Comparing Unidimensional and Multidimensional Models of Benefit Finding in Breast and Prostate Cancer

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Qual Life Res (2008) 17:771–781

DOI 10.1007/s11136-008-9348-z

Comparing unidimensional and multidimensional models


of benefit finding in breast and prostate cancer
Kathryn E. Weaver Æ Marı́a M. Llabre Æ
Suzanne C. Lechner Æ Frank Penedo Æ
Michael H. Antoni

Accepted: 9 April 2008 / Published online: 24 May 2008


 US Government 2008

Abstract examine if BFS domains are differentially related to


Purpose Psychosocial research with cancer patients has quality-of-life outcomes in both cancer and other medical
increasingly recognized that a large proportion report at patients.
least some positive changes or aspects of personal growth
that occur because of their illness—a construct referred to Keywords Psychological adjustment  Cancer 
as benefit finding. Although theory suggests that there may Measurement
be several domains to benefit finding, measurement
instruments such as the Benefit Finding Scale (BFS) are Abbreviations
typically considered to be unidimensional. BF Benefit finding
Method This study compared single and multiple factor SRGS Stress-Related Growth Scale
models of the BFS using group confirmatory factor analysis in PTGI Posttraumatic Growth Inventory
men with prostate cancer (n = 185) and women with breast BFS Benefit Finding Scale
cancer (n = 115) who were less than 2 years postdiagnosis. SES Socioeconomic status
Results In both samples, the multiple-factor model fit the CFA Confirmatory factor analysis
data significantly better than the single-factor model, and CFI Comparative Fit Index
factor loadings were equivalent between groups. Men with RMSEA Root Mean Square Error of Approximation
prostate cancer reported greater Personal Growth benefit SRMR Standardized Root Mean Squared Residual
finding and women with breast cancer reported greater AA/BC African American/Black Caribbean
Social Relationships benefit finding. Differential relation-
ships were observed between BFS factors and
sociodemographic and disease-related variables. Introduction
Conclusion Results suggest an alternate multidimen-
sional framework for the BFS that may be generalizable Despite the many stressors and negative psychological
across cancer populations. Future research should sequela associated with cancer (see for example, [1–3]),
there is increasing recognition that some individuals view
aspects of their illness as positive or beneficial. Terms used
to describe the identification of benefits following trau-
matic experiences include benefit finding (BF), stress-
K. E. Weaver  M. M. Llabre (&)  S. C. Lechner  F. Penedo 
related growth, found/discovered meaning, and posttrau-
M. H. Antoni
Department of Psychology, University of Miami, matic growth [4–8]. Cancer-related BF appears to be a
P.O. Box 24-8185, Coral Gables, FL 33134, USA relatively common phenomenon, with 50–80% of study
e-mail: mllabre@miami.edu participants identifying at least one positive life change
associated with breast [9] or lung cancer [10].
S. C. Lechner  F. Penedo  M. H. Antoni
University of Miami School of Medicine, Coral Gables, FL, BF theory suggests that positive life changes may occur
USA in several different domains. Tedeschi and Calhoun [4]

123
772 Qual Life Res (2008) 17:771–781

propose that traumatic events must be sufficiently disrup- enhanced coping skills) and is similar to the content of the
tive to evoke the cognitive activities that lead to growth. Personal Strength factor of the PTGI and the ‘‘enhanced
There is variability in both the domains and intensity of personal resources’’ and ‘‘enhanced or improved coping
negative changes reported by cancer survivors [11], sug- skills’’ domains identified by Park and colleagues [5]. The
gesting that BF may be most likely to occur in life domains Acceptance factor centers on the intentional acceptance of
in which the person has experienced disruption. In their things that cannot be changed, and the World View factor
conceptual model for understanding life crises and transi- focuses on changes in spirituality and perceptions of the
tions, Moos and Schaefer [12] further identify five groups purpose of others in the world. The Health Behaviors factor
of tasks that are typically encountered during a life crisis or is new to the 29-item version and is not covered by other
transition and propose that the importance of the different commonly used scales. The BFS also distinguishes
tasks varies depending on characteristics of the person and between BF with respect to family relationships and more
the crisis. The extent to which one engages in each of the general social relationships. This may be especially
tasks may influence outcomes in related areas. For exam- important in cancer patients, as intimate relationships
ple, the success with which a person sustains relationships might be strained by caregiving responsibilities or changes
with others or preserves a sense of competence and mastery in sexual functioning due to treatment side effects, result-
may influence perceptions of BF in related areas, i.e., social ing in more or less BF. Changed perceptions of family
relationships and personal strengths. relationships may or may not track with more generalized
Tedeschi and Calhoun [13] identified five distinct cate- perceptions of social support and closeness.
gories of posttraumatic growth (i.e., relating to others, new Although researchers have identified broad categories of
possibilities, personal strength, spiritual change, and BF and designed questionnaires to assess several dimen-
appreciation of life), and Park and colleagues [5] hypothe- sions (for review see [21]), in practice, BF is often treated
sized that stress-related growth encompassed three domains as a unidimensional construct. Even the PTGI, which has
(enhanced social resources, enhanced personal resources, five designated subscales, is often reduced to a single total
and new or improved coping skills). However, factor ana- score. Some psychometric data support the single dimen-
lytic studies have not always been consistent, suggesting sion approach to BF, as Cronbach’s alpha reliabilities for
both unidimensional and multidimensional models of BF, the scales are often high and exploratory factor analyses
even for the same scale (see for example, [5], [14]). Com- (EFA) have found that all items load strongly on a single
monly used BF measures, such as the Stress-Related general factor [5, 9, 13, 18–20]. However, other EFA
Growth Scale (SRGS) [5] and the Posttraumatic Growth analyses suggest multiple factor solutions for BF scales
Inventory (PTGI) [13], have been used to assess BF in used with medical and caregiving samples [22–24].
medical patients; however, there are several drawbacks to In their conceptual model, Moos and Schaefer [12]
this practice. As both of these scales were validated in identify personal and demographic characteristics as one of
samples of college students experiencing age-appropriate three sets of factors affecting the resolution of a crisis.
stressors and traumas, they may fail to assess domains of BF However, the results of studies linking demographic pre-
specific to medically ill populations. For instance, neither dictors such as gender, age, socioeconomic status (SES),
the SRGS nor the PTGI assesses the adoption of positive education, and ethnicity with BF in cancer patients vary by
health behaviors, one of the most prevalent forms of posi- cancer type and method of assessing BF. Although results
tive change among medical patients [15, 16]. from validation samples comprised of college students
In contrast, the Benefit Finding Scale (BFS) was suggest that women report greater BF than do men [5, 13],
developed for and validated in samples of medical patients, these findings are typically not replicated in samples of
specifically women with breast cancer. This measure was cancer patients [6, 25]. One exception was a study of mixed
adapted from Behr’s Positive Contributions Scale used cancer patients that found that women reported greater BF
with parents of disabled children [17], and additional items for four of five domains [26]. Some research suggests that
were written with the content directed toward the experi- younger cancer survivors report greater benefits than do
ence of individuals with cancer. A 17-item version of the older survivors [6, 26, 27], with one breast-cancer study
scale was initially developed in women with breast cancer reporting stronger associations for BF domains assessing
[18, 19], and a 22 item version with some unique items has relationships and appreciation for life [28]. Other studies
also been examined [20]. The 29-item version being used have found no relationship between age and BF in indi-
by this research team assesses six BF dimensions: Accep- viduals with breast cancer [9, 20, 29, 30].
tance, Family Relations, Personal Growth, World View, Studies of cancer patients also produce conflicting
Social Relations, and Health Behaviors. The Personal results regarding the impact of ethnicity and SES on BF. A
Growth factor focuses on identification of new strengths or few studies suggest that ethnic minority and low SES
positive qualities (i.e., patience, responsibility, strength, women perceive more benefits than white, non-Hispanic

