Treatment-Related Symptom Clusters in Breast Cancer: A Secondary Analysis

Article Outline

• Abstract
• Introduction
• Methods
  • Sample and Setting
  • Instruments
  • Data Collection Time Points
  • Combining Experimental and Control Groups from the Parent Study
  • Analyzed Symptoms
  • Data Analysis
• Results
  • Sample
  • Time 1 Symptom Custer (n=222)
  • Time 2 Symptom Clusters (n=199)
  • Time 3 Symptom Clusters (n=179)
• Discussion
• References
• Copyright

Abstract 

This study investigated treatment-related symptom clusters and the influence of selected demographic/clinical variables on symptom clustering in breast cancer patients across a treatment trajectory. A secondary analysis of 282 breast cancer patients receiving chemotherapy or radiotherapy was done to determine the clustering of oncologic treatment-related symptoms at selected time points of treatment. Two distinct clusters were identified: a psychoneurological cluster and an upper gastrointestinal cluster. The clustering of symptoms was generally stable across the treatment trajectory. The clustering, however, was weaker when the time lapse after the completion of treatment became longer. Demographic and clinical variables did not significantly influence symptom clustering. Psychoneurological symptoms had a tendency to occur together across the treatment trajectory, as did upper gastrointestinal symptoms. Effective symptom assessment/management strategies need to take into account this co-occurrence of symptoms. The findings from this study underscore the need for further investigation of the common biological basis of symptoms to attain more effective management of multiple symptoms.

Key Words: Symptom clusters, breast neoplasm, treatment, symptom management, symptom assessment, biological basis

Back to Article Outline

Introduction 

Cancer patients may experience several treatment-related symptoms at once.¹, ², ³, ⁴ When multiple symptoms are present, the collective consequences of the symptoms may result in more impairment than a single symptom, as multiple symptoms may exacerbate each other.⁴, ⁵ The purposes of this study were to identify treatment-related symptom clusters in breast cancer patients across three points in the treatment trajectory and to examine the influence of demographic/clinical variables on symptom clustering. A symptom cluster is defined as a stable group of interrelated symptoms occurring simultaneously.⁶ Identification of symptom clusters has the potential to lead to more efficient and effective symptom assessment and management strategies in cancer patients than the traditional approach to individual symptoms.

There has been a paucity of research identifying treatment-related symptom clusters in breast cancer, the most frequently occurring cancer in women in the United States.⁷ This cancer population is particularly useful in isolating treatment-related symptom clusters given that breast cancer patients have no specific symptoms induced by the cancer itself other than a breast mass in the absence of metastases. Thus, clusters of symptoms identified during therapy can be reasonably considered to be due to treatment. Therefore, breast cancer patients were selected for studying treatment-related symptom clusters. Research has suggested that some clinical and demographic variables (such as age, gender, employment status, comorbid conditions, disease stage, cancer type, treatment modality, treatment trajectory, and measurement time point) may influence symptom clusters.¹, ², ⁸, ⁹, ¹⁰, ¹¹, ¹² The influence of these variables was also examined in the analyses of symptom clustering.

Back to Article Outline

Methods 

Sample and Setting 

A secondary analysis of data collected from a randomized clinical trial of the effectiveness of a cognitive behavioral intervention on fatigue in cancer patients was conducted.¹³ Subjects in the primary study were a convenience sample (n=396) of oncology patients from two American cancer centers from 1999 to 2002. Patients in the primary study were initiating treatment for various cancer types, and planned to receive at least three cycles of chemotherapy (CTX), six weeks of radiotherapy (RTX), or concurrent RTX and CTX. In addition, they had not received prior treatment other than surgery for at least one month prior to enrollment, and had as a goal of treatment either cure or local control. Patients whose treatment plan included stem cell transplantation, interleukins, interferon, or tumor necrosis factor, who had a diagnosis of chronic fatigue syndrome or evidence of a psychiatric disorder, or who had received treatment for anemia or depression during the prior three weeks were excluded, along with patients who were enrolled in another psychoeducational intervention study. The approvals of the institutional review boards were obtained. For the present study, data from the breast cancer patients (n=282) were used with no further inclusion or exclusion criteria applied. The sample size was smaller than 282 in several analyses because of missing data.

Instruments 

Fatigue intensity was measured by one item from the General Fatigue Scale (fatigue in the past week).¹⁴ Items on the General Fatigue Scale assess the intensity of fatigue during various time frames (the present day, the past 48hours, the past week), the level of distress caused by fatigue, and the impact of fatigue on daily activities. Only one item (fatigue in the past week), however, was chosen for this study in order to establish consistency in the time frame/dimension measured and to salvage more cases for the analysis. The scale for fatigue ranges from 1 to 10 (no fatigue ∼ greatest possible fatigue).

