At the time a portion of this research was conducted, Sue P. Heiney was employed as Manager of Psychosocial Oncology at Palmetto Health Cancer Centers in Columbia, SC.
The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health.
The effects of a therapeutic group by teleconference for African American women with breast cancer have not been documented, although the benefits of therapeutic groups for European women are well established. African American women with breast cancer may experience social disconnection, a sense of being cut off from partners, family, and friends because of side effects of treatment and fatalistic beliefs about cancer. A therapeutic group by teleconference may counteract these problems and improve social connection.
A randomized trial design stratified by treatment type was used. Data were collected at baseline, at the end of the intervention, and 16 weeks from baseline. Repeated-measures, fixed-factor analyses of covariance were used for each outcome. The between-subject factors were group and replicate set, and the within-subject factor was time. Physical well being and educational level differed significantly between the 2 groups at baseline and were used as covariates.
The mixed-model analysis of the outcome variables revealed significant changes over time for knowledge (P ≤ .001), with higher scores on knowledge observed for the control group. Group-by-time interactions were observed for fatalism (P = .0276), fear (P = .0163), and social connection (P = .0174) as measured by the Social Well Being subscale from the Functional Assessment of Cancer Treatment-Breast Cancer Version. No group-by-time interaction was observed for social connection as measured by the Social Support Questionnaire. Social connection measured with the Social Well Being subscale improved significantly in the intervention group, whereas fatalism and fear significantly decreased.
Health disparities in African American (AA) women with breast cancer are well documented, including a black/white mortality differential.1 Factors like age and stage of diagnosis, differential/unequal treatment, and tumor biology partially explain the mortality difference.2-4 Another factor, social disconnection (a sense of being cut off from important relationships) has not been studied well but also may decrease survival.5 Conversely, the benefit of social connection in improving health, especially in AAs, is well documented.5-8 Social connection may influence behavior through role modeling, provision of information, practical support to participate in health activities, and emotional support to overcome fear of procedures or results of tests and treatments. Also, social connection improves quality of life and enhances recovery.9-11
After a cancer diagnosis, AA women may experience social disconnection because of cancer stigma, fatalism, treatment-related side effects (especially hair loss and fatigue), cultural role expectations, socioeconomic disadvantages, and family burdens.12-20 Thus, at a time of increased stress, AA women with breast cancer may be isolated from usual informational and emotional support, resulting in decreased social connection. Surprisingly, only a few interventions have been developed specifically for AA women after a breast cancer diagnosis, and none address social connection.21, 22 Given the health and quality-of-life benefits of social connection, it is critically important to develop and test interventions to improve social connection among AA women with breast cancer.23
Over the past 3 decades, more than 25 studies have established that therapeutic groups for women with breast cancer improve social connection.24-26 Unfortunately, the participation of AA women with breast cancer in these studies was <10%.27, 28 Only 1 study was identified that tested a therapeutic group for AA women with breast cancer,21 and no researcher tested a therapeutic group by teleconference. Although little research has been conducted on therapeutic groups by teleconference, researchers have demonstrated that this delivery modality improves mood and social support for physically disabled elderly,29 visually impaired elderly,30 and women with breast cancer.31 The value of the teleconference modality reported in those studies suggests the need for further research, especially among AA women with breast cancer. These women face many barriers to participation in face-to-face groups, including travel and family responsibilities. The current study builds on our pilot work and tests a therapeutic group by teleconference for AA women with breast cancer. To our knowledge, our study is the largest psychosocial intervention study to date of AA women with breast cancer.
MATERIALS AND METHODS
The objective of this study was to test the hypothesis that a therapeutic group delivered by teleconference would increase social connection compared with a control group receiving usual care. Our secondary aims were to decrease fear, isolation, and fatalism and to increase cancer knowledge.
The research tested an intervention based on our pilot,32 which included Yalom's Group Intervention Theory,33 the story work described by Heiney et al, and the therapeutic group work described by Fawzy and Fawzy.31, 34 We previously described the intervention and its evaluation, and a brief summary is provided below.35 We gave the study an acronym name, STORY (Sisters Tell Others and Revive Yourself) to aid in recruitment and study recognition and to define our study purpose. Our intervention was designed to serve as a temporary social connection, providing support and information when relationships within a woman's social network may be strained. Traditionally, in the AA culture, women rely on family and community as a source of strength, coping, and stability. However, a diagnosis of cancer may disrupt social connection because of cancer stigma, fatalistic beliefs about cancer, role expectations, and the side effects of treatment. Because of these negative influences on social connection, AA women with newly diagnosed breast cancer need an alternative social connection that can promote coping with the disease and reconnecting to family and community. The intent of the intervention is to improve social connection through story sharing, examination of hypothetical coping stories, and by providing information about a variety of topics to help cope with disconnection.
