Correlates of fatigue severity in patients with chronic myeloid leukemia treated with targeted therapy

Abstract

Purpose:

Tyrosine kinase inhibitors (TKIs) substantially improve survival for patients with chronic myeloid leukemia (CML), but fatigue associated with TKIs can negatively impact patients’ quality of life and adherence. This study sought to identify correlates of fatigue (e.g., sociodemographic characteristics, clinical characteristics, health behaviors) among patients with CML taking TKIs who reported moderate to severe fatigue.

Methods:

Adults with CML experiencing at least moderate fatigue were recruited for a pilot trial of a cognitive behavioral intervention to improve fatigue. Data collected pre-intervention were used to explore concurrent correlates of fatigue in univariate and multivariable models.

Results:

Participants (N=44, 48% female) were M=55.6 years old (SD=12.6) and had been diagnosed with CML M=5.2 years prior (SD=5.3). Participants had been taking their current TKI for M=2.5 years (SD=2.7). Most participants (64%) had previously been treated with ≥1 other TKI. More than three-quarters of participants (77%) reported severe fatigue. In univariate models, worse fatigue was associated with higher BMI (r=−0.36, p=0.018), prior treatment with other TKI(s) (r=−0.34, p=0.024), worse sleep disturbance (r=−0.51, p<0.001), and less physical activity (r=0.31, p=0.043). In a multivariable model, significant univariate correlates accounted for 39% of the variance in fatigue. Worse fatigue remained significantly correlated with higher BMI (β=−0.33, p=0.009) and more disturbed sleep (β=−0.45, p<0.001).

Conclusion:

Results may inform future research aiming to identify fatigued patients with CML at risk for experiencing more severe fatigue during TKI therapy. Identifying predictors of fatigue severity could aid clinicians in identifying which patients will benefit from referrals to supportive therapy.

Trial registration:

NCT02592447, October 30, 2015

Since 2001, targeted therapies such as tyrosine kinase inhibitors (TKIs) have substantially improved the treatment of chronic myeloid leukemia (CML), with 8-year relative survival increasing from less than 15% to over 85% [1]. Thus, patients with CML who are successfully treated with TKIs now have life expectancy similar to that of the general population. However, TKI therapy requires long-term, often indefinite, daily treatment. In addition, TKIs are associated with side effects that can significantly compromise patients’ health-related quality of life (HRQOL) and, in turn, affect treatment adherence [2,3].

One such side effect is fatigue, which is both highly prevalent and persistent during TKI therapy [4–6]. The vast majority of patients with CML who are treated with TKIs report some level of fatigue (82%) [4], and up to 68% of patients report fatigue that is moderate to severe [7,8]. In turn, worse fatigue is associated with worse HRQOL [7,9]. In fact, Efficace and colleagues [9] singled out fatigue as the primary factor limiting the HRQOL of patients with CML on long-term TKI therapy. Fatigue may also provide unique clinically relevant information, as a recent study showed that pre-treatment fatigue independently predicted worse treatment response among patients with CML starting TKI therapy [10]. Research has also shown that physicians tend to underestimate the severity of fatigue among patients with CML relative to patient reports [11]. Thus, obtaining patient self-reports of fatigue and integrating that information into clinical care may be a critical component of patient-centered care. In addition, identifying correlates of fatigue could help clinicians to better identify patients who are most at risk for worse fatigue and for potentially worse outcomes.

A recent study by Janssen and colleagues [8] explored correlates of fatigue severity among patients with CML on TKI therapy and found that worse fatigue was associated with younger age, female gender, more medical comorbidities, use of medications known to cause fatigue, and physical inactivity. These findings were consistent with the broader cancer survivorship literature, which has similarly identified age, female gender, and medical comorbidities as risk factors of worse cancer-related fatigue [12–15]. However, Janssen and colleagues did not find associations between fatigue and other clinical characteristics and health behaviors such as overweight/obesity and sleep disturbance that are broadly supported in the literature [12,14,16]. Thus, additional research may help to further elucidate correlates of fatigue severity among patients with CML taking TKIs.

