Randomized Pilot of a Self-Guided Internet Coping Group for Women
With Early-Stage Breast Cancer
Jason E. Owen, Ph.D., M.P.H.
Department of Psychology
Loma Linda University
Joshua C. Klapow, Ph.D.
Department of Psychology
Department of Health Care Organization and Policy
Center for Outcomes and Effectiveness Research and Education
University of Alabama at Birmingham
David L. Roth, Ph.D.
Department of Biostatistics
University of Alabama at Birmingham
John L. Shuster, Jr., M.D.
Department of Psychiatry
University of Alabama School of Medicine
Jeff Bellis, Ph.D.
Department of Health Behavior
University of Alabama at Birmingham
Ron Meredith, Psy.D. and Diane C. Tucker, Ph.D.
Department of Psychology
University of Alabama at Birmingham
Background: Internet-based methods for provision of psy-
chological support and intervention to cancer survivors hold
promise for increasing the public impact of such treatments.
Purpose: The goal of this controlled pilot study was to examine
the effect and potential mechanisms of action of a self-guided,
Internet-based coping-skills training group on quality of life
outcomes in women with early-stage breast cancer. Methods:
Sixty-two women completed baseline evaluations and were ran-
control condition. Results: No main effects for treatment were
cant interaction between baseline self-reported health status
and treatment, such that women with poorer self-perceived
health status showed greater improvement in perceived health
over time when assigned to the treatment condition. Linguistic
analyses revealed that positive changes across quality of life
variables were associated with greater expression of negative
emotions such as sadness and anxiety, greater cognitive pro-
cessing, and lower expression of health-related concerns. Con-
clusions: These results demonstrate the potential efficacy of
self-guided Internet coping groups while highlighting the limi-
tations of such groups.
(Ann Behav Med
An increasing body of research suggests that adjuvant psy-
chological therapies (APTs) for women with breast cancer im-
prove quality of life, yet a number of barriers to utilizing these
services have attenuated the potential public health impact of
these interventions (1,2). Among those who are interested, such
barriers include patient and institutional factors, including dis-
tend regular meetings (4), distance (5), busy home and work
schedules (6), substantial clinician time required to provide ser-
vices, the relative dearth of trained psycho-oncology practitio-
ners (7), and providers’ lack of awareness of existing commu-
nity resources (2). In addition, the overall impact of adjuvant
psychological therapies may be hindered by a relatively poor
understanding of the mechanisms by which such therapies exert
positive influences on quality of life (8). The study of mecha-
nisms of action may provide insight into possible associations
between components of therapy (e.g., provision of treatment el-
ements derived from social-cognitive theory) and subsequent
changes in outcome (9). The use of the Internet to deliver
adjuvant psychological therapies may uniquely address several
These results are part of a larger ongoing study of the Internet as a
source of support and information for women with breast cancer, sup-
ported by Grant No. DAMD17-00-1-0121 from the Department of the
Army Breast Cancer Research Program.
Reprint Address: J. E. Owen, Ph.D., Department of Psychology, Loma
Linda University, 11130 Anderson Street, Loma Linda, CA 92350.
© 2005 by The Society of Behavioral Medicine.
that can be used to examine potential mechanisms of action.
Although research on Internet-based interventions is na-
scent, several studies have documented the potential of the
Internet to increase the impact of psychosocial services for can-
cer survivors. Survivors who are unable to participate in tradi-
tional face-to-face support groups—those who live at great dis-
tance from the clinic (e.g., rural communities), have demanding
are too fatigued or physically debilitated to travel—may be
more likely to participate in Internet-based APTs (10). Consis-
tent with this hypothesis, results from a recent feasibility study
suggest that women with breast cancer express higher levels of
interest in Internet-based support groups than in face-to-face
groups (5). Several studies have suggested that Internet-based
support services, in addition to reducing barriers to obtaining
supportive care, may be efficacious for improving quality of life
outcomes (11,12). A randomized trial of an Internet discussion
group for women with breast cancer demonstrated reduced de-
pression, perceived stress, and cancer-related trauma among
participants in a 12-week facilitated group (13). As a result, a
number of community-based nonprofit organizations are now
providing Internet-based groups and forums (e.g., the Virtual
Wellness Community and the Association of Cancer Online Re-
sources ). In this study we sought to extend the findings of
previous studies by evaluating the efficacy of an online cop-
ing-skills training group for women with early-stage breast can-
cer. In contrast to previous studies, the groups described in this
study were peer based, and coping-skills exercises were pre-
sented in a self-guided format.