The therapeutic group intervention was led by 2 social workers who had experience working with oncology patients and leading support groups. They also received training based on the STORY group leadership manual.
The intervention was delivered by teleconference with participants interacting in real time with each other. The sessions lasted 90 minutes and were held weekly for 8 weeks; then, booster sessions were held every other week for 2 weeks. Session 1 focused on orienting group members about how the conference call would work, getting to know one another, and a brief overview of methods of coping. Sessions 2 through 8 had individual topics, and the order of those topics was determined by focus groups of AA patients and professionals.35 The topics for each session originated from the work of Fawzy and Fawzy and included the following: Session 2, feelings, fears, myths; Session 3, stress management and communication; Session 4, family relationships; Session 5, treatment; Session 6, intimacy and self-image; Session 7, recurrence; and Session 8, social network. Sessions 9 and 10 were booster sessions, and no new content was discussed. In total, 10 sessions were conducted. The sessions followed a script with 3 parts: personal story sharing by the participants, content provided by the group leaders, and examination of hypothetical coping stories. Throughout all sessions, the group leaders encouraged sharing of feelings, identification of common experiences, and connecting with group members.35 We previously reported on intervention integrity35; thus, as a check that each replicate set was exposed to similar leadership, we had 2 experienced group leaders to review the audiotape of each session. The group reviewers used a check sheet to verify that the content was followed and that appropriate leadership strategies were used to build commonality, connection, catharsis, and story—our intervention's active ingredients. If the leaders identified that the leadership script was not adequately followed, then the project coordinator met with the group leaders and pointed out discrepancies.
Usual Psychosocial Care (Description of Control)
Usual psychosocial care was defined as any support used by the patient in the course of cancer treatment. For individuals in the control group, no attempts were made to limit access to or use of psychosocial care available to breast cancer patients in the community or to national educational or supportive resources. Because both control and intervention participants may have attended support groups or received other psychosocial support, we carefully monitored all participant's use of psychosocial resources. At each assessment, participants completed the Psychosocial Support Use Scale (PSUS), which was adapted from the pilot based on Merle Mishel's Type of Information/Support Checklist (University of North Carolina Chapel Hill) and was used with permission. The PSUS contains 13 items to inventory psychosocial support services and programs used by the patient before and during the intervention period. The participants indicated which services or programs were used, and these were summed for a total score. There were no differences in support use at any time point when the control and intervention groups were compared. In addition, 35% of control participants (n = 93) reported attending a breast cancer support group between the baseline assessment and assessment 2 (which occurred after the intervention was completed). Overall, there were no differences in support use at any time point when the control and intervention groups were compared.
Patients were included in the study if they were US-born, English-speaking AA women aged ≥21 years who had been diagnosed with invasive ductal breast cancer within the past 6 months and who underwent lumpectomy. Study participants were restricted to those who underwent lumpectomy for study design and therapeutic reasons. We wanted to increase the internal validity of the study by decreasing variability because of different surgical treatment and its impact on the participants' physical status. Also, we wanted to minimize variability in the intervention sessions, because women who undergo lumpectomy have vastly different recovery experiences than those who undergo mastectomy (women who undergo mastectomy may have more pain and surgical side effects). These differences would have impacted the group climate and content of the sessions. Patients were excluded from the study if they had metastatic disease at diagnosis, a current diagnosis of psychosis or a major cognitive impairment, or a past or current diagnosis of cancer (except basal cell or squamous cell of skin), or if they were participating in another behavioral clinical trial. Participation in a medical treatment trial was not an exclusion factor.
We used a multiple-group, nonblinded, intention-to-treat, randomized trial design to compare the therapeutic group intervention with a control group receiving usual psychosocial care. A maximum of 10 participants were assigned randomly to each of 2 groups during a common time period, with 1 group receiving the intervention and the other receiving usual care. In 15 replicate sets, we enrolled 185 patients between September 13, 2006 and April 7, 2010. Because the study was conducted in replicate sets, final data collection for each set of participants occurred 16 weeks after the first assessment (baseline). The last replicate set began on April 26, 2010 and ended on July 12, 2010.
