Impact of Cancer-Related Symptom Synergisms on Health-Related Quality of Life and Performance Status

Article Outline

• Abstract
• Introduction
• Methods
  • Measurements
  • Statistical Analyses
• Results
  • Symptom Frequency
  • Profiles of Subgroups
  • Demographic and Clinical Characterization of the Subgroups
  • Differences in Outcomes of Symptom Subgroups
  • Impact of Symptoms on HRQOL and Performance Status
• Discussion
• Acknowledgments
• References
• Copyright

Abstract 

To identify the impact of multiple symptoms and their co-occurrence on health-related quality of life (HRQOL) dimensions and performance status (PS), 115 outpatients with cancer, who were not receiving active cancer treatment and were recruited from a university hospital in Sao Paulo, Brazil completed the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire-C30, the Beck Depression Inventory, and the Brief Pain Inventory. Karnofsky Performance Status scores also were completed. Application of TwoStep Cluster analysis resulted in two distinct patient subgroups based on 113 patient experiences with pain, depression, fatigue, insomnia, constipation, lack of appetite, dyspnea, nausea, vomiting, and diarrhea. One group had multiple and severe symptom subgroup and another had less symptoms and with lower severity. Multiple and severe symptoms had worse PS, role functioning, and physical, emotional, cognitive, social, and overall HRQOL. Multiple and severe symptom subgroup was also six times as likely as lower severity to have poor role functioning; five times more likely to have poor emotional; four times more likely to have poor PS, physical, and overall HRQOL; and three times as likely to have poor cognitive and social HRQOL, independent of gender, age, level of education, and economic condition. Classification and Regression Tree analyses were undertaken to identify which co-occurring symptoms would best determine reduction in HRQOL and PS. Pain and fatigue were identified as indicators of reduction on physical HRQOL and PS. Fatigue and insomnia were associated with reduction in cognitive; depression and pain in social; and fatigue and constipation in role functioning. Only depression was associated with reduction in overall HRQOL. These data demonstrate that there is a synergic effect among distinct cancer symptoms that result in reduction in HRQOL dimensions and PS.

Key Words: Cancer symptoms, health-related quality of life, performance status, symptom cluster, co-occurring symptoms

Back to Article Outline

Introduction 

Cancer-related symptoms have been associated with a decline in survival rates,¹ poor overall quality of life (QOL),² high interference with life,³ and poor functional status.⁴ These symptoms may be related to the disease itself, to the cancer treatment or to other origins. In patients with cancer, fatigue, pain, sleep disorders, weakness, distress, and lack of appetite have been identified as the most prevalent symptoms.², ³, ⁵, ⁶ In relation to treatment-related symptoms, patients with varied types and stages of cancer attributed different importance for each symptom when considering their impact on health-related quality of life (HRQOL).⁷

The symptoms generally occur at the same time and certainly there are synergic relationships among them that result in negative impact on QOL.⁸ Thus, the identification of the impact of simultaneous symptoms on HRQOL domains in cancer patients and the understanding of symptom synergism become particularly important as we look forward to the successful completion of cancer treatment, to the best strategies to decrease the impact of cancer and treatment-related symptoms, and to improving QOL. Appropriate symptom control, considering symptom interactions, should be a priority for patients with cancer. For some, managing symptoms while living with cancer, especially during cancer therapy, becomes a particular challenge.

In Oncology, the study of symptoms has focused primarily on single symptoms and, more recently, on overall symptom burden, leading to some symptom cluster identification.³, ⁶, ⁹ However, these studies were developed without considering the impact of symptom interactions on HRQOL. Only one study took into account this aspect, but it included only four symptoms in the cluster analysis: pain, fatigue, sleep, and depression; it did not estimate the impact of this cluster on HRQOL dimensions.¹⁰ It is not clear at the present time which symptoms interact to influence the HRQOL and performance status of patients who are experiencing cancer. Therefore, the purposes of this study were to identify the impact of multiple symptoms and their synergism on HRQOL dimensions and performance status.

Back to Article Outline

Methods 

A cross-sectional, descriptive study was performed after approval by the Human Ethics Committee at the Hospital das Clinicas, University of Sao Paulo, Brazil. One hundred fifteen outpatients were consecutively recruited from September 2004 to August 2006, after the screening of 364 patients with cancer. However, only 113 presented more than two symptoms. After the researcher explained the study purposes and procedures to qualified patients, written informed consent was obtained upon agreement to participate in the study. Patients completed the study instruments by interview. Some other clinical information was collected from the patient's medical records. The selection criteria were older than 18 years of age, no cancer treatment (radiation, chemotherapy, or surgery) at least 30 days before interview to ensure that symptoms were not treatment related, no pain unrelated to cancer (such as rheumatological source, accident), to be able to understand and speak Portuguese, and no presence of cognitive impairment. Patients who did not present these criteria were excluded.

Measurements 

The study instruments included a demographic and clinical questionnaire, the Karnofsky Performance Status scale (KPS),¹¹ the Beck Depression Inventory (BDI), the Brief Pain Inventory (BPI),¹² and the European Organization for Research and Treatment of Cancer Quality of Life-Core 30 (EORTC QLQ-C30).¹³

The demographic and clinical questionnaire included information about gender, age, monthly income, marital status, educational level, religion, practice of religion, and employment status. The clinical information collected was type of cancer, extent of disease, presence of metastatic disease, prior cancer treatment, and comorbidities.

