SEARCH

SEARCH BY CITATION

Keywords:

  • gero-oncology;
  • elderly cancer survivorship;
  • path analysis;
  • physical function

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

BACKGROUND:

The majority of cancer survivors are aged ≥65 years, yet, historically, cancer research has focused infrequently on older patients. The objective of this study was to examine predictors of physical function within a framework that integrates the gerontologic and oncologic needs of older cancer survivors.

METHODS:

Path analysis tested 759 women who were breast cancer survivors aged ≥70 years from the American Cancer Society Study of Cancer Survivors II to examine the cancer, aging, and personal characteristics that had an impact on symptoms and physical functioning.

RESULTS:

High levels of symptom bother (β = −.42) and comorbidities (β = −.21) were strongly associated with lower physical function. Comorbidity and social support (β = .21) indirectly influenced symptom bother through emotional status (β = −.35). The model demonstrated good fit with the data (chi-square statistic, 50.6; adjusted chi-square statistic, 2.8; P < .001; goodness-of-fit index, .98; root mean square error of approximation, .049 [confidence interval, .03-.05]).

CONCLUSIONS:

The current findings supported prior research indicating that the majority of older survivors of breast cancer are doing well, but there is a subset of survivors that requires ongoing attention to symptoms, comorbidities, emotional health, and social support to thrive after cancer treatment. Cancer 2012. © 2012 American Cancer Society.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

Approximately 11.9 million people in the United States are living with a history of cancer, and over 7 million survivors are in the group aged ≥65 years.1 The aging of the US population is expected to contribute to the predicted exponential increase in older cancer survivors in the near future. Surprisingly little research has examined issues of cancer survivorship in the elderly,2-5 but more comorbidities and poorer quality of life, functional status, and general health have been reported.6-10

A previously reported conceptual model11 proposed a holistic approach to gero-oncology survivorship research. It hypothesized direct and indirect relations of cancer-specific variables (eg, disease stage, treatment, length of survivorship, and symptoms), personal characteristics, lifestyle behaviors, and aging-related concerns (eg, psychosocial support and comorbidity) on health outcomes for older cancer survivors. The conceptual model was revised to reflect the chosen outcome measure and selected variables. The purpose of this study was to evaluate the revised model (Fig. 1) by examining the influence of factors associated with aging, cancer, symptom bother, and individual characteristics on the important outcome of physical functioning in older survivors of breast cancer.

thumbnail image

Figure 1. This is the revised model of elderly cancer survivorship based on the conceptual model reported by Bellury et al (see Bellury LM, Ellington L, Beck SL, Pett M, Stein K, Clark, J. Elderly cancer survivorship: an integrative review and conceptual framework. Eur J Oncol Nurs. 2011;15:233-242).11

Download figure to PowerPoint

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

Parent Study

The American Cancer Society Studies of Cancer Survivors (ACS SCS) were developed to address some identified gaps in prior survivorship research. ACS SCS-II, as described in detail previously,12 is a cross-sectional descriptive study drawn from 18 population-based Surveillance, Epidemiology, and End Results cancer registry databases in 14 states. On the basis of registry records, participants were stratified by cancer type and time since diagnosis (2 years, 5 years, and 10 years). The primary aim of SCS-II was to compare quality of life and psychosocial functioning across the phases of cancer survivorship. There was an intentional over-sampling of ethnic and racial groups. Established instruments with a history of reliability and validity in cancer patients were used, and additional instruments of variables without established measures were constructed and pilot tested. Completion of the battery of instruments required approximately 60 minutes. All instruments were available in Spanish.

Sample Characteristics

Balducci13 defined age 70 years as the “lower limit of senescence,” and Stava et al3 reported similarities in health profiles between survivors of breast cancer and survivors of other kinds of cancers. On the basis of these 2 reports, breast cancer survivors aged ≥70 years were chosen for inclusion in the current analyses. ACS SCS-II included 896 possible participants. Sample size calculation14 estimated that the minimum sample size needed for this analysis was 550 participants. After removing those with missing data, 759 participants were included. This secondary analysis of deidentified data was granted exempt status from both the Emory University and University of Utah institutional review boards.

