Fax: (813) 745-3906
Prolonged impact of chemotherapy on fatigue in breast cancer survivors
A longitudinal comparison with radiotherapy-treated breast cancer survivors and noncancer controls
Article first published online: 15 NOV 2011
Copyright © 2011 American Cancer Society
Volume 118, Issue 15, pages 3833–3841, 1 August 2012
How to Cite
Goedendorp, M. M., Andrykowski, M. A., Donovan, K. A., Jim, H. S., Phillips, K. M., Small, B. J., Laronga, C. and Jacobsen, P. B. (2012), Prolonged impact of chemotherapy on fatigue in breast cancer survivors. Cancer, 118: 3833–3841. doi: 10.1002/cncr.26226
- Issue published online: 20 JUL 2012
- Article first published online: 15 NOV 2011
- Manuscript Accepted: 31 MAR 2011
- Manuscript Revised: 22 MAR 2011
- Manuscript Received: 24 JAN 2011
- breast cancer;
In this study, the authors examined the influence of prior treatment on the course of fatigue in breast cancer survivors. Patients who received chemotherapy were expected to have greater fatigue than patients who received radiotherapy and noncancer controls 6 months after the completion of treatment, but they were expected to recover to levels similar to those of the other 2 groups 3 years later.
Patients with stage 0 through II breast cancer completed the Fatigue Symptom Inventory (FSI) and the Profile of Mood States Fatigue Scale (POMS-FAT) 6 months (T1) and 42 months (T2) after completing chemotherapy with or without radiotherapy (the CT group; n = 103) or radiotherapy only (the RT group; n = 102). An age-matched group of women with no history of cancer (the NC group; n = 193) was assessed over a similar interval.
A significant (P = .041) group × time effect for FSI severity scores revealed that fatigue worsened over time in the CT group but remained stable and lower in the RT and NC groups. There also were significant group effects for FSI days (P < .001) and POMS-FAT (P = .010) scores, indicating that fatigue was significantly greater across time in the CT group than in the NC group (POMS-FAT) or the RT and NC groups (FSI days).
Contrary to expectations, fatigue did not diminish over time in patients with breast cancer who received chemotherapy. This finding has important implications for patient education and for fatigue monitoring during follow-up. The authors concluded that future research should seek to examine possible mechanisms to explain the apparent prolonged impact of chemotherapy on fatigue in breast cancer survivors. Cancer 2012. © 2011 American Cancer Society.
Fatigue is among the most common symptoms reported by women who are treated for breast cancer.1 Studies using cross-sectional designs also demonstrate elevated fatigue after treatment completion in disease-free breast cancer survivors relative to noncancer controls2-7 and general population norms.8 Cross-sectional studies do not provide information about the course of post-treatment fatigue in breast cancer survivors. This limitation has been addressed in 7 longitudinal studies.9-15 In 4 of those studies,9, 10, 13, 15 assessments were conducted at various time points after the completion of treatment (eg, between 6 months and 70 months after treatment ended). This feature precludes the possibility of characterizing fatigue at specific time points in the post-treatment period.
All of the other longitudinal studies conducted assessments at standard time points but differed considerably in other respects. Nieboer et al14 assessed patients who were randomized to receive standard-dose or high-dose chemotherapy before treatment and 1 year, 2 years, and 3 years post-treatment. Their findings indicated that there were no changes in fatigue in either group and no differences in fatigue between groups. Fan et al11 assessed patients toward the end of chemotherapy and 1 year and 2 years later and also assessed noncancer controls at similar intervals. Patients were more fatigued than controls at the initial assessment and, although patients' fatigue improved over time, differences with controls still were evident 1 years and 2 years later. Geinitz et al12 assessed patients who received primarily radiotherapy before the start of radiotherapy and again 2 months and 2.5 years post-treatment. Their findings generally indicated no changes over time, with the exception of an increase on a global measure of fatigue between 2 months and 2.5 years post-treatment. Overall, those studies provided conflicting evidence about changes in fatigue among breast cancer survivors who received adjuvant therapy. In addition, none of those studies addressed the important issue of whether the course of fatigue differs based on whether breast cancer survivors receive chemotherapy or radiotherapy only.