The EORTC QLQ-C30 version 3.0 was used to assess the HRQOL and symptom severity.¹³ The QLQ-C30 consists of 30 items that incorporate nine multi-item scales: five functional scales/dimensions (physical, role, cognitive, emotional, and social); three symptom scales (fatigue, pain, and nausea and vomiting); a global health status and quality-of-life scale, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties). The scores for all scales range from 0 to 100, and they were calculated according to Fayers et al.¹⁴ A higher score represents a higher (better) level of functioning or QOL and a higher (worse) level of symptoms. In our sample, the EORTC QLQ-C30 showed internal consistency in functional scales (Cronbach's α=0.74) and symptom scales (Cronbach's α=0.72).

The depression scores were calculated based on the BDI that was translated and validated in Brazilian Portuguese.¹⁵ The BDI includes 21 items selected to represent the severity of the major symptoms and attitudes of depression. Each item is rated on a four-point scale ranging from 0 to 3. The total score was calculated as the sum of the 21 items and can range from 0 to 63, with higher scores indicating higher level of depression severity. Although there are some different cut points to classify BDI scores, in this study, the cut point of 10 was adopted to define an individual as a depression case.¹⁶ In the present study, Cronbach's alpha for the BDI was 0.83.

The BPI was used to assess worst pain, least pain, average pain and right now pain intensity. Each item was rated on an 11-point scale ranging from 0 (no pain) to 10 (pain as bad as you can imagine).¹²

Patient's performance status was evaluated using the KPS.¹¹

Statistical Analyses 

Data were stored and analyzed using SPSS version 13.0. Descriptive statistics were generated on the sample characteristics.

Cluster analysis was used to identify cancer subgroups with multiple symptoms based on their symptom severity. It was performed with 113 outpatients who experienced more than two symptoms, using the TwoStep Cluster component of SPSS, with log-likelihood distance measure. The optimal number of clusters was given on the basis of Schwarz's Bayesian Criterion and the Akaike Information Criterion. Only patients who experienced more than two symptoms simultaneously were included in this analysis. Cluster analyses were performed based on responses to the EORTC QLQ-C30 fatigue, insomnia, constipation, diarrhea, dyspnea, nausea and vomiting, and lack of appetite subscales, as well as on the BDI and worst pain of the BPI. Next, a discriminant analysis was calculated to check if participants were correctly categorized based on the TwoStep Cluster procedure. For this analysis, because the clusters were not expected to be equal in size, the highest proportional probability for group membership was used to determine the category.

After assessing the adherence to normal distribution with the Kolmogorov–Smirnov test, the subgroups revealed by the TwoStep Cluster analysis were compared in relation to demographic and clinical characteristics, symptom scores, and outcome measures (i.e., performance status and HRQOL domains) using Student's t-test or Mann–Whitney U test. The effect size (ES) was calculated to determine the clinical significance of the differences between cluster subgroups. The ES was calculated based on the subgroups' means differences divided by the pooled standard deviations and was classified based on Cohen's effect size cut points: small=0.20, medium=0.50, large=0.80.¹⁷

Univariate and multivariate logistic regression models using stepwise procedure were used to identify factors associated with poor performance status and overall HRQOL and its dimensions. In the current study, poor HRQOL was defined as a score in QLQ-C30 functioning scales lower than their median scores because we have neither cutoff points to establish poor HRQOL nor population-based reference values. Also, there are some critics to the use of percentiles as a cut point.¹⁸ The independent variables included were cluster subgroup (Subgroup 2=1 and Subgroup 1=0), gender (female=1, male=0), level of education (less than high school=1 and at least high school=0), partnership status (without partner=1 and with=0), presence of metastases (yes=1 and no=0), monthly family income (continuous variable), per capita family monthly income (continuous variable), number of months since cancer diagnosis (continuous variable), age (continuous variable), prior treatment with chemotherapy (no=1 and yes=0), radiation (no=1 and yes=0), surgery (no=1 and yes=0), and hormone therapy (no=1 and yes=0). After univariate regression, all variables with a P-value of Wald's test lower than 0.20 were included in the multivariate models, using the automatic stepwise selection procedure. Overall model fit was assessed by goodness-of-fit statistics with the Hosmer–Lemeshow test.

Classification and Regression Tree analyses (CART) were undertaken to identify which co-occurring symptoms would best determine reduction on HRQOL dimensions and performance status. CART is a procedure that produces a decision tree that identifies the effect of each symptom and symptom association on the outcome variables. The CART was performed considering Chi-square automatic interaction as the growing method. The minimum number of cases permitted in child nodes was 10, and in the parent nodes it was 30. In this analysis, the QLQ-C30 symptom scores were used without being transformed to the 0–100 range.

In all analyses, a P-value less than 0.05 indicates significance. All statistical tests were two-sided.