An additional aim of this study was to identify mechanisms
portive–expressive group therapy for cancer, emotional expres-
mood disturbance (14,15). However, distinctions between types
of expressed emotion are likely to be important. Giese-Davis et
al. suggested that participation in supportive–expressive group
therapy promotes expression of negative affect while suppress-
and colleagues (16) further postulated that degree of involve-
ment or “psychological work” undertaken in group therapies is
associated with improved emotional well-being among breast
cancer survivors. Use of Internet-based APTs, by recording ev-
ery interaction between the therapist and patient, provides a
unique opportunity to evaluate the effect of the intervention on
the patient’s behavior and, more important, to identify the rela-
tionship between the patient’s behavior and clinically meaning-
naires to determine the mechanisms by which psychological
therapies related to changes in outcome. Internet-based thera-
pies, by cataloguing written exchanges between group mem-
bers, offer the additional opportunity to examine word choices
as measures of psychological processes that may be associated
Pennebaker and colleagues pioneered research on the re-
lationships between linguistic variables associated with word
choices and psychological processes (17). Among individuals
asked to write about emotional or traumatic experiences, im-
proved health has been ascribed to four patterns of word choice:
(a) high use of positive emotion words (e.g., happy, joyful), (b)
moderate use of negative emotion words (e.g., sad, angry), (c)
increasing use of causal and insight words (e.g., because, there-
fore) (18), and (d) change over time in the use of personal pro-
nouns (19). Although there is some evidence that expressive
ments in physical symptoms (20), little is know about the rela-
tionship between word choices and changes in quality of life in
this population. In qualitative analyses, Klemm and colleagues
(21) reported that women with breast cancer, when compared
with men with prostate cancer, make greater use of the Internet
for emotional expression and support seeking. We have recently
applied linguistic analyses to online discussion groups to repli-
cate these findings and have demonstrated that breast cancer
taneously exhibit the positive word choice patterns just de-
tion words (22).
The study presented here employed a randomized, con-
trolled design to pilot the efficacy of a self-guided coping-skills
training and support intervention provided over the Internet to
women with early-stage breast cancer. The intervention, which
self-guided delivery of coping-skills training exercises pre-
sented through a series of Web pages, and education on symp-
tom management, was provided to women primarily receiving
eastern United States. Our overall goal was to determine wheth-
er the use of a self-guided rather than professionally facilitated
intervention could result in similar quality of life improvements
demonstrated in previous reports of face-to-face and online in-
terventions. We hypothesized that participation in the interven-
tion would result in significantly greater overall quality of life
ever, given positive outcomes and high quality of life in women
be significant interactions between baseline levels of distress
and subsequent effects of intervention, such that participants
with high levels of distress prior to the study would be most
likely to benefit from participation.
A secondary aim of the study was to evaluate linguistic pat-
terns predictive of improved quality of life among participants
in the intervention. Given the findings pertaining to word use
efforts, we hypothesized that greater levels of expression of fear
and sadness and greater evidence of cognitive processing in the
of life. In addition, because the intervention itself was designed
discussion of specific medical treatments and procedures, we
Volume 30, Number 1, 2005
Breast Cancer Online Support 55
further hypothesized that higher levels of expression of health-
related concerns (e.g., tumor characteristics, treatments, etc.)
could detract from use of the coping exercises and would be as-
sociated with less improvement in quality of life over time.
At the outset of the study, women with histologically con-
ble for participation in the study. After randomization, it was
discovered that a small number of participants were most likely
Stage 0 or Stage 3 given their self-reported medical histories.
Because these participants expressed a strong desire to be in the
study and believed that they had early-stage breast cancer, they
were included in the study (see Table 1). Women were not ex-
cluded on the basis of medical treatment, time since diagnosis,
or previous psychiatric history. Participants were recruited pri-
marily through direct patient contact with consecutively sched-
uled patients at a Hematology/Oncology outpatient clinic at a
Additional recruitment efforts are described elsewhere in detail
(5). Survivors who expressed an interest in participating in the
teristics of their disease and to administer informed consent. Of
the 154 survivors who expressed initial interest in participating
in the study, 23 (14.9%) elected not to participate after being
given further information about the study, 24 (15.6%) could not
be reached after repeated telephone calls and e-mail messages,
11 (7.1%) did not feel comfortable enough using a computer to
ticipation in a competing trial (see Figure 1). Those participants
who remained interested after speaking by telephone with the
primary investigator (n = 95) provided consent and later re-
ceived a baseline assessment by mail. Participants who com-
pleted the baseline assessment (n = 62; a 65.3% response rate
among eligible and interested participants) were randomized
into one of two conditions: a waiting-list control group (n = 30)
or an Internet-based discussion group (n = 32).
