Identification of breast cancer survivors with high symptom burden

Abstract

Background

While women diagnosed with breast cancer have increased survival when compared to other cancers, survivorship may include residual symptom burden from treatment and continuing endocrine therapies.

Objective

The objective of this study was to identify subgroups of breast cancer survivors experiencing similar symptom severity.

Methods

Participants were 498 women with breast cancer, not on active treatment. Symptom severity was self-reported using the MD Anderson Symptom Inventory. Target symptoms were included in a Latent Profile Analysis (LPA). Factors related to subgroup membership and differences in QOL and functioning were explored using logistic regression.

Results

Mean age was 60.11 years (SD = 11.32), 86.1% were White, and 79.1% receiving endocrine therapy. Target symptoms included fatigue (reported at ≥ 5 by 22.8% of women), sleep disturbance (24.8%) and trouble remembering (17.2%). Two subgroups were identified: low symptom severity (77.0% of women) and high (23.0%). Older women (OR .971, 95% CI: .952, .989) and employed women (OR .621, 95% CI: .404, .956) were less likely to be in the high subgroup; women with poorer performance status (OR 1.653, 95% CI: 1.188, 2.299) were more likely to be in the high subgroup. Women in the high subgroup reported lower QOL (p = .000) and greater interference with functioning (p = .000).

Conclusions

Two subgroups of women with distinct symptom severity were identified.

Implications for Practice

Identification of women at risk for high symptoms during survivorship may allow clinicians to intensify their approach to symptom management, thereby mitigating poor outcomes and impairments in QOL.

Introduction

Breast cancer is characterized by significant disease and treatment related symptoms, but with a generally favorable prognosis. In the U.S., breast cancer is the most commonly diagnosed cancer in women, with an estimated 1 in 8 women anticipated to be diagnosed with invasive breast cancer in the course of their lifetime and 281,550 new cases of invasive breast cancer in 2021.¹ The overall 5-year survival rate for women with breast cancer is 90% with an estimated 3.8 million breast cancer survivors alive in the United States.² While women diagnosed with breast cancer may have increased survival when compared to other cancers, that survivorship may include uncertainty and residual symptom burden from treatment side effects and continuing hormonal therapies following acute treatment.^{3, 4–7} Survivors of breast cancer remain at risk for disease recurrence, as well as symptoms related to previously received and ongoing therapies, including endocrine therapy.³ While cancer survivorship has been defined as the period beginning with cancer diagnosis and lasting through end of life, the latter phases of this continuum, from the end of acute treatment through the remainder of life, have received limited attention with regards to studying the burden associated with persisting symptoms.³ The risk of physical long-term and late effects after initial treatment for breast cancer is associated with type of treatment(s) received, duration and dose of treatment(s), specific type of chemotherapy, receipt and type of endocrine therapy, and age of the patient during treatment.² Potential treatment modalities for breast cancer include surgery, radiation therapy, chemotherapy, and endocrine therapies. Importantly, breast cancer survivors have often received multiple types of treatment, combinations of chemotherapeutic agents, and may continue to receive endocrine therapies for 5–10 years following acute treatment. For patients with metastatic disease, ongoing treatment is typically necessary.

There is a large body of literature devoted to the description of individual symptoms, co-occurring symptoms and, more recently, symptom clustering among survivors of breast cancer.^8–14 To date, most studies have focused on relationships between variables, such as the occurrence of association of specific symptoms in combination or relationships between demographic and clinical variables and symptom presentation.^{11, 12, 15–16} Other studies have focused on the association between individual symptoms and health outcomes including functioning and quality of life.^{8, 10, 11, 14, 16, 17} Prevalence is often studied as opposed to severity or distress, with an abundance of literature supporting a high prevalence of disturbing symptoms during breast cancer survivorship.^2–20 Survivors of breast cancer report pain, disturbed sleep, neuropathy, fatigue, distress, depression, and cognitive impairment.^2–20 Potential long-term and late effects may include some concerns specific to breast cancer including lymphedema following radical surgeries and bone fragility, cardiac health and negative body image, affecting from 31–67% of breast cancer survivors.¹⁷ Breast cancer survivors may also suffer from sexual dysfunction, premature menopause and exacerbation of menopausal symptoms related to endocrine therapies.¹⁸ Additionally, determination of which symptoms to include in an analysis is often a priori, using the literature and guided by theory, and often based on prevalence, as opposed to severity or distress associated with symptoms. Symptom burden refers to the combined impact of all disease- and treatment-related symptoms the ability of an individual to function as they did prior to the onset of the disease or the initiation of the treatment or the sum of the severity of symptoms and the impact of those symptoms on functioning.²¹ Importantly, even with the existing body of literature describing symptoms among breast cancer survivors, the symptom burden of breast cancer survivors following acute treatment is not well-described in the literature.