In the current study, fatigue was assessed at 6 months and 42 months after the completion of adjuvant therapy in breast cancer survivors and at a similar time interval in a noncancer comparison group (the NC group). Data were used to determine whether the course of fatigue differed between women with breast cancer who received chemotherapy (either alone or with radiotherapy) compared with women with breast cancer who received radiotherapy only and women who had no history of cancer. We reported previously that fatigue was greater in the 2 patient groups at the end of treatment compared with the NC group and that, 6 months later, the chemotherapy group (but not the radiotherapy group) reported more fatigue than the NC group.16 We also reported previously that the prevalence of fatigue based on a case definition17 declined from 22% at the end of treatment to 13% at 42 months post-treatment among patients who received chemotherapy and/or radiotherapy.18 On the basis of these and other findings,11 we hypothesized that fatigue in the chemotherapy group would diminish further over a 3-year follow-up period to a level similar to that in the radiotherapy and NC groups. Thus, we expected to detect a significant group-by-time interaction effect on fatigue. We also considered the possibility that fatigue would remain higher over time in the chemotherapy group relative to the other 2 groups, as demonstrated by a significant group effect.
MATERIALS AND METHODS
Participants and Procedures
Data for this report were drawn from a larger study of quality of life in women who were treated for breast cancer at the Moffitt Cancer Center and the University of Kentucky Chandler Medical Center. This study was approved by institutional review boards at the University of South Florida and the University of Kentucky. Patient eligibility criteria for the larger study were: a diagnosis of stage 0 through II breast cancer; scheduled to receive chemotherapy, radiotherapy, or both after surgery; no other history of cancer other than basal cell skin carcinoma; no prior chemotherapy or radiotherapy; and no conditions in which fatigue is a prominent symptom (eg, acquired immunodeficiency syndrome, multiple sclerosis, or fibromyalgia). Additional eligibility criteria for this report were that patients completed the Time 1 (T1) assessment (6 months after completing treatment) and had no recurrent breast cancer or new primary cancer by the Time 2 (T2) assessment (42 months after completing treatment). Patients were recruited from November 1999 until June 2006 and provided informed consent at time of recruitment. Assessments were conducted approximately 6 months (T1) and 42 months (T2) after the completion of chemotherapy (for patients who received chemotherapy only) or radiotherapy (for patients who received radiotherapy only or radiotherapy after chemotherapy).
Each patient who completed the T1 assessment was matched with a noncancer control (as described in detail elsewhere16). Noncancer participants were women who were within 5 years of age and resided in the same zip code as their patient match. Additional eligibility criteria were: no history of cancer (other than basal cell skin carcinoma) or other potentially life-threatening diseases and no conditions in which fatigue is a prominent symptom (as described above). Controls were recruited from December 2004 to January 2008. After providing informed consent, they were assessed at baseline (T1) and approximately 3 years later (T2).
In total, 309 patients and 252 controls completed the T1 assessment. At T2, 21 patients were excluded because of disease recurrence or death, and 9 controls were excluded because of breast cancer diagnosis, fibromyalgia diagnosis, or death. An additional 83 patients and 50 controls declined further participation or could not be contacted. Thus, this report is based on 205 patients and 193 controls who completed both assessments.
The difference in study completion rates between patients (205 of 288; 71%) and controls (193 of 243; 79%) was statistically significant (P = .037). Additional analyses revealed no significant differences (P > .05) on any T1 fatigue index or demographic variable between the 50 controls who withdrew or could not be contacted and the 193 controls who completed both assessments. Similarly, there were no significant differences on any T1 fatigue index or clinical variable assessed between the 83 patients who withdrew or could not be contacted and the 205 patients who completed both assessments. Patients who withdrew or could not be contacted had less education (P = .001) and lower household income (P = .001) but did not differ on other demographic variables.
Among patients, 102 received radiotherapy only (RT group), and 103 received chemotherapy. There were no significant differences between the 16 patients who received chemotherapy only and the 87 patients who received both chemotherapy and radiotherapy on any T1 fatigue index described below (P ≥ .25). Therefore, chemotherapy patients who did and did not receive radiotherapy were merged into a single chemotherapy group (CT group) for all further analyses. Study completion rates did not differ significantly for patients who received chemotherapy only (73%), chemotherapy plus radiation (74%), and radiation only (68%; P = .565).
Demographic characteristics were assessed by self-report at T2. Clinical characteristics were assessed in patients through medical record reviews.
At T1 and T2, participants completed the Fatigue Symptom Inventory (FSI),19 They rated their level of fatigue from 0 (not at all fatigued) to 10 (as fatigued as I could be) on the day they felt most fatigued in the past week (FSI most), on the day they felt least fatigued in the past week (FSI least), on average in the past week (FSI average), and right now (FSI now). These items were averaged to create a composite severity score (FSI severity).20 Participants also rated the degree to which fatigue interfered in the past week from 0 (no interference) to 10 (extreme interference) with 7 aspects of functioning and quality of life, and these ratings were averaged to create an interference score (FSI interference). Fatigue frequency was assessed as the number of days in the past week (range, 0-7 days) they felt fatigued (FSI days) and how much of the day (from 0 [none] to 10 [entire]) they felt fatigued (FSI time). Participants also completed the 7-item Profile of Mood States Fatigue Scale (POMS-FAT),21 which provides an index of fatigue in the past week.
