snell KIE, Harrell FE, Martin GP, Reitsma JB, et al. Calculating the sample size

required for developing a clinical prediction model. The BMJ. March 18, 2020;368.

48. Bower JE, Ganz PA, Irwin MR, Kwan L, Breen EC, Cole SW. Inflammation and Behavioral

Symptoms After Breast Cancer Treatment: Do Fatigue, Depression, and Sleep Disturbance

Share a Common Underlying Mechanism? Journal of Clinical Oncology. September 9,

2011;29(26):3517.

49. Cohen M, Levkovich I, Katz R, Fried G, Pollack S. Low physical activity, fatigue and depression

in breast cancer survivors: Moderation by levels of IL-6 and IL-8. Int J Psychophysiol.

December 1, 2020;158:96–102.

50. Collado-Hidalgo A, Bower JE, Ganz PA, Cole SW, Irwin MR. Inflammatory Biomarkers for

Persistent Fatigue in Breast Cancer Survivors. Clinical Cancer Research. May 1,

2006;12(9):2759–2766.

Journal Pre-proof

30

51. Bower JE, Ganz PA, Aziz N, Olmstead R, Irwin MR, Cole SW. Inflammatory responses to

psychological stress in fatigued breast cancer survivors: relationship to glucocorticoids. Brain

Behav Immun. March 2007;21(3):251–258.

52. Vasbinder A, Thompson H, Zaslavksy O, Heckbert SR, Saquib N, Shadyab AH, et al.

Inflammatory, Oxidative Stress, and Cardiac Damage Biomarkers and Radiation-Induced

Fatigue in Breast Cancer Survivors. Biol Res Nurs. 2022;24(4).

53. Alfano CM, Imayama I, Neuhouser ML, Kiecolt-Glaser JK, Smith AW, Meeske K, et al. Fatigue,

inflammation, and ω-3 and ω-6 fatty acid intake among breast cancer survivors. J Clin Oncol.

April 20, 2012;30(12):1280–1287.

54. Orre IJ, Reinertsen K V., Aukrust P, Dahl AA, Fosså SD, Ueland T, et al. Higher levels of fatigue

are associated with higher CRP levels in disease-free breast cancer survivors. J Psychosom

Res. September 2011;71(3):136–141.

55. Xiao C, Miller AH, Felger J, Mister D, Liu T, Torres MA. Depressive symptoms and

inflammation are independent risk factors of fatigue in breast cancer survivors. Psychol Med.

July 1, 2017;47(10):1733–1743.

56. Musselman DL, Miller AH, Porter MR, Manatunga A, Gao F, Penna S, et al. Higher than normal

plasma interleukin-6 concentrations in cancer patients with depression: preliminary findings.

Am J Psychiatry. 2001;158(8):1252–1257.

57. Kong BS, Kim Y, Kim GY, Hyun JW, Kim SH, Jeong A, et al. Increased frequency of IL-6-

producing non-classical monocytes in neuromyelitis optica spectrum disorder. J

Neuroinflammation. September 25, 2017;14(1).

58. Swardfager W, Lanctt K, Rothenburg L, Wong A, Cappell J, Herrmann N. A Meta-Analysis of

Cytokines in Alzheimer’s Disease. Biol Psychiatry. November 15, 2010;68(10):930–941.

59. Kuo HK, Yen CJ, Chang CH, Kuo CK, Chen JH, Sorond F. Relation of C-reactive protein to stroke,

cognitive disorders, and depression in the general population: systematic review and metaanalysis. Lancet Neurol. June 1, 2005;4(6):371–380.

60. Starkweather A, Kelly DL, Thacker L, Wright ML, Jackson-Cook CK, Lyon DE. Relationships

among psychoneurological symptoms and levels of C-reactive protein over 2 years in women

with early-stage breast cancer. Support Care Cancer. January 1, 2017;25(1):167–176.

61. Rogers LQ, Vicari S, Trammell R, Hopkins-Price P, Fogleman A, Spenner A, et al. Biobehavioral

factors mediate exercise effects on fatigue in breast cancer survivors. Med Sci Sports Exerc.

2014;46(6):1077–1088.