Two subscales (depression and confusion) of the Profile of Mood States-Short Form were used to measure the intensity of depressive mood and cognitive disturbance for the past two to three days.¹⁵ Each subscale is composed of five items and each item was scaled from 0 to 4 (not at all ∼ extremely). Cronbach's α for each subscale was 0.81 and 0.75, respectively, in this study.

Insomnia (for the past month) was measured by the Pittsburgh Sleep Quality Index.¹⁶ The Pittsburgh Sleep Quality Index is composed of 19 self-rated questions and five bed-partner or roommate rated questions. The latter five questions are designed for clinical purposes only and thus are not included in scoring. The response options vary across items. Cronbach's α for the global score of the Pittsburgh Sleep Quality Index was 0.74 in this study.

The Side Effect Checklist measured the intensity of sixteen other treatment-related symptoms (for the past week). Each item measures each symptom using a four-point Likert-type scale (1=not at all severe; 4=quite a bit severe). If patients do not have a specific symptom, then they are recorded as having zero points for that item. The Side Effect Checklist was derived from a previous Self-Care Diary.¹⁷ Its content validity was tested by the oncology clinical experts prior to data collection for the primary study.

For the main analyses, single item scores were used for 16 symptoms from the Side Effect Checklist, and for fatigue, from the General Fatigue Scale. The scale score was used for depressed mood, cognitive disturbance, and insomnia; the scale scores were computed as described in the documentation for each instrument.

The prevalence of each symptom was calculated as follows: For fatigue, patients with a score ≥4.0 were considered to have the symptom. For depressed mood and cognitive disturbance, patients with a total score of ≥5.0 were considered to have each symptom. For depressed mood, cognitive disturbance, and fatigue, cut-points were determined clinically. The cut-point for the insomnia scale was greater than 5.¹⁶ (The scale for insomnia ranges from 0 to 21.) The presence or absence of other symptoms was measured.

Data Collection Time Points 

Data were collected over the treatment trajectory (before and after CTX/RTX began). Data were collected at baseline (Time 1) and at two follow-up time points chosen because they were the time points found to have maximum fatigue levels in cancer patients.¹³ Baseline data (Time 1) were collected prior to therapy (CTX or RTX). For CTX patients, the two follow-up points were 48hours after the second (Time 2) and third CTX treatments (Time 3). For RTX patients, data were collected during the last week of RTX (a total of six weeks of treatment) (Time 2) and one month after completion of treatment (Time 3). It should be noted that although patients were not at the same points in relation to their initial diagnosis, patients within each treatment group (CTX or RTX) were at similar points in relation to the treatment given during the study.

Combining Experimental and Control Groups from the Parent Study 

Although the primary study demonstrated significant decreases in fatigue over time for the experimental group, the experimental and control groups were combined for these analyses because there was no statistically or clinically significant difference in fatigue level found between the two groups for the first two time points and no clinically significant difference (although statistically significant) at the third time point (4.1±2.2 vs. 4.7±2.1 on a 10-point scale).¹³

Analyzed Symptoms 

Twenty common symptoms found among breast cancer patients undergoing treatment were examined: cough, pain, diarrhea, nausea, vomiting, decreased appetite, constipation, urinary frequency, urinary burning, hot flashes, rectal irritation, swelling of an arm or leg, dyspnea, sore throat or mouth, skin damage at the treatment area, IV site irritation, fatigue, depressed mood, cognitive disturbance, and insomnia. The initial analyses showed that symptoms with an extremely low prevalence rate might interfere with the analysis process (e.g., restricted variance) and interpretation of results. Furthermore, the symptoms with a low prevalence do not reflect the key symptoms in a cluster. Uncommon symptoms (<10% prevalence) were, therefore, excluded from the analyses. At Time 1, four symptoms (urinary burning, IV site irritation, skin damage at the treatment area, vomiting) were excluded. At Times 2 and 3, two symptoms (urinary burning, IV site irritation) were excluded.