Back to Article Outline

Results 

Symptom Frequency 

For both the QLQ-C30 and the BPI, a symptom was considered to have occurred if its severity was rated higher than one. Depression was considered present when Beck Depression Inventory (BDI) total score was higher than 10. The mean number of symptoms experienced simultaneously was 9.03 (SD=3.14, range=1–15). Most of the patients (86.1%) experienced at least four symptoms simultaneously. Only four patients (3.5%) experienced just two symptoms, and 8.7% (n=10) had three symptoms. Pain was the most prevalent symptom, followed by fatigue, insomnia, and constipation (Table 1). The least prevalent symptom was diarrhea, followed by nausea and vomiting.

Table 1. Frequency and Severity of Symptoms for the Total Sample (n=115)

Beck Depression Inventory (BDI).

Profiles of Subgroups 

One hundred thirteen outpatients with cancer who were not undergoing active cancer treatment and who provided complete data on all nine symptoms and had more than two symptoms were included in the cluster analysis. We applied the TwoStep Cluster procedure of SPSS and found two clusters as the optimal number of clusters (Schwarz's Bayesian Criterion change=−104.31). The first comprised 57 (50.4%) patients with significantly lower scores for all symptoms. The second cluster included 56 (49.6%) outpatients who showed higher scores than the first subgroup (Table 2). This subgroup showed the most outstanding mean ratings in pain, insomnia, lack of appetite, and constipation. In addition, patients in the high-symptom subgroup had a significantly higher number of symptoms than those in the low-symptom subgroup (P=0.000). Individuals in the low-symptom subgroup had a mean of 4.12 symptoms (percentile 50th=4.0, SD=1.36, range, 2–7). In the high-symptom subgroup, the mean was 6.67 (percentile 50th=7.0, SD=1.10, range, 5–9).

Table 2. Symptom Severity Scores of the Patients Who Had More Than Two Symptoms (n=113) and the Two Subgroups

Symptoms	Total Sample (n=113)		Subgroup 1 (n=57)		Subgroup 2 (n=56)		P-Value	Effect Size
	Mean	SD	Mean	SD	Mean	SD
Pain	7.46	2.34	6.47	2.51	8.21	2.02	<0.001	−0.76
Depression	14.87	8.74	9.16	4.84	19.80	8.01	<0.001	−1.61
Fatigue	42.23	29.03	29.73	25.73	52.78	28.60	<0.001	−0.85
Nausea and vomiting	12.54	24.04	2.63	6.89	20.83	30.19	<0.001	−0.83
Dyspnea	24.19	32.81	15.79	28.94	32.74	34.52	0.002	−0.53
Insomnia	58.41	41.44	35.67	40.27	79.17	30.85	<0.001	−1.21
Lack of appetite	38.64	42.16	8.19	19.19	66.07	37.87	<0.001	−1.93
Constipation	37.17	41.96	12.28	24.91	60.71	41.25	<0.001	−1.42
Diarrhea	5.31	17.57	0.58	4.42	9.52	23.54	0.004	−0.53

Subsequent discriminant analysis revealed that the lack of appetite (Wilk's Lambda=0.51, F=105.59, P=0.000), depression (Wilk's Lambda=0.60, F=73.38, P=0.000), constipation (Wilk's Lambda=0.66, F=57.32, P=0.000), and insomnia (Wilk's Lambda=0.73, F=41.43, P=0.000) were the most important symptoms for the discrimination of the two subgroups. The distinction was practically complete. Applying the canonical discriminating function derived from the discriminant analysis, the cluster analysis yielded 98.2% correct identifications. Two patients of the low-symptom subgroup who were misclassified as high-symptom subgroup members caused the incomplete correct identification.

Demographic and Clinical Characterization of the Subgroups 

Table 3 summarizes the demographic and clinical characteristics of those patients who had more than two symptoms (n=113). No significant differences were found between patient subgroups for any demographic characteristics. In relation to clinical characteristics, a significant difference was only found for presence of metastases. A significantly higher percentage of patients in Subgroup 2 (high-symptom subgroup) had metastases (P=0.046).

Table 3. Demographic and Disease Characteristics of the Total Sample (n=113) and the Two Subgroups

Characteristics	Total Sample		Subgroup 1		Subgroup 2		χ2	P-Value
	n	%	N	%	n	%
Gender, female	69	61.10	38	69.10	30	53.60	2.02	0.150
Education, less than high school	78	69.00	38	67.90	40	71.40	0.17	0.680
Partnership status, with partner	72	63.70	34	60.70	37	66.10	0.35	0.560
Religion, Catholic	69	61.60	33	58.90	36	64.30	1.80	0.773
Type of cancer
Breast	44	38.94	29	50.88	15	26.79	12.84	0.050
Lung	27	23.89	9	15.79	18	32.14	–	–
Prostate	12	10.62	5	8.77	7	12.50	–	–
Genitourinary	15	13.27	8	14.04	7	12.50	–	–
Head and neck	8	7.08	2	3.51	6	10.71	–	–
Gastrointestinal	6	5.31	4	7.02	2	3.57	–	–
Other	1	0.88	0	0.00	1	1.79	–	–
Presence of metastases	73	67.60	33	58.90	40	76.90	3.98	0.046
Previous treatment
Chemotherapy	57	51.40	30	53.60	27	49.10	0.22	0.640
Radiation therapy	61	55.00	30	53.60	31	56.40	0.09	0.770
Hormonal therapy	22	19.80	13	23.20	9	16.40	0.82	0.360
Surgery	74	66.70	42	75.00	32	58.20	3.53	0.060
	Mean	SD	Mean	SD	Mean	SD	Test	P-value
Age (years)	57.24	13.46	58.68	13.64	55.77	13.23	1352.50	0.162
Family monthly income (US$)	563.68	375.40	601.95	389.53	527.62	361.63	1139.50	0.360
Individual monthly income (US$)	213.77	198.14	220.32	204.17	207.56	193.90	1440.50	0.791
Number of months since cancer diagnose	31.40	35.53	33.31	36.14	29.83	35.67	1224.50	0.870

