Invariance test of the Multidimensional Body Self-Relations ...

2 downloads 29 Views 218KB Size Report
Jun 12, 2010 - Breast cancer survivors reported a small yet significantly higher mean on ...... breast cancer survivors in their first season of dragon boating.
Qual Life Res (2010) 19:1171–1180 DOI 10.1007/s11136-010-9680-y

Invariance test of the Multidimensional Body Self-Relations Questionnaire: do women with breast cancer interpret this measure differently? Catherine M. Sabiston • Shayna Rusticus • Jennifer Brunet • Meghan H. McDonough • Valerie Hadd • Anita M. Hubley • Peter R. E. Crocker

Accepted: 20 May 2010 / Published online: 12 June 2010 Ó Springer Science+Business Media B.V. 2010

Abstract Purpose To examine whether the meaning and interpretation of body image are similar for breast cancer survivors and women without breast cancer. Method Women completed the Multidimensional Body Self-Relations Questionnaire—Appearance Scales as part of two studies. There were 469 women with breast cancer and 385 women without breast cancer. Invariance testing was conducted to examine whether the items assessing the body image dimensions were similar, whether the dimensions were interpreted similarly, whether the items were equally salient and meaningful, and whether there were mean differences on the body image dimensions across the two groups. Results The meaning and interpretation of body image dimensions related to appearance evaluation and appearance orientation were similar across the groups, yet some group differences were found for overweight preoccupation and body areas satisfaction (and not testable for self-classified weight). Breast cancer survivors reported a small yet significantly higher mean on appearance evaluation and lower mean on appearance orientation compared to the women without breast cancer. Conclusions Meaningful comparisons in body image across cancer and non-cancer women can be made using

C. M. Sabiston (&)  J. Brunet McGill University, Montreal, QC, Canada e-mail: [email protected] S. Rusticus  V. Hadd  A. M. Hubley  P. R. E. Crocker University of British Columbia, Vancouver, BC, Canada M. H. McDonough Purdue University, West Lafayette, IL, USA

two of the Multidimensional Body Self-Relations Questionnaire—Appearance Scales. The overweight preoccupation subscale could be used to assess body image but should not be used if group mean differences are desirable. Assessing satisfaction with body areas across these groups is not recommended and may introduce systematic bias. Keywords Women’s health  Clinical oncology  Body image  Quality of life Abbreviations Measures and subscales MBSRQ Multidimensional Body Self-Relations Questionnaire MBSRQ-AS Multidimensional Body Self-Relations Questionnaire—Appearance Scales AE Appearance evaluation AO Appearance orientation BAS Body areas satisfaction OP Overweight preoccupation SCW Self-classified weight Descriptives BMI Body mass index Analysis and fit indices ANOVA Analysis of variance MGCFA Multigroup confirmatory factor analysis MACS Mean and covariance structures ML Maximum likelihood RMSEA Root mean square error of approximation CFI Comparative fit index NNFI Non-normed fit index SRMR Standardized root mean square residual

123

1172

Advances in breast cancer diagnosis and treatment have resulted in improved survival rates [1]. With a greater number of women living with breast cancer, efforts to improve quality of life during survivorship are increasingly needed. Given the physical changes that result from treatment such as scarring, loss of breast(s), weight gain, hair loss, and changes in skin texture [1], coupled with the emotional challenges associated with sexuality and femininity perceptions [2–4], body image is an important component of quality of life in this population [3, 5]. Body image is a multidimensional construct that generally includes perceptions of body appearance, thoughts, and beliefs regarding body shape and appearance, attitudes reflecting how individuals feel about their body size and shape, and behaviors that embody actions related to appearance [6]. Nonetheless, most measures fall short of assessing multiple dimensions of body image, and few measures have been examined to see whether the meaning and interpretation of the assessments are similar across various subpopulations— such as breast cancer survivors and women without breast cancer. Meaningful comparisons of scores obtained on measures across groups cannot be made if such measures are deemed to be assessing different underlying constructs. Breast cancer-specific measures of body image are emerging [2, 4] and show promise in assessing general perceptions of the body within models of disease and illness. The existing measures place emphasis on sexuality concerns [4] or on multiple dimensions of the body affected by disease or illness [2]. Whereas the measures are useful for examining body image specific to cancer-related issues, they may fall short if one wants to compare body image prior to and following a cancer diagnosis or make meaningful comparisons to populations without a history of cancer. Cancer-specific measures also do not provide a tool to compare specific body image dimensions since they often focus on global assessments of satisfaction only. Such comparisons may be informative for tailored intervention strategies by identifying body image concerns that are significantly different for breast cancer survivors compared to women who have not had cancer. To accomplish this proposition, the same measures need to be used to assess body image across populations, and researchers need to be confident that these measures assess the same construct(s) across samples of women with and without a history breast cancer. Furthermore, using the same measures to assess body image across samples may reduce the issue of overgeneralization. It is well documented that women report higher levels of body image dissatisfaction compared to men [6, 7] and these body image concerns are reported throughout the life span [8]. Therefore, there is a certain level of body image that is likely to exist independent of age or trauma, and this heightened body image dissatisfaction could falsely be described as resulting from breast cancer when, in