123
Qual Life Res (2008) 17:771–781 773

women [18, 20]. Others find no differences between ethnic interest in the parent studies, we excluded participants with
groups [9] or no relationship between income or education a history of chronic illness associated with permanent
and BF [6, 25]. SES may be related to some measures or changes in the immune system or those who were pre-
dimensions of BF but not others [9, 28]. scribed medications with immunomodulatory effects.
Due to measurement limitations in existing research, Complete recruitment and participation information has
few decisive conclusions can be drawn from the research been reported elsewhere [31, 32].
linking sociodemographic and disease-related predictors
with BF. Conflicting findings may be due to differences in Breast cancer sample
the method (interview vs. questionnaire), domains of BF
assessment, and the population studied. If BF domains are Eligible participants included women between the ages of
differentially related to predictors, study results may be 18 and 70 with a primary breast cancer diagnosis within the
obscured when total scores, which combine or average BF previous 24 months who were at least 3 months postadju-
across domains, are used. Finally, studies of mixed cancer vant treatment (excluding tamoxifen and aromatase
types may combine men and women in the same samples, inhibitors). We excluded participants with metastatic breast
potentially obscuring results if BF factors are expressed cancer (stage IV) or a previous diagnosis of cancer.
differently by gender or disease type.
The purpose of this study was to refine the BFS Prostate cancer sample
instrument and address some basic questions regarding BF
measurement with this scale: (1) Is the BFS best concep- Men aged 45 years or older who had undergone radical
tualized as a unidimensional or multidimensional prostatectomy or radiation therapy for stage I or II prostate
instrument? (2) Can the factor structure of the BFS be cancer in the past 18 months were eligible. We excluded
generalized across men and women with different types of men reporting a previous diagnosis of nonskin cancer or
cancer? (3) Are demographic and disease-related variables current adjuvant treatment for prostate cancer.
(e.g., age, marital status, ethnicity, and type of cancer
treatment) differentially related to BFS dimensions? We Procedure
examined measurement models of the BFS in women with
breast cancer and men with localized prostate cancer using All participants provided informed consent prior to partic-
confirmatory factor analysis (CFA). We hypothesized that ipation in the study, and the university Institutional Review
multiple-factor models would provide a better fit for the Board approved all study procedures. Qualified participants
data in all samples and used multiple group CFA to completed psychosocial instruments via interview and self-
establish factorial invariance across the samples. Whereas report formats and a medical evaluation. Participants
the means of the latent factors were expected to differ received monetary compensation for their involvement. At
between groups, factor loadings were not. Hybrid models, the completion of the baseline appointment, participants
including demographic (i.e., age, ethnicity, SES, and were randomized to either a 10-week cognitive–behavioral
marital status) and disease-related variables (i.e., type of stress management intervention or a 1-day stress manage-
treatment), allowed us to determine if correlates differed by ment condition. The study used data from a subset of
BFS domain. patients who had completed the baseline assessment at the
time of analyses before recruitment was completed.

Method Measures

Participants Benefit finding

Data for the study came from two larger psychosocial We used the 29-item version of the BFS to measure per-
intervention studies [31, 32]. We recruited participants ception of growth from the experience of cancer. The
from Miami-Dade and Broward counties (Florida) through measure assesses six domains of BF—Acceptance, Family
direct recruitment and referrals from physicians, hospital Relations, Personal Growth, World View, Social Relations,
clinics, service agencies, tumor registries, community and Health Behaviors—using a 5-point Likert-type scale
health centers, conferences, and community events; (1 = not at all to 5 = extremely). Alpha reliability was
advertisements in the local press; and previous participant high (0.91–0.96) in samples of women with breast cancer
referral. Participants were required to be free of cognitive and men with prostate cancer for full and reduced versions
impairment and major psychopathology or substance of the measure, and previous researchers treated the mea-
dependence. As immune variables were outcomes of sure as unidimensional [18–20, 31, 33].