62. Petersen AMW, Pedersen BK. The role of IL-6 in mediating the anti-inflammatory effects of

exercise. J Physiol Pharmacol. November 1, 2006;57 Suppl 10(SUPPL. 10):43–51.

63. Henneghan A, Wright ML, Bourne G, Sales AC. A Cross-Sectional Exploration of CytokineSymptom Networks in Breast Cancer Survivors Using Network Analysis. Can J Nurs Res.

September 1, 2021;53(3):303–315.

Journal Pre-proof

31

64. Bower JE, Ganz PA, Irwin MR, Castellon S, Arevalo J, Cole SW. Cytokine genetic variations and

fatigue among patients with breast cancer. May 1, 2013;31(13):1656–1661.

65. Cameron B, Webber K, Li H, Bennett BK, Boyle F, de Souza P, et al. Genetic associations of

fatigue and other symptoms following breast cancer treatment: A prospective study. Brain

Behav Immun Health. January 1, 2020;10.

66. Bower JE, Ganz PA, Irwin MR, Crespi CM, Petersen L, Asher A, et al. Type I interferons,

inflammation, and fatigue in a longitudinal RNA study of women with breast cancer. Brain

Behav Immun. May 1, 2024;118:312–317.

67. Miaskowski C, Dodd M, Lee K, West C, Paul SM, Cooper BA, et al. Preliminary Evidence of an

Association Between a Functional IL6 Polymorphism and Fatigue and Sleep Disturbance in

Oncology Patients and Their Family Caregivers. J Pain Symptom Manage. October

2010;40(4):531.

68. Collado-Hidalgo A, Bower JE, Ganz PA, Irwin MR, Cole SW. Cytokine gene polymorphisms and

fatigue in breast cancer survivors: Early findings. Brain Behav Immun. November

2008;22(8):1197–1200.

69. Doong SH, Dhruva A, Dunn LB, West C, Paul SM, Cooper BA, et al. Associations between

cytokine genes and a symptom cluster of pain, fatigue, sleep disturbance, and depression in

patients prior to breast cancer surgery. Biol Res Nurs. May 4, 2015;17(3):237–247.

70. Karshikoff B, Sundelin T, Lasselin J. Role of Inflammation in Human Fatigue: Relevance of

Multidimensional Assessments and Potential Neuronal Mechanisms. Front Immunol. January

20, 2017;8(JAN).

71. Maurer T, Jaskulski S, Behrens S, Jung AY, Obi N, Johnson T, et al. Tired of feeling tired - The

role of circulating inflammatory biomarkers and long-term cancer related fatigue in breast

cancer survivors. Breast. April 1, 2021;56:103–109.

72. Bower JE, Wiley J, Petersen L, Irwin MR, Cole SW, Ganz PA. Fatigue after breast cancer

treatment: Biobehavioral predictors of fatigue trajectories. Health Psychol. November 1,

2018;37(11):1025–1034.

73. Inagaki M, Isono M, Okuyama T, Sugawara Y, Akechi T, Akizuki N, et al. Plasma interleukin-6

and fatigue in terminally ill cancer patients. J Pain Symptom Manage. February

2008;35(2):153–161.

74. De Raaf PJ, Sleijfer S, Lamers CHJ, Jager A, Gratama JW, Van Der Rijt CCD. Inflammation and

fatigue dimensions in advanced cancer patients and cancer survivors: an explorative study.

Cancer. December 1, 2012;118(23):6005–6011.

75. Duivon M, Lequesne J, Di Meglio A, Pradon C, Vaz-Luis I, Martin AL, et al. Inflammation at

diagnosis and cognitive impairment two years later in breast cancer patients from the CantoCog study. Breast Cancer Research. December 1, 2024;26(1):1–10.

Journal Pre-proof

32

76. Carroll JE, Nakamura ZM, Small BJ, Zhou X, Cohen HJ, Ahles TA, et al. Elevated C-Reactive

Protein and Subsequent Patient-Reported Cognitive Problems in Older Breast Cancer

Survivors: The Thinking and Living With Cancer Study. J Clin Oncol. January 10, 2023;41(2).

77. Bower JE. The role of neuro-immune interactions in cancer-related fatigue: Biobehavioral risk

factors and mechanisms. Cancer. February 1, 2019;125(3):353–364.