123

Qual Life Res (2010) 19:1171–1180

fact, it is more a result of the general discontent reported by women [8]. While breast cancer-specific measures of body image may be informative of how women feel about their bodies after a diagnosis and treatments for breast cancer, they do not provide a full context of body image concerns and thus do little to facilitate comparisons with the general population of women. The most widely used measure of body image is the Multidimensional Body Self-Relations Questionnaire [MBSRQ; 9] and the shorter scale called the MBSRQ— Appearance Scales (MBSRQ-AS). The MBSRQ-AS assesses five dimensions of body image focused generally on appearance, body, and weight satisfaction and value: appearance evaluation (i.e., feelings about and satisfaction with physical appearance), appearance orientation (i.e., investment in appearance and the importance of appearance), body areas satisfaction (i.e., satisfaction with specific aspects of appearance), overweight preoccupation (i.e., fat anxiety, weight vigilance, efforts aimed at weight and body shape), and self-classified weight (SCW; i.e., identification of being under-, normal-, or over-weight). Using methods such as invariance testing, it is important to determine whether these body image dimensions are interpreted the same way across groups (such as breast cancer survivors and women without breast cancer), or whether there is a systematic bias in the way one group responds to questions about their appearance, body, and weight [10–12]. Invariance testing has been conducted with the original MBSRQ across gender and age [11]. Many of the subscales were invariant [i.e., appearance evaluation, appearance orientation, and overweight preoccupation; 11], yet the authors recommended to avoid assessing body areas satisfaction across different ages to evade systematic biases in comparisons and subsequent implications. This evidence highlights the importance of testing the measurement constraints of study tools prior to making assumptions of equivalency and perhaps biased comparisons. The current study aims to test the invariance of the MBSRQ-AS to determine whether meaningful comparisons can be made using this scale to assess body image among breast cancer survivors and women who have not had breast cancer. If invariance was supported, then comparisons of mean scores between both groups were made.

Method Participants Breast cancer survivors These women (N = 469) ranged in age from 28 to 80 and reported body mass index (BMI) ranging from 15.2 to

Qual Life Res (2010) 19:1171–1180 Table 1 Demographic characteristics for the samples of breast cancer survivors and women without breast cancer

1173

Breast cancer survivors

Women without breast cancer

N

469

385

Age (M, SD)

57.1 ± 7.9 years

55.4 ± 13.5 years

Body mass index (M, SD)

25.8 ± 4.1 kg/m2

25.5 ± 5.2 kg/m2

Caucasian (%)

91.3

89.6

Education (%) \High school

3.0

3.4

High school

9.5

15.3

Some college or university

13.1

23.4

College diploma

33.0

10.4

University undergraduate degree

24.8

25.5

University graduate degree

16.6

21.3

53.1 kg/m2. For this sample, 91.3% of the women identified themselves as Caucasian, 1.9% as Asian, 0.4% as Aboriginal/First Nations, 0.2% as Hispanic, and 5.1% as other. The women tended to be well educated, with 41.4% holding at least a university undergraduate degree. See Table 1 for a summary of these demographic characteristics. Women reported their first breast cancer diagnosis between several months and 32 years prior to the study (M = 7.5, SD = 5.3 years) and treatments (excluding reconstructive surgery and hormonal therapies) completed between several months and 27 years ago (M = 5.9, SD = 4.9 years). They reported the following breast cancer treatments: lymph node dissection (n = 373, 80.0%), lumpectomy (n = 286, 61.4%), single (n = 192, 41.2%) or double (n = 43, 9.2%) mastectomy, reconstructive surgery (n = 103, 22.1%), chemotherapy (n = 268, 57.5%), radiotherapy (n = 297, 63.7%), and hormonal therapy (n = 237, 50.9%). Recurrence rate was 12.7% (n = 59). Women without breast cancer These women (N = 385) are a subset of data from a previous study on body image [11]. Women were selected from the original dataset (N = 840) if they were at least 35 years of age (with the exception of one woman aged 28 who was selected to match the one 28-year-old woman in the breast cancer group), did not report a cancer diagnosis in the last 3 years, and reported having no body scars. This latter criterion was used as a proxy indicator that women had likely not been treated for breast cancer since physical scarring inevitably occurs as a result of breast cancer treatment. They ranged in age from 28 to 89 years and ranged in BMI from 16.1 to 51.7 kg/m2. For this sample, 89.6% of the women identified themselves as Caucasian, 4.7% as East Asian, 1.3% as South Asian, 1.0% as Hispanic, 0.8% as Aboriginal/First Nations, and 2.3% as other. The group of women tended to be well educated, with 46.8% holding at least a university undergraduate degree.