Setting and Procedures
The study was conducted in 1 Southern state with recruitment emanating from a central recruitment office of the principal investigator. Our study was approved by the Palmetto Health Institutional Review Board (IRB) and by the IRB of each site that generated participants. We previously reported on our recruitment framework and procedures.36
Potential participants were identified from self-referral or from physician referral using a signed release of protected health information or partial waiver of protected health information from the IRB. Screening for eligibility included the date of confirmed pathologic diagnosis (either through the cancer registry or from the participants' medical records using a signed release of protected health information). This date was used to confirm the eligibility criteria of 6 months from diagnosis to study entry, which we defined as verbal consent to participate in the study. After potential participants exceeded the 6-month date, we discontinued efforts to enroll them in the study.
After patients were screened and their eligibility was verified, informed consent was obtained at a location of the participant's choice, usually her home. Because of the potential for low literacy, all consents were read to participants. Participants received a $20 gift card at each assessment as well as a small gift valued at about $3. The gift was in keeping with Southern culture when visiting in the home. Regardless of group assignment, participants received a small gift valued at $1 to $3 by mail over the 16 weeks of the intervention (gifts matched the intervention times: weekly for 8 weeks and every other week for 2 weeks).
Participants were evaluated at 3 time points 1) at baseline before the intervention began, 2) at the end of the intervention period, and 3) 16 weeks from baseline. Participants provided information on their health and medical history at baseline. We did not conduct a medical chart audit to ascertain medical history or data such as stage at diagnosis. Our inclusion/exclusion criteria required verification of certain medical information before enrollment. Outcome variable data were collected at all 3 time points.
After patients were screened for eligibility, consented, and completed the first assessment, they were assigned randomly to either the intervention group or the control group. The randomization was stratified by treatment type (chemotherapy only, radiation therapy only, or chemotherapy and radiation therapy combined). Randomization was completed by the statistician using a computer program. The enrollment, assessment, or treatment team members and the principal investigator were not involved in the randomization. Once randomization was completed, the assignment was sent to the project coordinator, who notified patients by mail and/or telephone of their assignment. See Figure 1 for the flow diagram of the study population. It is noteworthy that, using our recruitment model, we recruited 68% of the eligible women,36 which was a much higher proportion than that reported in most behavioral clinical trials.27, 28
We established 5 outcomes for our study: increased social connection, increased knowledge, decreased fear, decreased isolation, and decreased fatalism. The instruments were chosen for their use in cancer patients and if possible in AA samples.
Social connection was assessed with 2 measures: the Northouse Social Support Questionnaire (SSQ) and the Social Well Being subscale (SWB) from the Functional Assessment of Cancer Therapy-Breast Cancer Version (FACT-B) (version 3). Because of difficulty in locating culturally appropriate measures of social connection, we used 2 measures to operationally define this outcome.
Northouse Social Support Questionnaire
The Northouse SSQ is 24-item scale that measures the amount of perceived emotional connection with partners, family, and nonkin. Moderate construct validity was established when the SSQ was compared with the Family Environment Scale. Original testing indicate that internal consistency was 0.90,37 and subsequent testing indicated internal consistency of 0.9038and 0.87.39, 40 The scale was used previously in a study that included AA women with breast cancer.41 The Cronbach alpha for the current sample was 0.90.
Social Well Being subscale from the Functional Assessment of Cancer Therapy-Breast Cancer Version
We used the 7-item SWB subscale from the FACT-B as a second measure of social connection. This subscale was designed as a measure of social well being in patients to assess outcomes in clinical trials, but we used it as a proxy to measure social connection. The authors acknowledged the limitations of this proxy; however, at the time measurement decisions were made, this measure had been used in an AA sample,41 had demonstrated strong reliability and validity, and the items assessed information about relationships.42 Furthermore, the subscale could be interpreted independently. Lower scores on the SWB indicate poorer functioning. The reliability coefficient for the total scale has been reported as 0.90.42 The Cronbach alpha for the current sample was 0.797.
Cancer knowledge questionnaire
Breast cancer knowledge was measured with the 17-item Breast Cancer Knowledge Scale. This scale uses responses of “true,” “false,” and “don't know” to determine the amount of correct information the participant has about breast cancer symptoms and treatment side effects.43 A response of “don't know” is considered a wrong answer. The score is the percentage of correct answers and ranges from 0% to 100%. The Cronbach alpha was 0.77 from a sample of 163 women under the age of 65 years (C. J. Braden, personal communication). For the current sample, the Cronbach alpha was 0.61.
