This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Save to my personal folders Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Bordeleau, L. Articles by Goodwin, P. J. Search for Related Content PubMed PubMed Citation Articles by Bordeleau, L. Articles by Goodwin, P. J.

Quality of Life in a Randomized Trial of Group Psychosocial Support in Metastatic Breast Cancer: Overall Effects of the Intervention and an Exploration of Missing Data

Louise Bordeleau, John Paul Szalai, Marguerite Ennis, Molyn Leszcz, Michael Speca, Rami Sela, Richard Doll, Harvey M. Chochinov, Margaret Navarro, Andrew Arnold, Kathleen I. Pritchard, Andrea Bezjak, Hilary A. Llewellyn-Thomas, Carol A. Sawka, Pamela J. Goodwin

From the Departments of Medicine, Research Design and Biostatistics, Psychiatry, Oncology, Psychology, Patient and Family Counseling, Medical Oncology and Radiation Oncology, Center for the Evaluative Clinical Sciences, Community and Family Medicine, Division of Epidemiology at the Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Sunnybrook and Women’s College Health Science Center, Toronto-Sunnybrook Regional Cancer Center, University Health Network-Princess Margaret Hospital, and the Universities of Toronto, Toronto; Ottawa Regional Cancer Center, Ottawa; Hamilton Regional Cancer Center, Hamilton, Ontario; Tom Baker Cancer Center, Calgary; Cross Cancer Institute, Camrose, Alberta; Cancer Care Manitoba, Winnipeg, Manitoba, Canada.

Address reprint requests to Louise Bordeleau, MD, Mount Sinai Hospital, 1286-600 University Avenue, Toronto, Ontario, Canada M5G 1X5; email: lbordeleau{at}mtsinai.on.ca.

	ABSTRACT

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION REFERENCES

 
Purpose: To evaluate the effect of a standardized group psychosocial intervention on health-related quality of life (HrQOL) in women with metastatic breast cancer and to explore the effect of missing data in HrQOL analyses.

Patients and Methods: Between 1993 and 1998, seven Canadian centers randomly assigned 235 eligible women to participate in a weekly, 90-minute, therapist-led support group that adhered to principles of supportive-expressive (SE) therapy or to a control arm (no SE). All women received educational material and any type of medical or psychosocial care deemed necessary. HrQOL data were prospectively collected using the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire C30 (EORTC QLQ-C30) at baseline, 4, 8, and 12 months. The primary HrQOL analyses compared scores in the two study arms. Analyses were limited to women with appropriate baseline HrQOL information (n = 215).

Results: Baseline EORTC QLQ-C30 scores were not different between the two study arms (all P > .05). Primary analysis of all subscales failed to show a significant influence of the intervention on HrQOL (all P > .05). There was a significant deterioration over time in several functional scales of the EORTC QLQ-C30: global (P = .03), physical (P = .0002), role (P = .01), and cognitive functioning (P = .04); and in symptom scales: dyspnea (P = .007), appetite loss (P = .04), and fatigue (P = .003); these changes were independent of randomization allocation. Results were similar in additional analyses of overall HrQOL using a variety of approaches to handling missing data.

Conclusion: Supportive-expressive group therapy in patients with metastatic breast cancer does not appear to influence HrQOL, as measured by the EORTC QLQ-C30.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION REFERENCES

 
PSYCHOSOCIAL INTERVENTIONS have been increasingly used in the care of cancer patients. Interest in such interventions has increased since Spiegel et al¹ reported an unexpected survival benefit of group psychosocial support in women with metastatic breast cancer. Although psychosocial interventions have been reported to improve survival in a small number of prospective trials for other cancers (ie, malignant melanoma, hematologic malignancies, and gastrointestinal cancers),^2–4 such survival benefits have not been replicated in patients with metastatic breast cancer.^5–8

Furthermore, although these interventions can improve a number of psychosocial outcomes by reducing anxiety and improving mood scores,^7,9–13 there has been little systematic assessment, using well-validated, multidimensional questionnaires, of the effect of these psychosocial interventions on overall health-related quality of life (HrQOL) in metastatic breast cancer patients. Marchioro et al¹⁴ reported on the HrQOL outcome of 36 patients with nonmetastatic breast cancer who were randomly assigned to receive either psychological intervention or standard follow-up. Compared with the control group, cognitive psychotherapy and family counseling significantly improved HrQOL, as assessed on the Functional Living Index-Cancer (FLIC) questionnaire, over a follow-up period of 9 months. In the context of metastatic breast cancer, Edmonds et al¹⁵ reported that, after a 14-month follow-up in a randomized trial of group psychological therapy, no improvements in HrQOL were identified on the FLIC questionnaire. Small sample size (n = 66) may have contributed to these negative results.