See Table 1 for a summary of these demographic characteristics. Measures Multidimensional Body Self-Relations Questionnaire—Appearance Scales (MBSRQ-AS) The appearance evaluation (AE) subscale (7 items, scores ranging from 1 = definitely disagree to 5 = definitely agree) assesses feelings about physical appearance; higher scores indicate greater satisfaction with appearance. The appearance orientation (AO) subscale (12 items, scores ranging from 1 = definitely disagree to 5 = definitely agree) assesses investment in appearance; higher scores indicate more importance and attention placed on looks and more engagement in grooming activities. The body areas satisfaction (BAS) subscale (9 items, scores ranging from 1 = very dissatisfied to 5 = very satisfied) assesses satisfaction with discrete aspects of appearance; higher scores indicate contentment with more areas of one’s body. The overweight preoccupation (OP) subscale (4 items, scores ranging from 1 = very dissatisfied to 5 = very satisfied and 1 = never to 5 = very often) assesses fat anxiety, weight vigilance, dieting, and eating restraint; higher scores indicate greater weight preoccupation. The selfclassified weight (SCW) subscale (2 items, scores ranging from 1 = very underweight to 5 = very overweight) assesses perception and labeling of weight, from very underweight to very overweight [9, 13]. Since the SCW subscale has only two items, invariance testing could not be conducted on this subscale. Thus, only four of the five subscales of the MBSRQ-AS were investigated. Procedure Both studies used in the current analyses were approved by the University of British Columbia Behavioral Science

123

1174

Research Ethics Board and were conducted in line with the 1964 Declaration of Helsinki. All participants in this analysis provided informed consent. The data for the group of breast cancer survivors were collected from women attending a 10-year international celebration of dragon boating. This sample is part of a larger study examining self-perceptions and stress during breast cancer survivorship [14, 15]. The women were provided consent forms and an anonymous questionnaire package and were asked to mail the survey back to the researchers in a stamped and addressed envelope or to return it to the researchers when complete (who were on hand during the festival at a research booth). Participants completed a questionnaire package including the MBSRQ-AS and measures of selfperceptions and psychological well-being. The group of women without breast cancer were recruited using convenience sampling from the community. They responded to either a web-based survey (n = 236) or a paper and pencil survey (n = 139). The subset of participants used in this analyses had completed the full 69-item MBSRQ, in addition to four other body image and self-concept measures.

Data analysis Descriptive analysis Scale psychometric properties and descriptive statistics were calculated. Group comparisons were conducted to examine whether women with and without breast cancer differed on the following demographic variables: age, BMI, ethnicity, and education. Analysis of variance (ANOVA) was used to compare the two continuous variables (i.e., age and BMI), and chi-square (v2) was used to compare the two nominal variables (i.e., ethnicity and education). Levels of measurement invariance and model evaluation Following recommended guidelines [10, 12, 16] and procedures used to assess body image invariance [17], four levels of measurement invariance were examined to assess each of the MBSRQ-AS subscales in independent models. Configural invariance is an examination of whether the women in each group use the same conceptual framework to answer the items on the scales [10, 12, 16]. To accomplish this level of invariance, the number of factors (in each case one since the MBSRQ-AS subscales were tested in separate models) and the pattern of salient factor loadings are constrained to be equal. Configural invariance is the weakest level of, yet a prerequisite to determine, multigroup invariance.

123

Qual Life Res (2010) 19:1171–1180

The next level of invariance examined was metric invariance. In this step, one determines whether the factor loadings are the same as, or at least proportional to, one another across the groups. Metric invariance suggests that the same unit of measurement is being used for the items across the groups [18] and the groups interpret and respond to the measure in a similar way [10]. This allows one to use the measure to examine structural relationships or correlations with other measures across groups [12]. Importantly, metric invariance cannot rule out whether the scores on items may be systematically biased upward or downward for some groups. Thus, metric invariance demonstrates weak invariance and alone is not sufficient to interpret or compare mean performance across groups. The third level of invariance examined was scalar invariance. With this level, one determines whether the factor loadings and intercepts (or thresholds in ordinal data) are equal across the groups. Scalar invariance suggests that, regardless of their group membership, individuals who have the same value on the latent variable would obtain the same value on the observed variable. Several researchers have argued that evidence of scalar invariance demonstrates strong invariance and is necessary to make mean comparisons across groups [16, 19]. Lack of scalar invariance indicates that there is a bias in how the different groups respond to the items. It has been suggested that, in such cases, some groups may differ in how salient an item is as a marker of the construct [20]. Alternatively, groups may differ in the degree to which they may show certain response styles such as acquiescence [16]. Finally, residual or error variance invariance was examined as a test of strict invariance to determine whether the residual variances for all the items are equal across the two groups [12, 19]. Some researchers argue that this level of invariance may not be a requirement for establishing multigroup invariance [12, 19] since group differences in the residual variances only suggest there is a difference in the reliabilities of the item scores not differences pertaining to measurement bias [19]. Nonetheless, there is some evidence that residual variance invariance is important under conditions of inter-correlations among the item residuals (i.e., a violation of conditional independence) [21, 22]. In this case, item residuals might be a function of unpredictable fluctuations (i.e., random noise) and/or systematic effects of unintentionally measured variables that are also left unmodeled [21, 22]. Therefore, if residual invariance is not met, it could mean that there are different variables affecting the measures across the groups, or the same variables may be functioning differently across the groups [21, 22]. To test for all levels of measurement invariance for each subscale of the MBSRQ-AS, multigroup confirmatory factor analyses (MGCFA) based on mean and covariance

