Epidemiology of cancer-related fatigue in the Swedish twin registry

Authors

  • Michael J. Forlenza Ph.D., M.P.H.,

    1. Faculty of Health Sciences, Simon Fraser University, Burnaby, British Columbia, Canada
    Search for more papers by this author
  • Per Hall M.D., Ph.D.,

    1. Department of Medical Epidemiology and Biostatistics, Karolinska Institute, Stockholm, Sweden
    Search for more papers by this author
  • Paul Lichtenstein Ph.D.,

    1. Department of Medical Epidemiology and Biostatistics, Karolinska Institute, Stockholm, Sweden
    Search for more papers by this author
  • Birgitta Evengard M.D., Ph.D.,

    1. Department of Laboratory Medicine, Karolinska Institute at Karolinska University Hospital Huddinge, Stockholm, Sweden
    Search for more papers by this author
  • Patrick F. Sullivan M.D.

    Corresponding author
    1. Department of Genetics, Psychiatry, and Epidemiology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
    • Department of Genetics, CB 7264, 4109 Neurosciences Research Building, University of North Carolina, Chapel Hill, NC 27599-7264
    Search for more papers by this author
    • Patrick F. Sullivan is a consultant for GlaxoSmithKline on the epidemiology and genetics of neuropsychiatric illness.

    • Fax: (919) 966-3630


Abstract

BACKGROUND

Estimates of the prevalence of cancer-related fatigue (CRF) are wide, and data suggest that fatigue is more prevalent among cancer patients than among the general population. However, most studies examining the prevalence of CRF have been hospital-based or clinic-based studies, which often are subject to bias.

METHODS

Point prevalence and prevalence odds ratios of fatigue were estimated using data from a large, population-based cohort that was screened for fatigue and linked with national registry-based data about cancer. Prevalence odds ratios and 95% confidence intervals were calculated using logistic regression with general estimating equations.

RESULTS

Approximately 23% of cancer registrants reported abnormal fatigue in the previous 6 months, 19% reported abnormal fatigue that lasted for at least 1 month, 14% reported abnormal fatigue that lasted at least 6 months, and 11% reported abnormal fatigue that lasted at least 6 months and caused significant functional impairment. Individuals who were listed in the cancer registry within the last 5 years were more likely to report experiencing fatigue than individuals who were not listed. There was an elevated prevalence of fatigue among those who were registered with carcinomas of the lung, uterine cervix, colon-rectum, ovaries, and prostate. Both women and men who were listed recently in the cancer registry were more likely to experience any level of fatigue than the comparison group. However, a greater proportion of women experienced fatigue relative to men.

CONCLUSIONS

A greater proportion of individuals who were listed in a national cancer registry reported experiencing fatigue compared with individuals in the general population. Cancer 2005. © 2005 American Cancer Society.

Cancer-related fatigue (CRF) is the persistent feeling of overwhelming tiredness related to cancer or cancer treatment that interferes with usual functioning in daily life.1 It often is unrelated to physical activity and is not relieved by sleep or rest.2 CRF may be the most common side effect of cancer and cancer treatment and may persist long after treatment has been completed. Patients report that fatigue negatively impacts their quality of life in multiple domains.3

CRF has a number of correlates, including pain, depression, and sleep disturbance, and may result from cancer treatments, such as radiotherapy, chemotherapy, and treatment with biologic response modifiers, such as interferon-γ.2, 4, 5 Clinically, CRF may result in decreased adherence to therapy and may limit participation in treatment protocols. CRF frequently is most severe during and immediately after cancer treatment and tends to dissipate over time. However, a substantial number of cancer survivors still struggle with debilitating levels of fatigue long after treatment has ended and when they are considered disease free.6

Estimates of the prevalence of CRF in cancer patients are wide: from 4% prior to treatment to 91% immediately after the completion of treatment.7 Recent data suggest that fatigue is more prevalent among cancer patients than in the general population,8 especially in the presence of anemia.9 However, the majority of studies examining the prevalence of CRF have been hospital-based or clinic-based studies, which often are subject to ascertainment and selection biases.

A recently published evidence report10 cited only two population-based studies that provided estimates of the prevalence of CRF.3, 11 Vogelzang and colleagues3 reported that the prevalence of CRF ranged from 11% to 32% based on fatigue frequency, and Cella and colleagues11 reported that the prevalence of CRF was 17% based on proposed diagnostic criteria for CRF.12 However, both of those studies were uncontrolled and, thus, could not estimate the relative risk of fatigue in cancer patients compared with population-based controls.