This minimal investigation of the effect of psychosocial interventions on HrQOL is at odds with the fact that formal evaluation of HrQOL has become an important end point of multiple trials of biomedical treatment (eg, chemotherapy) in oncology. It is a concern because the effect of therapies on HrQOL in patients with noncurable, advanced cancer is fundamentally important for planning optimal treatment and supportive care.

HrQOL has been broadly defined as a person’s own sense of well-being, as derived from his or her current experience of life as a whole. Multidimensional HrQOL measurement in oncology examines the effect of disease and treatment on physical and emotional roles and social functioning and on physical symptoms.¹⁶ On the basis of the psychological benefits of group support reported by others and the nature of the intervention used in our trial, which addressed a wide range of issues in the lives of women with metastatic breast cancer (eg, social isolation, relationship with medical staff, disease-related symptoms, pain control via relaxation exercises, life values), we hypothesized that participation in supportive-expressive therapy might enhance overall HrQOL and, in particular, emotional and social functioning and pain control.

Here, we report the effect of supportive-expressive group therapy in women with metastatic breast cancer on HrQOL using the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire Core Module (EORTC QLQ-C30). Our trial, which was designed to replicate the original study by Spiegel et al,¹ examined whether participation in supportive-expressive group therapy led to prolonged survival in women with metastatic breast cancer. The main trial results have been published elsewhere.⁹ Although significant improvements in mood and in pain perception were observed, as measured by the Profile of Mood States (POMS) and linear analog self-assessment (LASA) scales, respectively, there was no significant survival difference between the intervention and control subjects. We also explore an important methodological issue that is often overlooked in HrQOL studies involving seriously ill patients: that of missing data. Missing questionnaires, which can potentially introduce bias and limit the power of HrQOL studies, complicated the analysis of our data. We describe the extent of missing HrQOL data, explore the effect of various assumptions about types of missing data, and test the effects of various approaches to handling missing HrQOL data.

To our knowledge, this article represents not only the first report of the HrQOL effect of a group psychosocial intervention in a prospective multicenter trial of women with metastatic breast cancer but also one of the first reports to comprehensively explore the potential effect of missing data on these results.

	PATIENTS AND METHODS

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION REFERENCES

 
Setting and Patients
The multicenter randomized trial was coordinated at the Samuel Lunenfeld Research Institute at Mount Sinai Hospital, University of Toronto. Other participating centers included the Hamilton Regional Cancer Center, Ottawa Regional Cancer Center, Manitoba Cancer Treatment and Research Foundation (Winnipeg), Tom Baker Cancer Center (Calgary), Cross Cancer Center (Edmonton), and the British Columbia Cancer Agency (Vancouver). Ethical approval was obtained from all participating centers.

Women were recruited between June 1993 and December 1997. In summary, intense recruitment drives, lasting 2 to 4 months, were held at each participating center to initiate groups. Women were recruited using three approaches: systematic evaluation of all women attending breast cancer follow-up clinics by a designed study nurse, referral of women to the study by clinic staff, and direct recruitment of patients by psychosocial investigators. Women with metastatic breast cancer were eligible if they met all of the following inclusion criteria: histologic confirmation of breast cancer at the time of diagnosis, presence of metastatic disease beyond the breast and ipsilateral axilla before study entry, consent of the most responsible treating physician, and written informed consent of the patient. Women were excluded if they met any of the following exclusion criteria: presence of brain or other CNS metastases or expected survival less than 3 months; active psychosis, untreated major depression, or severe personality disorder; inability to speak and read English; current participation in a long-term (more than 4 months) therapist-led support group; and residence more than 1 hour from the study center.