78. Schmidt ME, Wiskemann J, Steindorf K. Quality of life, problems, and needs of disease-free

breast cancer survivors 5 years after diagnosis. Qual Life Res. August 1, 2018;27(8):2077–

2086.

79. Dong ST, Butow PN, Costa DSJ, Lovell MR, Agar M. Symptom clusters in patients with

advanced cancer: a systematic review of observational studies. J Pain Symptom Manage.

September 1, 2014;48(3):411–450.

80. Zhu L, Ranchor A V., van der Lee M, Garssen B, Almansa J, Sanderman R, et al. Co-morbidity

of depression, anxiety and fatigue in cancer patients receiving psychological care.

Psychooncology. April 1, 2017;26(4):444–451.

81. Pfob A, Mehrara BJ, Nelson JA, Wilkins EG, Pusic AL, Sidey-Gibbons C. Towards Patientcentered Decision-making in Breast Cancer Surgery: Machine Learning to Predict Individual

Patient-reported Outcomes at 1-year Follow-up. Ann Surg. January 1, 2023;277(1):e144.

82. Révész D, Van Kuijk SMJ, Mols F, Van Duijnhoven FJB, Winkels RM, Hoofs H, et al.

Development and internal validation of prediction models for colorectal cancer survivors to

estimate the 1-year risk of low health-related quality of life in multiple domains. BMC Med

Inform Decis Mak. March 12, 2020;20(1):1–13.

83. Pfob A, Mehrara BJ, Nelson JA, Wilkins EG, Pusic AL, Sidey-Gibbons C. Machine learning to

predict individual patient-reported outcomes at 2-year follow-up for women undergoing

cancer-related mastectomy and breast reconstruction (INSPiRED-001). Breast. December 1,

2021;60:111–122.

84. Révész D, van Kuijk SMJ, Mols F, van Duijnhoven FJB, Winkels RM, Kant IjJ, et al. External

validation and updating of prediction models for estimating the 1-year risk of low healthrelated quality of life in colorectal cancer survivors. J Clin Epidemiol. December 1,

2022;152:127–139.

85. Haapakoski R, Mathieu J, Ebmeier KP, Alenius H, Kivimäki M. Cumulative meta-analysis of

interleukins 6 and 1β, tumour necrosis factor α and C-reactive protein in patients with major

depressive disorder. Brain Behav Immun. October 1, 2015;49:206–215.

86. Solmi M, Suresh Sharma M, Osimo EF, Fornaro M, Bortolato B, Croatto G, et al. Peripheral

levels of C-reactive protein, tumor necrosis factor-α, interleukin-6, and interleukin-1β across

the mood spectrum in bipolar disorder: A meta-analysis of mean differences and variability.

Brain Behav Immun. October 1, 2021;97:193–203.

Journal Pre-proof

33

87. de Jong N, Candel MJJM, Schouten HC, Abu-Saad HH, Courtens AM. Prevalence and course of

fatigue in breast cancer patients receiving adjuvant chemotherapy. Ann Oncol. June

2004;15(6):896–905.

88. Goodwin PJ, Stambolic V. Impact of the obesity epidemic on cancer. Annu Rev Med. January

14, 2015;66(1):281–296.

89. Van Vulpen JK, Sweegers MG, Peeters PHM, Courneya KS, Newton RU, Aaronson NK, et al.

Moderators of Exercise Effects on Cancer-related Fatigue: A Meta-analysis of Individual

Patient Data. Med Sci Sports Exerc. February 1, 2020;52(2):303–314.

90. Tjoe JA, Piacentine LB, Papanek PE, Raff H, Richards J, Harkins AL, et al. Team triathlon effects

on physiological, psychological, and immunological measures in women breast cancer

survivors. Support Care Cancer. December 1, 2020;28(12):6095–6104.

91. Lengacher CA, Reich RR, Paterson CL, Shelton M, Shivers S, Ramesar S, et al. A Large

Randomized Trial: Effects of Mindfulness-Based Stress Reduction (MBSR) for Breast Cancer

(BC) Survivors on Salivary Cortisol and IL-6. Biol Res Nurs. January 1, 2019;21(1):39–49.