Possible differences among groups on demographic variables were tested with chi-square tests for categorical variables and with 1-way analyses of variance for continuous variables. Variables for which there were significant (P < .05) group differences were entered as covariates in all subsequent analyses.
To evaluate effects of time, group status, and their interaction on fatigue, separate repeated-measures analyses of variance were carried out for each fatigue index with time (T1 and T2) as a within-subjects factor and group (CT, RT, or NC) as a between-subjects factor. Of principal interest was the group × time interaction effect. In the case of significant interactions, additional comparisons were conducted to evaluate changes over time within each group and differences among the groups at T2. For comparisons that yielded significant group differences at T2, the Cohen d was calculated to determine the effect size (small effect, d = 0.2; medium effect, d = 0.5; or large effect, d = 0.8).22 For analyses that did not yield significant interaction effects, the analysis focused on the group effect. In the case of significant group effects, comparisons were conducted to evaluate the magnitude of differences between pairs of groups across T1 and T2. For comparisons that yielded significant differences, the Cohen f was calculated to determine the effect size (small effect, f = 0.10; medium effect, f = 0.25; or large effect, f = 0.40).22
If a significant difference between groups was observed, then additional analyses were conducted to evaluate the clinical meaningfulness of the difference at T2. In these analyses, participants' fatigue scores were classified as “abnormal,” using criteria similar to those used in our prior research,16 if they scored >1 standard deviation (SD) above the mean of the NC group at T2 on the measure being evaluated. Scores of this magnitude would be expected to occur in <16% of individuals for normally distributed variables. A mean difference on a measure was then considered clinically meaningful if chi-square analyses indicated that the percentages of patients with abnormal scores at T2 differed significantly among the 3 groups.
A 2-tailed P value of .05 was used as the criterion for statistical significance. Rather than correcting the significance level for the number of comparisons performed, which has the potential to increase the probability of Type II error, we chose to limit the number of comparisons performed based on the outcome of significance tests of group × time effects and group effects. With 398 participants, power was approximately 0.8 with the stated P value to detect an interaction effect size of f = 0.08 and a group effect size of f = 0.14. Analyses were performed using the SPSS software package (version 17.0; SPSS Inc, Chicago, Ill).
Participant demographic characteristics are presented in Table 1. The CT group was significantly younger than the RT and NC groups. Also, mean time between assessments was significantly longer in the RT group than in the NC group. On the basis of these findings, age and time between assessments were used as covariates in all further analyses.
|CT Group||RT Group||NC Group||Pa|
|Age: Mean±SD, y||54.9±8.8||61.0±8.7||59.6±8.8||<.001|
|Body mass index: Mean±SD, kg/m2||28.1±6.0||27.2±5.7||27.7±6.0||.575|
|Time between T1 and T2 assessments: Mean±SD, y||3.07±0.20||3.11±0.21||3.05±0.10||.032|
|Household income, %|
|Menopausal status, %|
|Premenopausal or perimenopausal||18||14||17||.699|
|Marital status, %|
Clinical characteristics of the RT and CT groups are presented in Table 2. Patients in the CT group, as expected, were more likely to have undergone mastectomy, have more advanced disease, and be further in time from cancer diagnosis at T1 than patients in the RT group. The RT group was more likely to have received hormone treatment than the CT group.
|Characteristic||CT Group||RT Group||Pa|
|Time from diagnosis to T1 assessment: Mean±SD, mo||13.8±2.0||10.3±1.8||< 001|
|Hormone therapy, %|
|No. of treatments: Mean±SD||31.7±3.6||30.5±4.5||.057|
|RT dose: Mean±SD, cGy||6058±503||5930±652||.140|
|No. of chemotherapy cycles. %|
|Chemotherapy regimens, %|
|Anthracycline, cyclophosphamide, taxane||30|
|Cyclophosphamide, methotrexate, fluorouracil||11|
|Anthracycline, cyclophosphamide, fluorouracil||3|
|Anthracycline, cyclophosphamide, fluorouracil, taxane||1|
Means and SDs for fatigue scores for each group at T1 and T2 are presented in Table 3, and the results of statistical comparisons are presented in Table 4. There were no significant time effects for any of the fatigue indices (results not shown). Analyses of composite FSI severity scores yielded a significant group × time effect (see Fig. 1). FSI severity increased in the CT group (P = .007) from T1 to T2 but not in the RT group (P = .569) or the NC group (P = .892). Consistent with this pattern, FSI severity at T2 was higher in the CT group compared with the RT group (P = .009; d = 0.38) and the NC group (P = .017; d = 0.30).
