Fatigue in colorectal cancer patients: prevalence and associated factors

Los objetivos de este estudio fueron la identificación de la superioridad y los predictores de fatiga en pacientes con Cáncer Cuello rectal (CCR). Se trata de estudio seccional con 157 pacientes de ambulatorio con CCR (edad 60±11,7 años; 54% hombres; estadio cáncer IV 44,8%). La Escala de Fatiga de Piper Revisada fue utilizada para evaluar fatiga. Datos sociodemográficos, clínicos, depresión, funcionalidad, dolor y sueño fueron evaluados. La asociación entre variables fue realizada por regresión logística. Fatiga fue reportada por 26,8% pacientes. Por la regresión logística se identificaron tres predictores: depresión (OR: 4,2; 95%IC 1,68-10,39), funcionalidad (OR: 3,2; 95%IC 1,37-7,51) y perjuicio del sueño (OR: 3,2; 95%IC 1,30-8,09). Cuando todos predictores estaban presentes, la probabilidad de ocurrencia de fatiga fue del 80%; cuando ninguno estaba presente, la probabilidad fue del 8%. La especificidad y sensibilidad del modelo fueron, respectivamente, 81,9 y 58,6%. Conociéndose la probabilidad de fatiga, por medio de la evaluación de depresión, funcionalidad y perjuicio del sueño, se vuelve posible la implementación de estrategias de prevención y tratamiento en la clínica.


Introduction
For a long time, fatigue in cancer has been studied as an isolated symptom.There were studies that aimed to identify its prevalence, the evolution according to the course of the disease, different treatments for cancer and treatments for fatigue.As these pieces of knowledge seemed relevant, new studies tried to identify the factors that could predict the presence and intensity of fatigue.
Studies that investigate the tumor site as a risk factor for fatigue or focus on patients with a single disease site are important for the development of diagnostic strategies and targeted fatigue interventions.Few studies have investigated fatigue in colorectal cancer patients, and none has identified predictors of fatigue in these patients.
Thus, aiming to fill this gap, this study was undertaken to identify predictive factors of fatigue in patients with colorectal cancer.
The prevalence of fatigue is undoubtedly high.
In a systematic review that included 27 studies, the prevalence ranged from 4% to 91%, depending on the cancer population and assessment instruments (1) .Specifically in patients under cancer treatment, it ranged from 32% to 99% (2) .The intensity of the symptom is also considerable, with mean scores ranging from 2.6 to 6.9, depending on the cancer site and staging (3)(4) .
Fatigue seems to persist even after the "cure" of cancer, since 19% to 58% of treated people without evidence of cancer reported the symptom (5)(6) .
Fatigue is a deleterious symptom that causes suffering, distress, functional limitation and predicts overall survival and quality of life (7)(8)(9) .Many factors may be related to the occurrence or aggravation of the symptom and recent studies have investigated predictors of cancer-related fatigue in non-specific site groups, such as advanced cancer, outpatients and cancer survivors.It shows that, according to the disease stage, the associated factors may vary.It seems that, in outpatients, the predictors include a reasonable amount of socio-demographic variables (female gender, high education, having a regular job and living without a partner), besides depression and performance status (10)(11)(12) .On the other hand, fatigueassociated factors in patients with advanced cancer are basically related to other signs and symptoms present in terminally ill patients (e.g.pain, dyspnea, decreased appetite, depression, weakness sensation, anxiety, low hemoglobin levels and others) (3)(4)(13)(14) . Interesingly, among cancer survivors, fatigue predictors include not only performance and symptoms, but also cognitive impairment (stress, coping, self-efficacy) (5) .
From the description, a vast variety of physiological, physical, emotional, affective, socio-demographic and treatment-related predictive factors are noted.Yet, it was observed that many studies involved patients with cancer in various sites and that, in those that included patients with different tumors, the tumor site was not a predictive factor.Therefore, this paper specifically focused on colorectal cancer patients' fatigue experience, aiming to investigate its prevalence and independent associated factors.

