Patient reported outcomes (PROs) of which, CRF is a good representative, are a critical element of patient-centered care, since they provide a direct measurement of patient experiences  and also have a prognostic value . We have confirmed that CRF is a prevalent condition among BC survivors after finishing their treatment. In our sample, almost half of BC survivors reported a clinically significant score of fatigue. Contrary to the expected, objectively assessed PA and previous treatment with anthracyclines, hormone or radiotherapy were not associated with CRF. Objective tests for evaluating physical condition such as MWT, sit-to-stand and handgrip did not predict fatigue either. We found as correlates for having fatigue after treatment: being younger, higher weight and BMI, gait speed, previous therapy with trastuzumab, less time elapsed from diagnosis, reporting less walking time, having typical side effects and symptoms such as pain, dyspnea or appetite loss, having a poorer body image, having financial difficulties, cancer-related uncertainty and scoring lower in the functional scales of the EORTC-QLQ-C30 questionnaire. Though adherence to the Mediterranean diet did not predict CRF, the intake of nuts seemed to protect against fatigue. Some correlates could be interpreted as potential cause of fatigue such as weight or trastuzumab, whereas others seem to be symptoms that co-ocurred such as pain or dyspnea. It is very well known the presentation of fatigue as symptom cluster . But a role of these symptoms triggering or maintaining fatigue cannot be excluded.
Diagnosis of CRF depends on how active its search through systematic screening is. Despite of recommendations by guidelines , systematic assessment of fatigue is not being done during the follow-up of adult cancer survivors. As many others, our study shows considerable variability in fatigue suggesting that some patients are at a particular risk for experiencing fatigue whereas 9% did not report any fatigue at all. In order to estimate fatigue prevalence we considered fatigue as clinically significant, if it interfered with QoL (Supplementary Figure 1) . It is remarkable that the mean score of QoL in this cohort of BC survivors (68.9) is over the expected for the general Spanish population (66.8) and also for 50-59 year-old Spanish women (62.6) .
The ideal scale to evaluate CRF is not established and may be population dependant . Multidimensional scores offer a more comprehensive assessment by capturing multiple characteristics of fatigue. The choice of the PQ multidimensional scale was based on the fact that it was originally developed among Spanish-speaking cancer patients, showed good psychometric properties, exhibited good completion rates and had a known sensitivity to change . The comparison of PQ scores with other measures (FACIT, fatigue item in the EORTC QLQ-C30, the global EORTC QLQ score) revealed strong correlations. It is noteworthy that the mean score in the unidimensional fatigue item of the EORTC QLQ-C30 (22) was quite similar to the expected value for the general population in our country (23.9, considering the sample was constituted only by women) .
CRF cannot be entirely explained by characteristics of the disease and/or its treatment . Fatigue is multifactorial and can be influenced by social [7,31], medical , psychosocial , behavioural  and biological factors. In terms of socio-demographic factors, younger , unmarried patients [7,31] and those with a lower household income  have been reported to have a higher risk of reporting fatigue. Only age was significant among socio-demographic factors in this cohort. But old people and disadvantaged populations were not well represented in this sample.
The impact of cancer diagnosis and treatment on the development of fatigue is beyond doubt since fatigue is consistently higher among BC survivors than among their counterparts [26,31]. However, the mechanisms which mediate the appearance of fatigue have not been completely elucidated. In line with other studies, the longer the time elapsed from the diagnosis the less the fatigue reported by cancer survivors. Contrary to the expected, the role of chemotherapy and radiotherapy is controversial [15,26]. In this cohort, having been exposed to chemotherapy and specifically to anthracyclines or radiotherapy was not associated with fatigue. Neither was it, having been treated with hormonal therapy. Nevertheless, larger studies [12,31] and meta-analyses  have found an association with chemotherapy and radiotherapy. With regards to the finding that suggests a link between trastuzumab and CRF, there are few studies examining that association, rendering conflicting results [7,33]. Unexpectedly, we found a strong relationship that could not be explained by abnormal LVEF (mean value 62%) or because of concurrent concomitant treatment. Given the small number of patients receiving trastuzumab and the aforementioned conflicting results in the literature, this hypothesis should be confirmed.
Health-related quality of life (HRQOL) is a multidimensional concept encompassing physical, psychological, existential and functional aspects. Lower scores of HRQOL are found among people with fatigue. We confirmed the association identified by Abrahams  with all the EORTC QLQ-C30 functional scales (physical, role, social and cognitive functioning). Interestingly, several years post-treatment, constipation, diarrhea, appetite loss, and nausea, which represent typical side effects during cancer treatment, were more frequent in survivors with persistent fatigue. This finding had already been reported in other studies . Dyspnea has multiple causes including psychosocial factors. The observed associations between fatigue and dyspnea inspire about potential causal links and pathways of dyspnea and fatigue. Also fatigue and depressive symptoms go commonly together. That association of fatigue and depression is complex since fatigue can be a symptom of depression but fatigue can also precipitate depressed mood. Moreover, pretreatment levels of fatigue, anxiety or depression seem to predict post-treatment fatigue suggesting that biological, psychological or behavioural dysregulation was present before treatment . However, treatment strategies might only be effective for one of the two symptoms, supporting the idea that they are distinct.