Measures

Demographic variables included self-reported education, marital status, age, and race. Clinical variables, including length of survivorship and disease stage at diagnosis, were derived from cancer registry records. The number of self-reported treatment modalities received was used to represent the treatment variable. Participant selections from a list of 13 common medical conditions experienced in the past year were summed to represent the variable comorbidity.

The Multidimensional Scale of Perceived Social Support, a 12-item, self-report scale that has been validated among a variety of populations and used previously with cancer patients,15 was used as a measure of social support. The Medical Outcomes Study 36-item short form (SF-36)16 Mental Health subscale, which includes questions about mood and nervousness, was used as a measure of emotional status. The Modified Rotterdam Symptom Checklist was included as a measure of symptom bother and was an intermediate outcome measure in this study.17 The checklist contained a list of 30 symptoms and used a 4-point Likert evaluation of the extent of bother attributed to each symptom in the past 7 days. The 10-item Physical Function subscale of the SF-36 was chosen as the outcome measure. The Physical Function subscale asked whether current health limited normal, moderate, or vigorous activities. Both SF-36 subscales were reported as normed scores.

Analysis

Descriptive analyses were conducted based on the level of measurement (Tables 1-3). A full path model was then generated based on the revised conceptual model (Fig. 2). Path analysis was conducted using AMOS (version 18) with maximum likelihood estimation.18 Insignificant paths were removed from the full model through an iterative process based on parameter estimates, modification indices and theoretical considerations. Model fit was then analyzed using chi-square goodness-of-fit, root mean square error of approximation (RMSEA), incremental, predictive, and absolute fit indices.

Table 1. Demographics
CharacteristicNo. of Patients (%)
  1. Abbreviations: SD, standard deviation.

Age, y759 (100)
 Mean±SD [range], y77.6 ± 5.3 [70.0-97.8]
Marital status 
 Married345 (45.5)
 Widowed332 (43.7)
 Divorced/separated55 (7.3)
 Single, never married27 (3.6)
Race 
 White614 (80.9)
 Black68 (9)
 Hispanic27 (3.6)
 Other50 (6.6)
Education 
 ≤Eighth grade37 (4.9)
 Some high school112 (14.8)
 High school diploma250 (32.9)
 Vocational/some college190 (25)
 College graduate77 (10.1)
 Professional or graduate school93 (12.3)
Income540 (71.1)
 <$20,000221 (29.1)
 $20,000-74,999297 (39.1)
 ≥$75,00022 (2.9)
 Prefer not to answer129 (17)
 Total missing income219 (28.9)
Table 2. Clinical Sample Descriptions
DescriptorNo. of Survivors (%)
  1. Abbreviations: COPD, chronic obstructive pulmonary disease; SD, standard deviation.

Disease stage 
 Localized572 (75.4)
 Regional183 (24.1)
 Distant4 (0.5)
No. of treatments: Mean ± SD [range]2.7 ± 1.5 [0-7]
 Surgery690 (90.9)
 Hormone therapy446 (58.8)
 Radiation therapy424 (55.9)
 Chemotherapy286 (37.7)
 Bone marrow transplantation103 (13.6)
 Immunotherapy98 (12.9)
 Other13 (1.7)
Length of survivorship, y 
 2223 (29.4)
 5255 (33.6)
 10281 (37)
Comorbidity 
 No. of comorbidities: Mean ± SD [range]2.3 ± 1.8 [0-13]
 Arthritis422 (55.6)
 Hypertension368 (48.5)
 Osteoporosis165 (21.7)
 Chronic back pain156 (20.6)
 Heart problems123 (16.2)
 Diabetes122 (16.1)
 Neuropathy91 (12)
 Asthma, emphysema, COPD72 (9.5)
 Stomach or intestinal problems63 (8.3)
 Depression53 (7)
 Memory or concentration43 (5.7)
 Anxiety or nervousness39 (5.1)
 Stroke23 (3)
 None of the above89 (11.7)
 Other90 (11.9)
Table 3. Means, Standard Deviations, Ranges, Reliability, and Population Norms for Scale Measures
    Population Norm: Mean ± SD
ScaleMean ± SDRangeCronbach αAges 65-74 YearsAged ≥75 Years
  1. Abbreviations: MSPSS, Multidimensional Scale of Perceived Social Support; RSCL-M, modified Rotterdam Symptom Checklist; SD, standard deviation; SF-36 MH, Medical Outcomes Study 36-item short form Mental Health subscale; SF-36 PF, Medical Outcomes Study 36-item short form Physical Function subscale.