Data Analysis 

Common factor analysis with principal axis factoring was conducted at each time point. Symptoms with a salient loading on a factor were considered to be a symptom cluster. The squared multiple correlation method was used to estimate communality.¹⁸ The number of factors was estimated by examining: (a) Bartlett's Chi-square criteria;¹⁹ (b) parallel analysis using eigenvalues;²⁰ and (c) Velicer's Maximum Average Partial Test.²¹ The criteria to determine the best factor model were (a) at least three variables with loadings above 0.4 in a factor; (b) Cronbach's coefficient α (internal consistency) for each factor higher than 0.65 and item-total correlations greater than 0.20; (c) a parsimonious solution (no symptoms loading on multiple factors); (d) clinical/theoretical plausibility of factors; and (e) the highest hyperplane count (the number of zero or near zero loadings [0.10 to −0.10]). Plausibility was determined by compatibility with previous research/clinical findings related to pathophysiological mechanisms and concurrence of symptoms. An α coefficient of 0.65 or higher was chosen because of the heterogeneity of the sample for age, disease stage, treatment modality, and comorbid conditions.²² A minimum of three symptoms for each factor were required because of the statistical requirement for factor analysis.²³

The influence of the selected demographic/clinical variables (treatment modality, age, employment status, disease stage, and comorbid conditions) on symptom clustering was evaluated by two generalizability tests: coefficient κ and Cronbach's α. Both generalizability tests examined whether findings from the full sample were maintained across subgroups (patients were grouped by selected clinical/demographic characteristics). Coefficient κ²⁴ measures the agreement between patterns of salient and nonsalient loadings across two factors (i.e., factor structure congruence). Thus, it indicates the degree to which factors match, where one factor is based on the full sample of patients and the other factor is extracted from a given subsample of those patients. A κ>0.75 indicates excellent agreement; a κ between 0.40 and 0.75 represents fair to good agreement; and a κ<0.40 indicates poor agreement. Cronbach's α shows the generalizability of the associations of symptoms within a cluster (internal consistency) across subgroups.

Back to Article Outline

Results 

Sample 

Demographic/clinical characteristics of the sample are presented in Table 1. All patients in the sample (n=282) were female, with a mean age of 55 years (range, 30–83). The majority were married (70%), Caucasian (92%), and had post high school education (66%). Almost half (49%) were employed. More than half had at least one comorbid condition (56%), most frequently hypertension. Almost half of the women had early stage breast cancer (stages 0 and 1 [49%], with a few having stages 3 or 4 [11%]). During the study, 56% received RTX and 44% received CTX. None of the women received concurrent CTX and RTX.

Table 1. Demographic/Clinical Characteristics of the Sample (n=282)

Variables	Number of Patients (%)
Age (mean years±SD, range)	55.21 ± 12.1, 30–83
Race
Caucasian	258 (91.5)
Noncaucasian	24 (8.5)
Marital status
Married	198 (70.2)
Not married	79 (28.0)
Missing	5 (1.8)
Education
High school or less	89 (31.6)
Post high school education	185 (65.6)
Missing	8 (2.8)
Employment status
Employed	138 (48.9)
Unemployed	144 (51.1)
Treatment modality during the study
Chemotherapy	125 (44.3)
Radiation therapy	157 (55.7)
Disease stage
Stage 0	25 (8.9)
Stage 1	114 (40.4)
Stage 2	106 (37.6)
Stage 3	27 (9.6)
Stage 4	4 (1.4)
Missing	6 (2.1)
Comorbid conditions
At least one	157 (55.7)
None	124 (44.0)
Missing	1 (0.3)

Time 1 Symptom Custer (n=222) 

At Time 1, common factoring identified one symptom cluster characterized by depressed mood, cognitive disturbance, fatigue, insomnia, and pain (Table 2). Cronbach's α for the full sample was 0.68, with item-total rs ranging from 0.32 to 0.52.

Table 2. Factor Structure at Time 1(n=222)

The symptom cluster was generalizable to all subgroups of patients with respect to treatment modality (CTX vs. RTX), age, employment status, disease stage, and comorbid condition, with all κ>0.40 (Table 3). Internal consistency tended to fluctuate across the groups, with the greatest homogeneity (α=0.73) found for unemployed patients and the least (α=0.60) for employed patients (Table 3). Nevertheless, it was observed in correlation matrices that the relationships among the cluster symptoms within patient groups were generally strong and positive (0.12≤r≤0.54 in the full sample), although the relationship between pain and depressed mood and between pain and cognitive disturbance manifested weaker relationships in many groups (e.g., r=0.08 and 0.09, respectively, in the RTX group).

Table 3. Generalizability of Symptom Cluster to Subgroups at Time 1 (n=222)

		CTX (n=96)	RTX (n=126)	<65 Years (n=173)	≥65 Years (n=49)	Unemployed (n=113)	Employed (n=109)	Early Stage (n=113)	Late Stage (n=105)a	No Comorbidity (n=101)	Comorbidity (n=121)
α	Cluster I	0.69	0.65	0.68	0.66	0.73	0.60	0.66	0.67	0.66	0.70
κ	Cluster I	0.86	0.59	0.86b		0.74	0.71	0.52	0.42	0.71	0.63

Early stage included Stages 0 or 1 cancer patients, and late stage cancer included Stages 2, 3, or 4 cancer patients.