Tension-Anxiety subscale from the Profile of Mood States-Short Form
Fear was measured with 5 items comprising the Tension-Anxiety (TA) subscale of the Profile of Mood States-Brief (POMS-B).44 We used the TA subscale in our pilot, and it has been used extensively in cancer patients.45 Responses range from “not at all” to “extremely.” The responses range from 0 to 4, with higher scores indicating greater tension and anxiety. Cronbach alpha for the TA subscale was 0.90. Test-retest reliability of the subscale ranged from 0.65 to 0.74. Reliability for the overall short form in our pilot study with a 27% AA sample was 0.88.34 Repeated testing using factor analysis demonstrated validity.46 The Cronbach alpha for the current sample was 0.77.
University of California-Los Angeles Loneliness Scale, version 3
The 20-item University of California-Los Angeles Loneliness Scale measures perceptions of isolation from respondents with such items as, “I feel isolated from others.”47-49 Responses range from 1 to 4, possible scores range from 20 to 80, and higher scores indicate greater loneliness and isolation. Extensive validity testing has reported strong validity.48 The reliability reportedly ranges from 0.89 to 0.94 across samples. The scale has been used in a sample that included AA women with breast cancer in which the Cronbach alpha was 0.93.50 For the current sample, the Cronbach alpha was 0.84.
The Revised Powe Fatalism Inventory
Fatalism was measured using an 11-item scale revised by Mayo for a breast cancer sample51: the Powe Fatalism Inventory Revised. The original tool measures the defining characteristics of fatalism, God's will, and eventual death.52-58 Powe established validity and reliability at 0.84 to 0.87. Other researchers reported reliability as high as 0.89. Scores range from 0 to 11. Mayo et al reported that a score >5 was indicative of high fatalism and Gullate (2009)59 concurred. The Cronbach alpha for the current sample was 0.64.
The study objectives were to determine whether a therapeutic group by teleconference would increase social connection compared with a control group receiving usual care. We used an intention-to-treat analysis plan in which we retained all participant data for the final analysis regardless of trial participation. Our secondary objectives were to decrease fear, isolation, and fatalism and to increase cancer knowledge. Descriptive statistics were obtained for demographic and health data.
Our original power analysis was conducted using published data from the FACT total score.60 This analysis determined that a sample of 216 was estimated to yield 90% power for outcome measures at each time point using a 1-sided hypothesis test with .05 level of significance set as the level of significance. Preliminary data were not available for social connection as measured by the SSQ. We enrolled 185 participants by the end of the trial. Because we had a slightly smaller sample than planned, we conducted a post hoc power analysis of our primary outcome as measured by the SWB subscale from the FACT-B, including the covariates and the fixed effect of wave. By using results from Proc Mix in the SAS statistical software package (version 9.2; SAS Institute, Inc., Cary, NC), we calculated that we had 72% power to detect between-within interaction effects.
We enrolled 185 AA women from 26 of the 46 counties in the state. Our participants ranged in age from 24 years to 83 years (mean ± standard deviation [SD], 56 ± 11.06 years) and reported that they had from zero to 8 comorbid conditions (mean ± SD, 2.07 ± 1.57 comorbid conditions). For complete details of our study population, see Table 1. We attained overall high retention (93%; n = 185). Retention was slighter higher in the control group at 96% (n = 93) than in the intervention group at 91% (n = 92).
Table 1. Study Population Baseline Demographic and Health Characteristicsa
No. of Patients (%)
Control Group, n = 93
Intervention Group, n = 92
Some categories are greater than 100% because of rounding.
Twenty-seven participants chose not to answer; percentages are based on those who did answer.
To examine our randomization, we tested for baseline differences between the control group and the intervention group. At baseline, there were no significant differences between the control and intervention groups on age, partner status, income, comorbid conditions, time from diagnosis, rural versus urban location, attendance at church and related activities, or the number of cancer resources used. There was a significant difference by education (P = .0479), because the usual care group had more education. Also, there was a significant difference between groups on physical functioning (P = .0399), because the intervention group had worse physical functioning. No significant baseline differences were observed between the control and intervention groups for any of the outcome variables (social connection, fear, isolation fatalism, and knowledge).