Differences in Outcomes of Symptom Subgroups 

Functional Status. Individuals with high symptoms (Subgroup 2) showed significantly lower scores in KPS (Table 4) than those in the low-symptom subgroup. The high-symptom subgroup had a median of 70% (range, 50%–90%), and 57% showed a KPS score lower than 70%. These individuals could be classified as unable to work, able to live at home and care for most personal needs, with varying amounts of assistance required. Patients with low symptoms showed a median of 80% (range, 50%–100%). They could be defined as able to carry on normal activities and to work, without requiring special care. In this group, only one individual (2.2%) showed KPS=50%, four (8.9%) KPS=60%, and 13 (28.9%) KPS=70%. More than 58% of the group with low symptoms showed scores higher that 80%.

Table 4. Quality of Life and Performance Status Scores of the Total Sample (n=113) and the Two Subgroups

In the univariate analysis, females, individuals with high symptoms (Subgroup 2), those with presence of metastases and who have not received prior hormone therapy showed higher chance for poor performance status (KPS≤70%) (Table 5). However, in a multiple regression analysis, individuals with high symptoms showed 4.13 times more chance for poor performance status than those with low symptoms, and those with presence of metastases showed 7.83 times more chance (Table 5).

Table 5. Independent Predictors of Health-Related Quality of Life Dimensions and Performance Status Among 113 Cancer Patients, Using Logistic Regression Models^a

Outcomes	Univariate			Multiple			Hosmer–Lemeshow test
	OR	95% CI	P-Value	OR	95% CI	P-Value	P-Value	χ2
Overall HRQOL							–	–
Subgroups (Subgroup 2=1)	4.01	1.75, 9.15	0.001	4.01	1.75, 9.15	0.001
Family monthly income (US$)	0.99	0.99, 1.00	0.022	–	–
Physical HRQOL							–	–
Subgroups (Subgroup 2=1)	3.72	1.64, 8.44	0.002	4.43	1.78, 10.99	0.001
Presence of metastases (yes=1)	3.10	1.28, 7.49	0.012	–	–
Emotional HRQOL							–	–
Subgroup	4.82	2.05, 11.31	0.000	4.91	1.93, 12.51	0.001
Monthly per capita family income (US$)	0.99	0.99, 0.99	0.10	–	–
Age (years)	0.96	0.92, 0.98	0.007	–	–
Cognitive HRQOL							–	–
Subgroups (Subgroup 2=1)	2.71	1.18, 6.24	0.019	2.83	1.16, 6.92	0.022
Social HRQOL							0.52	3.29
Gender (female=1)	2.55	1.11, 5.83	0.027	–	–
Presence of metastases (yes=1)	3.16	1.29, 7.75	0.012	4.14	1.43, 11.95	0.009
Hormone therapy (no=1)	3.46	1.20, 9.97	0.022	3.98	1.10, 14.57	0.037
Subgroups (Subgroup 2=1)	2.38	1.09, 5.25	0.029	3.15	1.15, 8.64	0.026
Performance Status							0.99	0.11
Gender (female=1)	2.22	1.02, 4.83	0.045	–	–
Subgroups (Subgroup 2=1)	3.92	1.77, 8.66	0.001	4.13	1.57, 10.85	0.004
Presence of metastases (yes=1)	8.52	3.25, 22.37	0.000	7.83	2.63, 23.25	<0.001
Hormone therapy (no=1)	2.69	1.00, 7.25	0.050	–	–
Role Functioning							0.23	9.29
Subgroups (Subgroup 2=1)	3.41	1.56, 7.43	0.002	5.62	2.16, 14.59	<0.001
Months since cancer diagnosis	0.99	0.98, 1.01	0.170	0.98	0.97, 1.00	0.048

CI=confidence interval.

HRQOL. The highest HRQOL score for the total sample was in the cognitive dimension, followed by the social dimension. The lowest score was in the role functioning dimension, followed by the emotional dimension. The comparison of HRQOL scores between individuals in the low- and high-symptom subgroups demonstrated that the QOL scores in all subscales of the QLQ-C30 were statistically and clinically higher for patients with low symptoms (Subgroup 1) than for those in the high-symptom subgroup (Table 4). The largest difference was observed in the emotional dimension (ES=0.95), followed by the role functioning (ES=0.76) and physical dimensions (ES=0.74). The smallest difference was in the overall HRQOL (ES=0.56), followed by the cognitive dimension (ES=0.57). The small ES observed in the cognitive dimension and in the overall HRQOL suggested that, in this sample, although there were statistically significant differences, these differences were not so relevant clinically.