123
774 Qual Life Res (2008) 17:771–781

Disease-related and sociodemographic variables Preliminary data screening

Surgical (radical prostatectomy vs. radiation treatment for Six participants did not complete the BFS questionnaire
the prostate cancer sample; mastectomy vs. lumpectomy during the initial assessment, and 15 participants were
for the breast cancer sample) and adjuvant treatment missing items 26–29 due to the addition of these items after
(chemotherapy, radiation, and hormonal therapy) were assessments had begun. Prior to modeling, we examined
investigated as possible correlates. Age, education, income, descriptive statistics including the mean, standard devia-
marital status, and ethnicity were assessed via self-report. tion, skewness, and kurtosis for each BFS item in both
samples (see Table 2). No items violated assumptions of
univariate normality (skewness [ 2 or kurtosis [ 4).
Results
Confirmatory factor analysis of the BFS
Descriptive data
Analytic approach
Sociodemographic and disease-related sample characteris-
tics are shown in Table 1. Of the men with prostate cancer, We tested multiple factor models of the BFS in both
48.9% had a radical prostatectomy and 51.1% had radiation samples separately, made modifications to optimize fit, and
treatment. The majority of the women were diagnosed with then compared multiple- and single-factor models using the
stage I (40.8%) or stage II (43.7%) breast cancer, with chi-square difference test. A significant change in chi-
40.5% receiving mastectomy and 59.5% receiving lump- square would indicate that the multiple-factor model is a
ectomy. Of the women with breast cancer, 69.6% received superior fit for the data compared with the single-factor
chemotherapy and 76.8% radiation treatment. model. Group CFA analyses directly compared the breast

Table 1 Sociodemographic
Prostate CA Breast CA Test of association
and disease-related
(n = 186) (n = 115)
characteristics of the samples
Age [years (SD)] 65.05 (7.78) 49.71 (7.84) F = 271.93, P \ .001
Time since diagnosis 15.81 (7.68) 15.75 (5.07) F = 0.00, P [ .05
[months (SD)]
Income (%) v2 (4) = 11.99, P \ .05
\$20,000 18.3 5.3
$21,000–40,000 25.3 37.9
$41,000–60,000 19.3 16.8
$61,000–80,000 8.1 6.3
[$80,000 29.0 33.7
Ethnicity (%) v2 (2) = 10.25, P \ .05a
Non-Hispanic White 58.1 64.6
Hispanic 18.3 23.6
African American/Black 23.1 8.2
Caribbean
Other (American Indian, Asian, 0.5 3.6
or Multiethnic)
Education (%) v2 (3) = 4.52, P [ .05
\Grade 12 9.7 4.6
HS graduate or GED 20.4 15.4
Some college 25.3 31.8
SD standard deviation, HS high C4 Year college degree 44.6 48.2
school, GED General
Marital status (%) v2 (3) = 6.92, P [ .05
Educational Development
a Married or equivalent 75.8 64.6
This chi-square value reflects
the comparison for the first three Widowed 3.2 1.8
ethnic groups only due to the Divorced or separated 15.6 21.8
very small numbers in the other Never married 5.4 11.8
category

123
Qual Life Res (2008) 17:771–781 775

Table 2 BFS item means (Having cancer…)


Item Mean (SD) Mean (SD) Skewness Kurtosis
Prostate Breast CA
CA

(1) Has led me to be more accepting of things 3.12 (1.23) 3.51 (1.11) -.34 -.95
(2) Has taught me to adjust to things I cannot change 3.46 (1.17) 3.53 (1.11) -.61 -.44
(3) Has helped me take things as they come 3.56 (1.09) 3.57 (1.08) -.72 -.15
(4) Has brought my family closer together 3.24 (1.33) 3.10 (1.23) -.26 -1.04
(5) Has made me more sensitive to family issues 3.26 (1.35) 3.45 (1.08) -.49 -.78
(6) Has made me more grateful for each day 3.71 (1.26) 4.35 (.97) -1.06 .16
(7) Has taught me that everyone has a purpose in life 3.25 (1.40) 3.32 (1.46) -.38 -1.20
(8) Has shown me that all people need to be loved 3.35 (1.41) 3.60 (1.45) -.54 -1.04
(9) Has confirmed my faith in God 3.24 (1.58) 3.62 (1.50) -.46 -1.32
(10) Has made me realize the importance of planning for my family’s future 3.61 (1.35) 3.57 (1.30) -.70 -.65
(11) Has made me more aware and concerned for the future of all human beings 3.21 (1.35) 3.21 (1.31) -.25 -1.07
(12) Has taught me to control my temper 2.52 (1.37) 2.47 (1.32) .39 -1.12
(13) Has taught me to be patient 2.82 (1.35) 2.93 (1.31) .02 -1.26
(14) Has made me a more responsible person 2.89 (1.40) 2.45 (1.33) .17 -1.30
(15) Has given my life better organization 2.75 (1.41) 2.26 (1.31) .30 -1.25
(16) Has made me more productive 2.39 (1.39) 2.15 (1.27) .54 -1.05
(17) Has led me to deal better with stress and problems 2.91 (1.35) 2.71 (1.31) .13 -1.22
(18) Has helped me to budget time better 2.64 (1.37) 2.32 (1.28) .36 -1.20
(19) Has led me to have more friends 2.05 (1.24) 2.07 (1.17) .89 -.32
(20) Has led me to meet people who have become some of my best friends 1.94 (1.24) 1.97 (1.25) 1.08 -.07
(21) Has renewed my interest in participating in different activities 2.33 (1.28) 2.43 (1.30) .47 -1.02
(22) Has contributed to my overall emotional and spiritual growth 2.88 (1.45) 3.36 (1.30) -.07 -1.33
(23) Has helped me become aware of the love & support available from other people 3.42 (1.21) 3.78 (1.17) -.50 -.75
(24) Has helped me realize who my real friends are 3.06 (1.41) 3.74 (1.34) -.44 -1.15
(25) Has helped me become more focused on priorities, with a deeper sense of purpose in life 3.30 (1.27) 3.75 (1.25) -.55 -.74
(26) Has helped me become a stronger person, more able to cope effectively with future life 3.32 (1.25) 3.66 (1.20) -.54 -.62
challenges.
(27) Has led me to eat a healthier diet 2.93 (1.45) 3.07 (1.25) -.10 -1.26
(28) Has led me to exercise more regularly 2.68 (1.37) 2.64 (1.25) .18 -1.20
(29) Has led me to live a healthier lifestyle 3.06 (1.30) 3.16 (1.18) -.20 -.99
SD standard deviation