Methods
Cross-sectional study, involving 157 Brazilian adults with primary colon or rectal cancer.Inclusion criteria were: age 18 or more, being treated or followed in outpatient oncology services (inpatients were not included).To have or not fatigue was not an inclusion criteria.These subjects composed a convenience sample and were recruited from four outpatient oncology services at Sao Paulo city (1 public, 2 private and 1 both), July/2006 to July/2007.The sample representation from each service ranged between 15.3% and 33.1%.The number of patients that refused to participate was considered low (8 patients).
Oncologists identified the subjects at the participating oncology services.After a medical appointment, the researcher recruited the patients, checked the inclusion criteria and invited them to participate in the study.All participants signed the informed consent term and the ethics committees of each of the institutions involved approved the study.Socio-demographic data and data related to the cancer, its treatment, concomitant medication and clinical signs/symptoms were obtained through interview and medical chart analysis.Pain and sleep disturbance were assessed on a numerical scale (0-10 scale).Performance status (ability to perform daily activities) was assessed using the Karnofsky Performance Status Scale (KPS).

Depression was assessed through the Beck Depression
Inventory (BDI).It is a self-report instrument that consists of 21 items to evaluate depressive symptoms.Each item is scored from 0 to 3 and the total BDI score varies between 0 and 63.Scores between 16 and 20 account for dysphoria and above 21 for depression (15) .In the current study, Cronbach's alpha equaled 0.83.
The Piper Fatigue Scale (PFS) -revised is a 22-item self-report instrument with three dimensions (behavioral, affective and sensorial/psychological).Each item is scored on a numerical scale (0-10).The dimension scores and total score vary between 0 and 10 and are obtained by adding up the score of each item and dividing this by the number of items on each dimension, or by 22 when calculating the total score.This instrument was validated for use in Brazilian cancer patients and the internal consistencies for the total scale (Cronbach's alpha=0.94)and its dimensions (Cronbach's alpha between 0.84 and 0.94) were very good in this study sample (16) .The correlation between fatigue intensity on the numerical scale (0 to 10) and the Piper Fatigue Scale-revised (0 to 10) was also very good (Pearson r=0.748; p<0.001).Fatigue was analyzed as a nominal variable and the cut-off point was established at score 4 (<4 = no fatigue; ≥4 = with fatigue), based on the consensus of the National Comprehensive Cancer Network (2007) (17) and on the percentile distribution of the Piper scores, based on the present study findings.
When the subjects finished answering the self-report instruments, the researcher checked for missing data and, in case data were missing, the researcher handed the items back to the subjects so they could complete the instrument.Nevertheless, missing data occurred.On the BDI, a total of two items were blank and the mean score of that item all other subjects had answered was used to replace the missing data.On the PFS, three items were blank and the patient's mean subscale score of all items in the same dimension was used to replace the missing data.Missing socio-demographic, clinical, treatment data were left blank.
In Table 1, a significant portion of subjects with altered BMI (57.6%) can be observed.Of 157 subjects, 110 (70.1%) had colon tumors (data not shown in Table ), and the others had rectal cancer.Many of the subjects had advanced stage disease (III and IV) and almost all received chemotherapy and/or radiotherapy.Nearly half of the patients (55%) had co-morbidities and 20% used antidepressants and/or anxiolytics.Scores compatible with dysphoria and depression were present in 15.4% of the subjects.
Among those who reported fatigue, the mean score was 5.8 (SD=1.3), the median 5.4 and the minimum and maximum scores 4.2 and 9.6.

Associated and predictive factors
In univariate analysis, fatigue showed association with five out of twenty investigated variables: to be assisted at a public hospital, pain, sleep disturbance, depression and performance status.These five variables The sensitivity of the model identified in this study was 58.6%.The specificity level equaled 81.9%.These data indicate that, by using the Fatigue Prediction Table , constructed in this study, the chance to correctly identify a fatigued person is approximately 60%, against approximately 82% to correctly identify a person who is not fatigued.