Intervention
Potentially eligible women were assessed for medical eligibility. They then completed a battery of psychosocial questionnaires and were assessed by group leaders in a semistructured baseline interview to confirm psychological eligibility.¹⁷ Baseline information was obtained on initial diagnosis and treatment and on medical status and treatment after development of metastases.

Randomization was performed centrally, stratified for study center and for the presence of visceral metastases (any v none). A 2:1 randomization ratio (intervention/control) was used. This ratio was necessary to ensure that the support groups were of adequate size; it was incorporated into the sample size calculations.¹⁷ Women randomly assigned to the intervention arm participated in a weekly 90-minute therapist-led support group that adhered to the principles of supportive-expressive therapy.^18,19 Groups were formed initially when at least 12 eligible and consenting women were available for randomization at each center; women randomized to the intervention after groups were formed were added to ongoing groups on an individual basis. Each group consisted of six to 10 women and two leaders; groups continued indefinitely. The leaders were psychiatrists, psychologists, social workers, or nurse clinicians who were experienced in leading group therapy. The supportive-expressive therapy (described in a detailed manual¹⁹) was intended to foster support among group members and to encourage the expression of emotions about cancer and its effects on their lives.⁸ Relationships with family, friends, and the health care team were discussed, the effect of the disease and its treatment were evaluated, and life goals and priorities were re-evaluated. A relaxation exercise was conducted at the end of each seminar. Women were encouraged to practice this at home and to contact each other outside of group meetings. Women were asked to attend the group sessions for at least 1 year or longer if the sessions continued to be of benefit (most continued until they were too ill to attend). Leaders received standardized training before they began leading the groups (a detailed review of the treatment manual, a 2-day workshop, and review of videotapes of the therapy). They also participated in 2-day workshops every 9 to 12 months and received monthly reviews of videotapes of randomly selected sessions. Attendance at group-therapy sessions averaged 66.7% (3,475 of 5,208 possible person-sessions). The major reasons for nonattendance were illness and medical treatments. Participation in off-study support groups after randomization was similar in the two study arms (8.2% intervention, 10.4% control; P = .50).

Women randomly assigned to the control arm did not participate in a support group. Every 6 months, all women received educational materials about breast cancer and its treatment, relaxation, and nutrition. All study participants could receive any medical or psychosocial treatment deemed necessary. Details of the enrollment and the study intervention were published elsewhere.¹⁷

Data Collection
The primary outcome for the trial was survival. HrQOL and a group of defined psychological measures including the POMS were secondary outcomes. As outlined above, we hypothesized that HrQOL (global quality of life [QOL] and social and emotional functioning) would be enhanced by the intervention.

HrQOL was measured using the self-administered EORTC QLQ-C30 (version 1).²⁰ The EORTC QLQ-C30 is a 30-item questionnaire composed of five multi-item functional subscales (physical, role, emotional, social, and cognitive functioning); three multi-item symptom subscales (fatigue, pain, and emesis); a two-item global QOL subscale; and six single-item subscales (financial effect, dyspnea, sleep disturbance, appetite, diarrhea, and constipation). The psychometric properties and responsiveness of the questionnaire in patients with breast cancer have been reported.²¹ The questionnaires were given to patients by the research assistant during baseline assessment, and they were mailed out at planned assessment times (4, 8, and 12 months after randomization).

Data Management and Analysis
Preliminary Steps. For each point in the longitudinal data collection, we specified a timeframe for inclusion of HrQOL questionnaires. Baseline questionnaires were defined as those completed during the 4 months before randomization (n = 215). For each subsequent assessment time, QOL questionnaires completed within 2 months before or following the planned time point were included. When more than one questionnaire was completed in the time frame, double counting was avoided by selecting the questionnaire completed closest to the planned assessment time.

Each EORTC QLQ-C30 single-item and multi-item subscale was scored and linearly transformed according to algorithms recommended by the EORTC. Using two-sample t tests for continuous variables and Pearson’s ² test for categorical variables, the intervention and the control arms were compared in terms of clinical characteristics at the time of first development of metastases and at randomization, and in terms of baseline HrQOL scores.