The purpose of this analysis was to explore sociodemographic, clinical, and health behavior correlates of fatigue severity and the relationship between fatigue severity and HRQOL in a sample of patients with CML on TKI therapy. This analysis focused on patients who reported at least moderate fatigue severity. Data used for this study were collected at baseline as part of a feasibility randomized controlled trial (RCT) of an evidence-based intervention for fatigued patients with CML taking TKIs. The primary results of this trial were reported previously [17].

Methods

Participants and Procedures

This study was approved by the Chesapeake Institutional Review Board (Pro00023476) and registered on clinicaltrials.gov (NCT02592447). Eligible participants were: 1) ≥18 years old; 2) able to speak and read English; 3) diagnosed with chronic phase CML; 4) not treated for another cancer except non-melanoma skin in the past five years; 5) receiving care at Moffitt Cancer Center; 6) taking the same oral TKI for ≥three months; 7) reporting new onset or worsening fatigue after starting a TKI; 8) reporting at least moderate fatigue (≥4 average rating on the Fatigue Symptom Inventory 0–10 scale); and 9) without a clinical history of a disease that could account for fatigue (e.g., multiple sclerosis).

Preliminary eligibility was determined through electronic medical record (EMR) review and consultation with the hematologic oncology team. A trained member of the research staff screened potentially eligible patients for full eligibility criteria via telephone or in person during an outpatient clinic appointment. Eligible and interested patients signed informed consent and completed a baseline assessment of self-report questionnaires in hard copy or online via a secure web-based survey link. Participants were then randomly assigned to the experimental intervention or waitlist control group. A detailed description of the experimental intervention has been previously published [17]. For these analyses, only data collected at baseline (i.e., pre-treatment) were evaluated, so that fatigue scores were not confounded by intervention effects. The data that support the findings of this study are available from the corresponding author upon reasonable request.

Measures

Sociodemographic and clinical characteristics.

Sociodemographic information including age, gender, race, ethnicity, marital status, education, and household income was collected using a standardized self-report form. Clinical characteristics (e.g., date of CML diagnosis, current and previous TKI therapies and start dates) were extracted from patients’ EMR. Body mass index (BMI) was calculated as weight (kilograms) divided by height (meters) squared using values extracted from patients’ EMR on the closest available date to study consent. For descriptive purposes, BMI was categorized according to guidelines from the Center for Disease Control and Prevention as underweight (<18.5), normal/healthy (18.5–24.9), overweight (25.0–29.9), or obese (≥30.0) [18].

Fatigue.

The 13-item Functional Assessment of Chronic Illness Therapy-Fatigue scale (FACIT-Fatigue) assessed fatigue over the past week [19]. Item responses are on a five-point Likert-type scale from 0 (not at all) to 4 (very much). Items were summed to produce a total fatigue score. Possible scores range from 0–52, with lower scores indicating worse fatigue. Among patients with hematologic malignancies, a total score of 30 or less is a valid threshold indicating severe fatigue [20].

HRQOL.

The 27-item Functional Assessment of Cancer Therapy-General (FACT-G) includes four subscales assessing physical, social/family, emotional, and functional well-being over the past week [21,22]. Item responses are on a five-point Likert-type scale from 0 (not at all) to 4 (very much). Items were summed to create well-being subscale scores and a total HRQOL score. Possible total HRQOL scores range from 0–108, with higher scores indicating better HRQOL.

Health behaviors.

Participants self-reported their health behaviors including alcohol consumption in the past month and lifetime tobacco use, defined as having smoked ≥100 cigarettes in their lifetime [23]. Participants reported sleep-rest patterns using the 7-item Sleep-Rest subscale of the Sickness Impact Profile (SIP-SR), and items were summed with higher scores indicating more dysregulated sleep-rest patterns [24,25]. Participants reported physical activity during the past week using the 7-item International Physical Activity Questionnaire Short Form (IPAQ-SF) by indicating how many days they engaged in walking, moderate activity, and vigorous activity, and for how long each day they engaged in these activities [26,27]. Physical activity values at each intensity were weighted by its energy requirements, as defined by metabolic equivalents (METs): walking=3.3 METs/minute, moderate=4.0 METs/minute, and vigorous=8.0 METs/minute [28]. Total physical activity was calculated as the sum of METs per week for all physical activity. For descriptive purposes, participants were categorized as minimally, moderately, or highly physically active according to the IPAQ-SF scoring guidelines [28].