The objectives of this report were to estimate the point prevalence of fatigue among patients who were registered in a national cancer registry and to answer several related questions: 1) Are individuals listed in the Swedish Cancer Registry (SCR) more likely to experience any type of fatigue than those who are not listed in the SCR? 2) Among those listed in the SCR, does the prevalence of fatigue differ by registered cancer site? 3) What are the effects of gender on the proportion of SCR registrants reporting fatigue, and are they similar to the general population? We pursued these questions in the context of a large, population-based cohort that was screened for fatigue and linked with national registry-based diagnostic data about cancer.

MATERIALS AND METHODS

Participants

The Swedish Twin Registry (STR) is the largest population-based twin registry in the world and contains information on > 160,000 twins. The STR, as a large, population-based cohort, represents an ideal resource for epidemiologic research, and numerous epidemiologic studies have linked the STR and the SCR in an effort to understand various behavioral, environmental, and heritable risk factors for cancer.13–15 It is noteworthy that research has demonstrated that twins are not at excess risk for cancer due to their twin status, making generalizability to the population at large plausible.16

The Screening Across the Lifespan Twin (SALT) cohort study was an effort to contact and interview by telephone all twins in the STR who were born in 1958 and earlier.14, 17 The full study began in March 1998 and continued through December 2002. Recruitment procedures are described in detail elsewhere.14 In brief, approximately 1000 pairs of twins were selected randomly each month for interviews and were sent an introductory letter describing the study. Twins were then contacted by telephone and were interviewed about 2 weeks later. All participants provided informed consent for participation.

The interview was conducted by trained interviewers using a computer-assisted telephone interviewing system. Data were collected on demography, birth and family information, zygosity, common illnesses and medications, and health-related behaviors. Participants also were screened for a number of diseases and disorders, including chronic fatigue. Fatigue questions were asked only of those ages 42–64 years to control for the potentially confounding effects of aging.18 The overall response rate for the interview was 76%.

Assessment of fatigue

A screening module for fatigue was developed as part of the SALT interview and was based on the Centers for Disease Control and Prevention (CDC) consensus criteria for chronic fatigue syndrome.19 Participants were classified as fatigued if they answered yes to the question, “have you felt abnormally tired during the last 6 months?” Those who endorsed this item were then asked about both the continuity and the duration of fatigue. Participants were classified further as “impaired” if they considered themselves “too tired to live a normal life,” if their fatigue had caused social problems, or if their fatigue had caused at least a 25% work incapacity.

Data linking

The SCR receives information on all newly registered patients from six regional registries and is the central repository for all cancer registration data in Sweden. Overall reporting to the SCR is estimated at 96% of all diagnosed patients (retrieved November 15, 2004 from http://www.sos.se/epc/english/cancereng.htm). Tumor site codes are available for the entire registry as International Classification of Diseases (7th revision) codes.

Each individual born in Sweden has a unique identification number given at birth that is used in all health care registries throughout Sweden. This number was used to identify those individuals participating in SALT who also were registered in the SCR. Once they were identified, cancer-relevant data were extracted from the SCR and merged with fatigue-related data from the SALT database. Individuals in the cancer registry who reported fatigue prior to their first diagnosis of cancer were excluded from the analyses of cancer-related fatigue to minimize misclassification biases resulting from undetected disease. Individuals with benign (i.e., in situ) carcinomas of the breast and cervix and those with nonmelanoma skin carcinomas also were excluded from this analysis.

Measures: Fatigue-related definitions

Based on the CDC consensus criteria for chronic fatigue syndrome,19 we created several definitions of fatigue with increasing severity: Fatigue was defined as the presence of self-reported, abnormal tiredness in the last 6 months. Prolonged fatigue was defined as the presence of fatigue with a duration ≥ 1 month. Chronic fatigue-A (CF-A) was defined as the presence of fatigue with a duration ≥ 6 months, and chronic fatigue-B (CF-B) was defined as CF-A with impairment.

Statistical analyses

We conducted statistical analyses using SAS software (version 8.02; SAS Institute Inc., Cary, NC) and SPSS software (version 10.1.1; SPSS Inc., Chicago, IL). We determined between group differences in covariates with a series of two-sided t tests (continuous variables) and chi-square tests (categorical variables). To account for the clustering of twins within pairs, prevalence odds ratios (POR) and 95% confidence intervals (95% CI) were calculated using logistic regression (PROC GENMOD) with generalized estimating equations (GEE)20 and were adjusted for covariates as appropriate. The GEE approach was introduced by Liang and Zeger21 as an extension of generalized linear models. The GEE procedures account for clusters of correlated data and increase the efficiency of estimation by correcting the standard errors of the within-patients effects. Without GEE, standard errors for between-patients effects tend to be underestimated, and standard errors for within-patients effects tend to be overestimated.22