HrQOL Analyses. Because there were no between-group baseline differences in transformed HrQOL scores, we used absolute rather than change scores in a mixed model for repeated measures (PROC MIXED) in SAS (SAS/STAT User’s Guide, version 6. Cary, NC, SAS Institute, 1990).

Given the occurrence of missing HrQOL questionnaires, a series of exploratory analyses was conducted to establish the robustness of our results. There is no accepted standard for the analysis of a trial with missing HrQOL data, and the results of some, but not all, HrQOL analyses in other cancer studies examining the effect of missing data have varied, depending on the assumptions made about the mechanism of missing data and the method of analysis used.^22,23 Therefore, we used a two-step strategy to proceed with our exploratory analyses.

First, compliance with questionnaire completion was examined. Missingness was classified in terms of missing items (ie, specific item not completed on an otherwise completed form) or missing forms (ie, entire form not completed). The reasons for these missing assessments (apart from death and dropout) were not systematically collected during the course of the study.

Three mechanisms of missing HrQOL data have been described in the literature using a terminology that can be somewhat confusing.²⁴ "Data missing completely at random" are independent of previous, current, and future EORTC scores. This type of missingness may be the result of administrative error, for example, and does not bias the analysis. "Data missing at random" depend on previously observed EORTC scores but not current or future scores. "Data missing not at random" depend on future unobserved EORTC scores. These latter two types of missingness (dependent on previous or future scores) are often attributed to disease and treatment-related morbidity or mortality, which are frequently encountered in patients with advanced cancer. If these two types of missingness are ignored, they can lead to biased conclusions.

Exploratory multivariate logistic regression models were generated using clinically relevant covariates to test for factors predictive of missingness at each time point. Linear regressions were then performed to identify the predictors of actual global QOL scores that could be used to impute missing values in the exploratory analyses.

Second, we carried out a series of analytic approaches, namely, complete case analysis (repeated measures analysis of variance) and available case analysis (mixed model for repeated measures and summary statistics), to compare HrQOL in the two study arms. Analysis of available data using mixed model for repeated measures was our primary approach. Then novel imputation methods incorporating predictors of global QOL scores and predictors of missing data were performed using Solas 3.0 (Statistical Solutions Ltd, Cork, Ireland). These methods included four single imputation techniques (last value carried forward, hot deck, group means, and predicted mean) and two multiple imputation techniques (propensity score-based and predictive model-based methods). These methods are described in a footnote to Table 4.¹⁶ The purpose of these analyses was to examine whether results of our primary analysis changed when missing data were handled in different ways. In all analyses, significance level was set at P = .05 (two-tail), and no adjustment was made for multiple comparisons.

	RESULTS

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION REFERENCES

Participant characteristics are summarized in Table 1. There were no significant differences between the study arms in these baseline characteristics at the time of first development of metastasis and at randomization. As previously reported, median survival times were 17.9 and 17.6 months in the intervention and control arms, respectively (P = .72).⁹ The nonsignificant effect of the intervention on survival remained after adjustment for a number of known prognostic variables; that is, the presence or absence of progesterone or estrogen receptors, time from first metastasis to randomization, age at diagnosis, nodal stage at diagnosis, and use or nonuse of adjuvant chemotherapy (P = .22).

Prediction of global QOL scores. Multivariate linear regression techniques were used to identify predictors of global QOL scores both at baseline and at follow-up. Relevant clinical factors such as previous EORTC global QOL scores, ECOG performance status, age, current chemotherapy, presence or absence of visceral disease, randomization allocation, and long-term survivorship after randomization (ie, 2 years) were included in the models. ECOG performance status at baseline (P = .009) was related to the baseline global QOL scores. The relationship of age was borderline (P = .06), global QOL being marginally better in women more than 50 years old. Follow-up global QOL scores at all measurement points were strongly predicted by previous global QOL scores (P .0008). In addition, a higher (better) global QOL score at 8 months was independently predicted by long-term survivorship after randomization (to the end of study follow-up in October 2000; P = .005). At 12 months, the presence of visceral disease (P = .0002) and administration of chemotherapy during the corresponding time window (10 to 14 months; P = .01) were both associated with lower (worse) global QOL scores, whereas older age (P = .02) was associated with higher global QOL scores. These predictors were used to impute values for missing data in the exploratory analyses described below.