Time 2 Symptom Clusters (n=199) 

Promaxian rotation yielded two viable symptom clusters (Table 4). Cluster I was composed of upper-gastrointestinal symptoms (nausea, vomiting, and decreased appetite), whereas Cluster II consisted of depressed mood, cognitive disturbance, fatigue, insomnia, pain, and hot flashes. Internal consistency for the clusters was 0.73 and 0.69, respectively, with item-total rs ranging from 0.40 to 0.63 for Cluster I and from 0.29 to 0.53 for Cluster II. The Time 2 symptom clusters were submitted to second-order common factoring as based on unit-weighted scores and were found to be independent from each other (correlation between the clusters=0.28; overlapping variance=8%), revealing that each cluster conveyed a significant proportion of unique and reliable variance (55% for Cluster I; 51% for Cluster II).

Table 4. Factor Structure at Time 2 (n=199)

Table 5 illustrates the generalizability of the symptom clusters in terms of internal consistency and structural congruence across demographic and clinical subgroups of patients. Although the coefficient α for Cluster I was relatively low in the RTX group (α=0.61), the bivariate correlations between symptoms within this subgroup were all statistically significant. The lower α for Cluster I may be a consequence of relatively constricted symptom variance (e.g., SD=0.40 for vomiting in the RTX group vs. 0.97 in the CTX group). For Cluster II, the lowest α was for employed subgroup, whereas it was highest for the unemployed and patients with later stage disease. Pain and hot flashes had the lowest loadings for Cluster II (0.47 and 0.44, see Table 4) and they tended to have lower correlations with various symptoms in groups of patients whose α was lower. Symptoms with higher loadings (depressed mood, cognitive disturbance, fatigue, insomnia) maintained statistically significant correlation with each other even in the groups with a low α. The low α for certain groups indicates that the homogeneity of symptom experience is somewhat variant among patients in those groups. For example, patients in groups with higher αs for Cluster I (e.g., CTX group) were more likely to experience all three respective symptoms (nausea, vomiting, and decreased appetite) at the same time than were patients in group with lower α (RTX group). Generalizability of the factor structure was further evident for both clusters across all patient groups, with κ ranging from 0.57 to 0.82 for Cluster I and 0.73 to 1.0 for Cluster II.

Table 5. Generalizability of Symptom Cluster to Subgroups at Time 2 (n=199)

		CTX (n=95)	RTX (n=104)	<65 Years (n=151)	≥65 Years (n=48)	Unemployed (n=99)	Employed (n=100)	Early Stage (n=100)	Late Stage (n=94)a	No Comorbidity (n=86)	Comorbidity (n=113)
α	Cluster I	0.67	0.61	0.72	0.70	0.64	0.78	0.67	0.72	0.68	0.76
	Cluster II	0.72	0.69	0.67	0.72	0.73	0.64	0.66	0.73	0.71	0.70
κ	Cluster I	0.82	0.82	0.82		0.57	0.82	0.60	0.82	0.68	0.68
	Cluster II	0.75	0.87	1.0		1.0	1.0	0.73	1.0	0.88	0.88

Time 3 Symptom Clusters (n=179) 

Promaxian rotation also identified two viable symptom clusters at Time 3 (Table 6), these having been retained from overextraction to a three-factor model. (Note: Overextraction is a procedure commonly applied to disentangle primary common factors from underidentified secondary factors, where simple structure is resolved by rotating more factors than are ultimately retained.²⁵) The third factor (dropped factor) included only the two symptoms of urinary frequency and shortness of breath, which did not meet the criteria for the factor model.

Table 6. Factor Structure at Time 3 (n=179)

The third factor was dropped and it is not presented.

Cluster I consisted of nausea, vomiting, and decreased appetite and Cluster II of depressed mood, cognitive disturbance, fatigue, insomnia, and pain. Coefficient α for the respective clusters was 0.81 and 0.77, with item-total rs ranging from 0.64 to 0.67 and 0.44 to 0.64. Moreover, the clusters were quite independent (correlation between clusters=0.45; overlapping variance=20%) and conveyed substantial proportions of uniquely interpretable and reliable variance (48% for Cluster I; 44% for Cluster II).