To form small groups for each condition, participants were
recruited in six cohorts of up to 20 survivors. Participants in
each cohort were randomized when the cohort size reached 20
or when the first participant in the cohort had been enrolled for
approximately 4 weeks. Participants were than randomized us-
ing a random number generator to either the wait-list control
condition or the online coping program. Investigators were not
received a follow-up assessment in the mail. Wait-listed partici-
pants were then contacted by telephone to verify interest in re-
sulted in initial treatment group sizes of 5 to 13.
by e-mail and provided with a password for secure access to the
Web site (http://health.psy.uab.edu/survive) and brief instruc-
tions for using the Web site. The Web site for the online coping
port Language, offered a bulletin board for asynchronous group
discussion, a dictionary of medical terminology, a database of
breast cancer resources and Web sites, information and coping
advice for management of common physical symptoms such as
structured coping-skills training exercises (each of which was
presented across a series of Web pages). Each coping exercise
was designed to be completed by participants over the course of
uals used by face-to-face APTs (23,24). Consistent with the
aims of other manualized interventions for cancer survivors,
these coping exercises were primarily intended to facilitate the
identification and expression of cancer-related emotions (e.g.,
and to promote cognitive efforts to more positively reappraise
stressful cancer-related experiences. Coping-skills training ex-
ercises included identification of active and passive coping
styles (e.g., using hypothetical breast-cancer-specific scenarios
to define and recognize active behavioral, active emotional, and
avoidant coping efforts); communication with family and friends
(e.g., focusing on verbal and nonverbal cues that the participant
uses to express needs or changes in mood); identification of re-
ing automatic thought records associated with salient daily
stressors); stress-management training (e.g., use of deep breath-
ing and progressive muscle relaxation); assertiveness training
(e.g., discussing experiences and difficulties associated with
communicating information needs and preferences to health-
fying salient current problems, brainstorming potential solu-
tions, and implementing and evaluating a chosen solution). To
encourage participants to interact with the coping exercises and
to ensure consistency across groups, we also developed a series
of 39 prompts that were automatically sent by e-mail in regular
intervals over the course of the 12-week program. Each prompt
summarized a coping-skills training exercise that could be
found on the Web site and suggested that participants post a
message to the group to discusstheir experiences witheachcop-
ing exercise. Thus, treatment groups were self-guided rather than
facilitated by a group leader. When submitting messages to other
members of their group, participants were identified only by first
name or an alias to protect confidentiality.
Health-related quality of life. Health-relatedqualityoflife
was assessed with the Functional Assessment of Cancer Ther-
apy–Breast Cancer Form (FACT–B) and a quality of life ther-
mometer. The FACT–B is a 27-item questionnaire that utilizes
5-point Likert scales to evaluate overall quality of life and indi-
vidual domains that contribute to the overall score: social well-
being, physical well-being, emotional well-being, functional
well-being, and breast cancer-specific symptoms (25). This in-
strument has adequate internal consistency (overall α = 0.90,
subscale αs = 0.63–0.86) and good concurrent validity with
56Owen et al.
Annals of Behavioral Medicine
Eastern Cooperative Oncology Group (ECOG) performance
status (26). The measure has also been demonstrated to be sen-
sitive to longitudinal change in persons with cancer. The
EuroQol–5D “feeling thermometer” is a single-item visual ana-
logue scale designed to measure self-rated overall health (27).
Participants were asked to rate their overall health on a scale
ranging from 0 (least desirable state of health you can imagine)
to 100 (perfect health). The measure has good test–retest reli-
ability, concurrent validity, and sensitivity to change (28).
Distress. Psychological distress was assessed using the
Impact of Events scale (IES). The IES is a 22-item, Likert-type
scale designed to measure the intrusiveness of and avoidance of
cancer-related thoughts and stimuli (29). The instrument has
good internal consistency (Cronbach’s α = 0.79–0.92) and has
been shown to be sensitive to the effects of psychosocial inter-
Physical well-being. Physical well-being was assessed us-
ing the Memorial Symptom Assessment Scale (MSAS). The
MSAS is a 32-item inventory designed to measure prevalence,
monly reported by those living with cancer (31). Adequate reli-
ability has been reported (0.84–0.88), and the instrument has
both good content and good construct validity. The instrument
Expectations/satisfaction with web site. Comfort with the
Web site and perceived difficulty navigating the site was mea-
sured with thirty 7-point Likert items adapted from the Web
Analysis and Measurement Inventory (32) and 9 open-ended
questions assessing how the Web site could be improved. Wom-
en were also asked to describe their expectations of the online
coping group with regard to whether they would like to primar-
ily give support, obtain support from others with breast cancer,
or both give and receive support.
Quality of participation. Written messages posted to the
online coping group were stored on the host server and were
scored using two separate programs to determine level of ex-
pressions of emotion, cognitive processing of the cancer experi-
ence, and health and treatment-related concerns. Use of words
related to emotional expression and cognitive processing was
measured using Linguistic Inquiry and Word Count (LIWC)
(17,33). LIWC compares each word contained in a text file with
words or word stems classified into 74 categories, and it calcu-
lates the percentage of the total number of words that are con-
categories ranges from 0.86 to 1.00, and external validity of the
emotional expression and cognitive processing word categories
is good (33). In this study, 7 categories were retained for analy-
sis: expression of any affect, positive emotions, negative emo-
tions, anxiety, sadness, anger, and cognitive mechanisms. Cor-
dova et al. (34) described the factor structure of the LIWC
cognitive mechanisms variable as comprising two specific fac-
tors related to uncertainty and logic. We applied the factor load-
ings reported by Cordova et al. to standardized scores on the 6
categories composing the cognitive mechanisms domain to de-
rive scores for two cognitive factors of uncertainty and logic.