Characterization of the nature of the missing HrQOL data. At all time points, mean global QOL scores were correlated with the total number of HrQOL assessments completed. That is, patients with higher global QOL scores were able to complete more assessments than patients with lower global QOL scores. Factors such as long survival (to the end of study follow-up in October 2000) and imminent death (within 2 months of planned assessment) were also found to be predictors of missing data at 8 and 12 months, respectively, based on a multivariate logistic regression. That is, imminent death was a significant predictor of poor compliance with HrQOL questionnaire completion, whereas prolonged survival was a predictor of good compliance. Taken together, these observations indicate that poor overall HrQOL and imminent death were significant predictors of missing HrQOL data. This explanation of missing data is quite plausible in a patient population with advanced breast cancer; women with worsening health status secondary to disease progression or treatment toxicity were more likely to miss assessments.²³ Because data that are missing due to ill health or imminent death can lead to bias,²⁵ the potential effect of these missing data was addressed in the exploratory analyses described in the next section.

Evaluation of the effect of missing data on study outcomes. A number of exploratory analyses, summarized in Table 4, was then performed to assess the potential effect of missing data on the results of our analysis of the effect of the study intervention on the global QOL subscale of the EORTC QLQ-C30. With complete case analysis, data are assumed to be missing independent of previous scores, general health, or impending death. We have shown this not to be true, as missingness in this study was dependent on both previous and future EORTC scores. These results are presented for completeness, but we do not believe they are valid or generalizable. Available case analysis uses all available data to generate summary statistics (means, SD) and to perform a mixed model for repeated measures (ie, our primary analysis described above). Imputation methods were also used to replace missing data with plausible values. These methods allow missing data to be dependent on previous or future HrQOL, or both, depending on the method used to arrive at an imputed value (eg, long-term survival is used to reflect future HrQOL in the two multiple imputation methods). Four single imputation and two multiple imputation methods were used. The multiple imputation techniques likely represent the most robust analyses because they take into account both predictors of EORTC scores and predictors of missing data.

As can be seen in Table 4, all analyses yielded results that were similar to those of the main analysis. That is, no significant effects of supportive-expressive therapy on global QOL were identified (all P > .05). Global QOL outcomes were not affected by the method used to handle missing data in the analysis. Furthermore, all methods revealed a significant deterioration in global QOL scores over time except for the complete case analysis. The latter analysis used only the subgroup of 25 patients with complete data at all four time points and, by definition, excluded all women who died during the study or were too ill to complete the questionnaires. This led to severely limited power and probable bias.

Overall, there was no evidence in any of these analyses that the intervention affected HrQOL, as measured by the global QOL subscale of the EORTC QLQ-C30. Given the robustness of our results across many different analytic approaches, the conclusion that global QOL is not affected by supportive-expressive therapy in the women participating in this study does not appear to be influenced by missing data.

	DISCUSSION

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION REFERENCES

 
We have reported the effect of a group psychosocial intervention on HrQOL in women with metastatic breast cancer, explored missing HrQOL data, and presented the results of statistical analyses that handle missing data in different ways. The EORTC QLQ-C30 was chosen to measure HrQOL in this trial because it is well validated and has been widely used by a number of cooperative clinical trials groups for the study of biomedical interventions. Its usefulness as a measure of the effectiveness of psychosocial interventions was unproven. This is the first report examining HrQOL using this questionnaire in a large, multicenter, randomized trial of psychosocial intervention of patients with metastatic breast cancer, and it is one of the first to comprehensively examine the effect of missing data and the different approaches to statistical analysis.

Our HrQOL analyses failed to show a significant influence of the psychosocial intervention on global QOL, or any functional/symptom subscale of the EORTC QLQ-C30 questionnaire. However, a significant deterioration over time was observed in a number of functional scales (global, physical functioning, role functioning, and cognitive functioning) and in several symptom scales (dyspnea, appetite loss, and fatigue), regardless of study arm. This deterioration is expected in a patient population with advanced metastatic breast cancer and provides some evidence of the responsiveness of the EORTC QLQ-C30 to clinically important change in this population.