Although coefficient α for each cluster fluctuated across groups (Table 7), positive strong correlations between any two symptoms within each cluster were consistently observed for each subgroup of patients (0.57≤r≤0.60 for Cluster I in the full sample; 0.25≤r≤0.72 for Cluster II in the full sample). In subgroups with a lower α, however, insomnia tended to have lower correlations with cognitive disturbance or pain than in the full sample (e.g., rs=0.19 and 0.26, respectively, in the early stage group; rs=0.25 and 0.32 in the full sample).

Table 7. Generalizability of Symptom Clusters to Subgroups at Time 3 (n=179)

		CTX (n=78)	RTX (n=101)	<65 Years (n=135)	≥65 Yearsa (n=44)	Unemployed (n=93)	Employed (n=86)	Early Stage (n=91)	Late Stage (n=84)b	No Comorbidity (n=77)	Comorbidity (n=102)
α	Cluster I	0.75	0.74	0.82		0.70	0.85	0.75	0.81	0.82	0.80
	Cluster II	0.75	0.78	0.77	0.74	0.81	0.73	0.72	0.81	0.75	0.79
κ	Cluster I	0.57	0.57	1.0		1.0	0.68	1.0	0.82	0.57	0.82
	Cluster II	0.56	0.05	1.0		0.72	0.75	0.68	0.26	0.05	0.75

Low κs are in bold type.

Coefficient κs for Cluster I ranged from 0.57 to 1.0, indicating that Cluster I was generalizable across all subgroups (Table 7). On the other hand, the Cluster II was not generalizable for patients who were receiving RTX, were of late stage, or who did not have comorbid conditions. Not only were κs low in these subgroups (<0.40), but gastrointestinal and psychoneurological symptoms tended to load on the same factor. This phenomenon may be related to heterogeneity in the time lapse after the treatment. At Time 3, CTX patients had received treatment two days before, whereas RTX patients had completed their treatment one month before.

Back to Article Outline

Discussion 

Across the treatment trajectory, symptoms had a tendency to group into two distinct clusters: a psychoneurological cluster and an upper-gastrointestinal cluster (Table 8). The psychoneurological cluster was present both before (i.e., Time 1) and after initiating CTX or RTX (i.e., Times 2 and 3) whereas the upper-gastrointestinal cluster was present after initiating treatment (CTX or RTX). It was not surprising that the upper-gastrointestinal cluster (nausea, vomiting, and decreased appetite) was absent before treatment (i.e., Time 1), as gastrointestinal symptoms in breast cancer occur mainly after beginning CTX. Sarna and Brecht²⁶ also reported the grouping tendency of gastrointestinal symptoms (nausea frequency and severity, and appetite loss) in their study of female lung cancer patients. However, as patients in their study appear to be at various time points in the treatment trajectory, it is not clear whether the gastrointestinal symptoms were related to treatment or other factors. Nevertheless, because of the physiological associations among upper-gastrointestinal symptoms, it is understandable in breast cancer patients to find the upper-gastrointestinal cluster following cancer treatment (especially CTX).

Table 8. Symptom Clusters Across Time Points

The present study provided empirical evidence of psychoneurological symptom clustering. The clustering tendency of psychoneurological symptoms, however, is not a unique phenomenon found in this study. Bender et al.²⁷ in a study of breast cancer patients found that fatigue, cognitive impairment, and mood problems formed a cluster with some other psychoneurological symptoms (e.g., pain or difficulty sleeping) in three different samples: after surgery and prior to cancer therapy (n=40); after completion of CTX or RTX (n=88); and in women with metastatic breast cancer at various points in their treatment schedules (n=26). Difficulty sleeping and pain were present in the cluster only after surgery.²⁷ This is not consistent with our findings in which insomnia and pain consistently grouped with other psychoneurological symptoms. These differences can be attributed to the difference in treatment trajectory, symptoms investigated, and analytic technique used.

Other studies also have suggested a concurrence of psychoneurological symptoms.², ³, ⁴, ⁵, ¹¹, ²⁸ For example, Given et al.³ found that 18% of breast cancer patients in their sample experienced all three symptoms of fatigue, pain, and insomnia, and 33% had two of the three symptoms. Bower et al.²⁸ found that fatigue was correlated with pain, sleep disturbance, and depression in breast cancer survivors.

The symptoms in the psychoneurological cluster had a relatively high prevalence rate in our sample. The prevalence rates for fatigue and insomnia ranged from 58% to 74% across time points; for pain, 40%–55% across time points; for depressed mood, 24%–25% across time points; and for cognitive disturbance, 27%–34% across time points. These symptoms have been reported to occur frequently in cancer patients regardless of the type of cancer or treatment trajectory.¹, ²⁷ Findings from our study and others underscore the importance of vigilant assessment and management for these symptoms in caring for the breast cancer patient. The grouping tendency of psychoneurological symptoms can be used to improve assessment and management for such prevalent symptoms in cancer patients.