and prostate cancer groups. By specifying parameters to be Mean Squared Residual (SRMR) to evaluate the statistical
equal between groups, we examined measurement invari- fit of the model to the data. Generally, a nonsignificant chi-
ance and equivalence of factor means. Formal testing of square, CFI greater than 0.95, a RMSEA less than 0.06, and
measurement invariance typically involves examining a SRMR less than 0.09 are considered to indicate good
factor loadings and item intercepts. However, invariance of fit [36]. We used full information maximum likelihood
all these parameters is not a precondition for comparison of (FIML) to estimate all models. This approach allowed us to
latent factor means under conditions of partial measure- utilize all available data and provided unbiased parameter
ment invariance [34]. The approach of initially fitting the estimates when missingness could be predicted from other
model separately in the individual groups is recommended variables in the model.
by Thompson and Green [35], as fitting the model initially
in the combined sample might mask structural differences Multiple-factor models
between the groups.
We used Mplus statistical software (version 3.01) for all We grouped items using content analysis guided by theory
analyses and the chi-square goodness-of-fit statistic, and previously identified BF dimensions. Six factors were
Comparative Fit Index (CFI), the Root Mean Square Error identified: Acceptance, Family Relations, Personal Growth,
of Approximation (RMSEA), and the Standardized Root World View, Social Relations, and Health Behaviors. We

123
776 Qual Life Res (2008) 17:771–781

removed one item a priori whose content was excessively Table 3 Comparisons of goodness-of-fit indicators of nested single-
general (item 2: has contributed to my overall emotional factor and multiple-factor models of the benefit-finding scale
and spiritual growth). After examining this initial model, Model df v2 CFI RMSEA SRMR
we identified six items for which modification indices
Prostate cancer
suggested multiple factor loadings (items 6, 10, 19, 21, 25,
26). We examined models allowing these items to load on Single factor 207 1019.70* .76 .15 .08
more than one factor, confirmed that they did in fact have 6 Factor 192 333.23* .96 .06 .04
two or more factor loadings greater than 0.4, and subse- D v2 15 686.48*
quently removed them from the model. Following the Breast cancer
addition of two correlated errors that were suggested by the Single factor 206 570.44* .70 .13 .10
modification indices and were between items with very 6 Factor 191 263.79* .94 .06 .06
similar content (taught me to be patient/taught me to D v2 15 306.65*
control my temper & made me more responsible/given my A significant change in chi-square indicates worse model fit in the
life better organization), the multiple-factor model fit the single-factor models
prostate cancer data reasonably well [v2(192) = 333.226, CFI Confirmatory Fit Index, RMSEA Root Mean Square Error of
P B 0.0001; CFI = 0.96; RMSEA = 0.06; SRMR = Approximation, SRMR Standardized Root Mean Squared Residual.
* P \ 0.001
0.04]. Although the chi-square was significant, other fit
indices had acceptable values. With minor modifications
(see Fig. 1), the six factor model also fit the breast cancer significantly better than did the single factor-model in both
data reasonably well [v2(191) = 263.789, P \ 0.0001; samples (see Table 3).
CFI = 0.94; RMSEA = 0.06; SRMR = 0.06]. Many of
the correlated errors in both samples were between items Group analyses
that were adjacent on the scale, and the addition of these
parameters did not substantially change the factor loadings. Group CFA models assessed whether BF could be measured
the same way in women with breast cancer and men with
Comparing multiple-factor and single-factor models prostate cancer. A model specifying invariance of factor
loadings and item intercepts, but allowing the error vari-
A single-factor model of BF that specified all items to load ances and factor means, variances, and intercorrelations to
on a single general factor of BF was a poor fit in both vary, fit the data relatively well [v2(416) = 705.303,
sample (CFIs \0.77 and RMSEAs [0.11). A direct com- P \ 0.0001; CFI = 0.94; RMSEA = 0.069; SRMR =
parison between models was done by specifying factor 0.065]. Improvement in fit was achieved with the addition
intercorrelations to be one, implying a single-factor model. of three group-specific correlated errors [v2(413) =
This approach allowed us to directly test the difference in 660.035, P \ 0.0001; CFI = 0.95; RMSEA = 0.064;
model fit between multiple- and single-factor models using SRMR = 0.064]. These were the same correlated errors of
the chi-square difference test for nested models. As small magnitude that were necessary to fit the data in the
hypothesized, the multiple-factor model fit the data cancer samples separately. Unstandardized parameter

Family Personal Health


Acceptance Relations Growth Social Relations World View Behaviors

1.00 0.89 1.00 1.04 1.00 1.32 1.00 1.00


0.96 1.22 0.93

1 2 3 4 5 20 23 24 27 28 29
1.00 1.07 1.03 0.97
1.00 1.02 1.13 1.15 1.12 1.00 1.08
7 8 9 11
12 13 14 15 16 17 18

PCa= .27
BCa= .60 BCa= .30 BCa=.23 PCa=.18
PCa= .29 PCa= .11

Fig. 1 Unstandardized parameter estimates for a multiple-group Comparative Fit Index = 0.95; root mean squared error of approx-
confirmatory factor analysis (CFA) model comparing breast cancer imation = 0.064; standardized root mean squared residual = 0.064;
and prostate cancer. All parameters are statistically significant BCa breast cancer parameter, PCa prostate cancer parameter
(P \ .05). Model-fit statistics: v2 (413) = 660.035, P \ .0001;