RESULTS

Point Prevalence of Fatigue

Our analytic sample consisted of 30,525 individuals who had complete data on all variables: 29,422 individuals (96.4%) were not registered in the SCR, and 1103 individuals (3.6%) had ≥ 1 entries in the cancer registry. Approximately 52% of the sample was female. The mean age at the time of the SALT interview was 53.7 ± 5.8 years, and the average number of years of education was 11.1 ± 3.1 years. Among those who reported fatigue, the average age at the onset of fatigue was 50.4 ± 7.4 years (Table 1). Among those who were listed in the SCR, the average age at registration was 47.0 ± 9.5 years. The average age of fatigue onset among those in the SCR was 54.7 ± 5.4 years, and the average age of fatigue onset among those who were not listed in the SCR was 50.3 ± 7.5 years.

Table 1. Sample Characteristics
CharacteristicOverall sampleIn cancer registry n = 1103, 3.6%Not in cancer registry n = 29,422, 96.4%
FatigueaNo fatigueFatigueaNo fatigue
  • SD: standard deviation.

  • a

    Fatigue was defined as any self-reported, abnormal tiredness in the preceding 6 months.

  • b

    Chi-square (1, 1103) = 5.0 (P = 0.02).

  • c

    Chi-square (1, 29,422) = 379.0 (P < 0.0001).

  • d

    F (1, 29,420) = 167.8 (P < 0.0001).

  • e

    F (1, 29,337) = 23.0 (P < 0.0001).

  • f

    F (1, 6056) = 86.3 (P < 0.0001).

  • g

    F (1, 1101) = 5.17 (P = 0.02).

Total no. of patients (%)30,525 (100.0)256 (23.2)847 (76.8)5802 (19.7)23,620 (80.3)
No. of female patients (%)15,971 (52.3)213 (85.2)b583 (68.7)b3660 (63.1)c11,533 (48.8)c
Mean age in yrs at interview ± SD53.7 ± 5.855.9 ± 5.156.4 ± 5.452.7 ± 5.6d53.8 ± 5.8d
Mean yrs of education ± SD11.1 ± 3.111.1 ± 3.210.8 ± 3.111.3 ± 3.0e11.1 ± 3.1e
Mean age in yrs at onset of fatigue ± SD50.4 ± 7.454.7 ± 5.4f50.3 ± 7.5f
Mean age in yrs at first cancer registration ± SD47.0 ± 9.548.2 ± 9.0g46.7 ± 9.7g

There was a greater proportion of individuals who reported all levels of fatigue among those who were registered in the SCR compared with those who were not registered (Fig. 1). Of the 1103 individuals in the SCR, 23.2% reported experiencing any fatigue, 19.4% reported prolonged fatigue, 13.7% reported CF-A, and 11.2% reported CF-B. Of the 29,422 individuals who were not registered in the SCR, 19.7% reported experiencing any fatigue, 15.4% reported prolonged fatigue, 10.6% reported CF-A, and 7.9% reported CF-B. Among those who reported any fatigue, the prevalence difference (95% CI) between those in the SCR and those not in the SCR was 3.5% (95% CI, 0.96–6.02; P = 0.0043). Among those who reported prolonged fatigue, the prevalence difference (95% CI) between those in the SCR and those not in the SCR was 4.1% (95% CI, 1.7–6.6; P = 0.0004). Among those who reported CF-A, the prevalence difference (95% CI) between those in the SCR and those not in the SCR was 3.6% (95% CI, 1.3–5.8; P = 0.0006). Finally, among those who reported CF-B, the prevalence difference (95% CI) between those in the SCR and those not in the SCR was 3.8% (95% CI, 1.6–5.9; P = 0.00001).

Figure 1.

This graph illustrates the point prevalence of any fatigue (the presence of self-reported, abnormal tiredness in the last 6 mos), prolonged fatigue (the presence of fatigue with a duration ≥ 1 mo), chronic fatigue-A (the presence of fatigue with a duration ≥ 6 mos), and chronic fatigue-B (chronic fatigue-A with impairment).

Associations between Level of Fatigue and Cancer Registration Status

To determine the associations between differing levels of fatigue and being listed in the SCR, we used logistic regression (PROC GENMOD) with GEE to calculate both unadjusted and adjusted PORs with 95% CIs. Compared with individuals who were not in the SCR, individuals in the SCR were significantly more likely to report any fatigue (adjusted POR, 1.23; 95% CI, 1.06–1.42), prolonged fatigue (adjusted POR, 1.26; 95% CI, 1.08–1.47), CF-A (adjusted POR, 1.24; 95% CI, 1.04–1.48), and CF-B (adjusted POR, 1.35; 95% CI, 1.11–1.63) adjusting for age at SALT interview, years of education, and gender (Table 2).