Word use related specifically to cancer treatment was deter-
et al. (22). Finally, general concerns related to health were mea-
sured using the automated Gottschalk–Gleser scales (PCAD2000)
for comparison with word-based dictionaries, but PCAD2000
reliability and validity of the computerized scales is promising
(37). Although the computerized scales provide a number of
only the health concerns variable because it has been demon-
strated to be sensitive to the effects of chemotherapy in women
with breast cancer (38).
Using SAS Version 8.00, general linear modeling was em-
ployed for all multivariate analyses of continuous dependent
variables except where otherwise noted. Potential differences
between the two groups on baseline values of all demographic,
medical, and dependent variables were tested using t tests for
continuous variables and chi-square frequency analyses for dis-
crete variables. Multilevel modeling using SAS PROC MIXED
differed across cohorts. Separate intercepts were estimated for
each cohort using a random error term to take into account vari-
ability in cohort-specific intercepts. Significant clustering with-
in treatment cohorts was not observed, and all subsequent anal-
yses employed standard fixed parameter regression models.
Because our sample was composed of participants with rel-
atively high quality of life, we employed an analytic approach
described by previous researchers (39,40) used to test the hy-
pothesis that the intervention group had a stronger impact on
participants most expected to benefit from the intervention (i.e.,
participants with greater self-reported distress). Accordingly,
the interaction between group assignment and baseline value of
the dependent variable onto the Time 2 scores for the dependent
variable. Effects of demographic and medical variables that
were associated with baseline values of dependent variables
were controlled by including them as covariates in these analy-
ses. Baseline values for each dependent variable also were used
as covariates in each model. Main effects of group assignment
were only interpreted when the interaction term was not signifi-
cant (p > .05). To compare our results with those reported else-
were calculated using Cohen’s d (41) and applying the Hedges
guistic variables derived from text were calculated for each day
in which a participant submitted a message to the online group.
Volume 30, Number 1, 2005
Breast Cancer Online Support 57
12-week intervention. Pearson correlations were used to evalu-
ate the relationship between linguistic variables derived from
Figure 1 provides a detailed account of patient enrollment,
allocation, follow-up, and attrition across the course of the
study. Two hundred forty-three consecutive women living with
breast cancer were assessed for eligibility and offered participa-
tion in the study (5). Of these women, 154 expressed interest in
being involved in the study, but only 62 (25.5%) met eligibility
criteria, consented to participate in the study, completed the
baseline assessment battery, and were randomized into one of
the two study arms (treatment, n = 32; control, n = 30). Nine
women were lost to follow-up, leaving 26 participants in the
treatment condition and 27 participants in the control condition
who completed both baseline and follow-up assessments.
Demographic characteristics of the participants are shown
in Table 1. There were no significant baseline differences be-
tween the two groups for any of the demographic or dependent
variables. Participants in the treatment condition were signifi-
cantly more likely than those assigned to the control group to
have received some treatment with a chemotherapeutic agent,
χ2(1, N = 62) = 4.1, p = .04. Participants’motivations for enroll-
in coping with their cancer (74.6%), (b) to provide support to
others (23.7%), and (c) to receive support from others (1.7%).
plaints in the week prior to entering the study were fatigue
(72%), sadness (60%), irritability (57%), worry (57%), poor
concentration (57%), and difficulty sleeping (53%).
Use and Evaluation of the Web Site
Follow-up questionnaires asked participants in the treat-
ticipants found the coping exercises to be most helpful (68.2%
of participants), followed by the coping group/bulletin board
(59.1%) and information about coping with specific symptoms
exercises or the discussion group was helpful. Participants aver-
aged 35.5 total logins to the Web site, 52.2 hits to the bulletin
board, 9.5 postings sent to the bulletin board, 73.4 uses of the
coping exercises, 3.4 hr spent logged in to the Web site, and 8.2
of each component of the intervention over time is displayed in
Effect of Intervention on Quality of Life
Baseline and follow-up scores on all primary dependent
variables are provided in Table 2. No significant main effects of
group assignment were observed for the primary outcome mea-
sures. Trivial or small effect sizes for the main effect of group
were observed for cancer-related intrusive thoughts, symptom
prevalence, self-reported health status, and breast-cancer-spe-
cific concerns. Interactions between group assignment and
baseline quality of life, emotional well-being, breast-cancer-
specific concerns, distress, and symptom prevalence were not
observed. However, a significant interaction emerged between
group assignment and baseline health status (EuroQol-5D) on
follow-up health status, F(1, 39) = 16.4, p < .001, such that par-
ticipants with a low self-reported health status at baseline who
were assigned to the treatment condition (M baseline health sta-
tus = 58.8, M final health status = 81.8) exhibited greater im-
provements in health status over time than did participants as-
signed to the wait-list control group (M baseline health status =
64.4, M final health status = 76.7; see Figure 3). Among those
with better health status at baseline, no significant changes over
time were observed for participants in either the treatment (M
baseline health status = 90.3, M final health status = 88.0) or
control groups (M baseline health status = 90.6, M final health
status = 89.0). Post hoc analyses revealed that time since diag-
nosis was significantly correlated with baseline health status (r
= .26, p = .039) such that women who were farther out from di-
agnosis reported significantly higher levels of health status.