The number of breast cancer survivors will continue to increase with aging of the baby boom cohort, increased use of effective cancer screening, and improvements in treatments.²² As the population of breast cancer survivors expands, clinicians will require knowledge about expected symptom burden and methods for measuring symptom burden persisting in survivorship. Descriptions of the symptom burden and identification of risk factors for severe symptoms are needed to assist clinicians in targeting those at risk for increased symptom severity during survivorship and related poor outcomes.

Theoretical Framework

This study was guided by the Dynamic Symptoms Model, which describes the complex nature of the symptom experience over time, incorporating the potential for distinct symptom experiences that may be related to antecedent variables and may influence outcomes and allowing for identification of symptoms that co-occur in classes of individuals with varying frequency and severity.²³ The model incorporates antecedents (demographic, physiologic, psychological, social, spiritual, and environmental) factors that may precede symptoms, the symptom experience (temporality, intensity, quality, distress, and appraisal), symptom trajectories (growth parameters of change over time across multiple symptoms), and consequences of symptoms including outcomes such as quality of life (QOL), functioning, adjustment, cost, morbidity, survival, and perception of health.²³ Symptom management strategies are included in the model as potential influencers of the symptom experience, trajectories and consequences.²³ Consistent with the model, this study includes specific antecedents (demographic and clinical variables) and outcomes (quality of life and functional impairments). The Dynamic Symptoms Model offers a framework for studying the relationships among unique symptom experiences, antecedents, and outcomes with the assumption that symptoms may co-occur in individuals in such a way that those individuals can be subclassified based on their unique symptom experience. The model accommodates the unique symptom experience of the individual while describing the synergistic relationship among symptoms within class membership and allows for determination of whether class membership is predicted by antecedents and whether class membership precedes outcomes. Advances in statistical modeling have allowed for newer methodological approaches to studying classes of symptom experiences, identifying homogeneous classes of persons who share common symptom presentations. Identification of such classes may allow for targeting women who are more likely to experience severe symptoms in survivorship for symptom management education and enhanced symptom management. To our knowledge, only one study has identified subgroups exhibiting similar symptom experiences in breast cancer survivorship that included a cohort of breast cancer survivors transitioning from active treatment to survivorship.²⁴ Importantly, symptoms in this study were selected based on prevalence reported in the literature, as opposed to severity in the sampled cohort.

The purpose of this study was to examine the symptom burden of breast cancer survivorship in the years following active treatment, exploring potential classes of women experiencing similar symptom burden. The specific aims were to (1) examine the symptom burden of breast cancer survivorship, (2) determine subgroups of breast cancer survivors experiencing similar symptom severity, (3) determine if membership in differing symptom burden classes is determined by various demographic and clinical variables and (4) determine if differing symptom burden profiles are associated with overall QOL and functioning. Understanding the symptom burden of breast cancer survivorship may elucidate common etiologies, which may inform the development of targeted symptom interventions.