Table 2. Unadjusted and Adjusted Prevalence Odds Ratios with 95% Confidence Intervals for Levels of Fatigue by Cancer Registration Status
Fatigue levelaIn cancer registry (n = 1103)Not in cancer registry (n = 29,422)POR (95% CI)
No.%No.%UnadjustedAdjustedb
  • POR: prevalence odds ratio; 95% CI: 95% confidence interval.

  • a

    Fatigue was defined as any self-reported, abnormal tiredness in the preceding 6 months. Prolonged fatigue was defined as the presence of fatigue >1 month but <6 months. Chronic fatigue-A was defined as the presence of fatigue >6 months. Chronic fatigue-B was defined as the presence of chronic fatigue with impairment.

  • b

    Adjusted for age at interview, years of education, and gender.

No fatigue84776.823,62080.3ReferentReferent
Any fatigue25623.2580219.71.25 (1.09–1.44)1.23 (1.06–1.42)
Prolonged fatigue21419.4451715.41.35 (1.16–1.44)1.26 (1.08–1.47)
Chronic fatigue-A15113.7311210.61.35 (1.14–1.61)1.24 (1.04–1.48)
Chronic fatigue-B12411.223237.91.49 (1.23–1.81)1.35 (1.11–1.63)

Next, we stratified our sample by the length of time between their original cancer registration and the SALT interview, hypothesizing that fatigue would be most prevalent among those listed with a more recent diagnosis (Table 3). Approximately 41% of the SCR registrants were registered in the SCR within the 5 years prior to the SALT interview. Compared with individuals who were not registered in the SCR, recent SCR registrants were significantly more likely to report any fatigue (adjusted POR, 1.81; 95% CI, 1.48–2.22), prolonged fatigue (adjusted POR, 1.80; 95% CI, 1.45–2.24), CF-A (adjusted POR, 1.63; 95% CI, 1.27–20.9), and CF-B (adjusted POR, 1.76; 95% CI, 1.33– 2.31). In contrast, there were no significant associations between the level of fatigue and being listed in the SCR for > 5 years, suggesting that the effect was carried largely by those with a more recent SCR registration. Consistent with this, further analyses showed that, compared with individuals who were registered for > 5 years, individuals who were registered with the SCR for < than 5 years were significantly more likely to report any fatigue (adjusted POR, 2.01; 95% CI, 1.51–2.67), prolonged fatigue (adjusted POR, 1.95; 95% CI, 1.44–2.64), CF-A (adjusted POR, 1.65; 95% CI, 1.17–2.33), and CF-B (adjusted POR, 1.66; 95% CI, 1.14–2.41).

Table 3. Unadjusted and Adjusted Prevalence Odds Ratios with 95% Confidence Intervals for Levels of Fatigue by Cancer Registration Status Stratified by Length of Time in the Swedish Cancer Registry
Fatigue levelaIn cancer registryNot in cancer registryPOR (95% CI)
No.%No.%UnadjustedAdjustedb
  • POR: prevalence odds ratio; 95% CI: 95% confidence interval.

  • a

    Fatigue was defined as any self-reported, abnormal tiredness in the preceding 6 months. Prolonged fatigue was defined as the presence of fatigue >1 month but <6 months. Chronic fatigue-A was defined as the presence of fatigue >6 months. Chronic fatigue-B was defined as the presence of chronic fatigue with impairment.

  • b

    Adjusted for age at interview, years of education, and gender.

In SCR < 5 yrs      
 No fatigue31869.923,62080.3ReferentReferent
 Any fatigue13730.1580219.71.78 (1.46–2.17)1.81 (1.48–2.22)
 Prolonged fatigue11525.3451715.41.88 (1.52–2.33)1.80 (1.45–2.24)
 Chronic fatigue-A7817.1311210.61.76 (1.38–2.25)1.63 (1.27–2.09)
 Chronic fatigue-B6414.123237.91.91 (1.46–2.50)1.76 (1.33–2.31)
In SCR > 5 yrs      
 No fatigue52981.623,62080.3ReferentReferent
 Any fatigue11918.4580219.70.93 (0.77–1.14)0.89 (0.73–1.09)
 Prolonged fatigue9915.3451715.41.02 (0.82–1.26)0.93 (0.75–1.16)
 Chronic fatigue-A7311.3311210.61.09 (0.85–1.39)0.99 (0.78–1.27)
 Chronic fatigue-B609.323237.91.21 (0.93–1.58)1.08 (0.83–1.42)