Statistical Analyses

Power analyses for the parent study determined that a sample of 48 participants was adequate for a feasibility RCT of an evidence-based intervention for fatigued patients with CML taking TKIs [17]. For this secondary analysis, the sample was characterized at baseline using descriptive statistics (e.g., means, standard deviations, frequencies, percentages). Bivariate relationships were evaluated between patients’ fatigue and sociodemographic characteristics, clinical characteristics, and health behaviors using Pearson’s r and point biserial correlations. Variables that were significantly associated with fatigue were simultaneously entered into a multivariable regression model to assess associations while controlling for the effects of other relevant variables. Finally, relationships between fatigue and HRQOL were evaluated using Pearson’s r correlations. Significant associations were determined with p<0.05.

Results

Information about rates of screening and recruitment are described in a previous report of this study’s primary outcomes [17]. In total, 48 adults with CML consented to participate in this study. Four individuals did not provide baseline data, and thus 44 participants were included in analyses.

Patient Characteristics

In total, 77% of participants (n=34) reported severe fatigue, and the average fatigue score fell below the threshold indicating severe fatigue (M=23.4, SD=8.5, range 10–47). Table 1 shows the sociodemographic characteristics, clinical characteristics, and self-reported health behaviors of the sample.

Table 1.

Sociodemographic and clinical characteristics of the sample and univariate associations with fatigue.

	Statistic	Univariate Correlation with Fatigue
Age, years; M (SD), range	55.6 (12.6), 29–82	0.15
Female gender; n (%)	21 (48%)	0.11
White race; n (%)	38 (86%)	0.01
Not Hispanic/Latino; n (%)	44 (100%)	NA
Married; n (%)	35 (80%)	−0.04
College graduate; n (%)	24 (55%)	0.11
Annual household income >$40,000; n (%)	24 (67%)	0.22
Comorbidity Index Score; M (SD), range	5.6 (1.4), 3–10	−0.16
BMI, kg/(m2); M (SD), range	31.8 (6.1), 20–50	−0.36*
BMI category; n (%)		NA
Underweight (<18.5)	0 (0%)
Normal (18.5–24.9)	5 (12%)
Overweight (25–29.9)	9 (20%)
Obese (30+)	30 (68%)
Years since diagnosis, years; M (SD), range	5.2 (5.3), 0–25	−0.02
Current TKI therapy; n (%)		NA
Dasatinib	16 (36%)
Nilotinib	11 (25%)
Bosutinib	9 (20%)
Imatinib	5 (12%)
Ponatinib	3 (7%)
Time on current TKI therapy, years; M (SD), range	2.5 (2.7), 0–13	0.08
Have previously been on a different TKI therapy; n (%)	28 (64%)	−0.34*
Fatigue; M (SD), range	23.4 (8.5), 10–47
Overall HRQOL; M (SD), range	64.0 (14.7), 31–93	0.55***
Functional well-being	13.8 (4.7), 6–25	0.50***
Physical well-being	16.7 (5.1), 6–26	0.72***
Emotional well-being	16.0 (4.5), 4–23	0.37*
Social/family well-being	17.5 (5.4), 4–27	0.07
Tobacco use (lifetime); n (%)		−0.15
Yes	23 (52%)
No	21 (48%)
Alcohol consumption (past month); n (%)		0.12
Yes	26 (59%)
No	18 (41%)
Sleep-rest pattern; M (SD), range	209.4 (143.6), 0–499	−0.51***
Physical activity, total MET minutes; M (SD), range	1851.2 (2599.3), 0–12,240	0.31*
Physical activity category; n (%)		NA
Minimally physically active	23 (52%)
Moderately physically active	13 (30%)
Highly physically active	8 (18%)

Notes. BMI, body mass index; HRQOL, health-related quality of life; kg, kilogram; M, mean; m², meters squared; MET, metabolic equivalents; n, frequency; NA, not applicable; SD, standard deviation; TKI, tyrosine kinase inhibitor. Tobacco use was defined as having smoked ≥100 cigarettes in lifetime. Alcohol consumption was limited to the past month.