The logistic regression (Table 5) again confirmed the effect of symptom subgroup on overall HRQOL and on all HRQOL dimensions. In univariate analyses, the symptom subgroup was a predictor of poor HRQOL in all dimensions. In the physical dimension, the univariate regression showed that symptom subgroup and presence of metastases were independent risk factors for poor physical HRQOL, but multiple analyses revealed that only symptom subgroup was a factor associated with higher chance of poor physical HRQOL, independently of all other variables. In the emotional dimension, symptom subgroup, age, and monthly family per capita income were associated with poor emotional HRQOL. But, in a multiple model, only individuals with high symptoms showed more chance for poor emotional HRQOL. The symptom subgroup was the only one independent risk factor for poor cognitive HRQOL in the univariate and multiple regressions, when all variables were included in the model. Individuals with high symptoms showed 4.1 times more chance for poor overall HRQOL than those with low symptoms.

The poor social HRQOL, in the univariate analyses, was associated significantly only with gender, presence of metastases, symptom subgroup, and prior treatment with hormone therapy. In a multiple regression analysis, gender was not associated anymore. The chance associated with presence of metastases and high-symptom subgroup was adjusted by prior treatment with hormone therapy. The odds ratio of presence of metastases decreased from 4.50 (95% confidence interval [CI]=1.59–12.71) to 4.14 (95% CI=1.43–11.95), and of the symptom subgroups from 3.33 (95% CI=1.25–8.85) to 3.15 (95% CI=1.47–8.64).

Impact of Symptoms on HRQOL and Performance Status 

Although it was observed that patients with high severity and number of symptoms had worse HRQOL and higher chance for poor HRQOL and performance status, it was not clear what symptom or symptoms were the most important to reduce HRQOL and performance status. Thus, a CART was used to identify the symptom or symptom synergism that determined the reduction on these outcomes.

The nine symptoms were initially considered for assessing their effect on overall HRQOL, HRQOL dimensions, and performance status. However, only some symptoms appeared in each tree, and affected each one of the outcome variables differently. Fig. 1 depicts these trees.

• 
  • View Large Image
  • Download to PowerPoint
• 
  • View Large Image
  • Download to PowerPoint

• Fig. 1 

  Classification and Regression Tree (CART). (a) CART derived tree of performance status; (b) CART derived tree of physical HRQOL; (c) CART derived tree of role functioning; (d) CART derived tree of cognitive HRQOL; (e) CART derived tree of social HRQOL; and (f) CART derived tree of general HRQOL.

When performance status and physical HRQOL were selected as the outcomes of interest, the most important symptom predictor was fatigue. Patients with moderate or severe fatigue (score ≥3) showed a significantly larger reduction on performance status (F=26.87, P=0.000) and physical HRQOL (F=39.75, P=0.000) scores than those without fatigue (score=1) or with mild fatigue (score=2). For those who had mild fatigue (score=2), pain was the next most important predictor of reduction on performance status and physical HRQOL. Patients who also had severe pain (worst pain score >9) had significantly worse physical HRQOL and performance status than those with pain <9 (Fig. 1a, b). These two symptoms explained 38% of the performance status model's variance and 48.5% of the physical HRQOL model's variance, which indicates that these models were satisfactory.

Moderate or severe fatigue was also the most important factor associated with reduction on role functioning (Fig. 1c). Meanwhile, for those without fatigue or with only mild fatigue, reduction on role functioning was not observed, except for those with mild fatigue who also had constipation (score≥2), in which case the role functioning decreased significantly in comparison with those without constipation (F=7.72, P=0.024). The proportion of explained variance by the role functioning model was 34.2%.

The cognitive functioning/HRQOL was predicted primarily by fatigue. For those with severe fatigue (QLQ-C30 score >3), only fatigue was associated with reduction on cognitive HRQOL. But, for patients who had lower fatigue severity (score ≤ 3), there was no observed effect of fatigue on HRQOL, except for those patients who also had insomnia. In these individuals, the presence of severe insomnia (score >3) resulted in significant reduction on cognitive HRQOL (F=17.75, P=0.000) (Fig. 1d).

The social HRQOL was principally predicted by pain. The presence of severe pain (score >8) was associated with significant reduction on social functioning. However, for those with pain intensity ≤8, there was no observed reduction on this HRQOL dimension, except for those who also had a depression score >17 in the BDI (P=0.000). These patients practically showed the same scores as those with severe pain (Fig. 1e).

Depression was the only one predictor of overall QOL. Cancer patients who had a depression score ≥21 in the BDI had significantly poorer HRQOL than those with a score ≤20 (Fig. 1f). The proportion of variance explained by this model was only 26.2%, which indicated that this model was incomplete.