Associations between Chronic Fatigue and Specific Cancer Site

We also investigated whether specific cancers were associated differentially with the prevalence of chronic fatigue. We considered seven types of cancer with sufficient numbers to enable statistical analysis: lung carcinoma, cervical carcinoma, colorectal carcinoma, ovarian carcinoma, prostate carcinoma, breast carcinoma, and malignant melanoma. Table 4 shows the crude and adjusted POR and 95% CIs for CF-A and CF-B among patients with these disease types. Compared with individuals who were without cancer, those registered with lung carcinoma had the highest prevalence odds of developing CF-A (adjusted POR, 3.28; 95% CI, 1.28–8.44) or CF-B (adjusted POR, 3.71; 95% CI, 1.35–10.15) despite the small numbers of individuals who were diagnosed with lung carcinoma in this sample. Men registered with prostate carcinoma were more than twice as likely to report CF-A (adjusted POR, 2.24; 95% CI, 0.98–5.10) and were significantly more likely to report CF-B (adjusted POR, 2.57; 95% CI, 1.03–6.42). Those registered with cervical carcinoma also were significantly more likely to report CF-A (adjusted POR, 1.22; 95% CI, 1.00–1.48) or CF-B (adjusted POR, 1.27; 95% CI, 1.03–1.56). There were no significant associations between chronic fatigue and colorectal carcinoma, ovarian carcinoma, breast carcinoma, or malignant melanoma. There were insufficient numbers of registered individuals for further stratification by length of time in the SCR.

Table 4. Unadjusted and Adjusted Prevalence Odds Ratios and 95% Confidence Intervals for Chronic Fatigue-A and Chronic Fatigue-B (with Impairment) by Selected Disease Sites
Disease siteICD-7 code(s)No. of patientsPOR (95% CI)a
Chronic fatigue-AChronic fatigue-B
UnadjustedAdjustedbUnadjustedAdjustedb
  • POR: prevalence odds ratio; 95% CI: 95% confidence interval; ICD-7: International Classification of Diseases, 7th edition.

  • a

    Chronic fatigue-A was defined as the presence of fatigue >6 months. Chronic fatigue-B was defined as the presence of chronic fatigue with impairment.

  • b

    Analyses among patients with cervical, breast, and ovarian carcinoma were adjusted for age at interview and years of education. Men were removed from the comparison groups. Analyses among patients with colorectal and lung carcinoma were adjusted for age at interview, years of education, and gender.

Lung carcinoma162223.16 (1.24–8.04)3.28 (1.28–8.44)3.49 (1.28–9.54)3.71 (1.35–10.15)
Cervical carcinoma171471.22 (1.01–1.48)1.22 (1.00–1.48)1.27 (1.03–1.57)1.27 (1.03–1.56)
Colorectal carcinoma152–154711.43 (0.73–2.79)1.50 (0.77–2.92)1.72 (0.85–3.49)1.84 (0.91–3.75)
Ovarian carcinoma175271.14 (0.63–3.15)1.42 (0.63–3.18)1.81 (0.80–4.08)1.86 (0.83–4.18)
Prostate carcinoma177462.13 (0.94–4.82)2.24 (0.98–5.10)2.36 (0.95–5.86)2.57 (1.03–6.42)
Breast carcinoma1702830.96 (0.69–1.34)0.99 (0.71–1.39)1.15 (0.81–1.62)1.20 (0.85–1.70)
Melanoma190800.61 (0.28–1.29)0.60 (0.28–1.27)2.18 (0.81–5.87)2.23 (0.73–6.86)

Associations between Level of Fatigue and Cancer Status Stratified by Gender

In an effort to understand the factors driving the changes in the adjusted prevalence odds ratios between level of fatigue and cancer registration status, we calculated the crude and adjusted POR for levels of fatigue stratified by gender and length of time in the SCR (Table 5). Women who were registered in the SCR within the past 5 years were significantly more likely to report any fatigue (adjusted POR, 1.78; 95% CI, 1.40–2.26), prolonged fatigue (adjusted POR, 1.73; 95% CI, 1.34–1.2.23), CF-A (adjusted POR, 1.58; 95% CI, 1.18–2.12), or CF-B (adjusted POR, 1.75; 95% CI, 1.28–2.38) than women who were not listed in the SCR. However, there were no associations between fatigue and cancer registration in women who were registered in the SCR for > 5 years. This pattern of findings was similar among the men. Men who were registered in the SCR for < 5 years were significantly more likely to report any fatigue (adjusted POR, 1.83; 95% CI, 1.23–2.71) and prolonged fatigue (adjusted POR, 1.94; 95% CI, 1.27–2.94) and were more likely to report CF-A (adjusted POR, 1.72; 95% CI, 0.94–3.14) or CF-B (adjusted POR, 1.78; 95% CI, 0.97–3.27) compared with men without cancer, despite smaller sample sizes. Again, there were no associations between any level of fatigue and cancer registration in men who were registered in the SCR for > 5 years.