SF-36 PF40.1 ± 11.415.2-57.1.9143.6 ± 11.337.2 ± 12.2
SF-36 MH52.5 ± 8.816.4-64.1.7851.7 ± 9.250.4 ± 10.6
RSCL-M44.8 ± 9.730.0-112.5.87  
MSPSS68.2 ± 12.712.0-84.0.96  
thumbnail image

Figure 2. This is the full hypothesized path model (e1-e6 indicates error terms associated with each endogenous variable).

Download figure to PowerPoint

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

Descriptive Analysis

The majority of 759 participants were Caucasian (80.9%) who had at least a high school education (80.3%) and >$20,000 annual income (42%). The mean age (±standard deviation) was 77.6 ± 5.3 years. Most participants were either married (44.7%) or widowed (43.7%). Participants were divided fairly equally into the 3 length of survivorship groups. Most participants reported localized disease (75.4%) and surgery to remove the cancer (90.9%). Arthritis (55.6%) and hypertension (48.5%) were the most frequently reported comorbidities, and the average number of comorbid conditions reported was 2.3. The most frequently reported and most bothersome symptoms were tiredness (82.6%) and lack of energy (71.4%). The mean Mental Health score for our sample was 52.5, which was higher than the US population mean score reported for women ages 65 to 74 years (mean score, 51.7) or women aged ≥75 years (mean score, 50.4).19 The mean normed Physical Function score for our sample was 40.7, which was lower than the US population norm for women ages 65 to 74 years (mean score, 43.6) but higher than the US population norm for women aged ≥75 years (mean score, 37.2).19

Because of the large number of comparisons and large sample size, only correlations >0.3 were regarded as clinically significant (Table 4). Symptom bother had the strongest correlations with both physical function (correlation coefficient [r] = −.49) and comorbidity (r = .45). An analysis of variance indicated that widowed participants were significantly older than any other marital status group (F5753 = 18.2; P < .001). Analyses of variance or chi-square tests were run on each variable to determine statistically significant differences based on length of survivorship. Because no differences were identified, length of survivorship was not included in the path model.

Table 4. Correlations Between Variables
VariableAgeComorbidity (total no)SF-36 Physical FunctionRSC-M: SymptomsSF-36 Mental Health (emotional status)MSPSS: Social SupportTreatmentMarital StatusEducationRace
  • Abbreviations: MSPSS, Multidimensional Scale of Perceived Social Support; RSC-M, modified Rotterdam Symptom Checklist; SF-36, Medical Outcomes Study 36-item short form.

  • a

    Significant correlations (r > .3).

  • b

    Dichotomized as married or not

  • c

    Dichotomized as white or not.

Age1.0         
Comorbidity (total no).081.0        
SF-36 Physical Function−.18−.39a1.0       
RSC-M: Symptoms.09.45a−.49a1.0      
SF-36 Mental Health (emotional status).05−.24.17−.44a1.0     
MSPSS: Social Support−.05−.11.13−.13.231.0    
Treatment−.01.08−.01.12−.04.001.0   
Marital statusb−.28−.05.17−.03.02.22.001.0  
Education−.03−.08.10−.12.16−.00−.00.101.0 
Racec.03−.06.04.01.02.06.04.15.151.0

Path Analysis

The hypothesized model that was generated from the conceptual model was reduced systematically by eliminating insignificant paths (P > .05). Figure 3 presents this final model. The only variable that was eliminated through this process was race. A summary of direct, indirect, and total effects identified in the path model are reported in Table 5.

thumbnail image

Figure 3. This is the model reduced to significant paths with standardized regression weights (e1-e6 indicate error terms associated with each endogenous variable). r2 Indicates variance explained; GFI, goodness-of-fit index; CFI, comparative fit index; RMSEA, root mean square error of approximation; CI, confidence interval.