Establishing an association between physical condition and fatigue was one of most important objectives in the initial design of this prospective study. Our hypothesis consisted of assuming that more aggressive, more cardiotoxic therapies, being less active and having a poorer cardiorespiratory fitness would make up the perfect scenario for developing fatigue. Physical condition of the patient was therefore, evaluated. Physical condition is composed of a) body composition, estimated by BMI, waist and hip perimeter; b) muscular strength estimated through the test sit to stand and handgrip strength; c) cardiorespiratory fitness estimated through the MWT. BMI has consistently been associated with CRF [16,26]. Weight, BMI and hip perimeter were predictors of fatigue in this cohort. This finding, that is in line with previous reports, is interesting because contrary to other correlates, it is a modifiable feature. Contrary to our hypothesis we could not describe a significant association between the estimated VO2MAX and CRF, though there was a trend towards a negative association. In fact, the gait speed which is an important part of the formula used to calculate VO2MAX was associated with fatigue. The key to justify this discrepancy could be the final heart rate which is used in the formula and it is the only “heart rate” measure which does not correlate with fatigue in our sample. However, the fact that fatigue was not associated either with muscular strength confirms that in our sample, non-anthropometric indicators of physical condition were not related to CRF. Cardiorespiratory condition has been theoretically proposed as a cause of fatigue  but has not been extensively studied in this setting . It should be noted that estimated cardiorespiratory condition in this cohort was in the range of normative values (27.0-31.4 ml/kg/min for healthy women from 50 to 59 yo).
The strong association between heart rate and fatigue unveils a very interesting physiopathological explanation. Heart rate variability (HRV) gathers many of the factors that are discussed in the literature as involved in fatigue. Thus, diagnosis and treatment of BC is associated with therapy-induced cardiovascular injury and lifestyle perturbations, leading to increased activation of the sympathetic nervous system and decreased activation of the parasympathetic nervous system. This autonomic imbalance stimulates the hypothalamic-pituitary-adrenal-axis, the renin-angiotensin-aldosterone system and the endocannabinoid system leading to increased oxidative stress and increased inflammation . Some studies had previously shown that BC survivors with cancer related fatigue have reduced HRV and elevated norepinephrine levels. suggesting an autonomic dysregulation. HRV is a marker of autonomic dysfunction and its simplest measure is heart rate (HR) (and also, recovery heart rate), being the gold standard the interval R-R measured through EKG and Holter . HRV has been linked to a variety of psychological and physical illnesses including all-cause mortality. Resting heart rate itself is associated with increased risk of all-cause mortality in the general population and all-cause mortality and breast cancer-specific mortality in patients with breast cancer .
PA is one of the few strategies which have been proved helpful for ameliorating CRF. It is disconcerting that objectively measured weekly PA was not related to fatigue scores. Neither was it when exploring vigorous PA or sedentary time. PA is closely related to physical condition and heart rate variability. Some authors attribute the benefits of the exercise to some kind of attention or the benefits of self-management  more than to the physical training itself. An important fact that could explain our results is that our population is widely active. Canadian researchers also reported lack of association between PA and fatigue for active BC survivors (more than 181 min/wk) . It is especially interesting how this association differs when PA is subjectively evaluated. Other authors had previously reported an association of fatigue and subjective but not objective impairments.
Another unexpected interesting finding was the association of a higher nut intake with lower scores of fatigue. The association between fatigue and diet has been previously explored without consistent results . Nuts are good sources of unsaturated fatty acids, but also protein, fiber, potassium, magnesium and vitamin E and have been associated with cancer and all-cause mortality. Omega-3 fatty acids (FA) like eicosapentanoic and docosahexaenoic acids are considered anti-inflammatory whereas omega-6 FA have pro-inflammatory properties. Thus, increased intake of omega-3 relative to omega-6 was associated with reduced inflammation in an observational analysis in breast cancer survivors. However, a randomized clinical trial failed to demonstrate a benefit in fatigue in BCS, with higher intake of omega-3 FA . A post-hoc analysis in our sample showed an intriguing association between nut intake and indicators of HRV (basal and recovery HR).
In conclusion, a high prevalence of CRF interfering QoL in a population of mid-term BC survivors was found. Younger and obese BC survivors seem to be prone to suffer fatigue but previous cancer therapies do not justify it completely. Though CRF was strongly correlated with emotional distress, some objective correlated findings such as markers of adiposity and a higher heart rate among fatigued women would confirm a physiological substrate beyond the psychological predisposition. A deeper insight in heart rate variability and inflammatory mechanisms is warranted. The association of CRF with trastuzumab and (low) intake of nuts should be confirmed by further studies. Increasing awareness of health practitioners about the importance of assessing CRF is crucial. On the other hand, longitudinal studies are necessary to confirm the predisposing, trigger and perpetuating factors.
Strengths: Different and consistent measures of fatigue, objective evaluation of PA and objective evaluation of physical performance and anthropometry. The clinical data are obtained from the clinical records and therefore are not self-reported.
Limitations: The study was conducted in a single hospital and the sample corresponded to a well off, educated and active population and consequently results cannot be generalized. Some other limitations are: absence of the psychologist perspective in the design, lack of a healthy control, lack of information about pretreatment levels of fatigue, anxiety depression and about coping strategies. The results would have been enriched with some information about sleep patterns, cognitive dysfunction and social support.