Table 5. Unadjusted and Adjusted Prevalence Odds Ratios and 95% Confidence Intervals for Level of Fatigue by Cancer Registration Status Stratified by Gender and Length of Time in the Swedish Cancer Registry
Fatigue levelaIn cancer registryNot in cancer registryPOR (95% CI)
No.%No.%UnadjustedAdjustedb
  • POR: prevalence odds ratio; 95% CI: 95% confidence interval.

  • a

    Fatigue was defined as any self-reported, abnormal tiredness in the preceding 6 months. Prolonged fatigue was defined as the presence of fatigue >1 month but <6 months. Chronic fatigue-A was defined as the presence of fatigue >6 months. Chronic fatigue-B was defined as the presence of chronic fatigue with impairment.

  • b

    Adjusted for age at interview and years of education.

Women      
 In cancer registry < 5 yrs      
  No fatigue20666.211,53375.9ReferentReferent
  Any fatigue10533.8366024.11.63 (1.29–2.06)1.78 (1.40–2.26)
  Prolonged fatigue8828.3294219.41.65 (1.28–2.12)1.73 (1.34–2.23)
  Chronic fatigue-A6019.3203113.41.56 (1.17–2.07)1.58 (1.18–2.12)
  Chronic fatigue-B5216.7161410.61.70 (1.26–2.30)1.75 (1.28–2.38)
 In cancer registry > 5 yrs      
  No fatigue37780.711,53375.9ReferentReferent
  Any fatigue9019.3366024.10.78 (0.62–0.98)0.84 (.066–1.05)
  Prolonged fatigue7716.5294219.40.85 (0.66–1.08)0.88 (0.69–1.13)
  Chronic fatigue-A5511.8203113.40.89 (0.67–1.17)0.91 (.069–1.21)
  Chronic fatigue-B4910.5161410.61.01 (0.75–1.36)1.05 (0.78–1.14)
Men      
 In cancer registry < 5 yrs      
  No fatigue11277.812,08784.9ReferentReferent
  Any fatigue3222.2214215.11.60 (1.08–2.36)1.83 (1.23–2.71)
  Prolonged fatigue2718.8157511.11.83 (1.21–2.78)1.94 (1.27–2.94)
  Chronic fatigue-A1812.510817.61.73 (1.05–2.84)1.72 (0.94–3.14)
  Chronic fatigue-B128.37095.01.72 (0.94–3.14)1.78 (0.97–3.27)
 In cancer registry > 5 yrs      
  No fatigue15284.012,08784.9ReferentReferent
  Any fatigue2916.0214215.11.07 (0.72–1.59)1.15 (0.77–1.72)
  Prolonged fatigue2212.2157511.11.11 (0.71–1.73)1.16 (0.74–1.81)
  Chronic fatigue-A189.910817.61.33 (0.82–2.17)1.37 (0.84–2.23)
  Chronic fatigue-B116.17095.01.22 (0.67–2.24)1.26 (0.69–2.31)

Finally, comparisons between women and men who were listed in the SCR for < 5 years showed that women were more likely than men to report any fatigue (Table 6) (adjusted POR, 1.66; 95% CI, 1.04–2.64) and prolonged fatigue (adjusted POR, 1.65; 95% CI, 1.01–2.69), CF-A (adjusted POR, 1.52; 95% CI, 0.85–2.71), or CF-B (adjusted POR, 2.14; 95% CI, 1.09–4.18). There were no associations between any level of fatigue and being registered in the SCR for > 5 years.

Table 6. Unadjusted and Adjusted Prevalence Odds Ratios with 95% Confidence Intervals Comparing Women and Men in the Swedish Cancer Registry for Level of Fatigue Stratified by Length of Time in the Swedish Cancer Registry
Fatigue levelaWomenMenPOR (95% CI)
No.%No.%UnadjustedAdjustedb
  • POR: prevalence odds ratio; 95% CI: 95% confidence interval.

  • a

    Fatigue was defined as any self-reported, abnormal tiredness in the preceding 6 months. Prolonged fatigue was defined as the presence of fatigue >1 month but <6 months. Chronic fatigue-A was defined as the presence of fatigue >6 months. Chronic fatigue-B was defined as the presence of chronic fatigue with impairment.

  • b

    Adjusted for age at interview and years of education.