Back to Article Outline

Discussion 

Our results showed that patients with multiple and severe symptoms had worse performance status, and physical, emotional, cognitive, social, and overall HRQOL, and role functioning than those who had a small number of symptoms and with lower severity. In addition, we also observed that cancer patients with high-symptom number and severity, independent of their gender, age, level of education, and economic condition, had four times more risk for poor performance status, physical HRQOL, and overall HRQOL; three times more risk for poor cognitive HRQOL; three times more risk for poor social HRQOL; and almost five times more chance for poor emotional HRQOL.

Similar results were observed in other studies that showed that the number and severity of symptoms was inversely correlated with the SF-36 subscales physical functioning and role limitations;¹⁹ that the increase in the severity was correlated with poor performance status;²⁰ and that specific cluster symptoms were associated with interference with life.³

This study is the first to use CART to identify which symptom severity and symptom synergism negatively affects HRQOL dimensions and performance status. One of the most important findings of this study was obtained when we analyzed in detail what symptom or symptom combination had a higher impact on each HRQOL domain and performance status. We observed that different symptom co-occurrences were associated with reduction on each outcome variable; that these co-occurrences were severity dependent; and that the effect of some symptoms on specific HRQOL dimensions and performance status was mediated by other symptoms. In this scenario, fatigue was the most important predictor and symptom effect mediator for performance status, and physical, cognitive, and role functioning HRQOL dimensions. Fatigue mediated the effect of pain on physical HRQOL and performance status, the effect of insomnia on cognitive HRQOL, and the effect of constipation on role functioning. In relation to social HRQOL, pain was the most important predictor, and it mediated the effect of depression on this HRQOL dimension. Thus, according to our results, low performance status or physical functioning should suggest evaluation of the presence and severity of fatigue primarily, and also the presence of pain. Treatment of these symptoms may increase performance status and physical functioning. Although beyond the scope of the present study, future research could analyze the effect of pain treatment on reducing fatigue and vice versa, and also on improving HRQOL and performance status, to improve our knowledge about this synergism among symptoms.

In the same way, if a patient reports poor cognitive functioning, fatigue, and insomnia should be assessed. If a patient presents moderate fatigue and also severe insomnia, treatment for insomnia should be considered, because the co-occurrence of these symptoms and with these severities was associated with reduction in cognitive HRQOL.

A probable explanation for the relation between high fatigue, high insomnia, and low cognitive functioning is that they share the same mechanism. High fatigue and poor cognitive functioning were correlated with higher levels of interleukin-6 in leukemia patients.²¹ Also, sleep deprivation has been associated with significant increases in plasma levels of proinflammatory cytokines, such as interleukin-6 and tumor necrosis factor-alpha in healthy individuals.²² Perhaps these findings may explain why fatigue and insomnia were associated with reduction on cognitive functioning. Thus, if insomnia and fatigue have a deleterious effect on immune function, a successful treatment of these symptoms should have the opposite effect on immunity and also improve cognitive functioning. Future studies addressing the underling mechanism of symptom co-occurrence should be conducted so that it can be better understood.

These findings support the Unpleasant Symptom Theory proposed by Lenz et al.,⁸ which suggested that there was a cumulative effect among symptoms in diminishing functioning and QOL. Our results support this theory when we observed that patients with high symptoms and with higher severity had clinically and significantly worse HRQOL in all dimensions, and worse performance status, and when we observed that patients who had, for example, mild fatigue alone had better performance status and physical HRQOL than those who had fatigue together with severe pain. We observed that some symptoms occurred together, which suggests that probably one mediated the other's effect. Thus, we think that a better definition for symptom cluster is that a symptom cluster is a group of concurrent symptoms that may have a synergistic effect as a predictor of patient outcome.⁴ We understand that synergism refers to the interaction of two or more symptoms so that acting together they create an effect greater than the total power achieved by each symptom working separately.

It was also observed that fatigue was the most important predictor of performance status, and of the physical, role, and cognitive HRQOL dimensions. Similar results were observed in a study including patients with breast cancer after cancer treatment,² and in a study with patients with pancreatic cancer that observed that fatigue and pain were significantly associated with impairment of overall HRQOL.²³ However, in none of these studies was the interactive effect of symptoms analyzed, as was done in the present study. We observed that fatigue's effect over these outcome variables was mediated by other symptoms, such as pain and insomnia. If it was considered that a symptom cluster is concurrent symptoms that may have a synergistic effect as a predictor of patient outcome,⁴ we can consider pain and fatigue as a symptom cluster, and fatigue and insomnia another symptom cluster. Perhaps, if we treat one of these symptoms, it will result in improvement in the other symptom, and consequently improvement in HRQOL.

Unfortunately, these two symptoms did not explain most of the variance in the performance status, and physical, role, and cognitive functioning scores; it is necessary to assess the effect of other variables on these outcomes. Perhaps these other variables will be other symptoms because in the CART analyses, each variable is included and split until the data in each node is sufficiently well discriminated or until there are too few data in any node to support further analysis.²⁴ Unfortunately, as our study had a small sample size, it could not be split in many parent and child nodes. Thus, if the CART were repeated in a large sample, more symptoms could be included in these models, because each parent and child node could include more cases.