123
Qual Life Res (2008) 17:771–781 777

estimates for the cancer group model are shown in Fig. 1. in men with prostate cancer (standardized parameter esti-
Latent-factor intercorrelations ranged from 0.44 to 0.88. mates ranging from -0.29 to -0.50, all P \ 0.05). Further
We examined differences in the latent-factor means to analyses indicated that SES was differentially related to
determine whether men and women with cancer report BFS factors in men with prostate cancer, as the paths could
different levels of BF. The group analysis provides a test of not be constrained to be equal without harming model fit [D
the difference between the means of the latent factors, as v2(5) = 13.88, P \ 0.05]. SES was most strongly related
one group is specified to have a factor mean equal to zero, to World View, Personal Growth, Family Relations, and
and the difference between the group means is estimated Health Behaviors BF and less strongly related to Social
and tested in the second group. Men with prostate cancer Relationships and Acceptance BF in men with prostate
reported greater Personal Growth BF (mean differ- cancer.
ence = 0.27, P \ 0.05) and women with breast cancer
reported greater Social Relations BF factor (mean differ- Ethnicity
ence = 0.32, P \ 0.05).1 The factor means were not
significantly different for the other factors. The six latent factors of BF were regressed on two dummy
variables that coded for ethnicity [African American/Black
Demographic and disease-related predictors of benefit Caribbean (AA/BC) & Latino/Latina], with Non-Hispanic
finding Whites treated as the reference group. Latina women
reported greater Family Relations and World View BF. AA/
Analytic plan BC women reported greater Personal Growth and World
View BF. These results should be interpreted with caution
We used the multifactor measurement models of BF, because of the relatively small number of ethnic minority
within structural equation models, to examine sociodemo- women in the breast cancer group. Latino and AA/BC men
graphic and disease-related correlates of BF. As few with prostate cancer scored higher on all six factors of BF
studies have examined predictors of specific BF domains, when compared with Non-Hispanic White men.
these exploratory analyses were designed to provide
guidance to future studies utilizing a multidimensional BF Cancer treatment variables
approach. Significant collinearity was observed among
predictors (e.g., age, marital status, and ethnicity were all We tested the relationship between cancer treatment vari-
significantly related to SES). Thus, we used the breast and ables and BF separately in breast and prostate cancer
prostate cancer group model to examine each sociodemo- because of differences in types of treatment received.
graphic correlate of the BFS latent factors in a separate Treatment for prostate cancer (coded as radical prostatec-
model. Cancer treatment variables were examined sepa- tomy or radiation) was unrelated to any of the BF factors.
rately in the breast and prostate cancer samples, as there Neither surgery type (lumpectomy vs. mastectomy) nor
were differences in the types of treatments received. radiation treatment (received vs. not) was significantly
Unstandardized parameter estimates, which can be inter- related to any factor of BF in women with breast cancer.
preted as the mean difference in BFS factor scores for Women receiving chemotherapy treatment (compared with
dummy coded variables, are summarized in Table 4. those who did not) reported significantly greater Family
Relations, Personal Growth, and World View BF.
Age, marital status, and socioeconomic status
Comparison with total score approach
Age was unrelated to all BFS factors in both samples.
Married or partnered breast and prostate cancer survivors To determine whether conclusions regarding sociodemo-
reported significantly higher levels of Family BF compared graphic and treatment-related correlates of BF would differ
with those who were not, and married women with breast if we were to use the total score approach for the BFS, as
cancer also reported higher levels of World View BF. A suggested by a unidimensional model, we examined the
CFA model of SES, with income and education as indi- same variables as predictors of the BFS total score. BFS
cators, was examined in a combined model with the total score did not differ significantly between the men with
measurement model of BF. The SES latent variable was prostate cancer and the women with breast cancer. Con-
unrelated to any BFS factor in women with breast cancer sistent with multidimensional model findings, age was
but was significantly negatively related to all BFS factors unrelated to BFS total score in both samples. BFS total
scores were greater for Latino/Latina and AA/BC women
1
The estimates of the means were identical for the models with and with breast cancer and men with prostate cancer (mean
without the group specific correlated errors. differences ranging from 12.6 to 28.1 points, all P \ 0.05).

123
778 Qual Life Res (2008) 17:771–781

Table 4 Unstandardized
Acceptance Family Personal Social World Health
parameter estimates for Benefit
relations growth relations view behaviors
Finding Scale (BFS) factor
correlate analyses Age
Breast -.01 .00 -.02 .00 -.00 -.02
Prostate -.01 .00 .00 .00 .01 .00
Socioeconomic status
Breast -.01 .08 .11 -.04 -.12 .07
Prostate -.20* -.22* -.22* -.13* -.30* -.21*
Married/partnered
Breast .31 .62* -.04 .19 .44* .21
Prostate -.17 .87* -.04 .28** .29 -.06
African American/Black Caribbean
Breast .73** .45 .94* .38 1.51* .50
Prostate .40* .65* .95* .58* 1.05* 1.14*
Latino/Latina
Breast .08 .49* .22 .33** .87* .24
Prostate .60* .98* .99* .58* 1.11* 1.01*
PC Treatment -.04 -.03 .04 -.14 .00 .15
BC Surgery -.04 .04 -.14 .02 -.03 -.42
BC Radiation (y/n) -.01 -.25 -.15 -.11 .00 -.21
PC prostate cancer, BC breast
BC Chemotherapy .26 .49* .52* .15 .73* .30
cancer
(y/n)
* P \ 0.05, ** P \ 0.10