We previously reported on 1 aspect of intervention integrity: participants' highly positive perceptions of group value.35 In addition, we examined 2 other components of integrity, dosage and purity, to evaluate our intervention. Participant attendance (n = 92) at all 10 group sessions ranged from zero to 10, and the mean ± SD attendance was 6.98 ± 2.68 sessions (median, 8 sessions). This level of attendance indicates that the majority of participants received an adequate dose of the intervention, and only 2 participants received no intervention. We had 2 independent group reviewers to assess the purity of the intervention delivery. Our 21-item purity form, which was adapted from our pilot, evaluated (for each session) adherence to the structural and content components in the manual, leadership strategies, story, and therapeutic factors. The scale contains yes/no responses that evaluate whether or not the leader followed the script. Also, Likert-type items evaluate leaders' attention to group process. After the sessions, the group reviewers met to identify concerns over group leadership. The project coordinator discussed these with the group leaders if needed. Finally, we conducted a validity check on changes in the outcomes that could be attributed to the use of coping resources and support interventions strategies outside our intervention. At each assessment, participants completed the PSUS, which contains 13 items that inventory psychosocial support services and programs used by the patient during the intervention period. The participants indicated which services or programs were used, and these were summed for a total score. There were no differences in support use at any time point at which the control and intervention groups were compared.
The means of the outcome variables at all time points are provided in Table 2. T tests and chi-square analyses determined that, at baseline, the control and intervention groups differed only in terms of education and physical well being. These 2 variables were used as covariates in the repeated-measures, fixed-factor analyses of covariance for each for each outcome variable. We used ProcMix in SAS 9.2 for the analysis described below. The advantage of the use of ProcMix is that it takes into account the value at each measure at each time point. In this model, the between-subject factors were group (intervention or control) and replicate set (from 1 to 15), and the within-subject factor was time (baseline, assessment 2, and assessment 3). All statistical tests were evaluated at an alpha of .05. Table 3 provides results from the repeated-measures, fixed-factor analysis of each outcome with education and physical well being as covariates. The mixed-model analysis of the outcome variables revealed significant changes over time for knowledge (P = .001). Surprisingly, both the control group and the intervention group had significant increases in cancer knowledge over time (P = .0001), but the control group had higher scores on knowledge. Table 2 indicates that the control group had higher knowledge at baseline than the intervention group, and the mean scores of the control group continued to increase at assessments 2 and 3 and remained higher than the intervention group scores. Group-by-time interactions were observed for fatalism (P = .0276), fear (P = .0163), and social connection (P = .0174) measured by the FACT-B SWB subscale. No group-by-time interaction was observed for social connection measured by the SSQ (P = .0076). Social connection measured with the SWB subscale improved significantly in the intervention group, and fatalism and fear significantly decreased.
Table 2. Means and Standard Deviations for Outcomes of Assessments at Time 1, Time 2, and Time 3a
Time 1 (n = 92)
Time 2 (n = 82)
Time 3 (n = 83)
Time 1 (n = 93)
Time 2 (n = 86)
Time 3 (n = 89)
Abbreviations: FACT-B SWB, Functional Assessment of Cancer Treatment-Breast Cancer Social Well Being scale; SD, standard deviation; SSQ, the Northouse Social Support Questionnaire.
The Time 1 assessment was at baseline before the intervention began, the Time 2 assessment was at the end of the intervention period, and the Time 3 assessment occurred 16 weeks from baseline.
Social connection (SWB)
Social connection (SSQ)
Table 3. Repeated Measures Analysis Using ProcMix (Group, Time, and Group-by-Time Interaction) for Each Outcome Variable With Physical Well Being and Education as Covariates (n=185)
P for Wave
P for Group
P for Time
P for Group-by-Time
Abbreviations: FACT-B SWB, Functional Assessment of Cancer Treatment-Breast Cancer Social Well Being scale; SSQ, the Northouse Social Support Questionnaire.
Social connection (SSQ)
Social connection (FACT-B SWB)
To our knowledge, this is the largest controlled behavioral trial of AA women with breast cancer that has tested an intervention to improve social connection.61 Also, only 1 other study on therapeutic groups was identified that focused on a sample of AA women with breast cancer.21 Our study had twice the number of participants and offered a much more accessible intervention (teleconference). Our findings support our hypotheses that a therapeutic group by telephone would significantly increase social connection and decrease fear and fatalism compared with a control group. Our intent was for the intervention to provide a temporary support network for AA women with breast cancer when relationships with their natural network might be strained. We envisioned that, by addressing fear, fatalism, and silence, a triad of barriers related to breast cancer in AA women, the participants would gain confidence and coping skills. We anticipated that these skills could be generalized to the women's natural environment, as suggested by Yalom,33 and would improve the women's relationships with members of their social network. Our data support the effectiveness of a therapeutic group in improving social connection. Thus, we expected that the intervention group would improve and that the control group might decompensate, because control group participants would be exposed to these negative influences without the buffering support from the STORY intervention.