Methods

Participants

Outpatients with breast cancer (n = 1544) were enrolled from 37 Eastern Cooperative Oncology Group – American College of Radiology Imaging Network (ECOG-ACRIN) affiliated institutions over a 26-month period, including 5 academic centers and 32 community clinics. Patients were enrolled at any point in the disease- and treatment-trajectory. Participants in the ECOG-ACRIN study met the following eligibility criteria: 1) were at least 18 years old, 2) had a diagnosis of breast cancer, 3) provided informed consent, and 4) were judged by the study staff to have cognitive function adequate for completing patient-reported outcomes measures. Patients were excluded from the ECOG-ACRIN study who had inadequate cognitive function to complete the study measures (as judged by study screener). The protocol was approved by the Institutional Review Boards at all participating sites and all participants provided written informed consent. For the purposes of this secondary analysis, data from patients enrolled in the ECOG-ACRIN study who were not currently on active treatment, with the exception of endocrine therapies, were included to capture a subpopulation of breast cancer survivors. All patients included in the current analysis were being followed in breast cancer survivorship clinics. This secondary analysis was approved by the Institutional Review Board at The University of Texas MD Anderson Cancer Center.

Procedures

Patients were recruited to the parent study at survivorship clinic appointments. For consenting participants, basic clinical and demographic data were collected and the patient completed the MD Anderson Symptom Inventory (MDASI) and global QOL at this initial visit. In addition, patients completed the MDASI and global QOL items at a survivorship follow-up visit 28–35 days later.

Instruments

The MDASI measures patient reported symptom intensity and functional interference for 13 symptoms common to cancer and 6 items measuring interference with daily functioning. Core symptoms include pain, fatigue, nausea, disturbed sleep, distress, shortness of breath, cognitive difficulty, lack of appetite, drowsiness, dry mouth, sadness, vomiting, and numbness and tingling. Six additional symptoms were included in this study: diarrhea, constipation, mouth sores, skin rash, hair loss, and coughing. Participants are asked to rate the symptom at its worst in the previous 24 hours on an 11-point scale with 9 representing ‘not present,’ and 10 representing ‘as bad as you can imagine.’ For the six interference items, patients are asked to rate the degree to which symptoms interfered with various aspects of daily functioning (general activity, mood, work (including housework), relations with other people, walking, and enjoyment of life) during the last 24 hours on an 11 point scale, with 0 representing ‘did not interfere at all,’ and 10 representing ‘interfered completely.’ The MDASI can be averaged into 2 subscales: the mean core symptoms, the mean core symptoms plus any additional symptom items, and the mean interference. Symptom items may be used individually or in subsets if specified a priori. The interference subscale can also be separated into a mean activity-related interference (WAW (work, general activity, and walking)) and a mean mood-related interference (REM (relations with other people, enjoyment of life, and mood)). The MDASI has established content validity through qualitative interviews with patients with breast cancer as well as extensive cognitive debriefing of the items, known-groups validity established using performance status, tumor response and disease stage, sensitivity to change in quality of life, and adequate internal consistency and test-retest reliability, with Cronbach’s alpha ≥ 0.85 across all subscales in patients with breast cancer.^{25, 26}

A global QOL item asked patients to rate their QOL as excellent, good, fair, poor, or very poor.

Analysis

All analysis was completed by a biostatistician and psychometrician with expertise in the analysis of patient-reported outcomes and reviewed by the entire study team. Means, standard deviations, and evaluation of distributions were described for sample data, including age, race, employment status, ECOG performance status (range 0–4), prior chemotherapy/immunotherapy/hormonal therapy, prior radiation therapy, current hormonal therapy, overall QOL. Symptoms measured included pain, fatigue, nausea, disturbed sleep, being distressed, dyspnea, cognitive difficulties, anorexia/cachexia, drowsiness, dry mouth, sad/depressed, vomiting, numbness/tingling, diarrhea, constipation, sore mouth, rash/pruritus, hair loss, and coughing. Interference items included interference with general activity, mood, work, relationships with others, walking, and enjoyment of life. The scale of the symptoms/interference in the MDASI is in the range from 0 (not present/did not interfere) to 10 (as bad as you can imagine/interfered completely). The scale of overall QOL is in the range from 1 (very poor) to 5 (excellent).