Moderate effect sizes were detected for the intervention for
overall quality of life (FACT–B) and emotional well-being
(FACT Emotional Well-Being subscale), although main effects
of group assignment were not significant at the p < .05 level.
A repeated measures analysis of variance was employed to
test changes over time in quality of life among all participants.
Over time, participants in both groups exhibited improved
life, F(1, 51) = 11.0, p = .0017; and fewer physical symptoms,
F(1, 51) = 6.30, p = .0153. Collapsing across groups, there was
no significant effect of time for emotional well-being, breast
cancer-specific symptoms, or intrusiveness of cancer-related
Quality of Participation and Changes
in Quality of Life
expression, cognitive processing, and health-related concerns
relations between linguistic variables (derived from messages
posted to the online discussion group) and change scores for the
primary dependent variables were calculated. These results are
ety, anger, or overall negative affective states, was associated
with improvements in overall quality of life (r = .49, p = .023).
Greater expressions of anxiety, sadness, and overall negative af-
fect were associated with improved emotional well-being (rs =
.50–.61, ps = .005–.02). Similarly, expressing higher levels of
associated with reductions in intrusive cancer-related thoughts
(rs = .47–.60, ps = .004–.02). Emotional expression was not sig-
nificantly associated with change over time in health status,
physical symptoms, or breast-cancer-related concerns. Depth of
improved emotional well-being (r = .54, p = .012) but was not
associated with overall quality of life, health status, intrusive
58Owen et al.
Annals of Behavioral Medicine
Participant enrollment, allocation, follow-up, and attrition.
Demographic Characteristics of Participants
M ± SD% M ± SD%
Age, SD (years)
Median annual household income ($)
Race (% White)
Marital status (% married)
Distance to clinic (miles)
Employment status (%)
Time since diagnosis (months)
Breast conserving surgery (%)
Modified radical mastectomy (%)
Treated with radiation (%)
Treated with chemotherapy (%)
Clinical stage of disease (%)
Ductal carcinoma in situ
Positive lymph nodes outside of breast (%)
No positive nodes
One or more positive nodes
Use of complementary or alternative medicines (%)
51.3 ± 10.5
65,000 ± 24,407
15.3 ± 2.0
52.5 ± 8.6
65,000 ± 114,148
15.8 ± 2.2
55.2 ± 66.3 107.5 ± 301.5
31.8 ± 37.6 23.2 ± 28.4
an = 30.bn = 32.
*Significant between-group difference, p < .05.
number of messages posted by participants to the support group; Coping = mean number of coping skills training pages opened by participants;
Symptom = mean number of times participants viewed pages associated with information about symptom management; Resource = mean number of
definitions of cancer-related medical terms.
Use of the SURVIVE Web site over the course of the 12-week intervention study by women initially randomized into the treatment
cancer-related thoughts, symptom prevalence, or breast-can-
cer-related concerns. More frequent mention of health-related
concerns in postings to the online discussion groups was associ-
.027), health status (r = –.60, p = .006), intrusive cancer-related
p = .049). More frequent use of words related to cancer treat-
ment was associated with declines in emotional well-being (r =
–.55, p = .01). Use of words related to health and cancer treat-
ment was not associated with stage of disease or treatment vari-
ables. It should be noted that significance levels were not ad-
justed for experiment-wise type I error rate, so individual
correlation coefficients should be interpreted with caution. Fi-
nally, clusters of linguistic variables related to cognitive pro-
cessing (i.e., uncertainty and logic), affective processing (i.e.,
anxiety, sadness, anger, and positive emotions), and health-re-
lated concerns (i.e., health and cancer treatment) were entered
into multiple regression models predicting change scores on the
dependent variables. As a set, these linguistic variables strongly
larly for emotional well-being (R2= .91), F(8, 11) = 6.26, p =
3.17, p = .04 (see Table 3).
As reported elsewhere (5), nearly 45% of women who were
to do so. Only a small percentage of the participants were in-
volved in support services of any kind at the time of study entry,
suggesting that this Internet-based group improved the accessi-
bility of supportive care options to this survivor population.