Download figure to PowerPoint

Table 5. Standardized Effects of Study Variables on Physical Function
 Standardized Effects (β Weights)
VariableIndirectDirectTotal
Married.02.02
Age−.06−.06
Education.04.04
Social support.01.07.09
Comorbidity−.16−.21−.38
Treatment−.03−.03
Emotional status.15−.08.07
Symptom bother−.42−.42

The final path model indicated that higher symptom bother was associated significantly and negatively with physical function, indicating that higher levels of symptom bother were associated with worse physical function. High comorbidity was associated with high symptom bother and low physical function. Better emotional status was related to lower symptom bother. Emotional status was directly influenced by education, age, social support, and comorbidity, such that more education, older age, fewer comorbidities, and higher reported social support were related to better emotional status. The number of treatments had no significant, direct effect on physical function. Age exerted weak direct effects on treatment, in that older survivors reported fewer treatments. Being married was directly associated with higher social support, which, in turn, was associated significantly with higher emotional status. The r2 values ranged from .01 for treatment to .32 for symptom bother. Although 32% of the variance in symptom bother was explained by the predictors, 27% of the variance in physical function was explained by all of the other endogenous variables in the model.

Model Fit

The chi-square goodness-of-fit statistic was significant (P < .001), which reflected the large sample size. However, the adjusted chi-square and other fit indices indicated a good fit of the data (chi-square statistic, 50.6; adjusted chi-square/degrees of freedom, 2.8; P < .001; goodness-of-fit index, .98; RMSEA, .049 [CI, .03-.05]) (Table 6).20

Table 6. Fit Indices
Fit IndexValue (CI)Acceptable Rangesa
  • Abbreviations: CI, confidence interval.

  • a

    See Byrne BM. Structural Equation Modeling with AMOS: Basic Concepts, Applications, and Programming. 2nd ed. New York: Routledge Academic; 2010.20

Adjusted chi-square statistic2.80<3
Root mean square error of approximation.049 (.033-.065)<.05
Normed fit index.94>.90
Incremental fit index.96>.90
Comparative fit index.96>.90
Goodness-of-fit index.98>.90
Expected cross-validation index
 Default model.138 (.11-.17)Lowest value
 Independence model1.122 (1.00-1.25) 
 Saturated model.119 (.119-.119) 
Akaike information criteria  
 Default model256.70Lowest value
 Independence model901.38 
 Saturated model343.44 

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

In this study, we tested a holistic gero-oncology model of elderly survivorship. The path analysis using the ACS SCS-II data indicated acceptable fit statistics with the conceptual model. Length of survivorship was not related to physical function or other study variables. Our analysis, however, did indicate that symptom bother and comorbidity were similar across length-of-survivorship cohorts. To our knowledge, this stability across length-of-survivorship time frames represents a new finding. Other studies have reported that function appeared to decline and then gradually improve during the first 2 years post-treatment,21-23 a time frame that was not included in the current study. In survivors >2 years post-treatment, an impact from length of survivorship on function has not been demonstrated,7, 24-26 a conclusion consistent with the current findings.

It is noteworthy that the relations between age and the predictor variables were weaker than the relations between comorbidity, physical function, emotional status, symptoms, and social support. The impact of heterogeneity of health status among the elderly13, 27-31 because of genetic, environmental, and lifestyle factors was supported by the lack of strong direct effects of age on the endogenous variables. Alternately, the minimal effect of age may be because of the restricted age range of the sample.

Patient complexity is addressed minimally by current research paradigms.32-35 Simultaneously testing the wide range of variables present in the model was a methodological approach to address complexity. The retention of all variables except race in the path model demonstrated that a comprehensive approach to elderly cancer survivorship is reasonable and that these specific variables are important to an understanding of cancer survivorship in older groups.

The incidence of comorbidity was similar to population reports and reports from other breast cancer survivor studies.3, 5, 36 Increased comorbidity was associated with increased symptoms and decreased physical function. It has been demonstrated that comorbidity is predictive of change in both physical and mental health7 and is associated with increased symptoms, decreased survival, decreased function, and decreased treatment tolerance.6, 24, 37-39 Findings from this study supported prior reports of similar or better psychosocial functioning among older cancer survivors compared with noncancer peers.2, 40-42

An unexpected finding was the weak, inverse relation between emotional status and physical function, such that better emotional status was associated with slightly worse physical function (P < .05). Emotional status had direct and indirect effects on physical function in opposite directions. The total effect was weak but positive in the expected direction. This confounded finding points to limitations in linear modeling. For example, high symptom bother has been related to poor emotional functioning, depression, and morbidity,43, 44 which may mean that emotional status is better conceptualized as an outcome rather than as a predictor variable. It is probable that emotional status, physical function, symptom bother, and comorbidity are bidirectional, complexly interrelated, and not adequately explicated by linear models.