In cancer registry < 5 yrs      
 No fatigue20666.211277.8ReferentReferent
 Any fatigue10533.83222.21.59 (0.98–2.57)1.66 (1.04–2.64)
 Prolonged fatigue8828.32718.81.59 (0.98–2.56)1.65 (1.01–2.69)
 Chronic fatigue-A6019.31812.51.62 (0.93–2.85)1.52 (0.85–2.71)
 Chronic fatigue-B5216.7128.32.21 (1.14–4.30)2.14 (1.09–4.18)
In cancer registry > 5 yrs      
 No fatigue37780.715284.0ReferentReferent
 Any fatigue9019.32916.01.25 (0.79–1.97)1.26 (0.80–1.99)
 Prolonged fatigue7716.52212.21.43 (0.86–2.37)1.45 (0.87–2.41)
 Chronic fatigue-A5511.8189.91.21 (0.69–2.12)1.23 (0.70–2.15)
 Chronic fatigue-B4910.5116.11.81 (0.92–3.56)1.84 (0.94–3.63)

DISCUSSION

This study represents one of the largest population-based assessments of CRF and, to our knowledge, is the first to present estimates of the relative prevalence (i.e., PORs) for experiencing fatigue among those listed in a national cancer registry as being diagnosed with cancer. Our results show that approximately 23% of individuals listed in the cancer registry report experiencing abnormal fatigue in the last 6 months, 19% report abnormal fatigue that lasted for at least 1 month, 14% report abnormal fatigue that lasted at least 6 months, and 11% report abnormal fatigue that lasted at least 6 months and caused significant functional impairment in their daily lives. These data show that estimates of the prevalence of fatigue are highly dependant on how fatigue is defined and that the personal experience of fatigue can be a function of both fatigue duration and fatigue severity.

Our prevalence estimates generally agree with those of Cella et al.11 and Vogelzang et al..3 Minor differences are likely due to differing definitions of fatigue as well as differing sampling time frames. For example, Cella et al. defined CRF as present if cancer patients and survivors had ≥ 6 symptoms every day or nearly every day during the same 2-week period12 in the past month, whereas Vogelzang et al. asked about the frequency of fatigue at any time during the disease course and treatment. In contrast, our definitions of fatigue focused on the duration of fatigue within the prior 6 months and whether or not it was associated with impairment in daily life. Efforts aimed at establishing standard diagnostic criteria for CRF should help clarify this important issue.

Our estimates of the relative prevalence for fatigue indicate that individuals listed in the cancer registry are more likely to experience fatigue than individuals not listed in the registry, especially if their cancer was registered within the last 5 years and assuming that registration closely follows diagnosis. The increased prevalence odds of reporting fatigue among SCR registrants who were registered within the past 5 years ranged from 63% to 81%, depending of the severity of fatigue. This is not surprising and is in agreement with the published literature from clinic-based studies. It is noteworthy that these results suggest that an increased proportion of patients diagnosed with cancer experience fatigue for at least several years after diagnosis or treatment, highlighting the need for more detailed, prospective studies of CRF fatigue across the entire cancer trajectory.

It did not appear that the relative prevalence of fatigue was increased in patients who were registered with the SCR longer than 5 years ago. However, the effects of cancer on fatigue may have been diluted in our sample because of the extended period between registered diagnosis and interview that, in some patients, was > 30 years. As survivors progress beyond the 5-year mark the likelihood that they will develop prolonged or chronic fatigue should approach that of the general population, and subsequent episodes of fatigue may be due more to aging, comorbid disease, or lack of conditioning than cancer-specific disease or treatment effects. In addition, long-term survivors of cancer are a select group, and it is possible that those who are most likely to experience fatigue have a greater disease burden and are more likely to die early. Longitudinal studies currently are lacking that include multiple assessments and sufficient follow-up to provide adequate answers to these questions.

This also is a first attempt at estimating the relative prevalence of fatigue by different cancer sites. Although there are small numbers of patients when stratifying by specific site, results show significantly greater prevalence of fatigue among those registered with carcinomas of the lung and cervix and a nonsignificantly greater prevalence for carcinomas of the colorectum, ovaries, and prostate. Although it is tempting to speculate about why individuals with these cancers may be more susceptible to experiencing chronic fatigue than individuals with other cancers, we have little disease-related data and no treatment data to inform our explanations.

There also were too few registered individuals to stratify by length of time since initial registration. There may be both disease-specific and treatment-related mechanisms that explain the differences in the relative prevalence of fatigue. It was surprising that there were no differences in the prevalence of chronic fatigue between women registered with breast cancer and women without breast cancer, because others have shown increased fatigue among breast cancer patients and survivors.23, 24 However, Bower et al.6 found that breast cancer survivors were less fatigued than age-matched women in the general population but were more fatigued than women who participated in the National Surgical Adjuvant Breast and Bowel Project Breast Cancer Prevention Trial.25 This may have been due to intensive early-detection efforts, in which women with breast cancer were diagnosed at an earlier stage of disease; thus, the contribution of disease processes to fatigue were decreased. This would not be the case with cervical carcinoma, because patients those with carcinoma in situ were excluded from the analysis.