Future inquiry is indicated into the etiology behind this clustering, and how this clustering tendency can be used to guide clinical practice. Several researchers have proposed the possibility that a common etiology, in particular cytokines, may be responsible for symptom clustering in cancer patients.²⁹, ³⁰, ³¹ Also, we cautiously suggest that the psychoneurological symptom clustering in cancer patients can also be attributed to the psychological distress in response to cancer diagnosis and scheduled treatments. At this point, it is not clear yet whether or to what extent psychological distress or biological change in response to cancer or its treatments is responsible for symptom clustering in cancer patients.

Recently, a few research teams reported a slightly different symptom clustering from our findings in advanced cancer patients. Using cluster analysis, Walsh and Rybicki³² found that fatigue in advanced cancer (various cancer types) was included in the anorexia–cachexia cluster, but not in the psychoneurological cluster. Another team³³ used factor analysis to identify symptom clusters and reported that pain, fatigue, sleep disturbance, lack of appetite, and drowsiness composed what they called the sickness symptom cluster, whereas distress and sadness composed the emotional cluster. Fifty percent of the patients had a stage III or IV cancer (various cancer types) in their study.³³

Whether such differences in symptom clustering between our study and theirs should be attributed to the differences in the sample or in statistical decisions involved during analyses is not clear. Nonetheless, such differences can also indicate that etiology or a biological mechanism of a certain symptom may change in the context of illness trajectory and, therefore, interventions (e.g., pharmacological approaches) to a symptom should be adjusted as well. For example, fatigue in late stage cancer patients can be a part of cachexia syndrome, whereas fatigue in early stage cancer patients before initiating treatments can be a part of psychological distress syndrome. In other words, the clustering pattern of symptoms can provide a clue regarding etiology of a certain symptom and can guide clinicians in choosing interventions for that symptom. Such possibilities need to be further examined.

In the present study, hot flashes had strong relationships with other psychoneurological symptoms at Time 2. It is not clear why this symptom was included in the psychoneurological cluster only at Time 2. Although some studies have found that menopausal symptoms, such as hot flashes, were associated with other neurological symptoms, the findings from those studies were not consistent across the treatment trajectory.²⁷, ²⁸, ³⁴ Several studies have reported that menopausal symptoms were related to fatigue (one of the neurological symptoms) in breast cancer patients who had finished CTX.²⁸, ³⁴ However, another study of symptom clusters²⁷ did not find that hot flashes were associated with other neurological symptoms in breast cancer patients who had completed CTX or RTX. In contrast to these studies, our analyses found that hot flashes grouped with other neurological symptoms during cancer therapy. This finding needs to be replicated in another sample. Also, it may be useful to identify patients experiencing the psychoneurological cluster with hot flashes and to examine those patients' characteristics.

In our study, symptoms were factor analyzed. The factors (or symptom clusters) indicate the tendency for symptoms to cluster. This does not indicate that individuals will experience all of the clustered symptoms. For example, some symptoms had a relatively low prevalence in contrast to the other symptoms within the cluster. At Time 2, 10% of the sample experienced vomiting compared to 47% who experienced nausea and 40% who had decreased appetite. These disparate prevalence rates across symptoms in a cluster suggest that symptoms in this cluster did not always occur together, but yet still are related. Because of the use of antiemetics, many cancer patients undergoing cancer treatments were free from vomiting but they still had nausea and decreased appetite. Also, women undergoing RTX were less likely to have these symptoms. Therefore, the three symptoms had a tendency to occur together during or following cancer therapy but symptom manifestation was different across patients.

The interpretation of symptom concurrence in the psychoneurological cluster was complicated because numerous clinical/individual factors may influence the manifestation of symptoms in this cluster, such as the type of drugs given to patients. At this time point, the concept of symptom clusters in oncology is not fully developed, and thus it has not been determined whether symptoms in a cluster should always occur together. Nevertheless, it is valuable to identify a group of patients who experience symptoms in a similar pattern (e.g., patients with all symptoms in the cluster) and examine their clinical/individual characteristics. Symptom manifestations in cancer patients can be better understood by such efforts, and the concept of symptom clusters can be further clarified. In fact, our research team is currently investigating such clinical subgroups of patients using the empirically identified clusters on the same data.

There are several limitations imposed by using the existing data set for this study.