In this study, higher symptom bother was associated with more comorbidity and worse emotional status. To a lesser degree, symptom bother was related to more treatment types and older age. Although the analysis of specific symptom patterns was beyond the scope of this project, the predominance of fatigue and pain-related symptoms may reflect priority symptoms or a symptom cluster that could be investigated in this population. Fatigue, pain, insomnia, and depression are most commonly suggested as clustered symptoms among older adults during treatment or within the first year post-treatment.45-48 Our findings suggest that similar symptoms persist throughout survivorship.

Limitations

The conceptual model pointed toward the inclusion of more specific cancer-related variables, yet Stein et al49 concluded that measures of psychological long-term and late effects are particularly underdeveloped. Cancer-related variables (length of survivorship and treatment) did not contribute significantly to the outcome in this model of elderly survivorship. The self-reported treatment variable, which did not include any medical record data, was a serious limitation. Findings related to treatment should be interpreted cautiously, and future research could attempt to include clinical sources of treatment information.

Assumptions relative to temporal order, theoretical causation, nonrecursive paths, cross-sectional data, and causal direction are inherent in linear models. Thus, as complexity increases, the overall benefit of linear models may be questioned. Causal direction in path analysis is an a priori decision. It is possible that, although the associations between variables were substantiated, the direction of the association may be reversed. For example, the relation between social support and emotional status was confirmed, so it is possible that emotional status could contribute to the variation in social support, or vice versa.

Racial disparities have been consistently documented relative to cancer incidence, mortality, physical function, comorbidity, and treatment; but information specific to older survivors is limited.37, 42, 48, 50, 51 Although the sampling criteria for the parent study ensured a population-based sample, and 19% of our sample were minority participants, race was the only variable to be eliminated from the path model because of lack of significance. The race dichotomy (Caucasian and non-Caucasian) was created because of the small number of participants in the minority categories and may have confounded the differences identified within the racial and ethnic groups. Therefore, further examination of the role of race and ethnicity in older survivorship is warranted.

The conceptual model proposed by Bellury et al11 suggested the inclusion of lifestyle behaviors (body mass index, physical activity, and smoking) that have been associated with health.52-54 These variables were not included, because of individuals who had insufficient data on body mass and physical activity. It is likely that those characteristics would contribute to the holistic understanding of the gero-oncology survivorship paradigm. In addition, large numbers of individuals who had missing data precluded the use of income as a socioeconomic indicator in this analysis.

A limitation in this study was the low-to-moderate variance in physical function that was explained by the direct and indirect effects in the model. The addition of modifiable personal characteristics, clinically defined treatment and stage data, as well as refined measurement of weighted comorbidity may add to the variance explained. Reconceptualizing emotional status as a mediating variable or an outcome variable rather than as a predictor may explain the unexpected findings and increase the variance explained by the model. Thus, as descriptive data relative to older survivors accumulates, the model will need to be re-evaluated to conform to newly generated knowledge in this field.

Finally, the decision to restrict this initial model testing to a sex-specific, diagnosis-specific population of women breast cancer survivors limits generalizability of these results. The literature supports sex differences in somatic symptom reporting, pain experience, insomnia, and depression55-59; therefore, this women-only population may have a greater incidence of symptom bother and stronger associations between symptoms and emotional status compared with a mixed sex or men-only population. Follow-up studies should include testing the model on other disease-specific subgroups, mixed diagnostic groups, and comparisons with other age groups to validate the gero-oncology survivorship paradigm.

In conclusion, the current findings indicated that physical function in older breast cancer survivors was significantly impacted by social support, emotional status, comorbidities, and symptom bother. These findings support a gero-oncology survivorship paradigm, the need for a conceptually integrated research agenda, and a comprehensive approach to health policy and clinical care for older cancer survivors.

FUNDING SOURCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES

L. Belllury received an American Cancer Society Doctoral Scholarship in Cancer Nursing (DSCNR-07-220-03).

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. REFERENCES