Our methodology of using mixed models is a strength of this study, because as it incorporates baseline differences between groups. Previous interventions to decrease fatalism have focused on education about cancer and demonstrated that cancer education decreased fatalism.57 Our results are somewhat paradoxical; the intervention group had only small gains in cancer knowledge (less than the control group). We note that our examination of the use of support resources, such as nurse navigator or educational materials, did not reveal any differences between the control group and the intervention group. The higher gains in the control group could be attributed to their having higher levels of education than the intervention group. Also puzzling is that previous research reported an association between increased fatalism and higher education and income levels in a sample of community participants who were participating in a colon cancer screening program.55 In our sample, the control group had higher education; but their fatalism scores continued to increase over time. Our findings suggest that a support group may change fatalistic beliefs about cancer through a different mechanism than education. Possibly, the decreased fatalism may be attributed to interaction with other women who had a similar diagnosis and were progressing through treatment. Further research is needed to understand the role that social connection plays in fatalistic beliefs about cancer. Our intervention also decreased fear in the intervention group.
This study has multiple strengths. First, we paid close attention to cultural appropriateness in all phases of our study, from design to implementation, involving a community advisory group of AA cancer survivors and professionals. We included their feedback in the procedures for recruitment, consent, data collection, and the STORY intervention. Furthermore, because of community and professional involvement in the trial, our sample represented a group with lower socioeconomic status than many previous studies and included both rural and urban participants. The participants in our trial represent individuals with high health disparities who have received only minimal attention in other psychosocial studies. Because of our stringent inclusion criteria, our sample was homogeneous. Also, by determining potential covariates, we controlled for their influence on outcomes.
The current study also has several limitations. The results may be generalized only to AA women who undergo lumpectomy. The study had a lower level of power than desired. Although significant, the effect size for the social connection measure (FACT-G SWB subscale) was small. However, data regarding the FACT-G as a measure of trial effectiveness notes that, for this overall scale, 5 to 6 change points would indicate a minimally important difference and, thus, is clinically important.62 Change points for the subscales of the FACT-B were not examined62 for comparison with our results. We determined that the difference in mean score on the FACT-G SWB in the intervention group from baseline to the endpoint was 1.4. By using the distribution-based analyses described by Eton and colleagues, we observed that the 1.4 change likely represents a minimally important difference and is clinically significant. Our recruitment and retention were quite high but did not meet our original recruitment goal. A larger sample size may have yield higher power and more significant differences. Further research through cooperative trials is needed to better explore the benefits of the intervention.
Instrumentation was also a limitation. We had difficulty in locating culturally sensitive and low-literacy instruments with which to measure our concepts. Furthermore, we had difficulty finding instruments that matched the operational definition of our concepts and that had been used with AA samples. Thus, we had slightly lower than desirable reliability measures on our instruments and, in some cases, had to eliminate instruments from any analysis because of very low reliability. In addition, many cancer education scales measure knowledge about breast cancer prevention rather than treatment and side effects. The items on the Cancer Knowledge Questionnaire63 tested general knowledge about breast cancer and may not have adequately discriminated for the information presented in our sessions. Another example of measurement limitation is the use of the SSQ. The repeated-measures analysis of changes in SSQ scores over time between the control and intervention groups was not significant (P = .0076), and the narrow range in means and the large SDs may explain in part this lack of significance. At the time we designed the study, the SSQ seemed to fit conceptually with our outcome (social connection) and was chosen although it had not been used in clinical trials. Future researchers may be able to use our data for power calculations to help determine the clinical importance of results.
Our intervention has overall health implications, because it is known that social connection increases health practices. Thus, our intervention has the potential to continue to benefit women long past their participation. Our study could be improved by measuring health behaviors like adherence to treatment in addition to psychosocial measures. Further research is needed in this population to establish better instrumentation and to extend our research findings.
This project was supported by award R01CA107305 from the National Cancer Institute.
We thank Linda Hazlett, PhD for data management and Rudy Parrish for the original power analysis.