|Index||CT Group||RT Group||NC Group|
|FSI Severity||FSI Most||FSI Least||FSI Average||FSI Now||FSI Days||FSI Time||FSI Interference||POMS- FAT|
On the basis of this significant interaction, analyses were undertaken of the 4 components of the FSI severity score. Group × time effects were significant for FSI most and FSI least scores (see Figs. 2 and 3). FSI most scores increased from T1 to T2 in the CT group (P = .017) but not in the RT group (P = .147) or the NC group (P = .220). FSI most scores for the CT group at T2 did not differ from those for the RT group (P = .102) or the NC group (P = .198). FSI least scores also increased in the CT group from T1 to T2 (P = .032) but not in the RT group (P = .358) or the NC group (P = .130). Consistent with this pattern, FSI least scores at T2 were higher in the CT group compared with the RT group (P = .001; d = 0.43) and the NC group (P = .008; d = 0.28). Although the group × time effect for FSI now scores was not significant, the group effect was significant. Comparisons revealed that FSI now scores were higher across time in the CT group compared with the NC group (P = .022; f = 0.14) but not the RT group (P = .079). No significant group × time or group effect was observed for FSI average scores.
No significant group × time effect was observed for FSI days scores; however, the group effect was significant. The number of days participants felt fatigued was higher across time in the CT group than in the RT group (P = .016; f = 0.20) and the NC group (P < .001, f = 0.22). No significant group × time or group effects were observed for FSI time or FSI interference.
No significant group × time effect was observed for POMS-FAT scores; however, the group effect was significant. POMS-FAT scores were significantly higher in the CT group compared with the NC group (P = .009; f = 0.11) but not the RT group (P = .059).
Clinical Meaningfulness of Mean Differences
On the basis of the pattern of significant mean differences observed between groups, additional analyses were undertaken to determine the clinical meaningfulness of the differences in FSI severity, FSI days, and POMS-FAT scores at T2. The relation between group membership and rates of abnormal scores at T2 approached significance for FSI severity scores (P = .052); 29% of the CT group scored in the abnormal range on this measure compared with 17% of the RT group and 19% of the NC group. There was a significant relation between group membership and rates of abnormal scores at T2 for FSI days scores (P = .007); 37% of the CT group scored in the abnormal range on this measure compared with 25% of the RT group and 20% of the NC group. There also was a significant relation between group membership and rates of abnormal scores at T2 for POMS-FAT scores (P = .024); 25% of the CT group scored in the abnormal range on this measure compared with 14% of the RT and 14% of the NC group.
Analyses were undertaken to determine whether prescription of hormone therapy to patients at T1 was related to fatigue outcomes. There were no significant effects of group (any hormone therapy prescribed or no hormone therapy prescribed) or group × time on FSI severity scores (P ≥ .112). A similar pattern of nonsignificant results was obtained for an analysis limited to patients who were prescribed tamoxifen or no hormone therapy (P ≥ .080).
Analyses also were undertaken to determine whether administration of a taxane was related to fatigue outcomes in the CT group. There were no significant effects of group (taxane administered or not administered) or group × time on FSI severity scores (P ≥ .096). A similar pattern of nonsignficant results for FSI severity scores was obtained for analyses in which patients in the CT group were grouped into those who received 4 cycles or >4 cycles of treatment (P ≥ .108).
Contrary to expectations, fatigue did not diminish over the 3-year follow-up period in patients who received chemotherapy. Indeed, results indicated that the overall severity of fatigue and the ratings for most and least fatigue in the past week worsened over the 3-year period in the CT group, whereas they remained stable and lower in the RT and NC groups. On other indices (ie, current fatigue, days fatigued, and past week fatigue), fatigue did not change over time in the CT group and remained higher than in the NC comparison group across the follow-up period. Differences between the CT group and the RT group or the NC control group generally were indicative of medium-sized effects and were clinically meaningful. No differences in fatigue were evident between the RT and NC groups or between patients based on whether or not they were prescribed hormone therapy.