Many participants expressed difficulty finding or being able to
attend support services in their home communities, and average
travel time to receive breast cancer treatment or follow-up was
over 1 hr. Inherent to most studies of this nature, however, there
were barriers to participation (i.e., the consent process, lengthy
questionnaires to complete). Among those who expressed a
strong interest in participating in the study, only 49% provided
ally completed the baseline assessment. Thus, participation in
this trial may underestimate participation in community-based
programs, if information about such programs was provided
receive treatment and follow-up care.
Results of this pilot intervention provided mixed results for
ing group would result in improved quality of life. No signifi-
cant main effects were observed for the primary dependent vari-
ables. However, there was a trend toward greater improvement
in emotional well-being for treatment relative to control par-
ticipants, and effect sizes for overall quality of life, emotional
well-being, and breast-specific concerns were modestly sized.
In addition, participants with low health status at the begin-
ning of the study who were provided with access to the treat-
ment group demonstrated significantly greater improvement
Volume 30, Number 1, 2005
Breast Cancer Online Support61
Baseline and 12-Week Follow-Up Scores Across Outcome Measures by Treatment Group
F Value, Group × Baseline
Interaction Effect (df)
Effect SizeDomain MeasurePre M SDPost M SDPre MSD Post MSD
quality of life
0.33 (1, 45)
16.40 (1, 39)**
1.00 (1, 45)
1.63 (1, 47)
1.85 (1, 41)
23.15.8 23.9 6.3 22.2 6.522.214.171.124 (1, 48) 0.28
morial Symptom Assessment Scale;
an = 27.bn = 26.
*p < .10. **p < .001.
in self-reported overall health as measured by the EQ-5D thermometer
(Time 2 score – Time 1 score), significant at p < .001.
Interaction between group assignment and improvement
that women living with early-stage breast cancer who feel that
their overall health has been impacted by their cancer, particu-
larly those who have been more recently diagnosed, are most
likely to benefit from participation in this type of intervention.
Our efforts to systematically identify and offer participation to
eligible women may have resulted in a study sample that was
less distressed than samples obtained by other researchers (5).
This hypothesis is supported by our finding that the majority of
to other women and offers additional explanation for the ab-
type to women with identified distress could improve power to
evaluate the effects of intervention.
It is important to note that this was a self-guided interven-
tion in which the investigators monitored but did not participate
in the online discussions. This self-guided approach was se-
lected to address the perceived weaknesses of existing online
and face-to-face resources for cancer survivors: (a) Although
efits for participants, such services are not widely available in
the community for a variety of reasons addressed elsewhere (2),
and (b) although there are many existing Internet-based forums
that promote communication between cancer survivors, these
forums do not provide the types of structured cognitive–behav-
ioral or supportive–expressive exercises that have been demon-
line forums (e.g., widely available and minimizing the need for
intensive clinician involvement) while providing the kinds of
exercises that have been shown to result in quality of life im-
provements. Active facilitation, although more costly, could be
expected to increase the magnitude of effects observed in our
characterize the relationship between levels of participation in
psychosocial intervention for cancer and subsequent changes
over time in quality of life. Consistent with our hypotheses, im-
provements over time in a number of quality of life domains
were associated with three general patterns of word use: greater
ventilation of feelings associated with anxiety and sadness,
greater efforts to cognitively process the cancer experience, and
less expression of somatic and treatment-related concerns. Our
sion mirror previously reported relationships between question-
naire-derived estimates of emotional expression and adjustment
to breast cancer (14,44). Stanton et al. (45) provided a strong
theoretical framework for the relationship between emotional
approach coping and adjustment to stressful situations, suggest-
reappraisal of the stressor, promote social affiliation and sup-
port, and promote awareness of cues related to progress toward
personal goals. General expression of cognitive words, but not
specific linguistic indicators of cognitive processing, was asso-
ciated only with improved emotional well-being. Although sug-
experience may reduce mood disturbance, further research to
tive processing and coping efforts is necessary. Expression of
health-related concerns was associated with worse quality of
life outcomes but was not associated with clinical stage, sug-
gesting that effortful writing about these concerns reflects rumi-
nation or avoidance of alternative cognitive and emotional strat-
egies for adjusting to the challenges associated with cancer. The
high level of variance in change scores across dependent vari-
ables accounted for by the linguistic variables (R2s = .37–.91)
suggests that textual analysis of naturalistic interactions in on-
line support communities has substantial potential for improving
62 Owen et al.
Annals of Behavioral Medicine
Pearson Correlation Coefficients Between Linguistic Variables Associated With Messages Posted to Online Support Groups
and Change Scores on Dependent Variables
of LifeHealth Status
Intrusive Thoughts Symptoms
Specific ConcernsM SD Range
ment on the dependent variable; multiple R2for Time 2 values includes variance associated with Time 1 score for the dependent variable, with significance
tested for the set of linguistic variables after adjusting for Time 1 scores.