BFS total score was significantly related to SES in men the presence of 6 latent factors of BF: Acceptance, Family
with prostate cancer (standardized parameter estimate = - Relations, Personal Growth, World View, Social Relations,
0.44, P \ 0.001) but not women with breast cancer. Mar- and Health Behaviors. There is substantial overlap between
ried/partnered persons did not differ from unmarried the factors identified in this study and the factors identified
persons in BFS total score in either the prostate cancer or using exploratory factor analysis on a shorter version of the
breast cancer groups. Women who received chemotherapy measure with cancer caregivers [24]. In both the caregiving
reported greater BFS total scores (mean differ- study and the cancer samples used in the current study, a
ence = 15.77, P \ 0.05) compared with those who did not, multidimensional model of BF fit the data significantly
but there were no significant differences by type of surgery better then a unidimensional model. Multiple group anal-
or receipt of radiation treatment. There were no significant yses revealed that the factor loadings and item intercepts
differences by type of treatment for prostate cancer. Thus, were invariant across samples. These results suggest that
analyzing the BFS using a total score approach would have while samples may vary in the level of BF reported, the
masked differences between the breast cancer and prostate BFS has a similar multidimensional structure across the
cancer groups and between married and unmarried persons observed cancer patient groups.
and overgeneralized findings regarding race/ethnicity and
chemotherapy treatment in women with breast cancer. In Group differences in benefit finding
addition, the differences in the strength of association
between SES and BF domains would have been obscured Women with breast cancer reported higher levels of Social
in men with prostate cancer, although the direction was Relations BF and men with prostate cancer reported greater
consistent across factors. Personal Growth BF. This is best thought of as a group
comparison rather than a simple gender comparison, as the
two groups differed on disease, age, and other sociode-
Discussion mographic characteristics. Perhaps men with prostate
cancer perceive their illness and treatment as more of a
Study results support treating benefit finding, as measured personal challenge and/or struggle more with the crisis-
by the Benefit Finding Scale, as a multidimensional con- related task of preserving a satisfactory self-image and
struct. Confirmatory factor analysis of the BFS in men with consequently perceive greater growth in the Personal
prostate cancer and women with breast cancer established Growth domain. Women with breast cancer may

123
Qual Life Res (2008) 17:771–781 779

experience greater social disruption, utilize support from to measuring this construct may obscure significant rela-
friends and family as a primary coping strategy, and/or tionships by combining BF dimensions that are
have greater success with the crisis-related task of sus- differentially related to predictors and possibly outcomes.
taining relationships with important others, stimulating Future work should examine different domains of BF in
greater perceived growth in social relationships. Utilizing a relationship to the specific crisis-related disruptions and
total score approach would have obscured these differ- developmental tasks that individuals with medical illnesses
ences, In general, sociodemographic variables (SES, experience. This type of research has the potential to
marital status, and ethnicity) were more strongly associated advance conceptual models of BF by linking the life dis-
with BF in men compared with women. Psychosocial or ruption and/or the completion of adaptive tasks with
dispositional variables, such as coping, social support, specific domains of BF.
disclosure, and social constraint, may be more important
predictors of BF in women, especially given greater reports Implications for BFS scoring
of positive growth in social relationships.
The superior fit of multifactor models in both samples
Correlates of benefit finding studied here suggests that the total composite score
approach to this measure should be replaced by factor
Married or partnered persons with cancer reported greater scores using a 22-item version of the scale. It is recom-
Family Relations BF and World View BF, possibly due to a mended that factors be calculated by summing or taking the
greater number of intimate supportive relationships. Again, mean of the items, which assumes equal factor weights, as
these differences would not have been observed if BFS opposed to using weighted factors scores, which are unli-
total scores were used. Ethnic minority women reported kely to generalize across samples [37].
higher levels of Family Relations BF, World View BF, and
Personal Growth BF; and ethnic minority men reported Limitations and directions for future research
higher levels of all dimensions of BF. These dimensions
may be particularly sensitive to ethnic differences in family Several limitations of this study should be noted. First, the
structure, cultural values, use of coping strategies such as BFS was not designed specifically to assess all dimensions
acceptance, and spirituality. Lower SES was significantly, of BF in a comprehensive way. Whereas the six factors
but differentially, related to greater BF factor scores for identified are consistent with previous qualitative and
men with prostate cancer but not for women with breast quantitative research, it is possible that there are other
cancer. Low SES men may find the experience of being dimensions of BF not assessed by this scale (e.g., an
diagnosed with cancer to be more stressful or threatening expanded conceptualization of changes in spirituality,
because of difficulties accessing health care, limited health religiosity, or altruism). Future research should focus on
knowledge, and financial concerns. This greater perception identifying common and specific dimensions of BF across
of threat may translate to greater perception of benefits [6, populations (including trauma victims and medical
30]. Low SES individuals may also experience more patients), as well as developing more refined instruments.
traumatic or stressful life events and therefore may have The factor structure should be examined for invariance
more practice in searching for and/or finding benefits from across samples of individuals who have faced a variety of
negative experiences [20]. More research is needed to stressful and/or traumatic life events. This study provides a
understand how gender or disease type may interact with template for the evaluation of multifactor models of future
socioeconomic resources to influence BF and to confirm BF scales across samples of medical patients.
that the observed SES associations are not due to income, This study examines men and women with different
education, and/or ethnic differences between the groups or types of cancer, thus the effect of gender cannot be dis-
limited power to detect associations in the breast cancer tinguished from the effect of disease type. This may also be
sample. Future studies using a multidimensional approach a concern for studies of mixed cancer patients, as samples
are necessary to further elucidate the relationships among are likely to include a large proportion of individuals with
BF factors and correlates and ensure that differences common gender-specific cancers. Conclusive data regard-
observed in this study are not sample specific. ing the impact of gender of BF would be best generated
The differential relationships between dimensions of BF from studies examining gender-neutral cancers, such as
and predictive variables lends further support to treating colorectal or lung cancers. In this study, BF was assessed
BF as a multidimensional construct and may provide one using a self-report measure. Some researchers have inves-
explanation for the disparate findings in the research lit- tigated ways of obtaining more ‘‘objective’’ reports of BF,
erature linking predictors such as age, gender, ethnicity, either through collaborative reports by family member or
and SES with BF. Commonly used total-score approaches friends or behavioral outcomes (e.g., [5]). However,