Qual Life Res (2010) 19:1171–1180

structures (MACS) were conducted using PRELIS 2.51 and LISREL 8.5 [23]. To conduct these analyses with the ordinal data obtained from MBSRQ-AS responses, we followed Jo¨reskog’s approach and first set the item thresholds to be equal across the two groups and then the polychoric correlation and asymptotic covariance matrices were estimated. MGCFA with MACS was examined using robust maximum likelihood estimation methods. In each model, the variance of the latent variable was set to 1.0 for identification. Configural invariance was examined for each MBSRQAS subscale first by testing whether the items exhibited significant non-zero loadings on each factor. Although the v2 value is presented, it was not used to assess the overall model fit since v2 values have been found to be sensitive to sample size [24, 25]. Rather, model fit was evaluated by examining a combination of four commonly used goodness-of-fit indices [12]: the root mean square error of approximation [RMSEA; values less than 0.08 indicate acceptable fit; 26], the standardized root mean square residual [SRMR; values less than 0.08 indicate acceptable fit; 11, 25]; the comparative fit index [CFI; values of 0.95 or greater indicate acceptable fit; 19]; and the non-normed fit index [NNFI, also called the Tucker-Lewis Index; values of 0.95 or greater indicate acceptable fit; 19]. Based on these established fit indices, we claimed invariance if the RMSEA values B0.08 or CFI and NNFI values C0.95, and SRMR values B0.08. If configural invariance was met, it served as a baseline model for comparisons with more restricted models. If configural invariance was not tenable, an ordinal factor analysis using the logistic response function (POM) with full information maximum likelihood was conducted in PRELIS to explore whether the proposed unidimensional subscale may contain more than one factor [27, 28]. Since this type of factor analysis is an iterative process, a unidimensional model was tested first, followed by models with increasing number of factors. In all multidimensional models, an oblique (Promax) rotation was used to allow for correlations among the emergent factors within each subscale. As recommended by Stevens [29], a minimum factor loading of 0.40 was used to indicate a pattern of practical significance. Furthermore, an examination of the model fit statistics (likelihood ratio and goodness-of-fit v2, full information coverage ratio, and frequency of responses, see [27]) and the inter-correlations among proposed factors (i.e., low to moderate correlations were preferable) also informed decisions on how many factors to retain. Metric invariance was evaluated by testing whether the matrix of factor loadings was invariant across groups. If metric invariance was not found, testing was stopped. For those models that met the requirements of metric invariance, tests of scalar invariance were conducted. Scalar

1175

invariance was tested by constraining the vector of item intercepts/thresholds across groups and examining model fit. If evidence of invariance at this level was tenable, mean differences in observed scores could be compared, and such differences would reflect true differences between the groups on the latent variable. Specifically, ANOVAs were conducted with group membership (i.e., breast cancer survivor versus women without breast cancer) as the independent variable and the subscale mean score as the dependent variable. If scalar invariance was not supported, further tests (i.e., mean-level differences or residual variance invariance) were not conducted. Finally, setting equality constraints on the residuals in each model tested residual variance invariance. Metric, scalar, and residual variance invariance were tested by hierarchically nesting the models so that systematic comparison tests could be conducted [30]. While the degree of invariance across nested models has been assessed by v2 difference tests (Dv2), Cheung and Rensvold [16] suggest that the Dv2 test is susceptible to sample size and/or model complexity and holds little value in making decisions for multigroup invariance. Based on their recommendations [16], a change in CFI values (DCFI) was used to assess differences between the nested models, where values B|.01| indicated model invariance. The Dv2 was examined but was not used to make practical decisions on invariance of the MBSRQ-AS across groups of women with and without breast cancer.

Results Descriptive statistics Internal consistencies for the body image subscales ranged from a = 0.77 to 0.85 for the breast cancer survivor group and a = 0.82 to 0.88 for the women without breast cancer group, except for the OP subscale (see Table 2). The Cronbach’s alpha coefficient for the OP subscale was lower (a = 0.65 and 0.71, for breast cancer survivors and women without breast cancer, respectively). Results of the ANOVAs indicated that the two groups differed significantly on age, F(1, 848) = 5.28, p = 0.02, g2p \ 0.01. However, the small effect size1 indicates that these differences were not meaningful. There was no significant difference between the groups for BMI, F(1, 816) = 0.84, p = 0.36, g2p \ 0.01. Because of the small samples sizes for women who 1

Effect sizes are reported in addition to the statistical test results to indicate the magnitude of the effect size [42]. Kirk’s [43] criteria for interpreting omega-squared may be appropriately applied to interpreting partial eta-squared (g2p) as follows: trivial effect \ 0.01, small effect = 0.01 to 0.058, medium effect = 0.059 to 0.137, and large effect C 0.137.

123

1176

Qual Life Res (2010) 19:1171–1180

Table 2 Mean (M), standard deviation (SD), reliability coefficients, and ANOVA results for the subscales of the MBSRQ-AS Breast cancer survivors (N = 469)

Women without breast cancer (N = 385)

a

M (SD)

a

M (SD)

F

g2p

Appearance evaluation (AE)

.85

3.36 (0.71)

.88

3.25 (0.84)

3.90*

0.005

Appearance orientation (AO)

.85

3.39 (0.60)

.86

3.48 (0.61)

4.87*

0.006

Overweight preoccupation (OP)a

.66

2.49 (0.82)

.71

2.64 (0.91)





Body areas satisfaction (BAS)a

.77

3.45 (0.57)

.82

3.31 (0.67)





Self-classified weight (SCW)b

.83

3.53 (0.61)

.84

3.48 (0.68)