*p < .05. **p < .01.
Positive correlation coefficients indicate that a higher proportion of words related to the linguistic variable are associated positively with improve-
our current understanding of the mechanisms by which interven-
tions improve outcomes and warrants further investigation.
Several limitations of our study are noteworthy. First
and foremost, given relatively good baseline quality of life
its the power of the study to detect potential group differences
over time. Although comparable in size to similar studies
(10,11,40,46), a larger sample may be necessary to detect main
effects of treatment for variables that yielded modest effect
sizes. Second, participants were on average 29 months post-
diagnosis, and few of the women were in any form of active
treatment at the time of the study. Levels of distress among
women with early-stage breast cancer have been shown to de-
cline considerably by 4 months postdiagnosis (30), and women
with this disease have among the best longitudinal outcomes re-
lated to depression and anxiety in comparison with other cancer
types (47). Although we believe that broad inclusion criteria are
important to the development of group cohesion and opportuni-
tress (e.g., those with advanced disease, with fewer supports, or
in active treatment) are more likely to participate in and more
likely to benefit from Internet-based psychosocial treatments
(40). Characterization of the recruitment procedures and final
study sample is key to understanding the implications of the re-
of life who sought to provide social support to others; thus, ef-
fect sizes may underestimate effects of this intervention on
women with higher levels of distress. A full 24% of this sample
cal science, there are a great many people living today with a
cancer history. Continuing efforts to identify survivors with
psychosocial needs and to provide accessible, efficacious, and
effective care are needed to reduce their quality of life burden.
Future efforts to identify the relative effects of professional fa-
cilitation, structured coping skills training exercises, and peer
support will be important for the development of next-genera-
tion Internet-based support services.
(1) Greer S: Psychological intervention: The gap between research
and practice. Acta Oncologica. 2002. 41:238–243.
(2) Institute of Medicine National Research Council: Meeting Psy-
chosocial Needs of Women With Breast Cancer. Washington,
DC: National Academies Press, 2004.
(3) Gustafson DH, Wise M, McTavish F, et al.: Development and
with breast cancer. Journal of Psychosocial Oncology. 1993.
(4) Cunningham AJ, Edmonds CVI, Jenkins GP, et al.: A random-
ized controlled trial of the effects of group psychological ther-
apy on survival in women with metastatic breast cancer. Psy-
cho-Oncology. 1998. 7:508–517.
(5) Owen JE, Klapow JC, Roth DL, et al.: Improving the effective-
ness of adjuvant psychological treatment: The feasibility of
providing online support. Psycho-Oncology (in press).
(6) Fukui S, Kugaya A, Kamiya M, et al.: Participation in a psy-
chosocial group intervention among Japanese women with pri-
mary breast cancer and its associated factors. Psycho-Oncol-
ogy. 2001. 10:419–427.
(7) Jacobsen PB, Meade CD, Stein KD, et al.: Efficacy and costs of
dergoing chemotherapy. Journal of Clinical Oncology. 2002.
(8) Owen JE, Klapow JC, Hicken B, Tucker DC: Psychosocial
outcomes model. Psycho-Oncology. 2001. 10:218–230.
of life: A meta-analysis of psychosocial intervention compo-
nents. Health Psychology. 2003. 22:210–219.
(10) Winzelberg A: The analysis of an electronic support group for
individuals with eating disorders. Computers in Human Behav-
ior. 1997. 13:393–407.
port groups for breast carcinoma: A clinical trial of effective-
ness. Cancer. 2003. 97:920–925.
support on younger women with breast cancer. Journal of Gen-
eral Internal Medicine. 2001. 16:435–445.
(13) Winzelberg AJ, Classen C, Alpers GW, et al.: Evaluation of an
Internet support group for women with primary breast cancer.
Cancer. 2003. 97:1164–1173.
(14) Classen C, Koopman C, Angell K, Spiegel D: Coping styles as-
cer. Health Psychology. 1996. 15:434–437.
(15) Giese-Davis J, Koopman C, Butler LD, et al.: Change in
emotion-regulation strategy for women with metastatic
breast cancer following supportive-expressive group ther-
apy. Journal of Consulting and Clinical Psychology. 2002.
in patients with metastatic cancer. Psycho-Oncology. 2000.
individual difference. Journal of Personality and Social Psy-
chology. 1999. 77:1296–1312.
(18) Esterling BA, L’Abate L, Murray EJ, Pennebaker JW: Empiri-
cal foundations for writing in prevention and psychotherapy:
Mental and physical health outcomes. Clinical Psychology Re-
view. 1999. 19:79–96.
(19) Campbell RS, Pennebaker JW: The secret life of pronouns:
Flexibility in writing style and physical health. Psychological
Science. 2003. 14:60–65.