To answer Aim 1, to delineate the symptom burden of breast cancer survivorship, how patients rate symptom severity and interference with function was described. The prevalence and severity of individual MDASI symptoms and subscales (total symptom severity, total interference, WAW, REM, total MDASI) at baseline and follow-up was calculated. Descriptive statistics (mean, SD) were calculated for each symptom. The proportions of patients rating each symptom ≥ 7 (on the 0 to 10 scale) and ≥ 4 (on the 0 to 10 scale) was calculated. If the normality assumption was not met for t test, the Wilcoxon rank-sum test were used. Nominal p values were reported. Paired sample t test were used to compare differences in mean severity of individual MDASI symptoms and subscales at baseline and follow-up.

Target symptoms for the overall sample were identified from the symptom items on the MDASI using the symptoms most frequently reported at moderate to severe levels (greater than 15% of the sample) and those symptoms with the highest mean severity at baseline. The number of latent classes for the composite symptom score was determined using Latent Profile Analysis (LPA). Each latent class corresponded to a subpopulation that has its own symptom burden (symptom severity score on multiple symptoms). The categorical latent variables correspond to the person-oriented component, representing a category or class that describes subgroups of individuals who are relatively homogenous within that class and are heterogeneous across classes.²⁷ Latent classes were measured by target symptom severity items on the most frequently reported symptoms at baseline. Step-wise models were evaluated on the Bayesian information criteria (BIC), used to evaluate improvement tin model fit with the addition of classes. Smaller BIC values suggest a better model fit.²⁷ If the addition of a class resulted in a reduction in the BIC value relative to the BIC from the previously model (without the added class), then the new model was considered an improvement and the class was retained. The addition of classes continued until the BIC did not decrease with the addition of a class. Additionally, entropy was used to evaluate the probabilities of membership in each class for each individual. Entropy is a summary measure of classification based on these probabilities that ranges from 0–1.0.²⁷ The closer the entropy values are to 1.0, the better the classification. Finally, the best-fitting model was examined for the number of subjects in each class (greater than 15% of the sample) and visualized to determine of the predicted subgroups were clinically and theoretically relevant. The model that was retained produced the best fit indices relative to other tested models, with class proportions greater than 15% of the sample. Missing data were accommodated by MPlus version 6.0 through use of Full Information Maximum Likelihood.²⁷ After identifying the latent class solution that best fits the data, differences among the predicted classes for symptom burden were examined for potential correlates (age, race, employment status, stage of disease, treatment history, current hormonal therapy, performance status) using logistic regression.

After identifying the latent class solutions that best fit the data, differences among the predicted classes for symptom burden were examined for important outcomes including overall QOL and MDASI symptom interference subscales using independent-samples t tests and ANOVA for tests of mean differences and chi-square for tests of association among categories. Posterior probabilities and class assignments for each individual were saved into an output text file and extracted into the data set to be used for further analysis in tests of mean differences across the classes outside the model. While there is uncertainty in the model-predicted class assignment, high entropy and high posterior probabilities in the retained model suggest that mode-predicted class assignments could be considered observed variables. Class membership then became a correlate for these variables.²⁸ Adjustments were not made for missing data on the demographic and clinical variables. Therefore, the sample for each of these individual analyses was dependent on the largest set of complete data across groups.²⁹

Results

Sample demographics and clinical characteristics are reported in Table 1. Women in our sample had a mean age of 60.11 years (standard deviation = 11.32). The highest proportion of women in the sample were White (86.1%), not employed (53.8%), and receiving endocrine therapy at baseline (79.1%). 61.4% of women had prior radiation therapy and 71.7% had prior chemotherapy, immunotherapy, or endocrine therapy. Women in our sample had a median of 32.98 months since their breast cancer diagnosis.

Table 1.

Sample Characteristics for Breast Cancer Survivors (n = 498)

	Mean	SD
Age (years)	60.11	11.32
	Frequency	Percentage
Sex
Female	498	100.0
Race
White	429	86.1
Asian	10	2.0
African American	53	10.6
Unknown	6	1.2
Employment status
Employed, full-time	163	32.7
Employed, part-time	65	13.1
Not employed	268	53.8
Unknown	2	0.4
Current endocrine therapy	394	79.1
Prior radiation therapy	306	61.4
Prior chemotherapy/immunotherapy/endocrine therapy	357	71.7
Overall quality of life
Very poor	1	0.2
Poor	9	1.8
Fair	64	12.9
Good	257	51.6
Excellent	164	32.9
Unknown	3	0.6
ECOG
0	374	75.1
1	105	21.1
2	10	2.0
3	7	1.4
Unknown	2	0.4

Abbreviation: SD, standard deviation.