Our failure to identify beneficial effects of supportive-expressive group therapy on HrQOL may reflect the true absence of such an effect. It may also reflect certain characteristics of the EORTC QLQ-C30 questionnaire. This instrument was designed to evaluate the effect of cancer and its treatment on multidimensional QOL, and it has been used mainly in the context of clinical trials involving drug interventions; as such, it has a major focus on physical aspects of HrQOL. It is quite possible that it is insufficiently sensitive to detect change in aspects of QOL that are influenced by supportive-expressive therapy. The social and emotional functioning subscales, both of which measure attributes that should have been influenced by the intervention, contain only two and four items, respectively, and each item has only four response categories. This scaling may have been inadequate to detect beneficial effects of the intervention. This seems probable in the case of emotional functioning because beneficial effects of the intervention were readily identified using the POMS, a 65-item adjectival checklist that provides five response categories for each item and is highly responsive to clinically important changes in mood. The effects seen using the POMS were both statistically and clinically significant, the latter assessed using current definitions of clinical significance.²⁶ Similar arguments can be made for pain: benefits were seen when pain was measured using a 10-cm linear analog scale,⁹ but not when it was measured by a subscale (in the EORTC QLQ-C30) consisting of two items, each having four response categories. Thus, our conclusion that HrQOL as measured by the EORTC QLQ-C30 was not influenced by supportive-expressive group therapy does not undermine the importance of benefits demonstrated using other instruments; in fact, it highlights the need for careful selection of questionnaires to ensure that attributes likely to be influenced by the intervention are targeted and that responsiveness is adequate.

One additional explanation exists for our failure to identify HrQOL effects of supportive-expressive therapy. It is possible that missing HrQOL measurements introduced bias or reduced power. Because of this, we conducted a detailed analysis of reasons for, and effects of, missing data in our study. Although reasons for noncompliance with HrQOL questionnaires were not collected a priori, it is notable that the degree of compliance was similar in the two study arms. Retrospective analysis showed that one of the main reasons for reduced compliance was related to attrition from death and ill health. Data missing for these reasons should not be ignored because they are related to HrQOL outcomes. Therefore, various analytic models were used to explore the effect of various approaches to handling missing data on the HrQOL outcomes. Although some publications found that the use of different analytic models can lead to contradictory results, especially in HrQOL studies of patients with advanced cancer and significant missing data,^22,23 this was not the case in our study. No significant effects were identified using any analytic approaches.

The constancy of our HrQOL results using a wide range of analytic approaches provides evidence for robustness of our conclusion that HrQOL as measured by the EORTC QLQ-C30 is not influenced by supportive-expressive group therapy. It remains possible that a benefit was missed because the overall extent of missing data reduced power. However, we believe we had adequate power to identify clinically important effects. Sample size in this study was calculated on the basis of the estimated survival difference between the groups (primary outcome). A post hoc power calculation was performed to confirm that enough patients were included in the study to detect a clinically significant difference in HrQOL. This calculation was based on available cases (n = 215 patients), a 2:1 randomization ratio, and alpha = .05 (two-tail). We had 82% power to identify a 10-point difference in the global QOL subscale, and power approaching 100% to detect a 20-point difference between study arms. Given that an 8 to 10 point difference is usually considered clinically significant,²⁷ we believe that enough power was present to detect an HrQOL effect from the intervention if one did, in fact, exist. We recommend that the potential for an upper threshold for extent of missing data in clinical trials and its effect on HrQOL be explored, either through statistical modeling or through empiric observation in a spectrum of clinical trials. This may be difficult when data are missing because of death, particularly in trials such as this, when the majority of subjects die. As an initial step, investigation of reasons for missing data and the relationship of these reasons to HrQOL might be most useful, followed by an exploration of the effect of extent of missingness on estimates of HrQOL effects in a variety of interventions. This research may ultimately lead to guidelines for handling missing data in QOL analyses, and it would be useful to assure consistency of approach in this rapidly evolving field of research.

Details of enrollment procedures and outcome have been detailed in a separate publication.¹⁷ The population-based randomization rate estimate of 8.7% seen in this study compares favorably to published population-based estimates of enrollment rates in other studies. As a result, it is likely the generalizability of our study results, at least from a medical perspective, will be equivalent to most published trials of cancer therapy.