123
780 Qual Life Res (2008) 17:771–781

changes in internal states such as appreciation for life or factor analysis in men with prostate cancer and women with
loved ones or confidence in one’s abilities may not be able breast cancer. Multidimensional models were superior to
to be confirmed in a satisfactory way. In addition, we unidimensional models in both samples. These results
examined a limited number of sociodemographic correlates suggest that researchers may want to examine domain-
of BF cross-sectionally in this study, leaving the majority specific models of personal growth in cancer patients when
of the variance in all dimensions of BF unexplained. Future using the BFS. Future work is needed to more clearly
studies should examine the relationship between coping identify different dimensions of BF and to understand how
strategies (particularly positive reframing), social support, different dimensions of BF relate to both quality of life and
dispositional optimism, and personality variables and dif- medical outcomes over time.
ferent domains of BF. In addition, this study examined BF
at only one point in the disease trajectory (approximately Acknowledgements Kathryn Weaver is now a fellow in the Cancer
Prevention Fellowship Program, Office of Preventive Oncology,
15 months postdiagnosis). Future studies should examine National Cancer Institute, National Institutes of Health, Bethesda,
the trajectories of BF domains prospectively to understand MD, USA. This research was supported by grants NIMH
how BF develops in relation to disease processes. 5T32MH018917 (N. Schneiderman, PI), NCI 1P50CA84944 (Antoni,
Finally, this study does not address the relationship of PI), and NCI 5R03CA113096 (Lechner, PI).
BF with psychosocial adjustment. Whereas many studies
have linked the identification of benefits following medical
References
illness with positive psychological and medical outcomes
(e.g., [7], [15], [38]), some studies have found no or neg- 1. Cella, D. F., & Cherin, E. A. (1988). Quality of life during and
ative relationships (e.g., [9], [20], [30], [39]). Importantly, after cancer treatment. Comprehensive Therapy, 14, 69–75.
very few of these studies have examined multiple dimen- 2. Andersen, B. L., Anderson, B., & deProsse, C. (1989). Controlled
sions of BF. Future research should examine the prospective longitudinal study of women with cancer; II. Psy-
chological outcomes. Journal of Consulting and Clinical
relationships between BF domains and psychosocial and Psychology, 57, 692–697.
medical outcomes over time. It is possible that certain 3. Jacobsen, P. B., Widows, M. R., Hann, D. M., Andrykowski, M.
dimensions of BF will be more strongly related to positive A., Kronish, L. E., & Fields, K. K. (1998). Posttraumatic stress
psychological adaptation than others or that dimensions of disorder symptoms after bone marrow transplantation for breast
cancer. Psychosomatic Medicine, 60, 366–371.
BF will relate to psychosocial outcomes in opposite ways. 4. Tedeschi, R. G., & Calhoun, L. G. (2004). Posttraumatic growth:
Differential relationships between BF dimensions and Conceptual foundations and empirical evidence. Psychological
measures of well-being have been observed in both medi- Inquiry, 15, 1–18.
cal patient [23] and cancer caregiver samples [24]. 5. Park, C. L., Cohen, L., & Murch, R. (1996). Assessment and
prediction of stress-related growth. Journal of Personality, 64,
Replication of this pattern of results in cancer patients 71–105.
would confirm the utility of a multidimensional model of 6. Lechner, S. C., Zakowski, S. G., Antoni, M. H., Greenhawt, M.,
BF and provide one explanation for discrepancies in the Block, K., & Block, P. (2003). Do sociodemographic and disease-
research literature. related variables influence benefit-finding in cancer patients?
Psycho-Oncology, 12, 491–499.
7. Bower, J. E., Kemeny, M. E., Taylor, S. E., & Fahey, J. L. (1998).
Clinical implications Cognitive processing, discovery of meaning, CD4 decline, and
AIDS-related mortality among bereaved HIV-seropositive men.
From a clinical vantage point, the notion that BF is best Journal of Consulting and Clinical Psychology, 66, 979–986.
8. Affleck, G., & Tennen, H. (1996). Construing benefits from
conceptualized as multidimensional should encourage cli- adversity: Adaptational significance and dispositional underpin-
nicians to view patient reports of BF as such. That is, nings. Journal of Personality, 64, 899–922.
clinicians should not expect patients to report high levels of 9. Sears, S. R., Stanton, A. L., & Danoff-Burg, S. (2003). The
BF on every dimension and should take into account that yellow brick road and the emerald city: Benefit finding, positive
reappraisal coping, and posttraumatic growth in women with
only certain dimensions of BF are likely to change in a early-stage breast cancer. Health Psychology, 22, 487–497.
given patient. Should future research reveal that BFS 10. Sarna, L., Padilla, G., Holmes, C., Tashkin, D., Brecht, M. L., &
subscales are differentially related to psychosocial out- Evangelista, L. (2002). Quality of life of long-term survivors of
comes, interventions will need to take this into account. non-small-cell lung cancer. Journal of Clinical Oncology, 20,
2920–2929.
Although no interventions have been designed to enhance 11. Bellizzi, K. M., Miller, M. F., Arora, N. K., & Rowland, J. H.
BF per se, such studies are being conceptualized by (2007). Positive and negative life changes experienced by sur-
researchers in the field. Interventions should be tailored to vivors of non-Hodgkin’s lymphoma. Annals of Behavioral
target those aspects of BF most likely to result in Medicine, 34, 188–199.
12. Moos, R. H., & Schaefer, J. A. (1986). Life transitions and crises:
improvements in psychological and physical well-being. A conceptual overview. In R. H. Moos (Ed.), Coping with life
This study compared multidimensional and unidimen- crises: An integrated approach (pp. 3–34). New York, NY:
sional models of the BFS using multiple group confirmatory Springer.