Symptom Severity:

At baseline, the mean subscale scores for the core and interference subscales were 1.43 and 1.30 respectively (Table 2). The most severe core symptoms reported were fatigue, sleep disturbance, difficulty remembering, pain, distress, drowsiness, and numbness/tingling; vomiting, nausea, and lack of appetite were the least severe core symptoms. Moderate-to-severe levels of fatigue and sleep disturbance were found for 22.8% and 24.8% of participants. 17.2% of participants reported moderate-to-severe levels of difficulty remembering, 14.2% of participants reported moderate-to-severe distress, 14.4% reported moderate-to-severe pain, 14.4% reported moderate to severe numbness/tingling, and 12.8% reported moderate-to-severe drowsiness. Missing data were minimal at baseline; at most 2.8% for fatigue and much less for most items. At follow-up, the mean subscale scores for the core and interference subscales were 1.48 and 1.46 respectively, with the same list of symptoms reported as the most severe core symptoms (Table 3).

Table 2.

Descriptive Statistics on Baseline Symptom Items, Interference Items and Subscale Scores for Breast Cancer Survivors (n = 498)

	Mean	SD	Range	LCL	UCL	% = 0a	% 1–4b	% ≥5c	% ≥7d	% Missinge
										Baseline
Core symptoms (Rank Order)
Pain	1.62	2.62	10	1.38	1.87	58.9	26.5	14.4	9.0	1.2
Fatigue	2.55	2.70	10	2.30	2.80	29.7	44.6	22.8	10.6	2.8
Nausea	0.41	1.40	10	0.28	0.54	85.9	8.8	4.2	2.0	1.0
Sleep disturbance	2.36	2.89	10	2.09	2.63	40.6	33.3	24.8	11.4	1.2
Distress	1.53	2.37	10	1.31	1.75	50.8	34.1	14.2	7.2	0.8
Shortness of breath	1.08	2.00	10	0.90	1.27	63.9	26.6	8.6	3.4	0.8
Cognitive difficulties	1.93	2.42	10	1.70	2.15	38.2	43.6	17.2	6.4	1.0
Lack of appetite	0.64	1.68	10	0.49	0.80	78.7	14.0	5.8	2.4	1.4
Drowsiness	1.67	2.4	10	1.45	1.89	45.2	41.1	12.8	6.2	1.0
Dry mouth	1.49	2.48	10	1.26	1.72	56.8	29.0	12.4	7.6	1.6
Feeling sad	1.48	2.42	10	1.25	1.70	55.0	31.8	12.4	7.0	0.6
Vomiting	0.19	1.04	10	0.10	0.29	92.8	4.2	1.6	1.0	1.4
Numbness/tingling	1.57	2.64	10	1.32	1.81	56.8	27.4	14.4	7.8	1.2
Diarrhea	0.52	0.58	10	0.37	0.67	82.3	12.2	4.8	1.2	0.6
Constipation	1.11	2.25	10	0.90	1.32	67.5	22.4	8.8	5.0	1.2
Sore mouth	0.19	0.84	8	0.12	0.27	91.0	7.0	1.2	0.4	0.8
Rash/pruritis	0.38	1.37	10	0.25	0.50	87.6	8.6	3.0	1.8	0.8
Hair loss	0.75	2.20	10	0.54	0.95	81.1	11.8	6.2	5.4	0.8
Coughing	0.77	1.84	10	0.60	0.94	73.9	19.6	5.8	3.6	0.6
Interference items (Rank Order)
Work, including housework	1.59	2.52	10	1.36	1.82	54.8	30.0	14.2	7.6	0.8
General activity	1.42	2.39	10	1.20	1.64	60.6	26.4	12.0	6.0	0.8
Mood	1.37	2.34	10	1.15	1.58	59.2	28.6	11.4	5.8	0.6
Relations with other people	0.82	1.88	10	0.64	.099	73.3	19.2	6.6	3.0	0.8
Walking	1.47	2.59	10	1.23	1.71	61.8	23.6	13.0	8.2	1.4
Enjoyment of life	1.16	2.23	10	0.95	1.37	64.9	24.0	10.2	4.8	0.8
Subscale scores
Core symptom items	1.43	1.60	8.77	1.28	1.57
All symptom items	1.17	1.34	8.11	1.05	1.29
Mean interference (six items)	1.30	1.96	9.83	1.12	1.49
Mean WAW (walk-activity-work)	1.49	2.26	10	1.28	1.70
Mean REM (relate-enjoy-mood)	1.11	1.88	10	0.94	1.29