Findings that indicate greater fatigue in patients who receive chemotherapy are consistent with prior research. Several cross-sectional studies8, 23, 24 and longitudinal studies9, 16 have reported greater fatigue among breast cancer survivors who received chemotherapy compared with those who received no adjuvant therapy or other forms of adjuvant therapy. Similarly, several cross-sectional studies3, 6 and longitudinal studies11 have reported greater fatigue in breast cancer survivors who received chemotherapy compared with noncancer controls. The current study extends previous findings by identifying specific time points (ie, 6 months and 42 months after treatment completion) when differences are evident. It also extends previous research by providing 1 of the few planned comparisons between breast cancer survivors who received chemotherapy versus radiotherapy only.
The finding in the current study that fatigue severity worsened over time in chemotherapy-treated patients is not consistent with prior research. Other studies of the course of fatigue indicated either that it improved over time11 or that it remained stable14 in chemotherapy-treated breast cancer survivors. One possible explanation for why 3 different patterns of change have been detected in these 3 studies is differences in the research methodology. These differences include the measures used to assess fatigue and the timing of the post-treatment assessments. The demographic and clinical characteristics of the samples also differ. The average age of patients in the current study was 55 years compared with an average age of 46 years and a median age of 48 years in the other studies. In addition, patients in the current study were less likely to be treated with fluorouracil and were more likely to be treated with a taxane compared with patients in the previous studies. The latter treatment difference may not be relevant, because we observed no evidence in the current study to indicate that the administration of a taxane influenced the course of fatigue. Differences in supportive care also may be an explanation; however, data bearing on this possibility were not available across the 3 studies.
In contrast to the findings obtained for fatigue severity, group membership was not associated significantly with the degree to which fatigue interfered with functioning and quality of life. However, there was a marginally significant group × time interaction (P = .099), with a trend for fatigue interference to worsen in the CT group while remaining relatively stable in the other 2 groups. One possible explanation for the lack of greater differences on this measure may be the tendency for patients to accommodate their lifestyle as a means of coping with fatigue.25 In other words, patients may have become accustomed to the impact of heightened fatigue on their lives over the 36-month follow-up period, thereby blunting the extent to which it was experienced as disruptive.
Assuming replication of the finding that fatigue severity worsens over time in chemotherapy-treated patients, future research should seek to identify underlying mechanisms. One possible mechanism is weight gain. Weight gain commonly occurs among breast cancer patients who receive adjuvant chemotherapy,26 and few patients return to their pretreatment weight in subsequent years.27 Furthermore, there is evidence that an increase in weight from pretreatment to 6 months after treatment in patients who receive adjuvant chemotherapy is associated with an increase in fatigue.28
Several limitations of the current study should be acknowledged. First, patients who withdrew or could not be contacted had less education and household income compared with patients who completed both assessments. Thus, the education and income levels of the final patient sample did not reflect the larger group that completed the first assessment. Second, the sample was comprised primarily of white women and women with middle or greater household incomes. Consequently, our findings may not be generalizable to minority women and lower income women. Third, the study did not investigate possible mechanisms underlying the increase in fatigue severity observed among patients who received chemotherapy. Although weight gain is proposed as 1 possible mechanism, it could not be examined because of the lack of data on body mass index at T1. Likewise, we were unable to examine the possible contributory role of chemotherapy-induced menopause. Fourth, we cannot rule out the possibility that the heightened fatigue evident in the CT group in the post-treatment period may simply reflect the continued presence of pretreatment differences relative to the RT group and the NC group. However, results indicating that the severity of fatigue actually increased in the CT group in the post-treatment period seem to argue against this possibility.
Our findings have important clinical implications. Clinical practice guidelines recommend the continued monitoring of cancer patients for fatigue in the post-treatment period.29 Our results support this recommendation and suggest that continued monitoring is particularly important for breast cancer survivors who receive chemotherapy. Another implication concerns patient education. Cancer patients often are told that their fatigue will diminish gradually after treatment completion. Our results suggest that health care providers communicate to breast cancer patients, particularly those who are receiving chemotherapy, the possibility that fatigue may not improve over time and possibly may worsen. At the same time, patients should be informed of interventions that are effective against fatigue in the post-treatment period, such as cognitive behavior therapy30 and exercise.31
This research was supported by Grant R01CA82822 from the National Cancer Institute.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.
- 21Profile of Mood States Manual. San Diego, Calif: Educational and Industrial Testing Service; 1971., , .
- 22Statistical Power Analysis for the Behavioral Sciences, 2nd ed. Hillsdale, NJ: Lawrence Earlbaum Associates; 1988..
- 29The NCCN cancer-related fatigue clinical practice guidelines in oncology. J Natl Comp Cancer Netw. 2007; 15: 1054-1078., , , , , .