Although HrQOL, as measured by the EORTC QLQ-C30 questionnaire, was not influenced by supportive-expressive group therapy in our study, women with metastatic breast cancer clearly benefit from this intervention based on the positive psychosocial results previously reported.^9,13 As noted above, our results highlight the importance of questionnaire selection in research of this type. A cancer-specific instrument such as the EORTC QLQ-C30, which is intended to be used in a wide range of clinical trials involving various types and stages of cancer, may not be sensitive enough to detect important changes in emotional, social, or other domains when psychosocial interventions are delivered. Investigators should ensure that instruments tailored to detect changes in these specific aspects of QOL are included in future trials evaluating psychosocial interventions to make certain that important benefits of the intervention are not missed.

	ACKNOWLEDGMENTS

 
We thank the patients, physicians, nurses, group leaders, and data managers who participated in this study; in particular, Drs David Spiegel and Catherine Classen of Stanford University.

	NOTES

 
This research was funded by the Medical Research Council of Canada (now the Canadian Institutes of Health Research) and the Canadian Breast Cancer Research Initiative.

	REFERENCES

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION REFERENCES

 
1. Spiegel D, Bloom JR, Kraemer HC, et al: Effect of psychosocial treatment on survival of patients with metastatic breast cancer. Lancet 2:888–891, 1989[CrossRef][Medline]

2. Fawzy FI, Fawzy NW, Hyun CS, et al: Malignant melanoma. Effects of an early structured psychiatric intervention, coping, and affective state on recurrence and survival 6 years later. Arch Gen Psychiatry 50:681–689, 1993[Abstract]

3. Kuchler T, Henne-Bruns D, Rappat S, et al: Impact of psychotherapeutic support on gastrointestinal cancer patients undergoing surgery: Survival results of a trial. Hepatogastroenterology 46:322–335, 1999[Medline]

4. Richardson JL, Shelton DR, Krailo M, et al: The effect of compliance with treatment on survival among patients with hematologic malignancies. J Clin Oncol 8:356–364, 1990[Abstract]

5. Bagenal FS, Easton DF, Harris E, et al: Survival of patients with breast cancer attending Bristol Cancer Help Centre. Lancet 336:606–610, 1990[CrossRef][Medline]

6. Cunningham AJ, Edmonds CV, Jenkins GP, et al: A randomized controlled trial of the effects of group psychological therapy on survival in women with metastatic breast cancer. Psychooncology 7:508–517, 1998[CrossRef][Medline]

7. Edelman S, Lemon J, Bell DR, et al: Effects of group CBT on the survival time of patients with metastatic breast cancer. Psychooncology 8:474–481, 1999[CrossRef][Medline]

8. Gellert GA, Maxwell RM, Siegel BS: Survival of breast cancer patients receiving adjunctive psychosocial support therapy: A 10-year follow-up study. J Clin Oncol 11:66–69, 1993[Abstract]

9. Goodwin PJ, Leszcz M, Ennis M, et al: The effect of group psychosocial support on survival in metastatic breast cancer. N Engl J Med 345:1719–1726, 2001[Abstract/Free Full Text]

10. Classen C, Butler LD, Koopman C, et al: Supportive-expressive group therapy and distress in patients with metastatic breast cancer: A randomized clinical intervention trial. Arch Gen Psychiatry 58:494–501, 2001[Abstract/Free Full Text]

11. Fogarty LA, Curbow BA, Wingard JR, et al: Can 40 seconds of compassion reduce patient anxiety? J Clin Oncol 17:371–379, 1999[Abstract/Free Full Text]

12. Fukui S, Kugaya A, Okamura H, et al: A psychosocial group intervention for Japanese women with primary breast carcinoma. Cancer 89:1026–1036, 2000[CrossRef][Medline]

13. Spiegel D, Bloom JR, Yalom I: Group support for patients with metastatic cancer. A randomized outcome study. Arch Gen Psychiatry 38:527–533, 1981[Abstract]

14. Marchioro G, Azzarello G, Checchin F, et al: The impact of a psychological intervention on quality of life in non-metastatic breast cancer. Eur J Cancer 32A:1612–1615, 1996[CrossRef][Medline]