First, the effect of confounding variables on symptom clustering, such as CTX regimens, radiation dose, hormonal treatments, drugs used for symptom management, surgery type, and time lapse since surgery, were not controlled. Future studies with a larger sample size can evaluate the effect of these confounding variables. Second, the measurements asked patients to recall their symptoms over differing intervals. The difference in the recall time frame can limit the study because it may cause uncertainty about the time of concurrent duration. However, the inconsistency in the time frame across symptom measures was very small, except for the time frame of insomnia: Subjects were asked to rate their symptoms over the past week for most symptoms, the past two to three days for depressed mood and cognitive disturbance, and the past month for insomnia. Although patients were asked to recall slightly different time frames, patients are likely to recall their most recently occurring symptoms. Similarly, many other symptoms were measured by a single item, which is less reliable than a scale, and a single item may not completely cover the domain of a symptom. The use of a single-item measure remains as a limitation of the study. Lastly, the sample underrepresented minorities and later stage cancer patients.

Nevertheless, the findings from this study have implications for clinical practice and research. Systematic assessment of a cluster of symptoms for specific types of cancer would be beneficial to patients and clinicians because clinicians are apt to focus on individual symptoms and patients tend to underreport their symptoms.³⁵ Clinicians can use the concurrent tendency of symptoms in a cluster for assessing symptoms and providing preparatory information to patients regarding what to expect during cancer treatment in terms of symptom experience (symptom intensity, concurrent pattern of symptoms). The findings of this study also underscore the need for further investigation into the common etiology of symptoms and the possibility of collective symptom management.