(20) Stanton AL, Danoff-Burg S, Sworowski LA, et al.: Random-
ized, controlled trial of written emotional expression and bene-
fit finding in breast cancer patients. Journal of Clinical Oncol-
ogy. 2002. 20:4160–4168.
(21) Klemm P, Hurst M, Dearholt SL, Trone SR: Gender differences
(22) Owen JE, Klapow JC, Roth DL, Tucker DC: Use of the Internet
for information and support: Disclosure among persons with
breast and prostate cancer. Journal of Behavioral Medicine (in
(23) Fawzy FI, Fawzy NW: A structured psychoeducational inter-
vention for cancer patients. General Hospital Psychiatry. 1994.
Volume 30, Number 1, 2005
Breast Cancer Online Support 63
(24) Nezu AM, Nezu CM, Friedman SH, Faddis S, Houts PS: Download full-text
Helping Cancer Patients Cope. Washington, DC: American
Psychological Association, 1998.
(25) Cella D: F.A.C.I.T. Manual. Chicago: Center on Outcomes, Re-
search, and Education, Evanston Northwestern Healthcare and
Northwestern University, 1997.
(26) Brady MJ, Cella DF, Mo F, et al.: Reliability and validity of
the Functional Assessment of Cancer Therapy-Breast qual-
ity-of-life instrument. Journal of Clinical Oncology. 1997.
(27) Brooks R: EuroQuol: The current state of play. Health Policy.
(28) Llach XB, Herdman M, Schiaffino A, Dipstat A: Determining
correspondence between scores on the EQ-5D “Thermometer”
and a 5-point categorical rating scale. Medical Care. 1999.
(29) Horowitz M, Wilner N, Alvarez W: Impact of Event Scale: A
measure of subjective stress. Psychosomatic Medicine. 1979.
first year after diagnosis. Cancer. 1992. 69:817–828.
(31) Portenoy RK, Thaler HT, Kornblith AB, et al.: The Memorial
Symptom Assessment Scale: An instrument for the evaluation
of symptom prevalence, characteristics, and distress. European
Journal of Cancer. 1994. 30A:1326–1336.
(32) Kirakowski J, Claridge N, Whitehand R: Human centered mea-
sures of success in web design. Proceedings of the 4th Confer-
ence on Human Factors and the Web, Basking Ridge, NJ:
(33) Pennebaker JW, Francis ME, Booth RJ: Linguistic Inquiry and
Word Count: LIWC2001 Manual. Mahwah, NJ: Lawrence Erl-
baum Associates, Inc., 2001.
(34) Cordova MJ, Cunningham LLC, Carlson CR, Andrykowski M:
Social constraints, cognitive processing, and adjustment to
breast cancer. Journal of Consulting and Clinical Psychology.
(35) Gottschalk LA, Bechtel RJ: PCAD2000: Psychiatric content
analysis and diagnosis. Corona del Mar, CA: GB Software,
(36) Lebovits AH, Holland JC: Use of the Gottschalk-Gleser verbal
content analysis scales with medically ill patients. Psychoso-
matic Medicine. 1983. 45:305–320.
(38) Gottschalk LA, Holcombe RF, Jackson D, Bechtel RJ: The ef-
fects of anticancer chemotherapeutic drugs on cognitive func-
tion and other neuropsychiatric dimensions in breast cancer
patients. Methods and Findings in Experimental Clinical Phar-
macology. 2003. 2:117–122.
ventions for women with breast cancer: Who benefits from
what? Health Psychology. 2000. 19:107–114.
ment among African American breast cancer patients: One--
year follow-up results of a randomized psychoeducational group
intervention. Health Psychology. 2003. 22:316–323.
(41) Ray JW, Shadish WR: How interchangeable are different esti-
mators of effect size? Journal of Consulting and Clinical Psy-
chology. 1996. 64:1316–1325.
(42) Hedges LV, Olkin I: Statistical Methods for Meta-Analysis. Or-
lando, FL: Academic, 1985.
(43) Andersen BL: Psychological interventions for cancer patients
to enhance the quality of life. Journal of Consulting and Clini-
cal Psychology. 1992. 60:552–568.
in community cancer support groups: The role of emotional
suppression and fighting spirit. Journal of Psychosomatic Re-
search. 2003. 55:461–467.
(45) Stanton AL, Kirk SB, Cameron CL, Danoff-Burg S: Cop-
ing through emotional approach: Scale construction and vali-
dation. Journal of Personality and Social Psychology. 2000.
(46) Meyer TJ, Mark MM: Statistical power and implications of
meta-analysis for clinical research in psychosocial oncology.
Journal of Psychosomatic Research. 1996. 41:409–413.
(47) Spijker AV, Trijsburg RW, Duivenvoorden HJ: Psychological
ies after 1980. Psychosomatic Medicine. 1997. 59:280–293.
64Owen et al.
Annals of Behavioral Medicine