123
Qual Life Res (2008) 17:771–781 781

13. Tedeschi, R. G., & Calhoun, L. G. (1996). The post-traumatic patient, partner, and couple perspectives. Psychosomatic Medi-
growth inventory: Measuring the positive legacy of trauma. cine, 66(3), 442–454.
Journal of Traumatic Stress, 9, 455–471. 28. Bellizzi, K. M., & Blank, T. O. (2006). Predicting posttraumatic
14. Roesch, S. C., Rowley, A. A., & Vaughn, A. A. (2004). On the growth in breast cancer survivors. Health Psychology, 25, 47–56.
dimensionality of the stress-related growth scale: one, three, or 29. Katz, R. C., Flasher, L., Cacciapaglia, H., & Nelson, S. (2001).
seven factors? Journal of Personality Assessment, 82, 281–290. The psychosocial impact of cancer and lupus: A cross-validation
15. Updegraff, J. A., Taylor, S. E., Kemeny, M. E., & Wyatt, G. E. study that extends the generality of ‘‘benefit-finding’’ in patients
(2002). Positive and negative effects of HIV infection in women with chronic disease. Journal of Behavioral Medicine, 24, 561–
with low socioeconomic resources. Personality and Social Psy- 571.
chology Bulletin, 28, 382–394. 30. Cordova, M. J., Cunninghman, L. L. C., Carlson, C. R., & An-
16. Siegel, K., & Schrimshaw, E. W. (2000). Perceiving benefits in drykowski, M. A. (2001). Posttraumatic growth following breast
adversity: Stress-related growth in women living with HIV/AIDS. cancer: A controlled comparison study. Health Psychology, 20,
Social Science and Medicine, 51, 1543–1554. 176–185.
17. Behr, S. K., Murphy, D. L., & Summers, J. A. (1991). Kansas 31. Penedo, F. J., Molton, I., Dahn, J. R., Shen, B. J., Kinsinger, D.,
inventory of parental perceptions. Lawrence, KS: University of Traeger, L., et al. (2006). A randomized clinical trial of group-
Kansas. based cognitive-behavioral stress management in localized
18. Urcuyo, K. R., Boyers, A. E., Carver, C. S., & Antoni, M. H. prostate cancer: development of stress management skills
(2005). Finding benefit in breast cancer: Relations with person- improves quality of life and benefit finding. Annals of Behavioral
ality, coping, and concurrent well-being. Psychology and Health, Medicine, 31, 261–270.
20, 175–192. 32. Antoni, M. H., Lechner, S. C., Kazi, A., Wimberly, S. R., Sifre,
19. Antoni, M. H., Lehman, J. M., Kilbourn, K. M., Boyers, A. E., T., Urcuyo, K. R., et al. (2006). How stress management
Culver, J. L., & Alferi, S. M., et al. (2001). Cognitive-behavioral improves quality of life after treatment for breast cancer. Journal
stress management intervention decreases the prevalence of of Consulting and Clinical Psychology, 74, 1143–1152.
major depression and enhances benefit finding among women 33. Kinsinger, D. P., Penedo, F. J., Antoni, M. H., Dahn, J. R.,
under treatment for early-stage breast cancer. Health Psychology, Lechner, S., & Schneiderman, N. (2006). Psychosocial and so-
20, 20–32. ciodemographic correlates of benefit-finding in men treated for
20. Tomich, P. L., & Helgeson, V. S. (2004). Is finding something localized prostate cancer. Psycho-Oncology, 15, 954–961.
good in the bad always good? Benefit finding among women with 34. Byrne, B. M., Shavelson, R. J., & Muthen, B. (1989). Testing for
breast cancer. Health Psychology, 23, 16–23. equivalence of factor covariance and mean structures: The issue
21. Park, C. L. (2004). The notion of growth following stressful life of partial measurement invariance. Psychological Bulletin, 105,
experiences: Problems and prospects. Psychological Inquiry, 15, 456–466.
69–76. 35. Thompson, M. S., & Green, S. B. (2006). Evaluating between-
22. Tomich, P. L., & Helgeson, V. S. (2002). Five years later: A group differences in latent variable means. In G. R. Hancock &
cross-sectional comparison of breast cancer survivors with heal- R. O. Mueller (Eds.), Structural equation modeling: A second
thy women. Psycho-Oncology, 11, 154–169. course (pp. 119–169). Greenwich, CT: Information Age
23. Pakenham, K. I. (2005). Benefit finding in multiple sclerosis and Publishing.
associations with positive and negative outcomes. Health Psy- 36. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indices in
chology, 24, 123–132. covariance structure analysis: Conventional criteria versus new
24. Kim, Y., Schulz, R., & Carver, C. S. (2007). Benefit-finding in alternatives. Structural Equation Modeling, 6, 1–55.
the cancer caregiving experience. Psychosomatic Medicine, 69, 37. Kline, R. B. (1998). Principles and practice of structural equa-
283–291. tion modeling. New York, NY: The Guilford Press.
25. Andrykowski, M. A., Brady, M. J., & Hunt, J. W. (1993). Positive 38. Carver, C. S., & Antoni, M. H. (2004). Finding benefit in breast
psychosocial adjustment in potential bone marrow transplant cancer during the year after diagnosis predicts better adjustment 5
recipients: Cancer as a psychosocial transition. Psycho-Oncology, to 8 years after diagnosis. Health Psychology, 23, 595–598.
2, 261–276. 39. Mohr, D. C., Dick, L. P., Russo, D., Pinn, J., Boudewyn, A. C., &
26. Bellizzi, K. M. (2004). Expressions of generativity and post- Likosky, W., et al. (1999). The psychosocial impact of multiple
traumatic growth in adult cancer survivors. International Journal sclerosis: Exploring the patient’s perspective. Health Psychology,
of Aging and Human Development, 58, 267–287. 18, 376–382.
27. Manne, S., Ostroff, J., Winkel, G., Goldstein, L., Fox, K., &
Grana, G. (2004). Posttraumatic growth after breast cancer:

123

You might also like