LCL = lower 95% CL; UCL = upper 95% CL

^aPercent of patients scoring at the floor (score = 0 on the 0–10 scale)

^bPercent mild

^cPercent moderate to severe

^dPercent severe

^ePercent of missing data

Table 3.

Descriptive Statistics on Follow-Up Symptom Items, Interference Items and Subscale Scores for Breast Cancer Survivors (n = 498)

	Mean	SD	Range	LCL	UCL	% = 0a	% 1–4b	% ≥5c	% ≥7d	% Missinge
										Follow-Up
Core symptoms (Rank Order)
Pain	1.66	2.55	10	1.40	1.91	241	26.0	13.2	6.8	12.2
Fatigue	2.64	2.78	10	2.36	2.91	122	39.9	21.4	10.6	14.1
Nausea	0.54	1.65	10	0.38	0.71	73.7	9.4	4.0	2.2	12.9
Sleep disturbance	2.17	2.70	10	1.90	2.44	35.7	34.7	17.0	10.2	12.4
Distress	1.71	2.51	10	1.47	1.96	40.2	34.7	12.6	7.6	12.4
Shortness of breath	1.21	2.19	10	1.00	1.43	56.4	21.8	9.8	4.2	11.8
Cognitive difficulties	1.95	2.41	10	1.71	2.19	32.3	41.9	13.8	6.2	11.8
Lack of appetite	0.72	1.79	10	0.72	0.90	68.1	12.6	6.0	2.6	13.3
Drowsiness	1.85	2.51	10	1.61	2.10	39.8	32.7	15.4	6.2	12.0
Dry mouth	1.37	2.44	10	1.12	1.61	52.8	24.8	9.4	9.4	12.9
Feeling sad	1.62	2.50	10	1.37	1.87	44.4	29.9	13.6	7.4	12.0
Vomiting	0.23	1.08	10	0.13	0.34	81.3	5.0	1.6	1.0	12.0
Numbness/tingling	1.62	2.55	10	1.37	1.88	47.4	27.5	12.8	7.0	12.2
Diarrhea	0.55	1.54	10	0.40	0.70	70.9	12.4	4.8	2.0	11.8
Constipation	1.12	2.27	10	0.90	1.35	59.4	20.0	8.6	5.0	11.8
Sore mouth	0.19	0.89	10	0.10	0.27	80.7	6.4	0.8	0.8	12.0
Rash/pruritis	0.40	1.28	10	0.27	0.52	75.1	11.4	2.4	0.8	12.0
Hair loss	0.71	1.95	10	0.52	0.91	68.9	13.8	4.8	3.4	12.4
Coughing	0.74	1.84	10	0.56	0.93	65.5	16.4	5.8	3.4	12.2
Interference items (Rank Order)
Work, including housework	1.82	2.64	10	1.56	2.09	43.6	31.2	13.0	7.8	12.0
General activity	1.63	2.54	10	1.38	1.88	48.8	25.0	13.6	7.6	12.4
Mood	1.46	2.40	10	1.22	1.70	47.2	28.9	11.2	6.6	12.7
Relations with other people	0.95	2.10	10	0.74	1.16	62.4	18.6	6.8	4.4	12.0
Walking	1.57	2.53	10	1.32	1.82	49.6	25.4	12.4	7.0	12.4
Enjoyment of life	1.33	2.42	10	1.09	1.57	53.2	24.6	10.2	6.2	11.8
Subscale scores
Core symptom items	1.48	1.67	8.54	1.32	1.65
All symptom items	1.21	1.40	7.84	1.07	1.35
Mean interference (six items)	1.46	2.18	2.18	1.24	1.68
Mean WAW (walk-activity-work)	1.68	2.38	10	1.44	1.91
Mean REM (relate-enjoy-mood)	1.25	2.15	10	1.03	1.46