* p \ .05 a

Comparisons were not made because this scale did not achieve appropriate measurement invariance

b

Comparisons were not made because this scale consists of only two items, and thus could not be tested for measurement invariance

identified their ethnicity as something other than Caucasian, the v2 for ethnicity was run comparing women who were Caucasian versus other. Results of this analysis indicated no significant difference between the groups, v2 (1, n = 854) = 0.67, p = 0.41, Phi = 0.03.2 Finally, the results of the v2 for education indicated a small and significant difference, v2 (5, n = 846) = 68.16, p \ 0.001, Cramer’s V = 0.284 (see footnote 2). Follow-up analyses indicated that the women in the non-cancer group were significantly more likely to have completed some postsecondary education, whereas the women in the cancer group were significantly more likely to have completed a certificate or diploma program. Tests of invariance The results of the measurement invariance tests for all four subscales of the MBSRQ-AS are presented in Table 3. Configural invariance of the AE (appearance evaluation), AO (appearance orientation), and OP (overweight preoccupation) subscales was met since three fit statistics indicated an acceptable fit of the model (i.e., SRMR = 0.06, 0.07, 0.05; NNFI = 0.96, 0.98, 0.95; CFI = 0.98, 0.97, 0.98, respectively). Only the SRMR value met the criteria for acceptable model fit for the BAS (body areas satisfaction). Thus, configural invariance was not supported, and no further invariance testing of the BAS was performed. To explore why configural invariance did not hold for the BAS, an exploratory ordinal factor analysis was conducted separately for breast cancer survivors and women without breast cancer. The results indicated that two factors accounted for the variance of the items for each group (rather than a one-factor or three-factor solution), and the pattern of factor loadings was slightly different. For women without breast cancer, lower torso, mid torso, upper torso, muscle tone, and weight loaded on one factor and face, 2 Phi and Cramer’s V can be interpreted as 0.10 = small effect, 0.30 = medium effect, and 0.50 = large effect.

123

hair, and overall appearance loaded on a second factor. Height did not significantly load on either factor. For breast cancer survivors, items for mid torso, muscle tone, and weight loaded on one factor, and items for face, hair, upper torso, and overall appearance loaded on a second factor. Height and lower torso did not significantly load on either factor. Next, metric invariance for the AE, AO, and OP subscales was tested. For the AE, AO, and OP subscales, the DCFI indicated that metric invariance was supported (see Table 3). Then, scalar invariance for the AE, AO, and OP subscales was examined. For the AE and AO subscales, the DCFI indicated that the hypothesis of invariance was tenable. However, the OP subscale was not invariant at this level based on the DCFI and testing was stopped. Finally, residual variance invariance was tenable for the AE and AO subscales based on the stability of the CFI values. Mean-level differences Table 2 presents the mean and standard deviation for each subscale of the MBSRQ-AS. Since the AE and AO subscales met the requirements for scalar invariance, mean differences between groups were examined. Results indicated a statistically significant difference between the two groups for AE subscale, FAE (1, 848) = 3.90, p = 0.05, g2p = 0.01, and AO subscale, FAO (1, 848) = 4.87, p = 0.03, g2p = 0.01. However, the very small effect sizes indicated that these differences were small and were likely detected as statistically significant because of the large sample sizes. No group comparisons could be made on the OP or BAS given the lack of scalar measurement invariance across the two groups.

Discussion Body image has been associated with quality of life and psychological adaptation following breast cancer [3, 5, 31,

Qual Life Res (2010) 19:1171–1180

1177

Table 3 Goodness-of-fit indices for the Multidimensional Body Self-Relations Questionnaire—Appearance Subscales (MBSRQ-AS) for women with (N = 469) and without (N = 385) breast cancer Model

v2

df

RMSEA

CFI

NNFI

SRMR

DCFI

Invariance

Appearance evaluation Configural

152.12

28

.10

.98

.96

.06



Yes

Metric

186.54*

35

.10

.97

.96

.14

0.01

Yes

Scalar

210.84*

42

.10

.97

.97

.14

0.00

Yes

Residual variance

210.11

49

.09

.97

.97

.16

0.00

Yes

Configural Metric

429.21 464.18*

108 120

.08 .08

.97 .96

.98 .96

.07 .08

– 0.01

Yes Yes

Scalar

508.41*

132

.08

.96

.96

.08

0.00

Yes

Residual variance

533.35*

144

.08

.96

.96

.09

0.00

Yes

Appearance Orientation

Overweight preoccupation Configural

23.89

4

.10

.98

.95

.05



Yes

Metric

38.23*

8

.09

.97

.93

.10

0.01

Yes

Scalar

70.98*

12

.10

.93

.93

.11

0.04

No

54

.11

.93

.91

.07



No

Body areas satisfaction Configural

340.20

* Dv2 per Ddf was significant, p B .01

32]. Therefore, it is important that researchers have valid and reliable measures to assess body image dimensions that are consistently linked to health and well-being. The MBSRQ-AS subscales serve this purpose; however, the psychometric properties of this measure have not been evaluated in a sample of breast cancer survivors nor has the invariance of this measure across samples of cancer and non-cancer groups been examined. This study sought to examine measurement invariance for the MBSRQ-AS across a sample of women who have survived breast cancer and a community sample of women who reported no cancer diagnosis. Results of statistical analyses revealed that configural invariance was tenable for the measures assessing appearance evaluation, appearance orientation, and overweight preoccupation (i.e., AE, AO, and OP subscales). These findings suggest that the measures assess these same respective latent factors in both breast cancer survivors and women without breast cancer. Similar evidence has also been reported across the life span (i.e., age invariance) and for men and women (i.e., gender invariance) [11]. Furthermore, metric, scalar, and residual variance invariance were also tenable for the AE and AO subscales. These findings imply that correlations among these subscales, as well as general mean-level differences in appearance evaluation and appearance orientation, can be examined and interpreted between breast cancer survivors and women without breast cancer without systematic bias being introduced by the measure [10]. Finally, the reliabilities of these two subscales (i.e., AE and AO) did not differ based on group membership.