^*p<0.05;

^**p<0.01;

^***p<0.001.

Sociodemographic characteristics.

Participants were an average of 55.6 years old (SD=12.6; 39% <50, 39% 50–64, 23% 65+) and approximately half of participants were female (48%). The majority of participants identified as White (86%), and no participants identified as Hispanic/Latino. These demographics are similar to the larger population of people with CML living in the US as of January 1, 2018, according to the NCI SEER*Explorer web application [29], with the exception that a larger proportion of participants in the national CML population appear to be older (65+ years old). Most participants in this study were married (80%), and most had a college education (55%).

Clinical characteristics.

Participants’ average BMI was 31.8 (SD=6.1), which falls within the obese range; 68% of participants’ BMI fell within the obese range, 20% within the overweight range, and 12% within the normal range. Participants had been diagnosed with CML an average of 5.2 years prior to enrollment (SD=5.3), and participants were taking the following TKIs at the time of study participation: dasatinib (36%), nilotinib (25%), bosutinib (20%), imatinib (12%), and ponatinib (7%). Almost two-thirds of participants (64%) were on second-line TKI therapy, and the average length of time on their current TKI therapy was 2.5 years (SD=2.7).

Health behaviors.

Approximately half of participants (52%) reported a significant smoking history, and 59% of participants reported alcohol consumption in the past month. Participants endorsed moderate sleep disturbance (M=209.4, SD=143.6). Approximately half of participants were categorized as minimally active (52%).

Univariate Relationships with Fatigue

Table 1 shows the univariate associations between fatigue and participants’ sociodemographic characteristics, clinical characteristics, and health behaviors. Worse fatigue was associated with higher BMI (p=0.02). Participants on second-line TKI therapy or more had worse fatigue (M=21.2, SD=7.5) relative to participants on first-line therapy (M=27.2, SD=9.1; p=0.02). The difference in average fatigue scores between participants on first vs. second-line therapy or more was 6 points, which is double the minimally important difference (MID) considered clinically meaningful for the FACIT-Fatigue scale (i.e., 3 points) [30,31]. Worse fatigue was also associated with more sleep disturbance (p<0.01) and less physical activity (p=0.04).

Multivariable Relationships with Fatigue

BMI, prior TKI therapy, sleep disturbance, and physical activity were simultaneously entered into a multivariable regression model with fatigue as the dependent variable (Table 2). Collectively, the independent variables explained 39% of the variance in fatigue (F(4,39)=7.8, p<0.01). Higher BMI (β=−0.3, p=0.01) and more disturbed sleep (β=−0.4, p<0.01) remained significant correlates of fatigue, whereas prior TKI therapy (β=−0.2, p=0.22) and physical activity (β=0.2, p=0.08) were no longer significantly associated with fatigue.

Table 2.

Multivariable regression model predicting fatigue.

	B	SE	β	t value of parameter	Model Adjusted R2
Overall model					0.39
Independent variables
BMI, kg/(m2)	−0.45	0.17	−0.32	−2.69**
Have previously been on a different TKI	−2.81	2.23	−0.16	−1.26
Sleep disturbance	−0.02	−0.01	−0.40	−3.09**
Physical activity, total MET minutes	0.00	0.00	0.22	1.82

Notes. M, mean; MET, metabolic equivalents; N, frequency; NA, not applicable; SD, standard deviation;

^*p<0.05;

^**p<0.01;

^***p<0.001.

Relationship Between Fatigue and HRQOL

Table 1 shows the univariate associations between fatigue and HRQOL. Participants reported an average overall HRQOL score of 64.0 (SD=14.7). The difference in overall HRQOL scores between fatigued patients with CML in our sample and published norms for the general US adult population (M=80.1, SD=18.1) [32] was more than 16 points, far exceeding the MID considered clinically meaningful for the FACT-G scale (i.e., 5 points) [31,33,34]. Worse fatigue was associated with worse overall HRQOL (p<0.01) and worse functional (p<0.01), physical (p<0.01), and emotional (p=0.01) domains of well-being.