LCL = lower 95% CL; UCL = upper 95% CL

^aPercent of patients scoring at the floor (score = 0 on the 0–10 scale)

^bPercent mild

^cPercent moderate to severe

^dPercent severe

^ePercent of missing data

LPA analysis:

Target symptoms for the LPA included fatigue (reported at ≥ 5 by 22.8% of women), sleep disturbance (24.8%) and trouble remembering (17.2%). Fit indices for the step-wise LPA are presented in Table 4. A 2-class model was selected with the highest entropy. Two subgroups of women with distinct symptom severity were identified: low (77.0% of women) and high (23.0% of women). Class means for the target symptoms are presented in Table 5.

Table 4.

Model Fit Indices for Step-Wise Latent Profile Analysis

Model	Log Likelihood	BIC	Entropy	Posterior Probability	Class Proportions
1 class	−3516.25	7069.74	N/A	N/A	N/A
2 class	−3252.20	6566.47	0.910	0.984, 0.947	77.0%, 23.0%
3 class	−3154.66	6396.20	0.946	0.992, 0.940, 0.965	64.7%, 23.4%, 11.9%
4 class	−3112.62	6336.95	0.907	0.926, 0.970, 0.869, 0.958	15.7%, 60.9%, 11.3%, 12.1%

Abbreviation: BIC, Bayesian Information Criterion

Table 5.

Class Means for Target Symptoms For Breast Cancer Survivors (n = 498)

Class	Mean	SD
Fatigue
Low symptoms	1.56	0.11
High symptoms	5.88	0.36
Sleep disturbance
Low symptoms	1.09	0.14
High symptoms	6.68	0.27
Trouble remembering
Low symptoms	1.26	0.09
High symptoms	4.28	0.47

Abbreviation: SD, standard deviation.

Predictors of group membership:

Older women (OR .971, 95% CI: .952, .989) and employed women (OR .621, 95% CI: .404, .956) were less likely to be in the high subgroup and women with poorer performance status (OR 1.653, 95% CI: 1.188, 2.299) were more likely to be in the high subgroup. The odds of being in the high symptom subgroup were not statistically different between patients with varying prior treatments, diseases status, race, or use of hormonal therapies. Women in the high symptom subgroup reported lower overall QOL (p < .0001) and reported greater symptom interference with functioning (p < .0001) when compared to women in the low subgroup.

Discussion

The current study illustrates the occurrence of moderate-to-severe symptoms in a cohort of breast cancer survivors and the identification of distinct subgroups of survivors experiencing similar symptom severity. In our sample of breast cancer survivors, 3 symptoms were reported at moderate to severe levels by greater than 15% of our sample: fatigue (22.8%), sleep disturbance (24.8%), and trouble remembering (17.2%). The three target symptoms identified in our sample are consistently reported among the top symptoms experienced by breast cancer survivors in other samples.^19,30–34 Fatigue has been reported by breast cancer survivors in other studies.^30–36 Sleep disturbance in breast cancer survivors is well-documented, with as many as 90% of all survivors reporting sleep problems that persist long-term.^{19, 30–34} Changes in cognitive functioning, including trouble remembering, are commonly reported by survivors of breast cancer as well.^30–34 Our work contributes to the growing body of evidence demonstrating not only the prevalence, but also the severity of these symptoms in breast cancer survivors.