15. Edmonds CV, Lockwood GA, Cunningham AJ: Psychological response to long-term group therapy: A randomized trial with metastatic breast cancer patients. Psychooncology 8:74–91, 1999[CrossRef][Medline]

16. Fayers PM, Machin DE: Missing Data, in Wiley J (ed): Quality of Life Assessment, Analysis and Interpretation. New York, NY, John Wiley & Sons LTD, 2000 pp 224–246

17. Goodwin P, Leszez M, Quirt G, et al: Lessons learned from enrollment in the BEST study. J Clin Epidemiol 53:47–55, 2000[CrossRef][Medline]

18. Leszcz M, Goodwin PJ: The rationale and foundations of group psychotherapy for women with metastatic breast cancer. Int J Group Psychother 48:245–273, 1998[Medline]

19. Spiegel D, Spira J: Supportive-expressive group therapy: A treatment manual of psychosocial intervention for women with recurrent breast cancer; Stanford, CA, Psychosocial Treatment Laboratory, Stanford University School of Medicine, 1991

20. Aaronson NK, Ahmedzai S, Bergman B, et al: The European Organization for Research and Treatment of Cancer QLQ-C30: A quality-of-life instrument for use in international clinical trials in oncology. J Natl Cancer Inst 85:365–376, 1993[Abstract/Free Full Text]

21. Osoba D, Zee B, Pater J, et al: Psychometric properties and responsiveness of the EORTC Quality of Life Questionnaire (QLQ-C30) in patients with breast, ovarian and lung cancer. Qual Life Res 3:353–364, 1994[CrossRef][Medline]

22. Fairclough DL, Peterson HF, Cella D, et al: Comparison of several model-based methods for analyzing incomplete quality of life data in cancer clinical trials. Stat Med 17:781–796, 1998[CrossRef][Medline]

23. Fairclough DL, Peterson HF, Chang V: Why are missing quality of life data a problem in clinical trials of cancer therapy? Stat Med 17:667–677, 1998[CrossRef][Medline]

24. Rubin DB: Inference and missing data. Biometrika 63:581–592, 1976[Abstract/Free Full Text]

25. Diggle P, Kenward MB: Informative drop-out in longitudinal data analysis. Appl Stat 43:49–93, 1995[CrossRef]

26. Osoba D, Rodrigues G, Myles J, et al: Interpreting the significance of changes in health-related quality-of-life scores. J Clin Oncol 16:139–144, 1998[Abstract/Free Full Text]

27. Sloan J: Asking the obvious questions regarding patient burden. J Clin Oncol 20:4–6, 2002[Free Full Text]

Submitted April 10, 2002; accepted February 27, 2003.

This article has been cited by other articles:

				K. E Steinhauser, E. C Clipp, J. C Hays, M. Olsen, R. Arnold, N. A Christakis, J. H. Lindquist, and J. A Tulsky Identifying, recruiting, and retaining seriously-ill patients and their caregivers in longitudinal research Palliative Medicine, December 1, 2006; 20(8): 745 - 754. [Abstract] [PDF]

				Y.M. Chan, P. W.H. Lee, D. Y.T. Fong, A. S.M. Fung, L. Y.F. Wu, A. Y.Y. Choi, T.Y. Ng, H. Y.S. Ngan, and L.C. Wong Effect of Individual Psychological Intervention in Chinese Women With Gynecologic Malignancy: A Randomized Controlled Trial J. Clin. Oncol., August 1, 2005; 23(22): 4913 - 4924. [Abstract] [Full Text] [PDF]

				B. L. Andersen, W. B. Farrar, D. M. Golden-Kreutz, R. Glaser, C. F. Emery, T. R. Crespin, C. L. Shapiro, and W. E. Carson III Psychological, Behavioral, and Immune Changes After a Psychological Intervention: A Clinical Trial J. Clin. Oncol., September 1, 2004; 22(17): 3570 - 3580. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Save to my personal folders Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Bordeleau, L. Articles by Goodwin, P. J. Search for Related Content PubMed PubMed Citation Articles by Bordeleau, L. Articles by Goodwin, P. J.