92. Fagundes CP, Murray DM, Hwang BS, Gouin J-P, Thayer JF, Sollers JJ, et al. Sympathetic and

parasympathetic activity in cancer-related fatigue: more evidence for a physiological

substrate in cancer survivors. Psychoneuroendocrinology. September 2011;36(8):1137–47.

93. Kiecolt-Glaser JK, Bennett JM, Andridge R, Peng J, Shapiro CL, Malarkey WB, et al. Yoga’s

impact on inflammation, mood, and fatigue in breast cancer survivors: a randomized

controlled trial. J Clin Oncol. April 1, 2014;32(10):1040–1049.

94. Bower JE, Irwin MR. Mind-body therapies and control of inflammatory biology: A descriptive

review. Brain Behav Immun. January 1, 2016;51:1–11.

95. Van Gemert WA, May AM, Schuit AJ, Oosterhof BYM, Peeters PH, Monninkhof EM. Effect of

Weight Loss with or without Exercise on Inflammatory Markers and Adipokines in

Postmenopausal Women: The SHAPE-2 Trial, A Randomized Controlled Trial. Cancer

Epidemiol Biomarkers Prev. May 1, 2016;25(5):799–806.

96. Bower JE, Lacchetti C, Alici Y, Barton DL, Bruner D, Canin BE, et al. Management of Fatigue in

Adult Survivors of Cancer: ASCO–Society for Integrative Oncology Guideline Update. Journal

of Clinical Oncology. May 16, 2024;

97. Fabi A, Bhargava R, Fatigoni S, Guglielmo M, Horneber M, Roila F, et al. Cancer-related

fatigue: ESMO Clinical Practice Guidelines for diagnosis and treatment. Ann Oncol. June 1,

2020;31(6):713–723.

98. Mustian K, Lacchetti C, Zick S, Bower JE. Management of Fatigue in Adult Survivors of Cancer:

American Society of Clinical Oncology - Society for Integrative Oncology (ASCO-SIO) Guideline

Update Clinical Insights. https://doi.org/101200/OP2400372. July 2, 2024;

99. Bower JE, Bak K, Berger A, Breitbart W, Escalante CP, Ganz PA, et al. Screening, Assessment,

and Management of Fatigue in Adult Survivors of Cancer: An American Society of Clinical

Journal Pre-proof

34

Oncology Clinical Practice Guideline Adaptation. Journal of Clinical Oncology. June 10,

2014;32(17):1840–1850.

100. Vaz-Luis I, Masiero M, Cavaletti G, Cervantes A, Chlebowski RT, Curigliano G, et al. ESMO

Expert Consensus Statements on Cancer Survivorship: promoting high-quality survivorship

care and research in Europe. Ann Oncol. August 2022;

101. Di Meglio A, Vaz-Luis I. Systemic inflammation and cancer-related frailty: shifting the

paradigm toward precision survivorship medicine. ESMO Open. January 1, 2024;9(1).

102. Franzoi MA. Protocol of a master cohort study to evaluate the implementation of integrated

proactive supportive care pathways in oncology. JMIR Cancer. 2024;In press.

103. Blickle P, Haussmann A, Holzner B, Berger AK, Steindorf K, Schmidt ME. Providing the basis

for a patient-centred and effective screening for cancer-related fatigue (MERLIN study):

design of a longitudinal observational study. BMJ Open. September 28, 2023;13(9).

104. ISO 15189 knowledge hub [Internet]. Available at: https://www.ideagen.com/standards/iso15189-knowledgehub?creative=674749056955&keyword=en+iso+15189&matchtype=e&network=g&device=c

&utm_source=google&utm_medium=cpc&utm_campaign=FY24-ALL-QAL-ALL-DGC-GooglePPC-(Q-Pulse-QMS-Quality-Management-ISO-T-EMEA)&u

105. Randox | Global Healthcare [Internet]. [cited June 19, 2023]. Available at:

https://www.randox.com/

106. Rice P, Upasham S, Jagannath B, Manuel R, Pali M, Prasad S. CortiWatch: Watch-based

cortisol tracker. Future Sci OA. September 26, 2019;5(9).