Back to Article Outline

References 

1. Cooley ME, Short TH, Moriarty HJ. Symptom prevalence, distress, and change over time in adults receiving treatment for lung cancer. Psychooncology. 2003;12(7):694–708
   • View In Article
   • MEDLINE
   • CrossRef
2. Given CW, Given B, Azzouz F, Kozachik S, Stommel M. Predictors of pain and fatigue in the year following diagnosis among elderly cancer patients. J Pain Symptom Manage. 2001;21(6):456–466
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (82 KB)
   • CrossRef
3. Given B, Given CW, Azzouz F, Stommel M. Physical functioning of elderly cancer patients prior to diagnosis and following initial treatment. Nurs Res. 2001;50(4):222–232
   • View In Article
   • MEDLINE
   • CrossRef
4. Sarna L. Correlates of symptom distress in women with lung cancer. Cancer Pract. 1993;1(1):21–28
   • View In Article
   • MEDLINE
5. Dodd MJ, Miaskowski C, Paul SM. Symptom clusters and their effect on the functional status of patients with cancer. Oncol Nurs Forum. 2001;28(3):465–470
   • View In Article
   • MEDLINE
6. Kim H-J, McGuire DB, Tulman L, Barsevick AM. Symptom clusters: concept analysis and clinical implications for cancer nursing. Cancer Nurs. 2005;28(4):270–282
   • View In Article
   • MEDLINE
7. American Cancer Society. Breast cancer facts & figures (online document). 2005. Accessed March 6, 2006. Available at http://www.cancer.org/downloads/STT/CAFF2005BrF.pdf
   • View In Article
8. Akechi T, Kugaya A, Okamura H, Yamawaki S, Uchitomi Y. Fatigue and its associated factors in ambulatory cancer patients: a preliminary study. J Pain Symptom Manage. 1999;17(1):42–48
   • View In Article
9. Berger AM. Patterns of fatigue and activity and rest during adjuvant breast cancer chemotherapy. Oncol Nurs Forum. 1998;25(1):51–62
   • View In Article
   • MEDLINE
10. Gift AG, Jablonski A, Stommel M, Given W. Symptom clusters in elderly patients with lung cancer. Oncol Nurs Forum. 2004;31(2):203–210
    • View In Article
11. Jacobsen PB, Hann DM, Azzarello LM, et al. Fatigue in women receiving adjuvant chemotherapy for breast cancer: characteristics, course, and correlates. J Pain Symptom Manage. 1999;18(4):233–242
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (132 KB)
    • CrossRef
12. Hoskins CN. Breast cancer treatment-related patterns in side effects, psychological distress, and perceived health status. Oncol Nurs Forum. 1997;24(9):1575–1583
    • View In Article
    • MEDLINE
13. Barsevick AM, Dudley W, Beck S, et al. A randomized clinical trial of energy conservation for patients with cancer-related fatigue. Cancer. 2004;100(6):1302–1310
    • View In Article
    • MEDLINE
    • CrossRef
14. Meek PM, Nail LM, Jones LS. Internal consistency reliability and construct validity of a new measure of cancer treatment related fatigue: the General Fatigue Scale(GFS). [abstract] Oncol Nurs Forum. 1997;24:334
    • View In Article
15. McNair DM, Lorr M, Droppleman LF. Profile of Mood States, Revised (POMS). San Diego, CA: EdITS/Educational and Industrial Testing Service; 1981;
    • View In Article
16. Buysse DJ, Reynolds CF, Monk TH, Berman SR, Kupfer DJ. The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research. Psychiatry Res. 1989;28(2):193–213
    • View In Article
    • MEDLINE
    • CrossRef
17. Nail LM, Jones LS, Greene D, Schipper DL, Jensen R. Use and perceived efficacy of self-care activities in patients receiving chemotherapy. Oncol Nurs Forum. 1991;18(5):883–887
    • View In Article
    • MEDLINE
18. Gorsuch RL. Factor analysis. 2nd ed.. Hillsdale, NJ: Erlbaum; 1983;
    • View In Article
19. Geweke JF, Singleton KI. Interpreting the likelihood of ratio statistics in factor models when sample size is small. J Am Stat Assoc. 1980;75:133–137
    • View In Article
20. Horn JL. A rationale and test for the number of factors in factor analysis. Psychometrika. 1965;30:179–185
    • View In Article
    • MEDLINE
    • CrossRef
21. Velicer WF. Determining the number of components from the matrix of partial correlations. Psychometrika. 1976;41:321–327
    • View In Article
    • CrossRef
22. Spector PE. Summated rating scale construction: an introduction. Newbury Park, CA: Sage; 1992;
    • View In Article
23. Tabachnick BG, Fidell LS. Using multivariate statistics. 4th ed.. New York: HarperCollins College Publisher; 2001;
    • View In Article
24. Guadagnoli E, Velicer W. A comparison of pattern matching indices. Multivariate Behav Res. 1991;26:323–343
    • View In Article
25. Wood JM, Tataryn DJ, Gorsuch RL. Effects of under- and overextraction on principal axis factor analysis with varimax rotation. Psychol Methods. 1996;1:354–365
    • View In Article
    • CrossRef
26. Sarna L, Brecht ML. Dimensions of symptom distress in women with advanced lung cancer: a factor analysis. Heart Lung. 1997;26(1):23–30
    • View In Article
    • Abstract
    • Full-Text PDF (1212 KB)
    • CrossRef
27. Bender CM, Ergun FS, Rosenzweig MQ, Cohen SM, Sereika SM. Symptom clusters in breast cancer across 3 phases of the disease. Cancer Nurs. 2005;28(3):219–225
    • View In Article
    • MEDLINE
28. Bower JE, Ganz PA, Desmond KA, et al. Fatigue in breast cancer survivors: occurrence, correlates, and impact on quality of life. J Clin Oncol. 2000;18(4):743–753
    • View In Article
    • MEDLINE
29. Cleeland CS, Bennett GJ, Dantzer R, et al. Are the symptoms of cancer and cancer treatment due to a shared biologic mechanism? A cytokine-immunologic model of cancer symptoms. Cancer. 2003;97(11):2919–2925
    • View In Article
    • MEDLINE
    • CrossRef
30. Lee BN, Dantzer R, Langley KE, et al. A cytokine-based neuroimmunologic mechanism of cancer-related symptoms. Neuroimmunomodulation. 2004;11(5):279–292
    • View In Article
    • MEDLINE
    • CrossRef
31. Miaskowski C, Aouizerat BE. Is there a biological basis for the clustering of symptoms. Semin Oncol Nurs. 2007;23(2):99–105
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (97 KB)
    • CrossRef
32. Walsh D, Rybicki L. Symptom clustering in advanced cancer. Support Care Cancer. 2006;14(8):831–836
    • View In Article
    • MEDLINE
    • CrossRef
33. Chen ML, Tseng HC. Symptom clusters in cancer patients. Support Care Cancer. 2006;14(8):825–830
    • View In Article
    • MEDLINE
    • CrossRef
34. Broeckel JA, Jacobsen PB, Horton J, Balducci L, Lyman GH. Characteristics and correlates of fatigue after adjuvant chemotherapy for breast cancer. J Clin Oncol. 1998;16(5):1689–1696
    • View In Article
35. Chang VT, Hwang SS, Feuerman M, Kasimis BS. Symptom and quality of life survey of medical oncology patients at a Veterans Affairs medical center: a role for symptom assessment. Cancer. 2000;88(5):1175–1183
    • View In Article
    • MEDLINE
    • CrossRef