Women who were registered recently in the SCR were more likely to experience any level of fatigue than women in the comparison group, as expected. This is consistent with previous research suggesting that women are more likely to experience fatigue.26–28 However, this was not the case among women who were registered in the SCR for > 5 years, and it is possible that the high prevalence of fatigue among women in the general population overwhelms any additional effects of cancer on fatigue, thus removing any differences in relative risk. This seems likely, because our estimate of any level of fatigue among comparison women was relatively high at 24%. Similarly, our data also showed that a greater proportion of men who recently were registered with the SCR reported any level of fatigue compared with men who were not registered in the SCR. A greater proportion of men who were registered with the SCR for > 5 years also reported more fatigue, but the results were not statistically significant, due in part perhaps to the small sample size. Finally, our data indicated that a greater proportion of women experience fatigue relative to men. This effect is driven largely by the patients listed in the SCR for < 5 years. These findings also are in agreement with those of Loge et al.,28 who also reported that total fatigue scores were significantly higher among women compared with men.

Limitations

Our data set lacked important diagnostic, prognostic, and treatment-related variables that may have influenced the prevalence of fatigue among those listed in the SCR, as discussed above, and it is possible that experienced fatigue may vary by disease stage or treatment modality. For example, there is evidence that treatment with biologic response modifiers, such as interleukin 2 or interferon-α, leads to increased fatigue,29–31 and it is possible that patients receiving these treatments represent an important subgroup. However, fatigue also has been documented in patients receiving conventional chemotherapy32 and radiation.33 Inclusion of this type of information, if it had been available, may have allowed us to identify important subgroups for further analysis. An objective for future study may be to document the specific effects of differing treatment modalities and protocols on the incidence and prevalence of fatigue.

Our definition of fatigue in cancer registrants was based on diagnostic criteria for chronic fatigue syndrome.19 It is noteworthy that our definitions, although they tried to capture differing durations and severity of fatigue, did not necessarily assess CRF fully, and we lacked sufficient detail on important covariates, such as major depression and primary sleep disorders.34

In addition, our results likely are underestimates of population parameters of fatigue due to age censoring in our data set. That is, no one in our sample was older than age 64 years. Fully 64% of all cancers in Sweden are diagnosed in individuals age 65 years and older, and it is possible that our estimates of point prevalence and relative prevalence would increase if these older individuals were included in our sample.

Finally, the cross-sectional nature of these survey data do not allow us to infer causality. Furthermore, because our exposure variable, being in the cancer registry (and, by implication, having cancer), is likely to influence the average duration of fatigue, assumptions allowing the prevalence odds ratios to be interpreted as an estimate of the incidence rate ratio likely are violated.

Further Questions

Our data suggest that a larger proportion of individuals listed in a national cancer registry are likely to experience fatigue compared with the general population. It is noteworthy that our comparison group was large, population-based, and reported considerable levels of fatigue. The high prevalence of CRF supports the National Comprehensive Cancer Network recommendations for aggressive fatigue screening, follow-up, and treatment among cancer patients and survivors.1

Estimates of prevalence and prevalence odds do not address important mechanistic questions related to CRF: That is, what may be the underlying mechanisms of fatigue in cancer survivors? Are these mechanisms different than in populations without cancer? Like fatigue in the general population, CRF is a heterogeneous syndrome and may have multiple causes.2 These causes are not necessarily mutually exclusive and may differ in different patients and at different times in the cancer trajectory. The pathophysiologic mechanisms of CRF are understood poorly, although some work has been done.35–39 For instance, CRF has been linked to anemia resulting from disease processes and cytotoxic treatments. If a patient is anemic and fatigued, then they may respond favorably to pharmacologic treatment with erythropoietin.40, 41 However, not all cancer patients are anemic; the prevalence of anemia is variable by disease site and clinical definition. For example, with anemia defined as hemoglobin levels of 90.0–110.0 g/L or hematocrit at 0.27–0.33, estimates of anemia range from 32% to 90% of patients.42 Although anemia is an important clinical issue and should be assessed and treated, it must be realized that only a subset of fatigued cancer patients are anemic, and other mechanisms of CRF should be considered.

Due to a constantly increasing cancer incidence, early detection, and therapeutic improvements, cancer survivors are more prevalent than ever before. This fact underlines the necessity to identify underlying mechanisms of chronic fatigue and other late adverse health effects. Multidisciplinary, mechanism-based interventions should follow as more details regarding the risk factors and causes of CRF are uncovered, decreasing the substantial burden of fatigue among cancer patients and survivors.

Ancillary