While the OP subscale revealed configural and metric invariance, there was a lack of scalar invariance. This indicates that there is a bias in how the women without breast cancer and the breast cancer survivors responded to the items on this scale. This bias may be reflective of differences in the salience of items with respect to the construct [20]. For example, the items on the OP scale target fear of being or becoming fat, consciousness of changes in weight, being on a weight-loss diet, and frequency of trying to lose weight via diets or fasting. It is likely that weight issues and the use of weight-loss diets may hold different salience across the groups of women. In a recent study, breast cancer survivors reported concern for making healthy dietary choices as a common stressor following treatments [15]. It may be that this group places emphasis on health and less on weight loss when thinking of dieting—thus the items directed at ‘diets’ may not be as relevant or appropriate to address among breast cancer survivors. Alternatively, the bias may be reflective of group differences in certain response styles more generally [16]. For example, one group may agree or disagree to the items on the OP subscale to a greater extent than the other group (i.e., acquiescence). Further research is necessary to detail possible group differences on overweight preoccupation. Whereas baseline evidence of measurement invariance was confirmed for the three dimensions of the MBSRQ-AS discussed above, the body areas satisfaction (BAS) dimension did not demonstrate configural invariance. This finding suggests that the two groups did not share a common single factor pattern and the same concept was not being measured

123

1178

or configured in the same way for both groups. The nine items on the subscale assessed satisfaction with one’s facial features and complexion, hair (color, thickness, texture), lower torso (buttocks, hips, thighs, legs), mid torso (waist, stomach), upper torso (breasts, shoulders, arms), muscle tone, weight, height, and overall appearance. It could be argued that all items, with the possible exclusion of height, are directly affected by cancer surgery and treatments [33]. Thus, the perceptions of these body parts and how they are connected to one another or configured are likely different (while not necessarily poorer or enhanced) for women who have been treated for breast cancer compared to women who have not had cancer treatments. In follow-up analyses, a twofactor structure was found for both groups, which is consistent with previous research [11]. The most notable difference between the two factor structures was that the lower torso item loaded on one of the factors for the women without breast cancer group but loaded on neither factor for the breast cancer survivor group. The two factors were suggestive of physical appearance and body shape components of body image for the women without breast cancer group. These are common distinctions in body image [8, 34, 35]. For the breast cancer survivors, the first factor included physical appearance attributes but also the upper torso consisting of the breasts, and the second factor was representative of body shape. Further research, perhaps using qualitative methodologies, may be informative of these possible differences in the way body image satisfaction is constructed for women who have experienced breast cancer compared to women who have not shared these experiences of critical illness. In line with emerging research linking body image and quality of life [36, 37], the way in which these distinct body image factors may be differentially associated with women’s quality of life should also be explored. Because configural invariance was not met for this subscale assessing body areas satisfaction, and the two groups did not interpret or conceptualize the construct the same way, it is not appropriate to use this subscale in correlational analyses or to compare the mean scores on the subscale across the groups. Thus, the higher scores reported by the breast cancer survivor group in this study cannot be meaningfully interpreted. Based on existing evidence and the current study findings, researchers are not recommended to use this subscale with middle- or older-adult women [11] and breast cancer survivors. Indeed, the finding of a twofactor structure for the body areas subscale raises questions about the appropriateness of a single ‘satisfaction with body parts’ score for women. Consequently, certain breast cancerspecific body image measures [2, 4] may provide a more reliable and valid assessment of this aspect of body image. Breast cancer survivors reported slightly lower appearance orientation (AO) and slightly higher appearance evaluation (AE) scores compared to the women who have not had