Discussion
The identification of fatigue predictors permits the prevention, early diagnosis and development of specific actions for its control.Little was known on the prevalence of fatigue in colorectal cancer patients and there were no studies on fatigue predictors in this population, which was achieved in this search.
In general, the study sample is similar to other studies on colorectal cancer patients concerning sociodemographic, clinical and treatment data (18)(19) .The data was collected at four oncology services with different characteristics (public and private, hospitals and outpatient clinics) and this improved the representation of patients with colorectal cancer.Although a randomization of the sample was desirable, that was not feasible.Thus, the option of using four different oncology services increased external validity.Utilizing the estimated error for each of the independent predictive variables, the Fatigue Prediction Table (Table 3) could be constructed.This Table shows the probability of the individual to have fatigue based on the presence/absence of each variable.Patients who do not have any of these predictors have approximately 8% chance of having fatigue and those with the three independent predictors have approximately 80% chance.
It is observed that not only the number of factors, but also their nature influences the likelihood of having fatigue.
Depression increases the likelihood of having fatigue more than the other factors (Table 3).
were included in the multiple regression model (Table 2) and three independent predictors for fatigue were identified: sleep disturbance, depression and performance status.The risk for fatigue was higher for depression (OR=4.2) than for sleep disturbance (OR=3.2) and performance status (OR=3.2).
The cut-off score of 4 or higher, which refers to the percentile 25 of the fatigue score distribution and meets National Comprehensive Cancer Network criteria (17) , represents the fatigue present in 75% of the sample.In the literature review, it was noted that other authors also adopted the same cut-off score (20) , which is beneficial to standardize and compare assessments.In addition, having fatigue cut-off scores higher than 1, 2 or 3 contributes to assuring that fatigue has clinical relevance, which may not occur with lower cut-off scores related to minimum discomfort.
No studies on fatigue predictors in colorectal cancer patients were found, but the study data on fatigue prevalence and intensity were similar to those used in this search (18)(19)21) . Whenfatigue prevalence (26.8%) and intensity (mean=5.8;SD=1.3; median=5.4) in this research were compared to studies involving other cancer patients, it was noted that fatigue was less frequent than in studies on patients in palliative care and receiving adjuvant chemotherapy or radiotherapy (1)(2) , and higher than some cancer survivors' experience, who reported 26% (8) .A slightly higher prevalence of fatigue (33%), using the same cut-off score as this study, was observed among colorrectal cancer patients in the United Kingdom (22) .This higher prevalence might be due to the instrument used for fatigue assessment.
In a Turkish study that proposed an intervention for cancer-related fatigue among intervention group patients undergoing chemotherapy, the baseline fatigue score assessed through the Piper Fatigue Scale was 3.5 (23) .These researchers probably included scores from all patients, lowering the mean score when compared to the present study.
The univariate and multivariate analysis identified the risk and predictive fatigue factors.The only sociodemographic risk factor was the "public health service".
No study was found that analyzed the relationship between fatigue and health service, probably because the majority of published studies include subjects who live in developed countries, where the quality of health services is similar, independently of the provider.
Pain, sleep disturbance and depression were risk factors for fatigue.The mean scores of these symptoms among fatigued patients were higher than among nonfatigued patients.In the same manner, performance status was worse among fatigued than among nonfatigued subjects.Various studies in different scenarios also identified performance status, pain, depression and sleep alteration, among others, as factors related to fatigue (4,6,11,14) .
The discussion on the relation between fatigue and depression should be raised.Studies have attempted to identify cause-effect relations between these symptoms, but no conclusion has been reached.
What is clear is that there is a positive and moderate/ strong correlation between them and that the rate of comorbidity fatigue/depression is harmful.Yet, some may consider the assessment of depression in cancer patients through the BDI as doubtful, due to the number of items that investigate somatic issues, which might cause confusion between depression and cancer symptoms.But this instrument was chosen because of its wide acceptance and utilization in cancer research.
The variables gender, age, marital status, skin color, education, employment status, family income, clinical cancer site, medical treatments, body mass index, level of hemoglobin, presence of illnesses in co-morbidities, use of antidepressants or anti-anxiety drugs and colostomy did not relate to fatigue.There are studies that observed some of these variables as risk factors for fatigue, such as female gender (18) , young age (2,19) , the absence of regular job (5) , cancer treatment (2) , among others.In other studies, other variables were also not observed, such as cancer staging, presence of some co-morbidities, concomitant medications, etc. (4,24) .
The greater or minor likelihood of a factor to predict an outcome (fatigue) can be calculated using logistic regression.Thus, by evaluating sleep, depression and performance status, the probability of the patient having fatigue can be observed in the Fatigue Predictors Table ( rectal cancer was found and, among the three studies that used the logistic regression method involving patients with other tumors, only one mentioned the discriminatory capacity of the model = 0.88 (3) .In that study, conducted with men with cancers in several sites, the predictors were sadness, dizziness, pain, appetite loss, irritability and dyspnea (3) .
The risk factors "public health service" and pain were excluded from the logistic regression models.
Not every risk factor is a predictive factor, because regression analysis distinguishes those that are important for an outcome occurrence (fatigue), in an independent way, from those that seem to be important for the phenomenon, but which in fact are surrogate to others.
The findings of this search raised some issues: Based on this study, some progress can be pointed out and the implications for the practice are diverse.
The first nursing publication about fatigue in Brazil was on laryngeal cancer (25) .Now, for the first time in our country, the prevalence of fatigue was determined in patients with colorectal cancer, and a cut-off score for fatigue diagnosis was established.In an international scope, some results are innovative, as the independent fatigue predictors were not known in outpatients with colorectal cancer.This study presents the Fatigue Prediction