107. Jagannath B, Lin KC, Pali M, Sankhala D, Muthukumar S, Prasad S. A Sweat-based Wearable

Enabling Technology for Real-time Monitoring of IL-1β and CRP as Potential Markers for

Inflammatory Bowel Disease. Inflamm Bowel Dis. October 1, 2020;26(10):1533–1542.

108. Laliberte KE, Scott P, Khan NI, Mahmud MS, Song E. A wearable graphene transistor-based

biosensor for monitoring IL-6 biomarker. Microelectron Eng. June 1, 2022;262:111835.

Journal Pre-proof

1

TABLES

Table 1. Descriptive statistics on the distribution of clinical and behavioral characteristics of the cohort

at breast cancer diagnosis.

Variable

(all assessed at diagnosis

of breast cancer)

Whole cohort

(N=1208)

Reporting fatigue of clinical importance at

year-2 (EORTC QLQC-30)

Yes

(N=415; 34.4%)

No

(N=793; 65.6%)

Age, years

Mean (SD) 57.9 (11.1) 56.2 (11.4) 58.9 (10.8)

Min-Max 22.3–88.2 28.6–88.2 22.3–85.5

Missing - - -

BMI, Kg/m2

Mean (SD) 25.9 (5.2) 26.1 (5.5) 25.8 (5.1)

Missing 3 2 1

Menopausal status, N (%)

Premenopausal 413 (34.4) 163 (39.6) 250 (31.7)

Postmenopausal 788 (65.6) 249 (60.4) 539 (68.3)

Missing 7 3 4

Charlson comorbidity index, N (%)

0 867 (78.6) 283 (77.1) 584 (79.3)

>=1 236 (21.4) 84 (22.9) 152 (20.7)

Missing 105 48 57

Previous mental health problems, N (%)

No 983 (83.7) 321 (79.3) 662 (86.1)

Yes 191 (16.3) 84 (20.7) 107 (13.9)

Missing 34 10 24

Marital Status, N (%)

Not partnered 275 (23.8) 109 (27.5) 166 (21.9)

Partnered 880 (76.2) 287 (72.5) 593 (78.1)

Missing 53 19 34

Education level, N (%)

Primary school 192 (16.6) 67 (16.8) 125 (16.5)

High school 532 (46.0) 185 (46.5) 347 (45.8)

College or higher 432 (37.4) 146 (36.7) 286 (37.7)

Missing 52 17 35

Household income, N (%)

<1500 166 (15.2) 73 (19.5) 93 (12.9)

>=1500 and <3000 442 (40.4) 156 (41.6) 286 (39.8)

>=3000 486 (44.4) 146 (38.9) 340 (47.3)

Missing 114 40 74

Alcohol consumption behavior, N (%)

Less than daily consumption 970 (84.3) 332 (84.7) 638 (84.2)

Daily consumption 180 (15.7) 60 (15.3) 120 (15.8)

Missing 58 23 35

Tobacco use behavior, N (%)

Current smoker 204 (17.3) 108 (26.9) 96 (12.4)

Former smoker 240 (20.4) 76 (18.9) 164 (21.2)

Never smoker 733 (62.3) 218 (54.2) 515 (66.5)

Missing 31 13 18

Physical activity (MET-h/week)

Median (Q1-Q3) 14.0 (0.0–40.0) 12.0 (0.0–40.0) 16.0 (0.7–38.0)

Missing 31 7 24

Breast cancer stage, N (%)

Stage I 651 (54.1) 201 (48.8) 450 (56.8)

Stage II 439 (36.5) 166 (40.3) 273 (34.5)

Stage III 114 (9.5) 45 (10.9) 69 (8.7)

Journal Pre-proof

2

Missing 4 3 1

Axillary surgery, N (%)

None or sentinel node biopsy 733 (60.7) 243 (58.6) 490 (61.8)

Dissection 475 (39.3) 172 (41.4) 303 (38.2)

Missing - - -

Breast cancer surgery, N (%)

Conservative 898 (74.3) 294 (70.8) 604 (76.2)

Mastectomy 310 (25.7) 121 (29.2) 189 (23.8)

Missing - - -

Chemotherapy, N (%)