Discussion

This study describes correlates of fatigue severity in a sample of patients with CML on TKI therapy and with self-reported fatigue that was at least moderate in severity. Using cross-sectional univariate and multivariate models, we identified correlates of more severe fatigue including higher patient BMI and worse sleep disturbance. These characteristics were significantly associated with worse fatigue, even after accounting for other patient characteristics (i.e., history of prior TKI therapy, physical activity). Worse fatigue, in turn, was associated with worse HRQOL.

Our results extend the literature, which to our knowledge includes only one other evaluation of fatigue correlates in this population led by Janssen and colleagues [8]. This past study also identified less physical activity as a correlate of worse fatigue among patients with CML taking TKIs [8]. However, we did not replicate their findings that younger age, female gender, and more medical comorbidities were additional correlates of fatigue, and we found an association between higher BMI and worse fatigue whereas this past study did not. Of note, as part of our study inclusion criteria, all participants reported at least moderate levels of fatigue, whereas there was no inclusion criterion related to existing fatigue severity in the study led by Janssen and colleagues [8]. It is possible that patient factors such as age, gender, and medical comorbidities may be more predictive of the incidence of fatigue in general, but are less strongly associated with fatigue severity as it reaches moderate to severe intensity. This possibility requires further investigation. Our results also implicate prior lines of TKI therapy as a correlate of worse fatigue, suggesting that there may be a cumulative effect of multiple TKIs on fatigue. However, the contributions of both physical activity and prior TKI therapy were no longer significant in a multivariate model.

Consistent with the broader cancer survivorship literature [12,14,15] we identified higher BMI and more sleep disturbance as correlates of worse fatigue. Moreover, these relationships remained significant above and beyond the contributions of the other predictors in a multivariable model. BMI and sleep disturbance have not previously been identified as correlates of fatigue in patients with CML taking TKIs. However, given the preliminary nature of this study, these associations should be confirmed in studies including larger samples.

The average fatigue score in this sample exceeded the threshold for severe fatigue for patients with hematologic malignancies, and 77% of participants in our sample reported severe fatigue [20]. In addition, the average HRQOL score in this sample was lower than the average HRQOL score in a normative cancer population by a factor of more than 3 MIDs [31–34]. These findings are consistent with past work showing a clear relationship between worse fatigue and worse HRQOL among patients with CML on TKI therapy [7,9].

Study Limitations

This was a secondary analysis of a pilot RCT testing an evidence-based intervention for fatigued patients with CML taking TKIs. Our sample consisted of mostly self-identified non-Hispanic White individuals. Thus, these results may not generalize to more racially or ethnically diverse patient populations. Eligibility criteria for the larger RCT required participants to report at least moderate fatigue at the time of screening, and thus these results reflect correlates of fatigue among an already fatigued sample. However, up to 68% of patients with CML taking TKIs report moderate to severe fatigue [7]. Therefore, these results may be representative of the majority, but not all, patients with CML taking TKIs. Other factors that could impact fatigue, such as depression, were not assessed in this study and should be explored in future research in relation to fatigue and health behaviors among patients with CML taking TKIs. Our analyses were cross-sectional and thus direction of causality could not be determined. Finally, our sample size was not large enough to support exploration of relationships between specific TKI regimens and fatigue, which should be explored in future work.

Clinical Implications

Research shows that more severe fatigue predicts worse clinical outcomes among patients with CML, yet physicians tend to underestimate fatigue in this population [11]. By identifying more easily observable and quantifiable patient characteristics that are significantly associated with worse fatigue, clinicians may be better able to identify patients who could benefit from referrals to supportive therapy to amend fatigue and potentially buffer against worse outcomes. These preliminary results suggest that higher BMI and worse sleep disturbance are notable correlates of worse fatigue and could potentially be risk factors. Pending confirmation, these results suggest that patients with CML with high BMI and self-reported sleep disturbance should be prioritized for fatigue treatment referrals, as these patients may be at risk for worse fatigue relative to other patients with CML.

Conclusions

This cross-sectional analysis of fatigued patients with CML on TKI therapy provides preliminary evidence for correlates of worse fatigue, specifically higher BMI and worse sleep disturbance. Findings can be used to guide future research to confirm the demographic, clinical, and health behavior risk factors for worse fatigue in this patient population. This information could be used to aid clinicians in identifying patients who will benefit from referrals to supportive care for managing cancer-related fatigue.