123

Qual Life Res (2010) 19:1171–1180

cancer. The small effect sizes associated with these mean differences, however, suggest that the practical differences between the groups may not be meaningful. Nonetheless, there is qualitative evidence of body image satisfaction among breast cancer survivors resulting primarily from (1) not having a mastectomy or having reconstructive surgery, (2) having no history of body image concerns or placing little emphasis on, or self-worth in, breasts, and/or (3) being more comfortable with one’s body, in spite of cancer, compared to younger years when the physique was more salient [5]. Furthermore, the sample of breast cancer survivors was diagnosed and treated for cancer, on average, more than 5 years prior to the study. Studies on body image among cancer survivors suggest that time since treatment is a moderator in the experience of body image concerns [3, 38]. Specifically, body image is lowest in the months and early years following treatment and improves over time [3, 38]. Therefore, many women from the current study’s sample may be at a time in their survivorship trajectory where body image is no longer threatened. Finally, the breast cancer survivors may be reporting a higher evaluation and less importance placed on their appearance as a result of a shift in priority or self-referent focus. Similar to an aging population, these women may place more emphasis on what the body can do rather than what the body looks like [5, 7]. Having to deal with a health threat such as cancer may initiate a process of adaptation whereby life priorities are directed away from the physical self and toward social relationships, health, and physical functioning [5, 32, 39, 40]. Longitudinal research should be conducted to track body image and related covariates over time. There are some important limitations to discuss with regard to the current study. The data were collected from two separate samples with unique recruitment strategies. Furthermore, the group of women without breast cancer was identified using two criteria, including that they had not been diagnosed with cancer in the last 3 years and reported no physical scars. While we used the latter criteria as a proxy for never having had breast cancer, we cannot be certain that women in this group have never had breast cancer. In recognition of this limitation, additional tests of invariance of the MBSRQ-AS were conducted with a subsample of breast cancer survivors (i.e., diagnosed \ 3 years ago; n = 104) and the sample women without cancer since it could be confirmed that these two groups were nonoverlapping. To this end, the same procedures for invariance testing as described above were performed, and results were identical to those when comparing all breast cancer survivors and women without breast cancer.3 These findings provide support for the appropriateness of the inclusion 3

The results of the supplemental invariance testing are available from the first author.

Qual Life Res (2010) 19:1171–1180

criteria used herein and increase the credibility and confidence of the findings of this study. Additionally, the group of breast cancer survivors was a convenience sample of breast cancer survivors who were attending a 10-year international celebration of dragon boating. They may not be representative of a larger population, thus limiting the generalizability of these findings to other groups of breast cancer survivors. However, it is important to note that the personal and treatment characteristics of these participants are similar to those reported in past research with breast cancer survivors [41]. Finally, while the samples differed slightly on education levels, invariance testing procedures did not allow for these small differences to be controlled for. These limitations aside, the study findings highlight the importance of examining measurement invariance prior to making group-level comparisons. According to Rusticus and Hubley [11], overlooking these analyses ‘‘may impact the validity of conclusions drawn and potentially distort ensuing theory’’ (p. 840). The results of the current study demonstrate that the MBSRQ-AS appearance evaluation and appearance orientation dimensions can be used to (1) assess body image among breast cancer survivors and women without breast cancer and (2) make meaningful comparisons across groups. If these subscales are appropriate given intended research questions, they can be used with confidence with a breast cancer survivor group. However, if research questions are directed at specific body areas satisfaction, the respective subscale on the MBSRQAS does not appear to be an appropriate subscale to use given the lack of support for a unidimensional structure and measurement invariance. The overweight preoccupation subscale can be used to assess body image among women with and without breast cancer but should not be used to make comparisons across groups.

References 1. What is breast cancer? 2009 [cited; Available from: http://www. cancer.ca/canada-wide/about%20cancer/types%20of%20cancer/ what%20is%20breast%20cancer.aspx?sc_lang=en. 2. Baxter, N., Goodwin, P., McLeod, R., et al. (2006). Reliability and validity of the body image after breast cancer questionnaire. Breast Journal, 12, 221–232. 3. Bloom, J. R., Stewart, S. L., Chang, S., et al. (2004). Then and now: Quality of life of young breast cancer survivors. PsychoOncology, 13, 147–160. 4. Dalton, E. J., Rasmussen, V. N., Classen, C. C., et al. (2009). Sexual adjustment and body image scale (SABIS): A new measure for breast cancer patients. Breast Journal, 15, 287–290. 5. McDonough, M. H., Sabiston, C. M., & Crocker, P. R. E. (2008). An interpretive phenomenological examination of psychosocial changes among breast cancer survivors in their first season of dragon boating. Journal of Applied Sport Psychology, 20, 425–440.