Despite its limitations, it is important to highlight the importance of this study in bringing a new method to trying to identify clusters of symptoms, their interactions and effect on HRQOL and performance status. Our results also suggested that there was an association between pain and depression to determine social HRQOL. Perhaps, there is a symptom cluster that includes pain and depression. Other studies that assessed symptom clusters in cancer patients observed that pain and depression were in the same cluster, together with sleep disturbance, drowsiness, lack of appetite, difficulty remembering things, constipation, anxiety, etc.⁹, ²⁵, ²⁶, ²⁷ In our study, probably due to the small sample size and the fact that CART clusters patients and not variables, it was not possible to observe these other symptoms together with pain and depression. Perhaps, in the study that we are developing with a large sample size, we will observe that patients with specific pain and depression scores will also present these other symptoms as predictors of HRQOL.

Depression was the only one predictor of overall HRQOL. Studies including patients with breast cancer also observed that depression was a significant predictor of overall QOL.²⁸, ²⁹ In a study with patients with lung cancer, depression explained 26% of the variance in QOL.³⁰ The cut point on the depression score identified in our study as associated with worse overall HRQOL is the cut point used to define a patient as having depression.³¹ Thus, we can say that the presence of clinical depression was the only predictor associated with reduction on overall HRQOL.

Our results showed that different symptoms and symptoms synergism have different importance and impact on performance status and different HRQOL dimensions. Thus, we suggest that the design of future trials considers which symptoms are the most important for the outcome of primary concern (i.e., general HRQOL, physical HRQOL, performance status) and that stratification and statistical analyses take these specific symptoms and symptom synergisms into account.

As noted, our study showed some limitations. First, the sample size was relatively small if we consider the large number of cancer patients in Brazil and in the world. Thus, we cannot guarantee its external validity. In addition, the sample was heterogeneous in relation to type of cancer, prior cancer treatment and disease status. The results should be confirmed in a larger and more homogenous sample and multicenter study. Probably, in a large sample, other symptoms will be identified as independent risk factors for reducing HRQOL and performance status.

Despite its limitation, the present study enlightened us that the CART may be a useful tool to identify symptom interactions and to define the profile of patients with higher chance for reducing HRQOL, based on their symptoms. As it looks, the use of other cluster techniques, such as hierarchical or TwoSteps, are useful to split patients, but they do not show clearly the relationship among symptoms and their cut points. On the other hand, the patient profiles established with CART take into account the effect of these symptoms on specific outcomes, and this is not required in the other two techniques. Thus, clinicians and researchers have to consider their goals and to choose the best method for each situation.

Another limitation of this study was the symptom measurement. Symptoms have at least four dimensions: intensity, timing, level of distress, and quality.⁸ In the present study, symptoms and patients were divided into subgroups considering only the intensity. Thus, other studies considering the other symptom dimensions should be addressed to confirm the cluster results found in this study. In all CART models, the total variance explained by symptoms was small. This is probably because the HRQOL and performance status are not only determined by symptoms. But our results showed that symptoms have an important impact on HRQOL domains and performance status, principally when synergically associated.

Back to Article Outline

Acknowledgments 

The authors gratefully thank the physicians from the Hospital das Clinicas, School of Medicine, University of Sao Paulo, who helped them enroll patients in the study.