For
univariate analysis, Chi-Square test, Likelihood Ratio, Mann-Whitney test or T-Test were done.The variables that were associated to fatigue were inserted in the logistic regression model.The quantitative variables were analyzed through ROC curves to establish cut-off scores based on these study findings, and they were pain (cut-off score=6.0),sleep disturbance (cut-off score=5.0),performance status (cut-off score=80%), and depression (cut-off score=13).The variables were inserted in the logistic regression model, using forward stepwise.The Statistical Package for the Social Sciences-version 15.0 was used to perform the analyses and p<0.05 was preset for statistical significance.

Table 3 -
Fatigue Prediction Table: probability of fatigue occurrence in the presence of independent predictive factors* *0 = no; 1= Yes

Table 3
be questioned, since these may influence the presence and intensity of fatigue.Still, the inclusion of inpatients with colorectal cancer is desirable, considering that the results of the predictors and risk factors might be broader in this population.Such limitations need to be overcome in future studies.Despite these limitations, the researchers believe that the results can be generalized to other Brazilian or non-Brazilian cancer patients, as the patients included in this study showed a profile similar to patients included in other international studies.The identification of fatigue predictors in colorectal cancer outpatients from other countries could help to validate the findings of this study.Yet, future research might as well test the Fatigue Prediction Table.A cohort study would be the ideal method to validate the findings of this search and improve the fatigue diagnostic strategies in clinical practice.
Table, which permits investigating the probability of fatigue occurrence and contributes to symptom control.Sleep disturbance, depression and poor performance status together predicted 80% of fatigue cases in colorectal cancer situations, which is very good.Depression increased the risk of fatigue by four times, and sleep disturbance and performance status by three times.Due to the moderate sensitivity result of the model (58%), there is a considerable chance of having false negatives, which perhaps can be minimized by using a numerical scale from 0 to 10, in which score ≤4 indicates fatigue.