1179 6. Cash, T. F., & Pruzinsky, T. (2002). Body images: A handbook of theory, research, and clinical practice. New York: Guilford. 7. Reboussin, B. A., Rejeski, W. J., Martin, K. A., et al. (2000). Correlates of satisfaction with body function and body appearance in middle- and older aged adults: The activity counseling trial (ACT). Psychology & Health, 15, 239–254. 8. Thompson, K., Heinberg, L., Altabe, M., et al. (1999). Exacting beauty: Theory, assessment, method, and treatment of body image disturbance. Washington DC: American Psychological Association. 9. Cash, T. F., Winstead, B. A., & Janda, J. H. (1986). Body image survey report: The great American shape-up. Psychology Today, 30–37. 10. Horn, J. L., & McArdle, J. J. (1992). A practical and theoretical guide to measurement invariance in aging research. Journal of Experimental Aging Research, 18, 117–144. 11. Rusticus, S. A., & Hubley, A. M. (2006). Measurement invariance of the Multidimensional Body Self-Relations Questionnaire: Can we compare across age and gender? Sex Roles, 55, 827–842. 12. Steenkamp, J. E. M., & Baumgartner, H. (1998). Assessing measurement invariance in cross-national consumer research. Journal of Consumer Research, 25, 78–90. 13. Cash, T. F. (2000). The Multidimensional Body Self-Relations Questionnaire users’ manual. Available from: www.body-images. com. 14. Brunet, J., McDonough, M. H., Hadd, V., et al. (2009). The posttraumatic growth inventory: An examination of the factor structure and invariance among breast cancer survivors. PsychoOncology. Available online October 27, 2009. doi:10.1002/pon. 1640. 15. Hadd, V., Sabiston, C. M., McDonough, M. H., et al. (2010). Assessing sources of stress in breast cancer survivors involved in dragon boat: Determining stressor dimensions and links to types of treatment. Journal of Women’s Health (in press). 16. Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodnessof-fit indexes for testing measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal, 9, 233–255. 17. Rusticus, S. A., Hubley, A. M., & Zumbo, B. D. (2008). Measurement invariance on the Appearance Schemas InventoryRevised and the Body Image Quality of Life Inventory across age and gender. Assessment, 15, 60–71. 18. Rock, D. A., Werts, C. E., & Flaugher, R. L. (1978). The use of analysis of covariance structures for comparing the psychometric properties of multiple variables across populations. Behavioral Research, 13, 403–418. 19. Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurement invariance literature: Suggestions, practices, and recommendations for organizational research. Organizational Research Methods, 3, 4–70. 20. Chan, D. (2000). Detection of differential item functioning on the Kirton Adaptation-Innovation inventory using multiple group mean and covariance structure analyses. Multivariate Behavioral Research, 35, 169–199. 21. Deshon, R. P. (2004). Measures are not invariant across groups with error variance homogeneity. Psychology Science, 46, 137–149. 22. Wu, A. D., Li, Z., & Zumbo, B. D. (2007). Decoding the meaning of factorial invariance and updating the practice of multi-group confirmatory factor analysis: A demonstration with TIMSS data. Practical Assessment, Research and Evaluation, 12, 1–26. 23. Joreskog, K. G., & Sorbom, D. (2001). LISREL (Version 8.5). Structural equation modeling [Computer Software]. Chicago: Scientific Software International. 24. Brannick, M. T. (1995). Critical comment on applying covariance structure modeling. Journal of Organizational Behavior, 16, 201–213.

123

1180 25. Kelloway, K. E. (1995). Structural equation modeling in perspective. Journal of Organizational Behavior, 16, 215–224. 26. Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136–162). Beverly Hills, CA: Sage. 27. Joreskog, K. G., & Moustaki, I. (2006). Factor analysis of ordinal variables: A comparison of three approaches. Multivariate Behavioral Research, 36, 347–387. 28. Moustaki, I. (2003). A general class of latent variable models for ordinal manifest variables with covariate effects on the manifest and latent variables. British Journal of Mathematical and Statistical Psychology, 56, 337–357. 29. Stevens, J. (1992). Applied multivariate statistics for the social sciences. Hillside, NJ: Erlbaum. 30. Joreskog, K. G. (1971). Simultaneous factor analysis in several populations. Psychometrika, 36, 409–426. 31. Fobair, P., Stewart, S. L., Chang, S., et al. (2005). Body image and sexual problems in young women with breast cancer. PsychoOncology, 15, 579–594. 32. Sabiston, C. M., McDonough, M. H., & Crocker, P. R. E. (2007). Psycho-social experiences of breast cancer survivors involved in a dragon boat program: Exploring links to positive psychological growth. Journal of Sport & Exercise Psychology, 29, 419–438. 33. Ward, A., Kuta, J., & Sanborn, L. (2003). Breast cancer. In L. M. LeMura, S. P. Von Duvillard, et al. (Eds.), Clinical exercise physiology (pp. 405–419). Philadelphia, PA: Lippincott Williams & Wilkins. 34. Cash, T. F., & Pruzinsky, T. (2002). Body image: A handbook of theory, research, and clinical practice. New York: Guilford.

123

Qual Life Res (2010) 19:1171–1180 35. White, C. A. (2002). Body images in oncology. In T. F. Cash & T. Pruzinsky (Eds.), Body image: A handbook of theory, research, and clinical practice (pp. 379–386). New York: Gilford. 36. Muennig, P., Jia, H., Lee, R., et al. (2008). I think therefore I am: Perceived ideal weight as a determinant of health. American Journal of Public Health, 98, 501–506. 37. Cash, T. F., Jakatdar, T. A., & Flemming Williams, E. (2004). The Body Image Quality of Life Inventory: Further validation with college men and women. Body Image, 1, 279–287. 38. King, M. T., Kenny, P., Shiell, A., et al. (2001). Quality of life three months and one year after first treatment for early stage breast cancer: Influence of treatment and patient characteristics. Quality of Life, 20, 789–800. 39. Tedeschi, R. G., & Calhoun, L. G. (2004). Posttraumatic growth: Conceptual foundations and empirical evidence. Psychological Inquiry, 15, 1–18. 40. Tedeschi, R. G., & Calhoun, L. G. (1995). Trauma and transformation: Growing in the aftermath of suffering. Thousand Oaks, CA: Sage. 41. Cordova, M. J., Cunningham, L. L., Carlson, C. R., et al. (2001). Posttraumatic growth following breast cancer: A controlled comparison study. Health Psychology, 20, 176–185. 42. Zumbo, B. D., & Hubley, A. M. (1998). A note on misconceptions concerning prospective and retrospective power. Journal of the Royal Statistical Society, Series D: The Statistician, 47, 385–388. 43. Kirk, R. E. (1996). Practical significance: A concept whose time has come. Educational and Psychological Measurement, 56, 746–759.