Back to Article Outline

References 

1. Teunissen SC, de Graeff A, de Haes HC, Voest EE. Prognostic significance of symptoms of hospitalized advanced cancer patients. Eur J Cancer. 2006;42(15):2510–2516
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (146 KB)
   • CrossRef
2. Arndt V, Stegmaier C, Ziegler H, Brenner H. A population-based study of the impact of specific symptoms on quality of life in women with breast cancer 1 year after diagnosis. Cancer. 2006;107(10):2496–2503
   • View In Article
   • MEDLINE
   • CrossRef
3. Wang XS, Fairclough DL, Liao Z, et al. Longitudinal study of the relationship between chemoradiation therapy for non-small-cell lung cancer and patient symptoms. J Clin Oncol. 2006;24(27):4485–4491
   • View In Article
   • CrossRef
4. 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
5. Portenoy RK, Thaler HT, Kornblith AB, et al. Symptom prevalence, characteristics and distress in a cancer population. Qual Life Res. 1994;3(3):183–189
   • View In Article
   • MEDLINE
   • CrossRef
6. Cleeland CS, Mendoza TR, Wang XS, et al. Assessing symptom distress in cancer patients: the M.D. Anderson Symptom Inventory. Cancer. 2000;89(7):1634–1646
   • View In Article
   • MEDLINE
   • CrossRef
7. Osoba D, Hsu MA, Copley-Merriman C, et al. Stated preferences of patients with cancer for health-related quality-of-life (HRQOL) domains during treatment. Qual Life Res. 2006;15(2):273–283
   • View In Article
   • MEDLINE
   • CrossRef
8. Lenz ER, Pugh LC, Milligan RA, Gift A, Suppe F. The middle-range theory of unpleasant symptoms: an update. ANS Adv Nurs Sci. 1997;19(3):14–27
   • View In Article
   • MEDLINE
9. Walsh D, Rybicki L. Symptom clustering in advanced cancer. Support Care Cancer. 2006;14(8):831–836
   • View In Article
   • MEDLINE
   • CrossRef
10. Miaskowski C, Cooper BA, Paul SM, et al. Subgroups of patients with cancer with different symptom experiences and quality-of-life outcomes: a cluster analysis. Oncol Nurs Forum. 2006;33(5):E79–E89
    • View In Article
11. Karnofsky D, Abelmann W, Craver L, Burchenal J. The use of nitrogen mustard in the palliative treatment of cancer. Cancer. 1948;1:634–656
    • View In Article
    • CrossRef
12. Cleeland CS. Measurement of pain by subjective report. In:  Chapman CR,  Loeser JD editor. Issues in pain measurement. New York, NY: Raven Press; 1989;p. 391–403
    • View In Article
13. Aaronson NK, Ahmedzai S, Bergman B, et al. The European Organization for Research and Treatment of Cancer QLQ-C30: a quality-of-life instrument for use in international clinical trials in oncology. J Natl Cancer Inst. 1993;85(5):365–376
    • View In Article
    • MEDLINE
    • CrossRef
14. Fayers PM, Aaronson NK, Bjordal K, et al. on behalf of the EORTC Quality of Life Group The EORTC QLQ-C30 scoring manual. 3rd ed.. Brussels: European Organization for Research and Treatment of Cancer; 2001;
    • View In Article
15. Gorenstein C, Andrade L. Validation of a Portuguese version of the Beck Depression Inventory and the State-Trait Anxiety Inventory in Brazilian subjects. Braz J Med Biol Res. 1996;29(4):453–457
    • View In Article
    • MEDLINE
16. Williams JW, Noel PH, Cordes JA, Ramirez G, Pignone M. Is this patient clinically depressed?. JAMA. 2002;287(9):1160–1170
    • View In Article
    • MEDLINE
    • CrossRef
17. Cohen J. Statistical power analyses for the behavioral sciences. second ed.. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.; 1988;
    • View In Article
18. Fayers PM. Interpreting quality of life data: population-based reference data for the EORTC QLQ-C30. Eur J Cancer. 2001;37(11):1331–1334
    • View In Article
    • Full Text
    • Full-Text PDF (143 KB)
    • CrossRef
19. Gift AG, Jablonski A, Stommel M, Given CW. Symptom clusters in elderly patients with lung cancer. Oncol Nurs Forum. 2004;31(2):202–212
    • View In Article
20. Bradley N, Davis L, Chow E. Symptom distress in patients attending an outpatient palliative radiotherapy clinic. J Pain Symptom Manage. 2005;30(2):123–131
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (97 KB)
    • CrossRef
21. Meyers CA, Albitar M, Estey E. Cognitive impairment, fatigue, and cytokine levels in patients with acute myelogenous leukemia or myelodysplastic syndrome. Cancer. 2005;104(4):788–793
    • View In Article
    • MEDLINE
    • CrossRef
22. Shearer WT, Reuben JM, Mullington JM, et al. Soluble TNF-alpha receptor 1 and IL-6 plasma levels in humans subjected to the sleep deprivation model of spaceflight. J Allergy Clin Immunol. 2001;107(1):165–170
    • View In Article
    • Abstract
    • Full-Text PDF (96 KB)
    • CrossRef
23. Muller-Nordhorn J, Roll S, Bohmig M, et al. Health-related quality of life in patients with pancreatic cancer. Digestion. 2006;74(2):118–125
    • View In Article
    • MEDLINE
    • CrossRef
24. Ciampi A, Lawless JF, McKinney SM, Singhal K. Regression and recursive partition strategies in the analysis of medical survival data. J Clin Epidemiol. 1988;41(8):737–748
    • View In Article
    • MEDLINE
    • CrossRef
25. Okuyama T, Wang XS, Akechi T, et al. Japanese version of the MD Anderson Symptom Inventory: a validation study. J Pain Symptom Manage. 2003;26(6):1093–1104
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (390 KB)
    • CrossRef
26. Chen ML, Tseng HC. Symptom clusters in cancer patients. Support Care Cancer. 2006;14(8):825–830
    • View In Article
    • MEDLINE
    • CrossRef
27. Yun YH, Mendoza TR, Kang IO, et al. Validation study of the Korean version of the M.D. Anderson Symptom Inventory. J Pain Symptom Manage. 2006;31(4):345–352
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (629 KB)
    • CrossRef
28. Andritsch E, Dietmaier G, Hofmann G, Zloklikovits S, Samonigg H. Global quality of life and its potential predictors in breast cancer patients: an exploratory study. Support Care Cancer. 2007;15(1):21–30
    • View In Article
    • MEDLINE
    • CrossRef
29. Wong WS, Fielding R. Change in quality of life in Chinese women with breast cancer: changes in psychological distress as a predictor. Support Care Cancer. 2007;15(11):1223–1230
    • View In Article
    • CrossRef
30. Fox SW, Lyon DE. Symptom clusters and quality of life in survivors of lung cancer. Oncol Nurs Forum. 2006;33(5):931–936
    • View In Article
31. Geisser ME, Roth RS. Robinson ME Assessing depression among persons with chronic pain using the Center for Epidemiological Studies-Depression Scale and the Beck Depression Inventory: a comparative analysis. Clin J Pain. 1997;13(2):163–170
    • View In Article
    • MEDLINE
    • CrossRef