No 675 (55.9) 214 (51.6) 461 (58.1)

Yes 533 (44.1) 201 (48.4) 332 (41.9)

Missing - - -

Radiotherapy, N (%)

No 111 (9.2) 37 (8.9) 74 (9.3)

Yes 1097 (90.8) 378 (91.1) 719 (90.7)

Missing - - -

Hormonal therapy, N (%)

No 99 (8.2) 30 (7.2) 69 (8.7)

Yes 1109 (91.8) 385 (92.8) 724 (91.3)

Missing - - -

Anxiety, N (%)

Non-case 445 (37.3) 119 (28.8) 326 (41.7)

Doubtful case 324 (27.1) 107 (25.9) 217 (27.8)

Case 425 (35.6) 187 (45.3) 238 (30.5)

Missing 14 2 12

Depression, N (%)

Non-case 990 (83.0) 309 (74.8) 681 (87.3)

Doubtful case 121 (10.1) 51 (12.3) 70 (9.0)

Case 82 (6.9) 53 (12.8) 29 (3.7)

Missing 15 2 13

Fatigue

Mean (SD) 25.8 (23.9) 38.8 (25.9) 18.9 (19.5)

Missing 17 3 14

Fatigue of clinical importance, N (%)

No 938 (78.8) 236 (57.3) 702 (90.1)

Yes 253 (21.2) 176 (42.7) 77 (9.9)

Missing 17 3 14

Insomnia

Mean (SD) 40.7 (33.1) 51.0 (34.4) 35.2 (31.0)

Missing 19 3 16

Pain

Mean (SD) 14.1 (20.6) 21.7 (24.3) 10.1 (17.1)

Missing 13 3 10

Hot flashes, N (%)

No 859 (74.5) 274 (70.1) 585 (76.8)

Yes 294 (25.5) 117 (29.9) 177 (23.2)

Missing 55 24 31

SD: Standard Deviation; Q: Quartile; BMI: Body Mass Index; MET-h: Metabolic-equivalent of task-hour; HR: hormone receptor, HER2: Human

epidermal growth factor receptor 2. Total physical activity scored as a continuous variable according to the Global Physical Activity

Questionnaire (GPAQ)-16. Anxiety and Depression scored according to the Hospital Anxiety and Depression Scale: non-case (score 0-7),

Doubtful case case (8-10), case (11-21). Fatigue, insomnia, and pain scored using the EORTC QLQ-C30; Hot flashes assessed by the

Common Terminology Criteria for Adverse Events –CTCAE- v 4.0 (Yes= any grade).

Journal Pre-proof

3

Table 2. Descriptive statistics on the distribution of inflammatory markers at breast cancer diagnosis.

Variable (all assessed at

diagnosis of breast cancer) Whole cohort

(N=1208)

Reporting fatigue of clinical importance

at year-2 (EORTC QLQ-C30)

Yes

(N=415; 34.4%)

No

(N=793; 65.6%)

IL-1a, N (%)

Low 928 (77.3) 307 (74.3) 621 (78.9)

High 272 (22.7) 106 (25.7) 166 (21.1)

Missing 8 2 6

IL-1b, N (%)

Low 300 (25.0) 94 (22.8) 206 (26.2)

Middle low 305 (25.4) 113 (27.4) 192 (24.4)

Middle high 296 (24.7) 107 (25.9) 189 (24.0)

High 299 (24.9) 99 (24.0) 200 (25.4)

Missing 8 2 6

IL-2, N (%)

Low 1057 (89.5) 354 (86.8) 703 (90.9)

High 124 (10.5) 54 (13.2) 70 (9.1)

Missing 27 7 20

IL-4, N (%)

Low 1131 (95.8) 389 (95.3) 742 (96.0)

High 50 (4.2) 19 (4.7) 31 (4.0)

Missing 27 7 20

IL-6, N (%)

Low 304 (25.3) 93 (22.5) 211 (26.7)

Middle low 300 (24.9) 96 (23.2) 204 (25.8)

Middle high 299 (24.9) 97 (23.5) 202 (25.6)

High 300 (24.9) 127 (30.8) 173 (21.9)

Missing 5 2 3

IL-8, N (%)