Symptoms may be synergistic, with an increase in one symptom influencing another symptom, and may be related to various demographic or clinical predictors.²³ Fatigue, sleep disturbance and cognitive impairments are often reported as being associated and possibly synergistic.^20,30 In our sample, two subgroups of women experiencing distinct symptom severity for fatigue, sleep disturbance and trouble remembering were identified for breast cancer survivors, with antecedents and consequences to distinct symptom experiences identified, consistent with the Dynamic Symptoms Model.²³ In a recent report a symptom cluster of pain, fatigue, sleep disturbance, and depression was identified among early-stage breast cancer survivors with four distinct classes.²⁴ Differences in the identified symptom cluster and the number of classes may relate to differences in the stage of survivorship across the two samples, where our sample included women much further from active treatment. Among the demographic and clinical variables, older age and being employed decreased the odds of being in the high symptom severity subgroup and poorer performance status increased the odds of being in the high symptom severity subgroup. In our sample, older women experienced less severe symptoms. Previous studies are conflicting in the relationship between age and symptoms, with some studies demonstrating the younger survivors may have more severe symptoms and other demonstrating that older survivors may be especially vulnerable for high symptom burden.^{31, 38–40} Associations among symptoms, including fatigue and cognitive impairment, and inability to return to work following treatment for breast cancer have been reported, supporting our finding that women who were employed were less likely to be in the higher symptom severity subgroup.^38–40 We did not find an increase in the odds of being in the high symptom severity subgroup based on race, although associations between race and symptom experiences have been reported.^41–43 While we found that various treatment histories did not increase the odds of a higher symptom severity in our cohort, others have found receipt of chemotherapy related to higher symptom burden.^{30, 31} This discrepancy may be related to our inability to differentiate between type of chemotherapy, immunotherapy and/or endocrine therapy in our sample. Lower quality of life and increased symptom-related interference with functioning were reported by women in the higher symptom severity subgroup, which is consistent with previously reported findings.^{11, 44–45}

Our study was limited by a lack of data for several relevant variables of interest, including co-morbidities, menopausal status, symptom management, and previous sleep patterns and the inability to separate varying treatment regimen histories. In addition, we were limited to the symptom items included on the symptom inventory in the parent study. Some important symptoms (e.g. symptoms related to lymphedema) may not have been included on the measure and, therefore, were not included in our analysis. The use of a single-item measure for symptom severity has limitations, including the risk for increased measurement error when compared to measures using multiple items. However, single-item measures are reliable and valid for collecting symptom data and minimize participant burden when monitoring multiple symptoms over time. While data were collected from multiple sites across the United States, findings may not generalize to all populations of breast cancer survivors.

Identification of women at risk for high symptoms during survivorship may allow clinicians to intensify their approach to symptom management, thereby mitigating poor outcomes and impairments in QOL. Survivorship care should emphasize screening for and management of symptoms, including fatigue, sleep disturbance, and cognitive changes in women with breast cancer, acknowledging the potential for a synergistic effect. Recognizing the existence of a percentage of women who experience concurrent fatigue, sleep disturbance and trouble remembering at moderate to severe levels during survivorship is important, as not all women will need the same level of symptom management. This is particularly important as these women may be at increased risk for poorer outcomes, including decreased quality of life and impaired functional status. Potential interventions for at-risk women may include patient and caregiver education regarding self-care, careful and routine monitoring and assessment, and referral or pharmacological intervention. Evidence-based guidelines from national organizations such as the Oncology Nursing Society Putting Evidence into Practice (PEP) resources and the National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines for Supportive Care should be consulted for identifying and implementing potential symptom management interventions for fatigue, disturbed sleep, and cognitive changes in survivorship care that are derived from the existing evidence-base [46, 47]. Future research should focus on determining whether these classes are replicable and stable across the survivorship trajectory and continue to study potential antecedents of class membership, including sleep-related variables, co-morbidities, and genetic and molecular factors that may underlie co-occurring symptoms. Additional research is needed to determine whether mitigation of individual symptoms through symptom management and targeted approaches to addressing any potential modifiable risk factors such as unmanaged co-morbid conditions and sleep-related variables results in a reduction in the prevalence or severity of co-occurring symptoms.