Low 296 (25.1) 108 (26.5) 188 (24.3)

Middle low 296 (25.1) 103 (25.2) 193 (25.0)

Middle high 294 (24.9) 96 (23.5) 198 (25.6)

High 295 (25.0) 101 (24.8) 194 (25.1)

Missing 27 7 20

IL-10, N (%)

Low 915 (76.3) 310 (75.1) 605 (76.9)

High 285 (23.8) 103 (24.9) 182 (23.1)

Missing 8 2 6

IFNg, N (%)

Low 850 (82.6) 287 (81.1) 563 (83.4)

High 179 (17.4) 67 (18.9) 112 (16.6)

Missing 179 61 118

IL-1Ra*

Low 579 (50.1) 184 (46.0) 395 (52.2)

High 577 (49.9) 216 (54.0) 361 (47.8)

Missing 52 15 37

TNF-a, N (%)

Low 303 (25.1) 102 (24.6) 201 (25.3)

Middle low 305 (25.3) 103 (24.9) 202 (25.5)

Middle high 298 (24.7) 97 (23.4) 201 (25.3)

High 301 (24.9) 112 (27.1) 189 (23.8)

Missing 1 1 0

C-reactive protein, N (%)

Normal/low 462 (38.2) 149 (35.9) 313 (39.5)

Journal Pre-proof

4

Moderately elevated 571 (47.3) 199 (48.0) 372 (46.9)

High 175 (14.5) 67 (16.1) 108 (13.6)

Missing - - -

IL: Interleukin ; IFNg ; Interferon gamma ; IL1Ra : IL1 receptor antagonist ; TNF-a: tumor necrosis factor-alpha. The Cytokine Custom array

HS (CTK CST X, EV3881/EV3623), the Metabolic Syndrome array I (METS I, EV3755) and Metabolic Syndrome array II (METS II, EV3759/A)

were used for quantification of IL-1a, IL-1b, IL-4, IL-8, IL-10, IFNg, IL-1Ra (CTK), IL-6, TNFa (METSI), and CRP (METSII). If a significant

proportion of marker values across the cohort distribution fell below the sensitivity threshold for the respective assay, continuous values were

dichotomized as low vs high according to whether they were below vs above the sensitivity threshold, respectively. If the sensitivity threshold

for an individual assay was relatively low respective to the distribution, continuous values were categorized according to the quartile (Q)

distribution of the cohort as “low” (Q1), “middle low” (Q2), “middle high” (Q3), and “high” (Q4). *Sensitivity threshold not available, continuous

values dichotomized according to median (i.e., “high” vs “low” according to whether they were above vs below median, respectively). Ranges

and units for each category and variable are available in Supplementary Table 2.

Journal Pre-proof

5

Table 3. Clinical and bio-behavioral model of global fatigue of clinical importance 2 years after diagnosis.

Variable (all assessed at diagnosis of breast cancer) Odds

Ratio

95% CI p

Lower Upper

Fatigue of clinical importance at diagnosis*, Yes vs. No 3.99 2.81 5.66 <.0001

Age, continuous (per 1-year decrement) 1.02 1.01 1.03 0.0021

Tobacco use behavior, Former vs. Never 0.96 0.68 1.35 0.7991

Tobacco use behavior, Current vs. Never 1.81 1.26 2.58 0.0012

Insomnia*, continuous (per 10-unit increment) 1.08 1.04 1.13 0.0002

Pain*, continuous (per 10-unit increment) 1.12 1.04 1.21 0.0023

IL-6, middle low vs. low 1.27 0.87 1.86 0.2234

IL-6, middle high vs. low 1.15 0.78 1.69 0.4957

IL-6, high vs. low 2.06 1.40 3.03 0.0002

Intercept 0.49 0.22 1.05 0.0672

Naïve AUC (95% CI)

Optimism-corrected AUC

0.75 (0.72-0.78)

0.74

CI: Confidence Interval; AUC: Area Under the Curve. *Scored according to the European Organisation for Research and Treatment of Cancer

(EORTC) Quality of Life Questionnaire (QLQ)-C30 (a score of ≥40/100 indicates fatigue of clinical importance). Journal Pre-proof